Vous êtes sur la page 1sur 81

Arrt Cardiaque au Bloc Opratoire

Jean-Luc Hanouz
Ple Ranimations Anesthsie SAMU
CHU de Caen
hanouz-jl@chu-caen.fr

Mortalit pri-opratoire en Europe : 4% (1 22)


3%

Programm
Non programm

5%

Urgence

Risque lev

10%

Dcs

13%

84%
87%

16%
Pearse RM et al. Lancet 2012;380:1059-65
Pearse RM et al. Crit Care 2006;10:R81
Report of the National Confidential Enquiry into Patient Outcome and Death 2011

230 millions dactes chirurgicaux par an


(mortalit pri-opratoire de 1 plus de 20%)

3M

4M
8M
8M
3M

63 M

23M

15 M
4M
2M

En 2020 proportion des plus de 60ans > proportion des moins de 5ans

Weiser TG et al. Lancet 2008;372:139-44

Mortalit imputable lanesthsie dans le monde


(mta analyse de 87 tudes pidmiologiques)

Avant 1970
Dcs imputables
3,57 pour 104

1970 - 1980

1990 - 2000

Dcs imputables
0,52 pour 104

Dcs imputables
0,34 pour 104

Pays riches 0,32


Pays pauvres 1,01

Pays riches 0,25


Pays pauvres 1,41

Bainbridge D et al. Lancet 2012;380:1075-81

Mortalit imputable lanesthsie dans le monde


(mta analyse de 87 tudes pidmiologiques)

Avant 1970
Dcs imputables
3,57 pour 104

Evnements pour 10 000


ASA 1
ASA 2
ASA 3

5,57
14,08
93,69

ASA 4

617,97

1970 - 1980

1990 - 2000

Dcs imputables
0,52 pour 104

Dcs imputables
0,34 pour 104

Pays riches 0,32


Pays pauvres 1,01

Pays riches 0,25


Pays pauvres 1,41

Bainbridge D et al. Lancet 2012;380:1075-81

Mortalit imputable lanesthsie dans le monde


(mta analyse de 87 tudes pidmiologiques)

1,0 vnements pour 10 000

0,1 vnements pour 10 000

Bainbridge D et al. Lancet 2012;380:1075-81

Mortalit imputable lanesthsie


(USA 1999-2005 - Rpartition selon les ges des patients)

Mortalit imputable lanesthsie en France


(imputabilit partielle et totale)

1978 - 1982
Dcs imputables
3,38 pour 104
Totalement : 0,76
Partiellement : 2,62

Evnements pour 10 000


ASA 1

0,04

ASA 2

0,5

ASA 3

2,71

ASA 4

5,54

1996 - 1999
Dcs imputables
0,54 pour 104

54% de dcs jugs vitables

Totalement : 0,07
Partiellement : 0,47

Haut Comit de la Sant Publique. Rapport sur la scurit anesthsique. 1994, ditions Ecole Nationale de la Sant Publique
Lienhart A et al. Anesthesiology 2006;105:1087-97

Arrt circulatoire per-opratoire dans le monde

1,1 / 104
1,0 / 104

0,7 / 104
0,6 / 104

26 / 104
12 / 104
35 / 104
22 / 104

5,8 / 104
6,2 / 104

Evnement rare < 1 pour 10 000 anesthsie dans les pays riches

Bainbridge D et al. Lancet 2012;380:1075-81


Ellis SJ et al. Anesthesiology 2014;120:829-38
Rukewe A et al. Niger J Clin Pract 2014;17:28-31
Braz LG et al. Br J Anaesth 2006;96:569-75
Biboulet P et al. Can J Anaesth 2001;48:326-32

Arrt circulatoire per-opratoire dans le monde


1970 - 1980
Arrt circulatoire
18,72 pour 104
Pays riches 17,98
Pays pauvres 36,42

Evnements pour 10 000


ASA 1
ASA 2

1,93
11,12

ASA 3

59,36

ASA 4

1990 - 2000
Arrt circulatoire
7,19 pour 104
Pays riches 6,79
Pays pauvres 20,68

Bainbridge D et al. Lancet 2012;380:1075-81

75,19

Arrt circulatoire per-opratoire en France


(tude pidmiologie 1989 1995 monocentrique)

Incidence globale estime : 1,1 pour 104


Evnements pour 10 000
ASA 1
ASA 2
ASA 3
ASA 4

0,2
0,7
6,6
19,2

Biboulet P et al. Can J Anaesth 2001;48:326-32


Lienhart et al. Anesthesiology 2006;105:1087-97

Mortalit pri-opratoire en France


(donnes de lenqute mortalit SFAR INSERM - 1999)

Auroy Y, Thse dUniversit soutenue Paris en 2008


Lienhart et al. Anesthesiology 2006;105:1087-97

Mortalit pri-opratoire en France


(donnes de lenqute mortalit SFAR INSERM - 1999)

> Arrt cardiaque et dcs pri-opratoires


Les dfaut de soins le plus souvent identifis

Ranimation peropratoire (18%)


Evaluation et prparation propratoire (16%)
Soins postopratoires (13%)
Gestion des pertes sanguines (12%)
Gestion de linduction de lanesthsie (11%)
Ralisation de la technique danesthsie (8%)
Tenue des documents (6%)
Surveillance en SSPI (5%)
Monitorage peranesthsique (3%)
Auroy Y, Thse dUniversit soutenue Paris en 2008

Mortalit pri-opratoire en France


(donnes de lenqute mortalit SFAR INSERM - 1999)

Proportion de dcs jug vitables selon limputabilit de lanesthsie


Partielle
Totale

50%
92%

Auroy Y, Thse dUniversit soutenue Paris en 2008

Mortalit pri-opratoire en France


(donnes de lenqute mortalit SFAR INSERM - 1999)

Proportion de dcs jug vitables selon limputabilit de lanesthsie


Partielle

50%
92%

Totale
Ecart de soins rencontrs pour
lensemble des dcs

Ecart de soins rencontrs pour les


dcs avec facteur anesth. vitable

Ra. Peropratoire

46%

Ra. Peropratoire

66%

Soins PostOp.

42%

Soins PostOp.

70%

Phase PrOp.

38%

Phase PrOp.

Pertes sanguines

37%

Pertes sanguines

Induction

34%

Induction

97%

Technique

99%

Technique

26%

Auroy Y, Thse dUniversit soutenue Paris en 2008

96%
78%

Mortalit pri-opratoire en France


(donnes de lenqute mortalit SFAR INSERM - 1999)

> Arrt cardiaque et dcs pri-opratoires


Les causes racines identifies dans les dcs jugs vitables
Causes Organisationnelles (53%)

Garde et transfert de tches (100%)


Organisation et utilisation de la SSPI (87%)
Renfort non disponible(83%)
Pression production / programmation (71%)

Causes li au personnel danesthsie (48%)


Vigilance, Comptence, jugement (60%)

Causes structurelles (32%)


Structure ou service inappropris (97%) et ressources insuffisantes (61%)

Equipements (32%)
Tches (32%)
Ralisation (97%), systme documentaire (96%) et surcharge (75%)

Equipe (27%)
Mobilisation des ressources (100%)
Communication (61%)
Auroy Y, Thse dUniversit soutenue Paris en 2008

Donnes pidmiologiques rcentes


(362 767 actes de chirurgie non cardiaque entre 2005 et 2007)

> 262 arrts circulatoires peropratoires


Epidmiologie gnrale
7,22 vnements pour 10 000
Mortalit 24h : 44%
Mortalit 30 jours : 63%
Caractristiques associes la survenue de larrt circulatoire

Age
Classe ASA
Comorbidits propratoires
Dpendance propratoire
Risque chirurgical
Urgence
Etat de choc et troubles de conscience propratoire
Transfusion pr et peropratoire
Goswami S et al Anesthesiology 2012;117:948-50

Donnes pidmiologiques rcentes : facteurs de risques


(362 767 actes de chirurgie non cardiaque entre 2005 et 2007)

Age > 70 ans


ASA 3

Odds Ratio estims associs la survenue dun arrt circulatoire


peropratoire
1,17
3,71

ASA 4

8,75

Chirurgie risque

1,11

Urgence

2,04

Dpendance

2,33

Cardiaque
CGR 1-3
CGR 4-6
CGR 6-9
CGR > 10

1,48
2,51
7,59
11,40
29,80
Goswami S et al Anesthesiology 2012;117:948-50

Donnes pidmiologiques rcentes : facteurs de risques


(362 767 actes de chirurgie non cardiaque entre 2005 et 2007)

Age > 70 ans


ASA 3

Odds Ratio estims associs la survenue dun arrt circulatoire


peropratoire
1,17
3,71

ASA 4

8,75

Chirurgie risque

1,11

Urgence

2,04

Dpendance

2,33

Cardiaque
CGR 1-3
CGR 4-6
CGR 6-9
CGR > 10

1,48
2,51
7,59
11,40
29,80
Goswami S et al Anesthesiology 2012;117:948-50

Donnes pidmiologiques rcentes : cause anesthsique


(217 365 actes de chirurgie entre 1999 et 2009)

> 160 arrts circulatoires dans les 24h dune anesthsie


Epidmiologie gnrale

7,40 vnements pour 10 000


Mortalit 24h : 29% pour les cas imputables lanesthsie
14 cas (9%) sur 160 totalement attribuable lanesthsie
23 cas (14%) partiellement attribuable lanesthsie

Caractristiques associes

Age, sexe
Classe ASA
Type de chirurgie (abdomen, thoracique, rachis)
Urgence
Dure de lintervention
Horaire (15h 7h)
Technique danesthsie
Ellis SJ et al Anesthesiology 2014;120:829-38

Donnes pidmiologiques rcentes : cause anesthsique


(217 365 actes de chirurgie entre 1999 et 2009)

Hmodynamique
Technique (cathter central)
Technique (cathter central)
Voies ariennes
Voies ariennes
Voies ariennes
Voies ariennes
Voies ariennes
Voies ariennes
Voies ariennes
Voies ariennes
Hmodynamique
Neurologique (ALR)
Voies ariennes

Donnes pidmiologiques rcentes : cause anesthsique


(217 365 actes de chirurgie entre 1999 et 2009)

Donnes pidmiologiques rcentes : cause anesthsique


(217 365 actes de chirurgie entre 1999 et 2009)

> 160 arrts circulatoires dans les 24h dune anesthsie


23 (14%) Cas partiellement attribuables lanesthsie

Hmodynamique (70%)
Voies ariennes (13%)
1 cas (4%) de raction la protamine
5 cas (22%) survenu dans le cadre dune prise en charge en urgence

123 (77%) cas non attribuable lanesthsie

Traumatismes graves
Insuffisance hpatocellulaire / transplantation hpatique
Sevrage de CEC impossible
Complications techniques chirurgicales
Anvrysme de laorte thoracique ou abdominale
Chirurgie cardiaque
Sepsis, hmorragie...
Chirurgie carcinologique ou transplantation
Ellis SJ et al Anesthesiology 2014;120:829-38

Donnes pidmiologiques rcentes : rythme initial


(2 524 arrt circulatoires pri-opratoires (24h) multicentrique entre 2000 et 2008)

Proportion sur 2524 arrts circulatoires dans les 24h postopratoires


Lieu

SSPI
10%

Salle dopration 58%

Rythme
RACS

Services 32%

Asystolie (39%)

AESP (37%)

FV /TVSP (24%)

57%

57%

66%

Cause immdiate
Trouble du rythme
Hypotension
Insuf. Respiratoire

59%
52%
29%

Isch. Myocardique

7%

Voies ariennes

6%

Ramachandran SK et al Anesthesiology 2013;119:1322-39

Donnes pidmiologiques rcentes : survie


(2 524 arrt circulatoires pri-opratoires (24h) multicentrique entre 2000 et 2008)

Proportion sur 2524 arrts circulatoires dans les 24h postopratoires


Lieu

Salle dopration 58%

Sortie H

Services 32%

Asystolie (39%)

AESP (37%)

FV /TVSP (24%)

RACS

57%

57%

66%

Sortie H

31%

26%

42%

CPC 1

65%

62%

65%

Rythme

Survie 24h

SSPI
10%

46% 24h
32%

64% CPC 1 (pas de squelles majeures)

Ramachandran SK et al Anesthesiology 2013;119:1322-39

Donnes pidmiologiques rcentes : pronostic


(2 524 arrt circulatoires pri-opratoires (24h) multicentrique entre 2000 et 2008)

> Principaux facteurs pronostics indpendants lis la survie


Facteurs indpendants de mauvais pronostic

Hypotension
Mtastases maladies hmatologique
Insuffisance cardiaque congestive, ischmie myocardique aigu
Insuffisance rnale
Septicmie
Survenue en garde
Age

Facteurs de bon pronostic

Problme de voies ariennes


Survenue en salle dopration ou en SSPI ou en service scop
Troubles du rythme
Fibrillation ventriculaire
Ramachandran SK et al Anesthesiology 2013;119:1322-39

Donnes pidmiologiques rcentes : pronostic


(2 524 arrt circulatoires pri-opratoires (24h) multicentrique entre 2000 et 2008)

Ramachandran SK et al Anesthesiology 2013;119:1322-39

Evnements particuliers au bloc opratoire


(disposer de protocoles locaux connus, diffuss et accessibles)

> Trois circonstances particulires


Toxicit des anesthsiques locaux
Intralipides intraveineux
Intralipides 20% : 3 ml.kg-1
Medialipide : 6 9 ml.kg-1
Diminuer les doses dadrnaline (< 1 g.kg-1)
Ne pas administrer damiodarone, dinhibiteurs calcique, de
betabloqueurs
Choc anaphylactique grade 4
Adrnaline
Remplissage vasculaire : cristalloides isotoniques (30ml.kg-1)
Hyperthermie maligne
Dantrolne : 2,5 mg.kg-1

Dernires Recommandations : 2010

http://circ.ahajournals.org/

http://www.erc.edu/

Rsum des principaux changements


Basic Life Support
- Interrogatoire standardis des
appelants pour la reconnaissance
prcoce de larrt cardiaque
- Compressions thoraciques par les
tmoins guides par tlphone
- Utilisation doutils spcifiques pour
guider et amliorer la qualit des
compressions thoraciques

Electrical Therapy
- Minimiser le temps sans
compressions thoraciques autour
des choc lectriques
- Objectif = Temps darrt des
compressions thoraciques < 5
secondes
- Une phase de compressions
thoraciques de 3-4 min avant la
dlivrance du choc lectrique
externe nest plus recommande
- 3 chocs lectriques externes
successifs rservs aux patients
monitors et dj connects au
dfibrillateur
- Le dveloppement de la
dfibrillation automatique est
encourag

Adult Advance Life Support


- Minimiser le temps sans
compressions thoraciques
- Objectif de qualit des
compressions thoraciques
- Promouvoir les systmes de tlalertes dans la prvention de larrt
cardiaque
- Administration intra trachale des
drogues non recommande
- Atropine non recommande
- Adrnaline 1mg et Amiodarone
300 mg aprs chec du 3me choc
lectrique externe pour FV/TV
- Intubation par des personnes
entranes uniquement avec
utilisation du CO2 expir
- Importance de la ranimation post
arrt cardiaque
- Place renforce de la
coronarographie

1970 - 1980
Dcs imputables
3,38 pour 104
Totalement : 0,76
Partiellement : 2,62

Evnements pour 10 000


ASA 1

0,04

ASA 2

0,5

ASA 3

2,71

ASA 4

1996 - 1999
Dcs imputables
0,54 pour 104
Totalement : 0,07
Partiellement : 0,47

5,54

Algorithme de la Dfibrillation Automatique

Autres Recommandations
- Dvelopper laccs du public au
dfibrillateur automatique
- A lhpital le dfibrillateur automatique ou
semi-automatique doit tre accompagn
dune formation des personnels et dun
registre clinique

Dfibrillation : les recommandations 2010


Minimiser le temps darrt des compressions thoraciques (< 5 sec)

Compressions thoraciques pendant la mise en charge du dfibrillateur


Compressions thoraciques reprises immdiatement aprs le choc lectrique
Port de gants
Communication efficace dans lquipe

1 seul choc lectrique externe


Minimise le temps darrt des compressions thoraciques
Exception : patient monitor et connect au dfibrillateur (3 chocs accepts)

Energie dlivr 150 J


Soit augmentation soit stabilit de lnergie au fil des chocs

Matriel
Semi automatique
Electrodes adhsives

Algorithme de la ranimation avance

Ranimation avance ce qui change


Voies ariennes
Intubation par du personnel entran et sans arrt des compressions thoraciques
Utilisation du CO2 expir
Dispositif supra-glottique (ou rien) dans les autres cas

Ventilation contrle
Frquence = 10 cylces / min
Fraction inspire dO2 pour obtenir SpO2 entre 94 et 98%

Abord vasculaire
Priphrique ou intra-osseux
Voie pulmonaire non recommande

Les mdicaments
vers la simplification en labsence de preuves
Adrnaline : 1mg tous les 2 cycles (3(3-5 min)
Aprs le 3me choc si TV/FV persistante
Ds la voie veineuse en place pour les autres rythmes non accessible au choc

Amiodarone : 300 mg en bolus (+/- 150mg, +/- 900mg/24h)


pour la FV/TV persistante aprs le 3me choc lectrique externe
Lidocaine 1 mg/kg si amiodarone non disponible

Atropine
Non recommande en 2010

Sulfate de magnsium
Rserv la torsade de pointe

Fibrinolytiques
Rserv au diagnostic avr ou suspect dembolie pulmonaire

Principes de lhypothermie thrapeutique

Induction

Entretien et contrle

Rchauffement

Anesthesia-related Cardiac Arrest


Sheila J. Ellis, M.D., Myrna C. Newland, M.D., Jean A. Simonson, M.D., K. Reed Peters, M.D.,
Debra J. Romberger, M.D., David W. Mercer, M.D., John H. Tinker, M.D., Ronald L. Harter, M.D.,
James D. Kindscher, M.D., Fang Qiu, Ph.D., Steven J. Lisco, M.D.
ABSTRACT
Background: Much is still unknown about the actual incidence of anesthesia-related cardiac arrest in the United States.
Methods: The authors identified all of the cases of cardiac arrest from their quality improvement database from 1999 to 2009
and submitted them for review by an independent study commission to give them the best estimate of anesthesia-related cardiac arrest at their institution. One hundred sixty perioperative cardiac arrests within 24h of surgery were identified from an
anesthesia database of 217,365 anesthetics. An independent study commission reviewed all case abstracts to determine which
cardiac arrests were anesthesia-attributable or anesthesia-contributory. Anesthesia-attributable cardiac arrests were those cases
in which anesthesia was determined to be the primary cause of cardiac arrest. Anesthesia-contributory cardiac arrests were
those cases where anesthesia was determined to have contributed to the cardiac arrest.
Results: Fourteen cardiac arrests were anesthesia-attributable, resulting in an incidence of 0.6 per 10,000 anesthetics (95%
CI, 0.4 to 1.1). Twenty-three cardiac arrests were found to be anesthesia-contributory resulting in an incidence of 1.1 per
10,000 anesthetics (95% CI, 0.7 to 1.6). Sixty-four percent of anesthesia-attributable cardiac arrests were caused by airway
complications that occurred primarily with induction, emergence, or in the postanesthesia care unit, and mortality was 29%.
Anesthesia-contributory cardiac arrest occurred during all phases of the anesthesia, and mortality was 70%.
Conclusion: As judged by an independent study commission, anesthesia-related cardiac arrest occurred in 37 of 160 cardiac
arrests within the 24-h perioperative period. (Anesthesiology 2014; 120:829-38)

UCH is still unknown about the incidence of anesthesia-related cardiac arrest in the United States.
Since our original report was published in 2002, there
have been a number of articles and editorials exploring the
topic of anesthesia-related cardiac arrest.15 A review article
published in 2002 suggested that the overall perioperative
mortality rate for patients having American Society of Anesthesiologists (ASA) physical status I to V is approximately 1
per 500 anesthetics.2 The data further suggest that the anesthesia-related perioperative mortality rate is approximately 1
death per 13,000 anesthetics.
Another report found an overall frequency of perioperative cardiac arrests of 4.3 per 10,000 anesthetics. Cardiac
arrests primarily attributable to anesthesia were estimated
to be approximately 0.5 per 10,000 anesthetics. Mortality
attributable to anesthesia was approximately 1 in 100,000
anesthetics.4 This article was accompanied by an editorial
noting that without standardized methods of data collection and analysis, it is difficult to compare results between
institutions.5
Beginning in 2005, there were additional publications on
this topic.610 In 2008, a report looked at an unanticipated
day of surgery deaths in Department of Veterans Affairs

What We Already Know about This Topic


Much is still unknown about the actual incidence of anesthesia-related cardiac arrest in the United States
Using a single institutions quality improvement database, all
cases of cardiac arrest from 1999 through 2009 were identified and submitted for review by an independent study commission to estimate the incidence of anesthesia-related cardiac arrest

What This Article Tells Us That Is New


As judged by an independent study commission, anesthesiarelated cardiac arrest occurred in 37 out of 160 cardiac arrests
within a 24-h perioperative period

Hospitals.11 After review of 88 unanticipated day of surgery


deaths, the authors concluded that improved anesthesia care
may have prevented fatality in approximately 1 of 13,900
cases. Another study using data from the American College of
Surgeons National Surgical Quality Improvement Program
database from 2005 to 2007 (n = 362,767) found that an
intraoperative cardiac arrest occurs at a rate of approximately
7 per 10,000 noncardiac surgeries with a 30-day mortality
rate of 63%.12 In 2012, a review of 87 articles selected from
the world literature on anesthesia-related mortality over the

Presented in part at the Annual Meeting of the American Society of Anesthesiologists, San Francisco, California, October 15, 2013.
Submitted for publication July 31. 2013. Accepted for publication January 6, 2014. From the Department of Anesthesiology, University of
Nebraska Medical Center, Omaha, Nebraska (S.J.E., M.C.N., J.A.S., K.R.P., J.H.T., S.J.L.); Veterans Affairs Nebraska Western-Iowa Healthcare
System, and Department of Internal Medicine, University of Nebraska Medical Center, Omaha, Nebraska (D.J.R.); Department of Surgery,
University of Nebraska Medical Center, Omaha, Nebraska (D.W.M.); Department of Anesthesiology, The Ohio State University Wexner Medical Center, Columbus, Ohio (R.L.H.); Department of Anesthesiology, University of Kansas Medical Center, Kansas City, Kansas ( J.D.K.); and
Department of Biostatistics, College of Public Health, University of Nebraska Medical Center, Omaha, Nebraska (F.Q.).
Copyright 2014, the American Society of Anesthesiologists, Inc. Lippincott Williams & Wilkins. Anesthesiology 2014; 120:829-38

Anesthesiology, V 120 No 4 829

April 2014

Anesthesia-related Cardiac Arrest

past 60 yr found considerable differences in mortality related


to both economic development and ASA physical status with
ASA physical status from III to V having markedly increased
mortality rates.13
In 2013, a study looked at predictors of survival from
perioperative cardiopulmonary arrests. The authors used a
national in-hospital resuscitation registry to identify patients
18 yr or older who had a cardiac arrest in the operating room or within 24h postoperatively. Out of a total of
2,524 perioperative cardiac arrests reported from 234 hospitals, they found 1,458 intraoperative cardiac arrests and
536 that occurred in the postanesthesia care unit (PACU).
The remainder occurred in telemetry, critical care areas, or
general inpatient areas. Those arrests in the operating room
and PACU had better survival compared with the survival in
other perioperative locations.14
To address this lack of information about anesthesiarelated cardiac arrests, we asked an independent study commission, comprised of three anesthesiologists, an internist
and critical care specialist, and a surgeon, to review all of the
cases of cardiac arrest occurring within the 24-h perioperative period at our institution, to determine the incidence and
outcome of anesthesia-attributable and anesthesia-contributory cardiac arrests.

Materials and Methods


After obtaining approval from the University of Nebraska
Medical Center Institutional Review Board, Omaha,
Nebraska, we identified all cardiac arrests that occurred
within 24h after anesthesia that were reported to our
anesthesia database at The Nebraska Medical Center from
August 15, 1999 to December 31, 2009. We used the same
methodology as in our first report in that after we identified all cases of cardiac arrest from our anesthesia database,
we prepared an abstract of each case from anesthesia and
medical records. (See appendix 1 for details extracted from
the records to prepare the abstract.) We then submitted the
abstracts to an independent study commission representing
anesthesiology, internal medicine, and surgery. The use of an
independent study commission for review of deaths related
to anesthesia in the perioperative period was proposed by
Henry S. Ruth, M.D. (18991956; Professor, Department
of Anesthesiology, Hahnemann Hospital and Medical College, Philadelphia, Pennsylvania) in 1945.15 The independent study commissions would be modeled after maternal
mortality study commissions. It was thought that there
would be an insufficient number of deaths in Philadelphia
to warrant the use of a commission. The independent study
commission we created was asked to determine which cases
were anesthesia-attributable and which cases were anesthesia-contributory to the cardiac arrest.1 From these values,
we calculated an anesthesia-attributable cardiac arrest rate
and an anesthesia-contributory cardiac arrest rate. Cardiac
arrests were identified from the anesthesia database that was
developed from a quality assurance (QA) form included as
Anesthesiology 2014; 120:829-38

mandatory documentation with each anesthetic record. The


anesthesiology faculty member, anesthesiology resident, or
nurse anesthetist providing the anesthetic completed the
QA form. Required data on the QA form contained patient
demographics, anesthesia provider information, date, location, ASA physical status classification, and a 60-item
checklist of airway, cardiovascular, respiratory, neurologic,
regional, and miscellaneous events. Providers were encouraged to describe the event, treatment, and outcome, and
make any comments as to the cause or causes. Cardiac arrest
was defined as an event requiring cardiopulmonary resuscitation, which may include closed or open-chest cardiac compressions. All QA forms were collected daily with a copy of
the anesthesia record and reviewed for completeness by one
of the authors (M.C.N.) throughout the review period covered by this study. During the study period, a weekly mortality and morbidity conference was held as part of our quality
improvement process in which challenging cases or perioperative complications were discussed. Providers were expected
to complete a 24-h follow-up of all cases. This allowed documentation of early postoperative complications that may
have occurred in the PACU, the intensive care unit, or in the
patients room. For outpatients, a designated nurse called the
patient within 24h to determine whether the patient had
any concerns. Any problems identified in the postoperative
period were added to the database.
Each case of cardiac arrest was matched by a proximal
convenience method to four other cases receiving anesthesia on the same date and in a similar operating suite. During the study period, all anesthetic records were kept in files
maintained by the Department of Anesthesiology. Two cases
randomly filed by billing personnel immediately preceding
the cardiac arrest case and two immediately after the cardiac arrest case were identified as controls. For each study
case and each control case, data were obtained on patient
demographics (including age and sex), ASA physical status,
operative status (emergency or elective), surgical procedure
performed, time of day, and outcome. A copy of this data
collection form is provided in appendix 2. A case was considered an emergency if it was designated as such by the ASA
physical status classification. Additional pertinent information related to preoperative assessment, intraoperative
course, and anesthetic management was recorded. Surgical
procedures were classified by major categories according to
the Physicians Current Procedural Terminology, 4th edition,
as found in the ASA Relative Value Guide.16
The medical and anesthesia records of each patient
who had a cardiac arrest during anesthesia or in the 24-h
perioperative period were reviewed by at least one of
the authors (S.J.E., M.C.N., J.A.S., or K.R.P.) from our
Department of Anesthesiology. Abstracts were prepared
without assigning responsibility for the cardiac arrest.
Each abstract was assigned a three-digit code and submitted anonymously to an independent study commission
formed for this analysis.
830 Ellis et al.

PERIOPERATIVE MEDICINE

Members of the independent study commission included


the Chairman of the Department of Anesthesiology at The
Ohio State University Wexner Medical Center, Columbus,
Ohio, a Professor of Anesthesiology and Director of Liver
Transplant Anesthesiology at the University of Kansas Medical Center, Kansas City, Kansas, the former Chairman of the
Department of Anesthesiology and now Professor Emeritus at the University of Nebraska Medical Center, Omaha,
Nebraska, the Chairman of the Department of Surgery at the
University of Nebraska Medical Center, Omaha, Nebraska,
who was not at this institution during the period of this
study, and a Professor of Internal Medicine and Critical Care
at the Veterans Affairs Nebraska Western-Iowa Healthcare
System, Omaha, Nebraska, and the University of Nebraska
Medical Center, Omaha, Nebraska with no direct involvement in the care of these patients. Commission members
from outside the institution may provide a more unbiased
evaluation of these cases than if we used faculty exclusively
within our institution.
Commission members were asked to review abstracts
of all cases of cardiac arrest and were asked on their initial
review to provide their assessment of the primary cause of
cardiac arrest or death as due to (1) anesthesia, (2) surgery,
(3) patient disease or condition, (4) other, for example,
serendipity, electrical malfunction, fall, catastrophic failure of equipment, or (5) unable to decide from information provided. Commission members were also asked to
determine which of the following, anesthesia, surgery,
patient disease or condition, or other contributing cause,
appeared to be a contributing cause of cardiac arrest or
death. Consensus was determined when at least three of
the five commission members agreed on a cause of the cardiac arrest or death.
The cases identified as anesthesia-attributable or anesthesia-contributory on the first review were resubmitted to the
commission for a second, more restrictive review. The commission members were asked to make a choice of the role
of anesthesia in these cases using the following scale, briefly
summarized from our previous publication:
1. Anesthesia was the primary cause of the adverse event
(certainty >90%)
2. Anesthesia was the primary cause of the adverse event
(certainty 51 to 90%)
3. Anesthesia was an important contributing cause of the
adverse event (certainty >90%)
4. Anesthesia was an important contributing cause of the
adverse event (certainty 51 to 90%)
5. Anesthesia was neither the primary nor an important
contributing cause of the adverse event.
After the second review by the commission, cases were
assigned as anesthesia-attributable if the majority of the
members judged them to have anesthesia as the primary
cause (1 or 2 on the review scale) or anesthesia-contributory if the majority of the members judged them to have
Anesthesiology 2014; 120:829-38

anesthesia as an important contributing cause (numbers 3


and 4 on the review scale) of the cardiac arrest.
Statistical Analysis
The incidence, cause, and mortality of anesthesia-related
cardiac arrests were summarized. Matching of cases and
controls was based on the anesthesia time and location.
The characteristics of case and control groups were summarized using means and SDs for continuous variables, and
frequencies and percentages for categorical variables. A P
value of less than 0.05 was considered to be statistically
significant.

Results
There were 217,365 anesthetics administered during the
slightly more than 10-yr period of this study, from August
15, 1999 to December 31, 2009. A total of 160 cardiac
arrests within the 24-h perioperative period were identified
from an anesthesia QA database. The incidence of cardiac
arrest from all causes was 1 per 1358, or 7.4 per 10,000
anesthetics (95% CI, 6.3 to 8.6). To provide a comparison
group, cases experiencing cardiac arrest were matched with
four other cases that underwent anesthesia on the same day
and in the same location by a proximal convenience method.
The characteristics of the cases that had a cardiac arrest and
the controls are reported in table1.
Differences were found between cases and controls with
regard to age, sex, ASA physical status, emergency surgery
status, surgical procedure, length of operation, time of
day, and anesthetic technique by univariate analysis (all P
< 0.05). The cardiac arrest group was older (53.522.6 vs.
45.722.7 for controls), had a higher proportion of males
(61.3 vs. 46.3%), greater percentages of patients with higher
ASA physical status (ASA IV, 68.6 vs. 7.6%), greater
percentage of patients having emergency surgery (51.9 vs.
12.0%), more patients with thoracic or spine procedures
(35.0 vs. 14.7%), more patients with upper abdominal procedures (16.9 vs. 4.7%), longer length of operation (3.22.9
vs. 1.61.7h), more evening surgery (37.5 vs. 22.7%), and a
greater percentage had general anesthesia compared with the
control group (93.8 vs. 86.6%).
In the cardiac arrest group, there were only four cases
less than 1 yr of age and only 12 cases between 1 and 20 yr.
The largest group of cases, 59 (37.8%), was between 51 and
70 yr of age. Controls were evenly distributed in the 31 to
50 and 51 to 70 yr old age groups. Males comprised 61%
of the cardiac arrest cases. Three patients with ASA I and
eight patients ASA II experienced a cardiac arrest. Seventythree percent of cardiac arrests occurred in patients with
ASA physical status III and IV. A little over 60% of cases
with cardiac arrest occurred during regular working hours
from 07:00 to 15:00 and the remainder took place in late
afternoon and evening/nighttime hours. General anesthesia was the predominant anesthetic technique used in cases
with cardiac arrest.
831 Ellis et al.

Anesthesia-related Cardiac Arrest

Table 1. Characteristics of Cases and Comparison Group for Cardiac Arrest


Characteristics
Age (yr)
<1
110
1120
2130
3150
5170
7190
Sex
Female
Male
ASA physical status*
I
II
III
IV
V
Emergency vs. scheduled
Surgical procedures
Head/neck
Thoracic/spine
Upper abdomen
Extremity
Other
Lower abdomen
Length of operation (h)

1.5
1.53
>3
Time of day (24-h clock)
Day (07:0015:00)
Evening/night (15:0007:00)
Anesthetic technique, general vs. other

Cases (n = 160)

Controls (n = 640)

P Value

53.522.6
4 (2.6)
7 (4.5)
5 (3.2)
10 (6.4)
35 (22.4)
59 (37.8)
36 (23.1)

45.722.7
18 (2.8)
38 (6.0)
37 (5.8)
77 (12.1)
185 (29.1)
184 (29.0)
96 (15.1)

<0.0001

62 (38.8)
98 (61.3)

344 (53.8)
296 (46.3)

0.0008

3 (1.9)
8 (5.0)
39 (24.5)
77 (48.4)
32 (20.1)
83 (51.9)

86 (13.5)
237 (37.3)
264 (41.6)
48 (7.6)
0 (0.0)
77 (12.0)

<0.0001

11 (6.9)
56 (35.0)
27 (16.9)
18 (11.3)
6 (3.8)
42 (26.3)
3.22.9
66 (41.3)
41 (25.6)
53 (33.1)

144 (22.5)
94 (14.7)
30 (4.7)
129 (20.2)
116 (18.1)
127 (19.8)
1.61.7
442 (69.1)
117 (18.3)
81 (12.7)

100 (62.5)
60 (37.5)
150 (93.8)

495 (77.3)
145 (22.7)
554 (86.6)

<0.0001
<0.0001

<0.0001

0.0001
0.02

* The reference level of ASA physical status in the univariate conditional logistic regression analysis was level IV.
ASA = American Society of Anesthesiologists.

Figure1 is a flow diagram illustrating the results of the


review process used by the independent study commission
in reviewing the 160 cardiac arrest case abstracts. All 160
abstracts were sent to the commission members for the initial review. After the initial review, 12 cases received at least
three of five votes for anesthesia-attributable and nine cases
had three of five votes for anesthesia-contributory. All were
included in the cases sent back to the reviewers for the second review. An additional 24 cases had at least one vote for
anesthesia-attributable and 17 cases had at least two votes for
anesthesia-contributory. Many cases qualified in both categories because they had both one vote for anesthesia-attributable and two votes for anesthesia-contributory. A total of
53 abstracts were sent back to the reviewers for a second,
more restrictive review. Reviewers were asked not to consult
any notes from the first review but make a single judgment
as outlined in the Materials and Methods. After this second
review, there were now 14 cases determined to be anesthesia-attributable. Two of the original anesthesia-attributable
cases were moved to the anesthesia-contributory category.
Anesthesiology 2014; 120:829-38

An additional four cases were added to the anesthesia-attributable category making a total of 14. One case had been in
the anesthesia-contributable group and moved to anesthesiaattributable. Three cases had previously had two votes for
anesthesia-attributable and one case had previously only had
one vote for anesthesia-attributable, but now, all three cases
moved to the anesthesia-attributable group after the second
review. The original nine cases in the anesthesia-contributory group changed to eight after the second review because
one case moved from anesthesia-contributory to anesthesiaattributable. An additional 15 cases were added to the anesthesia-contributory group of 8 after the second review giving
a total of 23 cases in the anesthesia-contributory group. Sixteen cases of the 53 in the second review were judged to be
neither anesthesia-contributory nor anesthesia-attributable.
After the more restrictive second review, a total of 14 cases
were identified as anesthesia-attributable with a mortality rate
of 29% (95% CI, 8 to 58%), 4 of 14. These are listed in
table2. The adverse event leading to the arrest, the period of
the anesthetic in which it occurred, the anesthetic technique
832 Ellis et al.

PERIOPERATIVE MEDICINE

Fig. 1. Flow diagram of review process to identify anesthesia-attributable and anesthesia-contributory cases from 160
cases of perioperative cardiac arrest.

used, and the outcomes are also found in table 2. The 14


cardiac arrests attributable to anesthesia result in a cardiac
arrest rate attributable to anesthesia of 1 per 15,526 anesthetics or 0.6 per 10,000 anesthetics (95% CI, 0.4 to 1.1). Risk
of death due to anesthesia-attributable cardiac arrest was 1 in
54,341 or 0.2 per 10,000 anesthetics (95% CI, 0.1 to 0.5).
Mortality in the anesthesia-attributable cardiac arrest
group was 29% compared with a mortality rate of 70% in
cases of patient disease or condition or surgical/technical
factors causing cardiac arrest. Nine of the 14 cases involved
airway management either at induction or with extubation
of the trachea at the end of the case followed by inability to
mask ventilate or loss of airway on transport to or after arrival
in the PACU (two cases). Several problems with placement
of double-lumen tubes were noted. There were two cases of
cardiac arrest involving central venous access and one case of
hypotension after a small dose of intravenous narcotic. The
patient with bradycardia during subarachnoid block ultimately required insertion of an intravenous pacemaker. The
patient with a seizure after interscalene block had multiple
medical problems and on another occasion had experienced
a cardiac arrest during general anesthesia.
Anesthesiology 2014; 120:829-38

Cases were considered by the independent study commission to have anesthesia as an important contributory cause
of cardiac arrest for an anesthesia-contributory cardiac arrest
of 1 case per 9,450 or 1.1 per 10,000 anesthetics (95% CI,
0.7 to 1.6).
Table3 lists the adverse events in the 23 cases of anesthesia-contributory cardiac arrests. Seventy percent had cardiac
or cardiovascular complications which included myocardial
infarction, hypotension, ST segment depression, bradycardia, ventricular fibrillation, and myocarditis. There were
three cases (13%) of airway complications, all of which had
a cardiac arrest in the PACU. Two cases (9%) of anesthesiacontributory cardiac arrest had pulmonary edema. There was
one case of a type III protamine reaction and one case that
underwent surgery after an intracranial hemorrhage. Five
of the 23 anesthesia-contributory cardiac arrests occurred
in patients having emergency surgery and four of those five
patients died. Overall, 16 of the 23 cases of anesthesia-contributory cardiac arrests died resulting in a mortality of 70%
(95% CI, 47 to 87%). This is the same mortality rate found
for perioperative cardiac arrests not related to anesthesia
which are listed in table4.
Comments from the reviewers as to why anesthesia was
considered contributory to the cardiac arrest focused on
four areas: inadequate preoperative evaluation, issues with
intraoperative management, inadequate volume resuscitation during the case, and postoperative respiratory depression from narcotics administered during the case. In several
cases, the reviewers noted that the patient should have been
assigned an ASA physical status IV instead of an ASA physical status III.
Table 4 lists cardiac arrests that were not anesthesiarelated but attributable to patient disease/condition or
attributable to surgical/technical factors. Technical factors include complications during cardiac catheterization,
interventional radiology, or attempts at central venous
access. Mortality in these 123 cardiac arrests was 70% with
approximately 50% of patients in three categories: trauma,
complications associated with liver transplantation, and
inability to wean from cardiopulmonary bypass. The next
most common categories were surgery on the thoracic or
abdominal aorta (mortality 100%) and technical complications with 50% mortality. Two cases with perioperative
myocardial infarction as well as two cases with pulmonary
embolus died. There was one perioperative mortality in
which no definitive cause was determined.

Discussion
This report is similar to our previous study1 in that it
describes the findings of an independent study commission
that reviewed 160 cases of cardiac arrest in 217,365 anesthetics over a slightly more than 10-yr period of time from
August 15, 1999 to December 31, 2009. The denominator in this current study contains over 144,000 cases more
than the previous study. During this time period, a merger
833 Ellis et al.

Anesthesia-related Cardiac Arrest

Table 2. Adverse Events in Anesthesia-attributable Cardiac Arrests (n = 14)


Age (yr)

ASA PS

Location

Adverse Event Leading to Cardiac Arrest

Period

Outcome

Anesthesia
Technique

78

III

IP

Maintenance

Recovered

General

70

IV

IP

Emergence

Died

General

33

IVE

IP

Preinduction

Recovered

None

67

IV

IP

PACU

Recovered

General

42

OP

Emergence

Recovered

General

76

III

IP

Maintenance

Recovered

General

67

II

OP

Induction

Recovered

General

76

III

IP

Induction

Recovered

General

70

IV

IP

PACU

Died

General

10

53

IV

IP

Induction

Recovered

General

11

26

III

IP

Induction

Recovered

General

12

91

III

IP

Maintenance

Recovered

SAB

13

27

IV

IP

Induction

Died

Regional

14

59

II

IP

Hypotension after intravenous narcotic. Multiple


comorbidities.
CL pulled out during patient move to PACU bed.
Cardiac arrest during attempted replacement
of CL.
Hemothorax after CL attempt by anesthesia. Thoracotomy performed to control bleeding.
Loss of airway on transport to PACU after early
extubation.
Loss of airway on emergence due to bleeding and
laryngospasm after nasal trumpet placed.
Attempted DLT placement for thoracotomy after
laparotomy.
Loss of airway in patient with known cancer of
larynx. Unable to ventilate.
Breathing circuit misconnected after position
change. Unable to ventilate using circuit. Successful with self-inflating bag.
Extubated at end of long intra-abdominal operation.
Apneic on arrival in PACU. Reintubation.
Loss of airway with induction drugs. Difficult ventilation, bradycardia.
Difficulty in ventilating with DLT. DLT replaced with
single-lumen endotracheal tube.
Bradycardia and dysrhythmia. Ejection fraction
20%.
Seizure after interscalene block. Multiple comorbidities, previous cardiac arrest during general
anesthesia.
Known difficult airway but extubated at end of case.
Loss of airway. Unable to ventilate. Tracheostomy
done.

Emergence

Died

General

No.

ASA PS = American Society of Anesthesiologists physical status score; CL = central line; DLT = double-lumen tube; IP = inpatient operating rooms;
OP = outpatient operating rooms; PACU = postanesthesia care unit; SAB = subarachnoid block.

between the University Hospital and a neighboring community hospital was completed with a significant increase
in total numbers of cases attended by the Department of
Anesthesiology. The study commission found that 37 of 160
cardiac arrests were related to anesthesia. Fourteen cases were
judged to be anesthesia-attributable. This compares with the
anesthesia-attributable cardiac arrest rate of 1 per 14,591
anesthetics or 0.69 per 10,000 (95% CI 0.085 to 1.29) in
our first study. Nine of 14 (64%) of anesthesia-attributable
cases were related to airway management compared with 2
of 5 (40%) in our previous study. Airway problems included
difficulties either placing or ventilating through a doublelumen endotracheal tube, extubation and loss of airway on
emergence or in the PACU, or difficulty ventilating on induction of anesthesia. Two of the four deaths in this group were
related to airway management. Mortality in this group was
29%. This improved mortality rate contrasts with an 80%
mortality rate in our first study. The improved mortality rate
may be related to early recognition of the airway problem
contributing to the cardiac arrest and having sufficient help
available to rescue the patient. Other than one case of drug
reaction after an interscalene block and complications with
Anesthesiology 2014; 120:829-38

central venous access, all other patients in this category were


rescued successfully.
The study commission found that 23 patients were in the
anesthesia-contributory group. This compares with a somewhat higher anesthesia-contributory cardiac arrest rate of
1.37 per 10,000 anesthetics (95% CI, 0.52 to 2.22) in our
first study. This could be related to a smaller denominator in
the first study while having significant numbers of trauma,
complex surgeries, and other high-risk patients. All of the
patients in the anesthesia-contributory group in the current
study were with ASA physical status III or IV and presented
challenges with predominantly cardiovascular problems such
as recent myocardial infarctions, dysrhythmias, hypotension,
and pulmonary edema that were not easily correctable. Mortality was 70% in the anesthesia-contributory group. This is
in contrast to our previous study, where 70% of the patients
with anesthesia-contributory cardiac arrest survived. Their
complications were less life threatening and more easily
reversible problems of adverse drug events, vagal reactions,
and easily correctible dysrhythmias.
A recently published review article on quality and safety
in pediatric anesthesia discusses several institutions where
834 Ellis et al.

PERIOPERATIVE MEDICINE

Table 3. Adverse Events in Anesthesia-Contributory Cardiac Arrests (n = 23)


Adverse Events and Issues Leading to
Cardiac Arrest

No. Age (yr) ASA PS


1

47

III

92

IVE

75

IV

20

III

68

IV

86

IVE

44

III

41

III

70

IVE

10

64

III

11

57

IV

12

91

III

13

67

IV

14

74

III

15

49

III

16

38

IIIE

17

46

IV

18

86

III

19

IV

20

84 d

IVE

21

67

III

22

55

III

23

87

IV

Intraoperative hemorrhage and MI with inadequate volume resuscitation


during the case.
Bowel obstruction with recent MI. Etomidate used for induction and
probable inadequate volume resuscitation.
Surgery postintracranial hemorrhage and problems with intraoperative
management.
Respiratory arrest in PACU. Likely an ASA IV with metastatic cancer and
bowel obstruction. Respiratory depression after morphine
administration.
Severe type III reaction to protamine. Unanticipated reaction to 1/3 dose
of protamine.
Non Q-wave MI and unstable angina. Cardiac arrest after induction
dose of etomidate.
Hypotension during AICD change. Problems with intraoperative management.
Acute MI after superior laryngeal nerve block in morbidly obese patient.
Inadequate preoperative evaluation. ASA IV.
Sudden back pain and reintubation after offpump CABG. Cause uncertain.
Sudden bradycardia and hypotension. Active myocarditis. Inadequate
preoperative evaluation and of resting pulse of 107 beats/min before
elective hip replacement.
Dysrhythmia, desaturation, and hypotension at the end of insertion of
intravenous port during MAC anesthesia. Postoperative respiratory
depression secondary to narcotics.
Hypotension and bradycardia. Problems with intraoperative management.
Hypotension and bradycardia. Case cancelled after induction due to
discovery of skin lesion. Inadequate preoperative evaluation. Complete heart block developed while emerging from anesthesia.
Septic arthritis and hypotension. Inadequate volume replacement after
intraoperative loss of 3,000ml of blood.
Sudden onset pulmonary edema in patient with multiple comorbidities
and ejection fraction of 25% scheduled for esophagogastroduodenoscopy. Induction with etomidate and propofol. Inadequate preoperative evaluation.
Respiratory arrest within 15min of arrival in PACU. Postoperative respiratory depression secondary to narcotics administered throughout
case and within 30min of extubation in the operating room.
Recurrent episodes of hypotension. Attempted intravenous sedation
for placement of hemodialysis catheter and AV fistula. Problems with
intraoperative management.
Bradycardia and hypotension after extubation in OR. Reintubated in
OR. Problems with intraoperative management of elective total hip
replacement.
Respiratory arrest 30min after arrival in PACU. Likely cause respiratory
arrest secondary to narcotics.
Pulmonary edema during case resulted in occluded endotracheal tube.
Difficulties encountered during attempts at replacing. Problems with
intraoperative management.
Hypotension and bradycardia 15min after induction of anesthesia. ST
depression in lead II. Cardiology found stenotic lesion in anomalous
RCA. Successful CABG 9 days later.
Cardiac arrest 15min after induction of anesthesia for placement
of permanent pacemaker. Sudden VF in patient with sick sinus
syndrome.
Sudden onset of hypotension and severe ST segment depression after
administration of neostigmine and glycopyrrolate for reversal of muscle relaxants. Patient had undergone repair of femoral neck fracture.
Multiple medical problems.

Period

Outcome

Anesthesia
Technique

Maintenance

Died

General

Maintenance

Died

General

Maintenance

Died

General

PACU

Died

General

Maintenance

Died

General

Induction

Died

General

Maintenance Recovered

General

Induction

Died

General

ICU

Died

General

Maintenance

Died

General

Emergence

Died

MAC

Maintenance

Died

General

Emergence

Died

General

Maintenance

Died

General

Maintenance Recovered

General

PACU

Recovered

General

Maintenance

Died

General

Emergence Recovered

General

PACU

Recovered

General

Maintenance

Died

General

Induction

Recovered

General

Maintenance Recovered

General

Emergence

Died

General

AICD = automatic implantable cardiac defibrillator; ASA PS = American Society of Anesthesiologists physical status score; AV = arteriovenous;
CABG = coronary artery bypass grafting; ICU = intensive care unit; MAC = minimal alveolar concentration; MI = myocardial infarction; OR = operating room;
PACU = postanesthesia care unit; RCA = right coronary artery; VF = ventricular fibrillation.

Anesthesiology 2014; 120:829-38

835 Ellis et al.

Anesthesia-related Cardiac Arrest

Table 4. Cardiac Arrests Attributable to Patient Disease/Condition, or Surgical/Technical Factors

Causes of Arrest
Trauma: motor vehicle, gunshot wound, fall, pedestrian, other
End-stage liver disease and complications associated with liver transplantation
Inability to wean from cardiopulmonary bypass
Complications associated with cardiac surgery
Thoracic or abdominal aortic aneurysm surgery (seven ruptured)
Technical complications: surgical, special procedures, cardiac catheterization laboratory,
central venous access
Sepsis and/or multiple organ failure
Exsanguinating hemorrhage at operation associated with primary disease process
Complications associated with small bowel or kidney transplant
Complications associated with radical cancer surgery
Vagal reaction
Perioperative myocardial infarction
Pulmonary embolus
Pacemaker or implantable cardiac defibrillator related
Miscellaneous: drug reaction and cause undetermined

adverse event data are gathered from self-report, department quality improvement review, and other sources.17
Cases are then peer-reviewed by at least three anesthesiologists who were not involved in the case. Demographic
data on patients with anesthesia are provided by the
institutions so that estimates of incidence can be determined. An editorial discussing this article points out the
opportunity of using quality, patient safety, and process
improvement as an area of study.18 Clark also noted in
his editorial that collecting significant events is a problem
and may require manual tracking and analysis by experienced clinicians.
An analysis of factors associated with unanticipated day
of surgery deaths in Department of Veterans Affairs Hospitals was reported in 2008.11 The authors analyzed 815,077
elective surgical patients with ASA physical status I, II, or
III in the National Surgical Quality Improvement Program
database to identify patients who died on the day of surgery. They found that 0.08% or 646 patients died on the
day of surgery with the type of surgery the strongest predictor and aortic surgery the most risky. The authors did
a chart review of 88 deaths and found that opportunities
for improved anesthesia care were present in 13 of the 88
(15%). They also found that the time between the conclusion of surgery and final transfer of care into recovery was
a time in which many of the deaths occurred. We noted
this time period to be a factor in several of our anesthesiaattributable cardiac arrests. They suggested that a death
might have been prevented by improved anesthesia care in
approximately 1 per 13,900 cases. This number is similar
to that previously reported in a study of anesthesia-related
perioperative mortality.2
In comparison with our previous study of perioperative cardiac arrests, there are more airway-related complications in this latest 10-yr review compared with the
Anesthesiology 2014; 120:829-38

Number n (%)

Mortality (%)

123

70

24 (19.5)
18 (14.6)
18 (14.6)
6 (4.9)
10 (8.0)
14 (11.4)

100
61
100
67
100
50

6 (4.9)
5 (4.0)
4 (3.2)
4 (3.2)
3 (2.4)
2 (1.6)
2 (1.6)
2 (1.6)
3 (2.4)

83
80
75
100
0
100
100
0
33

previous 10 yr and a lower number of medication-related


complications. We have observed an increase in the number of airway complications on emergence from anesthesia
or in the PACU. Three of 23 (13%) of anesthesia-contributory cardiac arrests occurred from respiratory arrests
in the PACU. A report on perioperative cardiac arrest in
53,718 anesthetics over 9 yr from a Brazilian teaching
hospital found that all anesthesia-related cardiac arrests
were related to airway management and medication
administration.19
In our current study, the overall incidence of cardiac
arrest was 7.4 per 10,000 anesthetics. This compares with
our previous study in which the overall incidence of cardiac
arrest was 19.7 per 10,000 anesthetics. The distribution
and numbers of cardiac arrests attributable to patient disease/condition or surgical/technical factors remained fairly
similar between the two studies, but we had a much larger
denominator in the current study. This may explain the
change in incidence. The distribution of patients in the ASA
physical status groups has stayed approximately the same in
the two time periods. The top three categories for causes of
cardiac arrest remained the same: trauma, end-stage liver disease and complications associated with liver transplantation,
and inability to wean from cardiopulmonary bypass. Technical complications including surgical and special procedures
increased from 7 to 11.4%.
Limitations to this study include its representation of
perioperative cardiac arrests from a single institution. The
Nebraska Medical Center is a 600-bed tertiary referral center with approximately 20,000 surgical cases per year. It
includes a level 1 trauma center, an active solid organ transplantation program including both adult and pediatric liver,
small bowel, and kidney transplants, adult heart transplants,
high-risk obstetrical care, and neonatal and pediatric intensive care units. Patient mix is local and referral as well as
836 Ellis et al.

PERIOPERATIVE MEDICINE

some national/international patients. Our experience may


not be the same as other institutions.
We have maintained our database of adverse events for
more than 20 yr by reporting from providers including
faculty, residents, and certified registered nurse anesthetists. This is part of our QA and improvement program.
During the study period, adverse events were reviewed
weekly as part of our morbidity and mortality program.
It is always possible that not all events were captured. The
independent study commission did not have access to
original records. It is possible that this could have introduced unrecognized bias into the commissions interpretation of perioperative events.
A total of 37 cases of anesthesia-related cardiac arrest
were identified from 217,365 anesthetics over a 10-yr
period in an academic medical center. Fourteen cases were
determined to be anesthesia-attributable with an anesthesia-attributable cardiac arrest rate of 0.6 per 10,000 anesthetics (95% CI, 0.4 to 1.1) which is comparable with a rate
of 0.69 per 10,000 anesthetics (95% CI, 0.085 to 1.29) in
the previous 10 yr. Twenty-three cases were determined to
be anesthesia-contributory for a risk of anesthesia-contributory cardiac arrest of 1.1 per 10,000 anesthetics (95% CI
0.7 to 1.6) compared with a rate of 1.37 per 10,000 anesthetics in the previous 10 yr. The anesthesia-related cardiac
arrest rate has not changed significantly over a 10-yr period.
Many of the events occurred in the operating room after
the patient was emerging from anesthesia, during or after
transport to the PACU, and in the PACU. Airway management decisions, complications associated with vascular
access, preoperative patient assessment, and intraoperative
cardiovascular events are the areas where improvements
should be directed.

Acknowledgments
The authors acknowledge Benjamen Jones, B.S., Department of Anesthesiology, University of Nebraska Medical
Center, Omaha, Nebraska, and Ankit Agrawal, B.S., Department of Anesthesiology, University of Nebraska Medical
Center, for assistance with data entry and management.
Support was provided solely from institutional and/or
departmental sources.

Competing Interests
The authors declare no competing interests.

Correspondence
Address correspondence to Dr. Newland: 984455 Nebraska
Medical Center, Omaha, Nebraska 68198-4455. mnewland@
unmc.edu. Information on purchasing reprints may be
found at www.anesthesiology.org or on the masthead page
at the beginning of this issue. Anesthesiologys articles are
made freely accessible to all readers, for personal use only,
6 months from the cover date of the issue.

Anesthesiology 2014; 120:829-38

References
1. Newland MC, Ellis SJ, Lydiatt CA, Peters KR, Tinker JH,
Romberger DJ, Ullrich FA, Anderson JR: Anesthetic-related
cardiac arrest and its mortality: A report covering 72,959
anesthetics over 10 years from a US teaching hospital.
Anesthesiology 2002; 97:10815
2. Lagasse RS: Anesthesia safety: Model or myth? A review of
the published literature and analysis of current original data.
Anesthesiology 2002; 97:160917
3. Cooper JB, Gaba D: No myth: Anesthesia is a model for
addressing patient safety. Anesthesiology 2002; 97:13357
4. Sprung J, Warner ME, Contreras MG, Schroeder DR, Beighley
CM, Wilson GA, Warner DO: Predictors of survival following cardiac arrest in patients undergoing noncardiac surgery: A study of 518,294 patients at a tertiary referral center.
Anesthesiology 2003; 99:25969
5. Lagasse RS: Apples and oranges: The fruits of labor in anesthesia care. Anesthesiology 2003; 99:24850
6. Arbous MS, Meursing AE, van Kleef JW, de Lange JJ,
Spoormans HH, Touw P, Werner FM, Grobbee DE: Impact
of anesthesia management characteristics on severe morbidity and mortality. Anesthesiology 2005; 102:25768
7. Warner MA: Perioperative mortality: Intraoperative anesthetic management matters. Anesthesiology 2005; 102:2512
8. Cheney FW, Posner KL, Lee LA, Caplan RA, Domino KB:
Trends in anesthesia-related death and brain damage: A
closed claims analysis. Anesthesiology 2006; 105:10816
9. Lienhart A, Auroy Y, Pquignot F, Benhamou D, Warszawski
J, Bovet M, Jougla E: Survey of anesthesia-related mortality in
France. Anesthesiology 2006; 105:108797
10. Lagasse RS: To see or not to see. Anesthesiology 2006;
105:10713
11. Bishop MJ, Souders JE, Peterson CM, Henderson WG,
Domino KB: Factors associated with unanticipated day of
surgery deaths in Department of Veterans Affairs hospitals.
Anesth Analg 2008; 107:192435
12. Goswami S, Brady JE, Jordan DA, Li G: Intraoperative cardiac
arrests in adults undergoing noncardiac surgery: Incidence, risk
factors, and survival outcome. Anesthesiology 2012; 117:101826
13. Bainbridge D, Martin J, Arango M, Cheng D; Evidence-based
Peri-operative Clinical Outcomes Research (EPiCOR) Group:
Perioperative and anaesthetic-related mortality in developed
and developing countries: A systematic review and metaanalysis. Lancet 2012; 380:107581
14. Krishna Ramachandran S, Mhyre J, Kheterpal S, Christensen
RE, Tallman K, Morris M, Chan PS; American Heart
Associations Get With The Guidelines-Resuscitation Investi
gators: Predictors of survival from perioperative cardiopulmonary arrests: A retrospective analysis of 2,524 events
from the get with the guidelines-resuscitation registry.
Anesthesiology 2013; 119:132239
15. Ruth HS: Anesthesia study commissions. JAMA 1945; 127:5147
16. 2010 Relative Value Guide Book: A Guide for Anesthesia
Values, American Society of Anesthesiologists Committee on
Economics, Stead SW, Chair. Park Ridge, American Society of
Anesthesiologists, 2010, pp 173
17. Varughese AM, Rampersad SE, Whitney GM, Flick RP, Anton
B, Heitmiller ES: Quality and safety in pediatric anesthesia.
Anesth Analg 2013; 117:140818
18. Clark RM: The quality chasm is even bigger than we thought.
Anesth Analg 2013; 117:12734
19. Braz LG, Mdolo NS, do Nascimento P Jr, Bruschi BA, Castiglia
YM, Ganem EM, de Carvalho LR, Braz JR: Perioperative cardiac arrest: A study of 53,718 anaesthetics over 9 yr from a
Brazilian teaching hospital. Br J Anaesth 2006; 96:56975

837 Ellis et al.

Anesthesia-related Cardiac Arrest

Appendix 1. Data Collection Form for Items


to Be Included in Preparation of Abstracts

Appendix 2. Data Collection Form for Cases


and Controls

Case identification code:


Patient demographics:
Age
Sex
American Society of Anesthesiologists physical status
score
Emergency or scheduled operation
Surgical procedure
Length of operation (if helpful)
Time of day (day: 07:00 to 15:00, evening/night: 15:00
to 07:00)
Comorbid conditions
Obesity
Smoking
Hypertension
Diabetes mellitus
Coronary artery disease
Chronic obstructive pulmonary disease
End-stage renal disease
End-stage hepatic disease
Congestive heart failure
Myocardial infarction in past 6 months

1. Registration No. Case or Control Identifier No.


2. Date of operation
3. Time of operation
4. Sex

Pertinent facts related to preoperative assessment, intraoperative course, management of anesthetic, and resuscitation:
Outcome, if known:
Autopsy findings, if known:

Anesthesiology 2014; 120:829-38

5. Age in
a. Days
b. Months
c. Years
6. ASA physical status
7. Operation
a. Scheduled
b. Urgent
c. Emergency
8. Length of operation in

a. Minutes <60
b. Hours
9. Anesthetic

a. Local only
b. MAC
c. Regional
d. General

e. Regional and general
10. If cardiac arrest, was

a. Resuscitation successful

b. Resuscitation unsuccessful
ASA = American Society of Anesthesiologists; MAC = minimal alveolar concentration.

838 Ellis et al.

Predictors of Survival from Perioperative


Cardiopulmonary Arrests
A Retrospective Analysis of 2,524 Events from the Get With
The Guidelines-Resuscitation Registry
Satya Krishna Ramachandran, M.D., F.R.C.A.,* Jill Mhyre, M.D.,* Sachin Kheterpal, M.D., M.B.A.,*
Robert E. Christensen, M.D.,* Kristen Tallman, B.S.N., Michelle Morris, M.S., Paul S. Chan, M.D., M.Sc.,
for the American Heart Associations Get With The Guidelines-Resuscitation Investigators||

ABSTRACT
Background: Perioperative cardiopulmonary arrests are
uncommon and little is known about rates and predictors of
in-hospital survival.
Methods: Using the Get With The GuidelinesResuscitation national in-hospital resuscitation registry, we
identified all patients aged 18 yr or older who experienced
an index, pulseless cardiac arrest in the operating room or
within 24h postoperatively. The primary outcome was survival to hospital discharge, and the secondary outcome was
neurologically intact recovery among survivors. Multivariable logistic regression models using generalized estimating
equation models were used to identify independent predictors of survival and neurologically intact survival.
Results: A total of 2,524 perioperative cardiopulmonary
arrests were identified from 234 hospitals. The overall rate
of survival to discharge was 31.7% (799/2,524), including
41.8% (254/608) for ventricular tachycardia and ventricular fibrillation, 30.5% (296/972) for asystole, and 26.4%

* Assistant Professor, Research Coordinator, Department


of Anesthesiology, University of Michigan, Ann Arbor, Michigan.
Perianesthesia Nurse, Department of Nursing, University of Michigan. Assistant Professor, Saint Lukes Mid America Heart Institute,
Kansas City, Missouri. The American Heart Associations Get With
The Guidelines-Resuscitation Investigators are listed in appendix 1.
Received from the Department of Anesthesiology, University of
Michigan, Ann Arbor, Michigan. Submitted for publication October
23, 2011. Accepted for publication January 4, 2013. Supported by the
Department of Anesthesiology, The University of Michigan Health
System, Ann Arbor, Michigan. The American Heart Association, Dallas, Texas, provided funding for the collection and management of
the National Registry of Cardiopulmonary Resuscitation database.
The senior author (Dr. Chan) is a member of the Get With The
Guidelines-Resuscitation Investigators and is funded by the National
Institute of Health, Bethesda, Maryland, through K23#L102224.
Address correspondence to Dr. Ramachandran: Department
of Anesthesiology, University of Michigan, 1 H427 University
Hospital Box 0048, 1500 E Medical Center Drive, Ann Arbor,
Michigan 48109-0048. rsatyak@med.umich.edu. This article may
be accessed for personal use at no charge through the Journal
Web site, www.anesthesiology.org.

What We Already Know about This Topic


Perioperative cardiopulmonary arrests are uncommon events,
and their morbidity and mortality have not been well-studied
Using the Get With The Guidelines-Resuscitation national cardiopulmonary resuscitation registry, this study determined the
presentation, management, and outcomes of arrests occurring in the operating room and the postoperative period within
24h of surgery

What This Article Tells Us That Is New


Among patients with a perioperative cardiac arrest, one in
three survived to hospital discharge, and good neurological
outcome was noted in two of three survivors

(249/944) for pulseless electrical activity. Ventricular fibrillation and pulseless ventricular tachycardia were independently
associated with improved survival. Asystolic arrests occurring
in the operating room and postanesthesia care unit were associated with improved survival when compared to other perioperative locations. Among patients with neurological status
assessment at discharge, the rate of neurologically intact survival was 64.0% (473/739). Prearrest neurological status at
admission, patient age, inadequate natural airway, prearrest
ventilatory support, duration of event, and event location
were significant predictors of neurological status at discharge.
Conclusion: Among patients with a perioperative cardiac
arrest, one in three survived to hospital discharge, and good
neurological outcome was noted in two of three survivors.

ERIOPERATIVE cardiopulmonary arrests are


uncommon events, and their morbidity and mortality
have not been well-studied. The early postoperative period
poses additional risks to patients due to the proximate nature
of anesthesia and surgical insults. Thus, several features of
perioperative cardiac arrests are unique.1 As with other
specialized areas, such as the emergency room,2 perioperative

Copyright 2013, the American Society of Anesthesiologists, Inc. Lippincott


Williams & Wilkins. Anesthesiology 2013; 119:1322-39

Anesthesiology, V 119 No 6 1322

This article is featured in This Month in Anesthesiology.


Please see this issue of Anesthesiology, page 1A.

December 2013

PERIOPERATIVE MEDICINE

events may differ from cardiac arrests elsewhere in the hospital


in terms of resuscitation response times and underlying
etiology. Thus, survival outcomes may be different in these
locations than those seen in general in-patient units.
In prior studies, survival rates from perioperative arrests35
were higher than those reported in large multicenter in-hospital arrest studies.6 Other studies have reported on the incidence and risk factors for perioperative cardiac arrests,35 but
these have typically been single-institution studies with small
sample sizes (the largest study population of 223 patients),4
raising the question of generalizability. There remain significant limitations in our knowledge of perioperative arrests. For
instance, no prior study has described outcomes for cardiac
arrests occurring in the early postoperative period or variability of survival in different postoperative locations. Such
information may be important for anesthesiologists who are
often involved in the decision making for the postoperative
disposition of patients (floor status vs. telemetry vs. intensive
care). In addition, the relationship between process-of-care
measures (e.g., time to epinephrine, intubation, and defibrillation) and outcomes in the perioperative setting is scant.
To better address these gaps in knowledge related to perioperative cardiac arrests, we set out to study the presentation, management, and outcomes of arrests occurring in the
operating room (OR) and the postoperative period within
24h of surgery.

Materials and Methods


Study Design
To achieve the study goals, we analyzed data from the multicenter Get With The GuidelinesResuscitation (GWTG-R,
formerly known as the National Registry for Cardiopulmonary
Resuscitation)7 database, an American Heart Association sponsored prospective, multisite, observational registry, because of
its detailed collection of measures of care and outcomes for inhospital cardiac arrests. The members of the American Heart
Association GWTG-R Investigators are listed in appendix 1.
The study design of the GWTG-R has been described previously in detail.6 Briefly, a resuscitation event is defined as a
pulseless cardiopulmonary arrest that requires chest compressions and/or defibrillation. Data abstraction for each cardiac
arrest is performed by trained personnel at each participating
institution.7 Data accuracy within the GWTG-R is ensured
through periodic chart review, and the mean error rate has been
previously reported to be less than 2.4% for all data.8 To allow
for comparative analyses across multiple sites, data elements
within the registry are standardized using Utstein-style definitions to ensure uniformity of data collection.9 Oversight of
data collection and analysis, integrity of the data, and research
is provided by the American Heart Association.
The registry is currently the largest repository of information on in-hospital cardiopulmonary arrest from over 400
participating hospitals.10 Because the GWTG-R data are
deidentified and already exist, need for consent was waived
by the Adult Research Task Force of the National Registry of

Fig. 1. Exclusion criteria and final study cohort. PACU = postanesthesia care unit.

Cardiopulmonary Resuscitation and the Executive Database


Steering Committee of the American Heart Association.
Patient Population
Of 118,404 patients aged 18 yr or older who experienced
an index, pulseless cardiac arrest from February 24, 2000,
to August 3, 2008, we excluded 115,502 patients because
their cardiac arrest did not occur in the OR, postanesthesia
care unit (PACU), or any locations, within 24h after leaving
the PACU (fig. 1). An additional 378 patients were excluded
due to missing data on first pulseless rhythm or survival
outcomes. Our final study cohort comprised 2,524 patients
with perioperative cardiac arrests.
Study Outcomes
The primary outcome measure was survival to hospital discharge. We examined as a secondary outcome measure neurologically intact survival among patients surviving to hospital
discharge. Neurological outcome was assessed using previously
described cerebral performance category (CPC) scores,6 which
describes patients as having no major disability (CPC = 1),
moderate disability (CPC = 2), severe disability (CPC = 3),

Anesthesiology 2013; 119:1322-39 1323 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Table 1. Baseline Characteristics of Perioperative Cardiopulmonary Arrests by Survival to Discharge Status

Characteristic
Age
Male sex
White race
First documented rhythm
Asystole
PEA
PVT/VF
Duration of event
Length of hospital stay
Admitting diagnosis
Medical, cardiac
Medical, noncardiac
Surgical, cardiac
Surgical, noncardiac
Time of cardiac arrest
After hours
Weekend
Event location
Operating room
Postanesthesia care unit
Intensive care area
Telemetry/step-down
General in-patient area
Coexisting medical conditions
Congestive heart failure at admission
Previous congestive heart failure
Myocardial infarction at admission
Previous myocardial infarction
Respiratory insufficiency
Renal insufficiency
Hepatic insufficiency
Metabolic or electrolyte derangement
Baseline neurologic deficits
Acute stroke
Acute nonstroke neurological disorder
Pneumonia
Sepsis
Shock
Major trauma
Cancer
Immediate cause of arrest
Active myocardial infarction
Arrhythmia
Hypotension/hypoperfusion
Acute pulmonary edema
Acute pneumothorax
Acute pulmonary embolism
Acute respiratory insufficiency
Inadequate invasive airway
Inadequate natural airway
Invasive airway displacement
Metabolic abnormality
Invasive ventilation during or prior arrest

P Value

Unadjusted
Odds Ratio
(95% CI)

Missing
Data,
n (%)

65.118.1
980 (56.8)
1,253 (77.7)

<0.001
0.640
0.014

1.0 (0.91.1)
0.3 (1.11.6)

0 (0)
6 (0.2)
168 (6.6)

296 (37.0)
249 (31.2)
254 (31.8)
14.314.2
15.415.1

676 (39.2)
695 (40.3)
354 (20.5)
25.519.6
11.76.3

0.304
<0.001
<0.001
<0.001
<0.001

0.9 (0.81.1)
0.7 (0.60.8)
1.6 (1.32.0)

371 (12.8)
371 (12.8)
371 (12.8)
0 (0)
9 (0.3)

64 (8.1)
60 (7.6)
129 (16.4)
516 (65.6)

98 (5.7)
224 (13.1)
229 (13.4)
994 (58.0)

<0.001
<0.001
0.055
0.001

1.5 (1.02.0)
0.5 (0.40.7)
1.3 (1.01.6)
0.3 (1.11.6)

28 (0.9)
28 (0.9)
28 (0.9)
28 (0.9)

67 (8.4)
88 (11.0)

324 (18.8)
381 (22.1)

<0.001
<0.001

0.4 (0.30.5)
0.4 (0.30.6)

0 (0)
0 (0)

455 (56.9)
214 (26.8)
76 (9.5)
20 (2.5)
34 (4.3)

1,003 (58.1)
32 (18.8)
256 (14.8)
38 (2.1)
106 (6.1)

0.571
<0.001
<0.001
0.573
0.054

1.0 (0.81.1)
1.5 (1.21.9)
0.6 (0.50.8)
0.2 (0.72.1)
0.7 (0.51.0)

0 (0)
0 (0)
0 (0)
0 (0)
0 (0)

55 (6.9)
117 (14.6)
81 (10.1)
130 (16.3)
228 (28.5)
151 (18.9)
20 (2.5)
60 (7.5)
45 (5.6)
19 (2.4)
32 (4.0)
26 (3.3)
35 (4.4)
188 (23.5)
42 (5.3)
75 (9.4)

175 (10.1)
284 (16.4)
165 (9.6)
280 (16.2)
655 (38.0)
483 (28.0)
81 (4.7)
263 (15.2)
184 (10.7)
44 (2.6)
106 (6.1)
82 (4.8)
194 (11.2)
708 (41.0)
233 (13.5)
214 (12.4)

0.008
0.245
0.652
0.981
<0.001
<0.001
0.009
<0.001
<0.001
0.796
0.028
0.083
<0.001
<0.001
<0.001
0.027

0.7 (0.50.9)
0.9 (0.71.1)
1.0 (0.81.4)
1.0 (0.81.3)
0.7 (0.50.8)
0.6 (0.50.7)
0.5 (0.30.9)
0.5 (0.30.6)
0.5 (0.40.7)
0.9 (0.51.6)
0.6 (0.41.0)
0.7 (0.41.1)
0.4 (0.30.5)
0.4 (0.40.5)
0.4 (0.30.5)
0.7 (0.61.0)

0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)

40 (5.0)
528 (66.4)
290 (36.5)
6 (0.8)
6 (1.0)
14 (1.8)
227 (28.6)
13 (1.6)
36 (4.5)
6 (0.8)
32 (4.0)
713 (89.2)

129 (7.5)
959 (55.9)
1,025 (59.7)
30 (1.7)
14 (1.1)
47 (2.7)
513 (29.9)
38 (2.2)
37 (2.2)
10 (0.6)
218 (12.7)
1,688 (97.9)

0.021
<0.001
<0.001
0.051
0.800
0.139
0.493
0.339
0.001
0.614
<0.001
<0.001

0.7 (0.51.0)
1.6 (1.31.9)
0.4 (0.30.5)
0.4 (0.21.0)
0.9 (0.32.3)
0.6 (0.41.2)
0.9 (0.81.1)
0.7 (0.41.4)
2.2 (1.43.4)
0.3 (0.53.6)
0.3 (0.20.4)
0.2 (0.10.3)

14 (0.5)
14 (0.5)
14 (0.5)
14 (0.5)
744 (25.6)
14 (0.5)
14 (0.5)
14 (0.5)
14 (0.5)
14 (0.5)
14 (0.5)
0 (0)
(continued)

Survivors
(n=799)

Nonsurvivors
(n=1725)

63.215.7
446 (55.8)
615 (82.1)

Anesthesiology 2013; 119:1322-39 1324 Ramachandran et al.

PERIOPERATIVE MEDICINE

Table 1. Continued

Characteristic

Survivors
(n = 799)

Nonsurvivors
(n = 1,725)

Neurological status at admission


No major disability
Moderate disability
Severe disability
Coma or vegetative state

566 (78.2)
115 (15.9)
36 (5.0)
7 (1.0)

865 (57.1)
394 (26.0)
148 (9.8)
109 (7.2)

P Value

Unadjusted
Odds Ratio
(95% CI)

Missing
Data,
n (%)

<0.001
<0.001
<0.001
<0.001

2.7 (2.23.3)
0.5 (0.40.7)
0.5 (0.30.7)
0.1 (0.10.3)

92 (3.1)
92 (3.1)
92 (3.1)
92 (3.1)

PEA = pulseless electrical activity; PVT/VF = pulseless ventricular tachycardia or ventricular fibrillation.

and coma or vegetative state (CPC = 4). For this study, we


dichotomized patients as having neurologically intact survival
(CPC = 1) or survival with neurological disability (CPC > 1).
Of the patients who survived to hospital discharge (survivors),
only those with both admission and discharge CPC recorded
(n = 663 [82.9% of survivors]) were included in multivariate
analyses of predictors of neurologically intact survival.
Statistical Analysis
Baseline patient characteristics were compared between survivors and nonsurvivors with Pearson chi-square test for discrete
variables, t test for normally distributed continuous data, and
Wilcoxon rank sum test for nonnormally distributed variables.
We then constructed separate multivariable models to
identify predictors of survival to discharge and neurologically intact survival. Variables with a univariate association
with survival (P < 0.10) were considered for model inclusion. Candidate patient-level variables included admitting
diagnosis (medical, cardiac; medical, noncardiac; surgical,
cardiac; or surgical, noncardiac) and presence or absence of
coexisting medical conditions at the time of cardiac arrest
(respiratory, renal, or hepatic insufficiency; congestive heart
failure, metabolic or electrolyte derangements; pneumonia; neurological disorders; shock; sepsis; major trauma, or
cancer). Additionally, we controlled for variables related to
the cardiac arrest, including initial cardiac rhythm (asystole, pulseless electrical activity [PEA], pulseless ventricular
tachycardia [PVT], or ventricular fibrillation [VF]), duration of cardiac arrest, time of cardiac arrest (during work
hours or during after-hours periods [i.e., 5.00 pm to 8.00
am], and weekend events. Consistent with previous literature,8,10 shockable rhythms PVT and VF were analyzed
together as one rhythm type. For models assessing neurologically intact survival among survivors, we also included as a
binary covariate prearrest neurological status (baseline CPC
score 1 vs. other CPC score). As the data in the GWTG-R
database are derived from multiple sites of differing volume,
all models used generalized estimating equation methodology with an exchangeable correlation matrix to control for
patient clustering at the facility level. Collinearity was evaluated on all pairs of variables to assess for independence. The
magnitudes of the standard errors were used as additional
measures of collinearity. The model test of significance was

used to investigate model performance. The resultant chisquare statistic value is a measure of the relationship between
observed and expected frequencies. A P <0.05 in this test
denotes that the null hypothesis is rejected. If the test was
significant, linear relationships between predictors and log
odds of outcome for continuous variables were further investigated. Model overfitting was limited by ensuring that more
than 10 subjects per independent variable were included in
the model.
Secondary analyses were performed to explore the differences in outcomes relating to initial rhythm type and presenting location (intraoperative, PACU, telemetry/step-down,
intensive care unit [ICU], or general in-patient area). For this,
we assessed the unadjusted relationship between process-ofcare measures (time to administration of epinephrine, invasive airway placement, and defibrillation), event location,
and survival. Groups were compared using t test for normally
distributed continuous data and Wilcoxon rank sum test for
nonnormally distributed variables. Additionally, multivariate
models were developed for evaluating risk factors for survival to
discharge for each primary rhythm type. Finally, we examined
the relationship between the comorbid disease burden (defined
as the total number of preoperative comorbidities documented
at the time of the arrest) and outcomes (survival and good neurological status, defined as a CPC 1). For this analysis, we compared increasing number of coexisting medical conditions with
the binary outcomes of interest using Pearson chi-square test
and Fischer exact test as appropriate. Statistical analyses were
performed using Stata 10 (StataCorp LP, College Station, TX).
Missing data analyses were performed to compare
between cases included and excluded from the general
estimating equation models. Survival to hospital discharge
(27.7% vs. 32.4%; P = 0.056), admission CPC (66.9% vs.
63.1%; P = 0.451), and discharge CPC (67.7% vs. 63.6%;
P = 0.517) were similar in the missing data and included
patients groups. In order to explore and present the missing
datas potential influence on the estimates of risk, data
imputation was performed using the following methodology.
The number of missing values was assessed for key variables
used in the analysis. Only missing data for the predictors
were imputed. We employed multiple imputation with
IVEware Version 0.1 (University of Michigan, Ann Arbor,
Michigan) for missing data. IVEware is an imputation and

Anesthesiology 2013; 119:1322-39 1325 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Table 2. Independent Predictors of Survival to Discharge and Good Neurological Outcome in Perioperative Arrests
Survival to Discharge
Preimputation
Risk Factor

Odds Ratio (95% CI)

Acute nonstroke neurological event


1.02 (0.681.53)
Baseline depression in neurological status
0.64 (0.411.01)
Congestive heart failure during admission
0.58 (0.410.83)
Hepatic insufficiency
0.62 (0.361.08)
Hypotension/hypoperfusion
0.54 (0.430.67)
Metastatic or hematologic malignancy
0.53 (0.380.74)
Metabolic and electrolyte abnormality
0.67 (0.441.02)
Renal insufficiency
0.69 (0.530.88)
Respiratory failure
0.82 (0.651.03)
Septicemia
0.46 (0.280.74)
Active or evolving myocardial infarction
0.67 (0.421.07)
Inadequate natural airway
2.39 (1.204.76)
Arrhythmia
1.42 (1.171.73)
After hours
0.56 (0.420.88)
Arrest rhythm (PEA reference)
Asystole
1.00 (0.771.31)
PVT/VF (shockable rhythms)
1.60 (1.162.20)
White vs. nonwhite
1.22 (0.941.57)
Event location (general in-patient unitreference)
Operating room
1.38 (0.832.30)
PACU
2.03 (1.163.56)
Telemetry
1.22 (0.502.99)
Intensive care area
0.96 (0.521.77)
Weekend
0.62 (0.450.86)
Invasive ventilation in place prearrest
0.28 (0.170.45)
Illness Category (medical noncardiacreference value)
Medical cardiac
2.04 (1.133.72)
Surgical cardiac
1.65 (1.032.65)
Surgical noncardiac
1.44 (1.002.08)
Trauma
0.20 (0.090.46)
Age at hospital admission
0.98 (0.970.99)
Duration of event
0.96 (0.950.96)
Admission CPC score 1
Not included

Postimputation for Missing Data

P Value

Odds Ratio (95% CI)

P Value

0.918
0.055
0.003
0.092
<0.001
<0.001
0.064
0.003
0.089
0.002
0.097
0.013
<0.001
0.001

1.08 (0.721.62)
0.71 (0.471.06)
0.59 (0.410.85)
0.66 (0.381.15)
0.51 (0.410.63)
0.63 (0.470.83)
0.64 (0.430.96)
0.66 (0.520.85)
0.91 (0.731.13)
0.45 (0.280.72)
0.58 (0.360.93)
1.79 (1.003.22)
1.37 (1.131.66)
0.61 (0.520.8)

0.712
0.095
0.005
0.142
<0.001
0.001
0.029
0.001
0.401
0.001
0.025
0.051
0.001
<0.001

0.993
0.004
0.135

0.97 (0.781.21)
1.54 (1.201.98)
1.05 (0.841.33)

0.78
0.001
0.663

0.215
0.013
0.658
0.894
0.004
<0.001

1.07 (0.691.66)
1.57 (0.972.52)
1.02 (0.462.27)
0.70 (0.401.21)
0.53 (0.390.71)
0.44 (0.330.59)

0.758
0.064
0.958
0.2
<0.001
<0.001

0.019
0.039
0.048
<0.001
<0.001
<0.001

1.91 (1.153.19)
1.86 (1.222.84)
1.50 (1.092.06)
0.25 (0.120.51)
0.98 (0.970.99)
0.95 (0.950.96)
Not included

0.013
0.004
0.013
<0.001
<0.001
<0.001

CPC = cerebral performance category; PACU = postanesthesia care unit; PEA = pulseless electrical activity; PVT/VF = pulseless
ventricular tachycardia or ventricular fibrillation.

variance estimation software that creates single or multiple


imputations of missing values using the Sequential Regression
Imputation Method.11,12 IVEware also creates partial or
full synthetic data sets using the sequential regression
approach to protect confidentiality and limit statistical
disclosure and can combine information from multiple
sources by vertically concatenating data sets and multiply
imputing the missing portions to create larger rectangular
data sets. For our imputation, IVEware was used to impute
data through SAS version 9.2 (SAS Inc., Cary, NC). This
approach allowed us to handle complex data structures that
were created from a large number of variables with mixed
formats (dichotomous, categorical, continuous, counts, and
others). For this analysis, five imputations were performed
Anesthesiology 2013; 119:1322-39

and the datasets were assembled into one dataset so analysis


could be conducted. After imputation, Stata 10 was utilized
to analyze multivariate models for major outcome measures
using General Estimation Equation to account for clustering
at the facility level. The same syntax was used to recreate the
multivariate models after imputation to provide information
on the influence of missing data on risk estimates.

Results
Of 2,524 patients from 234 hospitals, 1,458 (57.7%) had a
cardiac arrest in the OR and the rest had arrests in the postoperative setting (fig. 1). Return of spontaneous circulation
occurred in 1,485 patients (58.7%), 1,151 patients (45.5%)

1326 Ramachandran et al.

PERIOPERATIVE MEDICINE

Table 2. Continued
Good Neurological Outcome
Preimputation
Risk Factor

Odds Ratio (95% CI)

Acute nonstroke neurological event


0.90 (0.292.78)
Baseline depression in neurological status
0.46 (0.211.00)
Congestive heart failure during admission
1.07 (0.373.04)
Hepatic insufficiency
0.83 (0.154.56)
Hypotension/hypoperfusion
0.71 (0.431.19)
Metastatic or hematologic malignancy
0.73 (0.361.47)
Metabolic and electrolyte abnormality
1.90 (0.794.55)
Renal insufficiency
0.63 (0.321.22)
Respiratory failure
0.90 (0.511.59)
Septicemia
0.24 (0.070.84)
Active or evolving myocardial infarction
1.19 (0.492.84)
Inadequate natural airway
0.48 (0.201.19)
Arrhythmia
0.85 (0.541.35)
After hours
1.32 (0.434.00)
Arrest rhythm (PEA reference)
Asystole
0.79 (0.431.45)
PVT/VF (shockable rhythms)
0.91 (0.491.70)
White vs. nonwhite
1.73 (0.943.18)
Event location (general in-patient unitreference)
Operating room
5.77 (1.7419.10)
PACU
5.64 (1.7717.95)
Telemetry
2.21 (0.627.95)
Intensive care area
6.79 (1.7726.05)
Weekend
1.65 (0.733.74)
Invasive ventilation in place prearrest
0.08 (0.030.21)
Illness Category (medical noncardiacreference value)
Medical cardiac
1.66 (0.584.79)
Surgical cardiac
1.88 (0.705.03)
Surgical noncardiac
1.88 (0.834.25)
Trauma
4.56 (0.2681.53)
Age at hospital admission
0.98 (0.960.99)
Duration of event
0.99 (0.971.00)
Admission CPC score 1
48.84 (23.53101.38)

survived to 24h after their cardiac arrest, and 799 (31.7%)


survived to hospital discharge. Neurologically intact survival
was observed in 473 (64.0%) of 739 survivors with valid
CPC scores at discharge.
Results are clustered around key areas relevant to perioperative arrests listed below: primary arrest rhythm, location-specific
differences, patient-level associations, event-level associations,
and neurological status at admission and outcomes. Univariate analyses are presented in table 1. Multiple adjusted analyses
presented in tables 2 and 3, and appendices 27 are provided to
compare estimates of risk factors before and after data imputation for missing variables. All estimates presented in the following sections refer to the postimputation data analyses. Patients
excluded due to any missing data were similar in baseline

Postimputation for Missing Data


P Value

Odds Ratio (95% CI)

P Value

0.855
0.05
0.905
0.832
0.198
0.378
0.152
0.171
0.712
0.026
0.704
0.113
0.5
0.626

0.94 (0.352.52)
0.61 (0.271.41)
1.03 (0.442.42)
0.77 (0.252.36)
0.84 (0.561.26)
0.86 (0.521.43)
1.38 (0.792.43)
0.64 (0.381.08)
0.94 (0.581.54)
0.47 (0.161.40)
1.05 (0.492.26)
0.44 (0.210.92)
0.78 (0.541.13)
0.68 (0.281.62)

0.909
0.249
0.941
0.653
0.387
0.568
0.257
0.093
0.812
0.177
0.899
0.029
0.195
0.38

0.449
0.76
0.081

0.94 (0.631.39)
1.04 (0.671.61)
1.52 (0.942.47)

0.75
0.866
0.088

0.004
0.003
0.223
0.005
0.229
<0.001

1.64 (1.164.39)
1.72 (1.074.55)
0.93 (0.233.74)
1.72 (0.515.85)
1.15 (0.642.08)
0.59 (0.350.99)

0.041
0.049
0.919
0.383
0.637
0.046

0.349
0.209
0.128
0.302
0.001
0.047
<0.001

2.27 (0.955.39)
1.88 (0.874.08)
1.65 (0.873.11)
3.69 (0.7318.70)
0.98 (0.970.99)
0.98 (0.970.99)
21.69 (12.5237.57)

0.064
0.11
0.122
0.115
0.003
0.001
<0.001

characteristics to patients in the final study cohort, except that


the excluded patients had lower rates of previous myocardial
infarction (4.1% vs. 6.8%, P = 0.045), septicemia (5.4% vs.
9.0%; P = 0.019), arrhythmia (50.8% vs. 59.3%; P = 0.042),
and metabolic derangement (5.1% vs. 10.0%, P = 0.001).
Patients excluded for missing data on primary rhythm did not
differ significantly from those without missing data with regard
to survival to discharge (35.4% vs. 33.3%; P = 0.362) and good
neurological outcome (67.7% vs. 63.3%; P = 0.306).
Primary Arrest Rhythm
Asystole was the most commonly encountered rhythm, but
survivors were more likely to have a shockable initial cardiac
arrest rhythm (31.8% vs. 20.5%; P < 0.001) compared with

Anesthesiology 2013; 119:1322-39 1327 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

nonsurvivors (table 1). After adjusting for several patientlevel and event-level variables, PVT/VF alone was significantly associated with survival to 24h postarrest and survival
to discharge (tables 2 and 3; appendices 2 and 3). Adjusted
analyses of survival outcomes stratified by first documented
pulseless rhythm are described in table 3 (and appendices
46). There were no rhythm-specific differences in neurological outcomes on adjusted models (table 2, appendix 7).
In adjusted sub-analyses, arrest location, increasing age, and
longer duration of arrest were common independent associations with worse survival to discharge across all the three
primary rhythms (appendices 46).
Location-specific Differences
There were significant differences in survival and neurological
outcome by hospital location of arrest. The majority of arrests
occurred in the OR (1458/2524). Survival rates were highest in PACU arrests (214/536; 39.8%) followed by telemetry
(20/58; 35.1%), OR arrests (455/185; 31.2%), general inpatient areas (34/140; 24.3%), and ICU locations (76/332;
23.0%). Arrests in the OR, PACU, and ICU were associated
with significantly shorter time to epinephrine, whereas OR
and PACU arrests were associated with significantly shorter
times to invasive airway placement. There were no locationspecific differences in the time to defibrillation (table 4).
On adjusted analyses, arrest location was not associated
with survival to discharge. However, significant locationspecific differences were observed in adjusted sub-analyses
of survival to discharge within each arrest rhythm. Among
patients exhibiting asystole as the first documented pulseless
rhythm (appendix 5), intraoperative (adjusted odds ratio,
1.4; 95% CI, 1.02.6; P = 0.047) and PACU (adjusted
odds ratio, 2.0; 95% CI, 1.0 4.1; P = 0.044) locations were
significantly better survival to discharge compared to general
in-patient locations (fig. 2). ICU location was associated with
worse survival following PEA arrests (adjusted odds ratio,
0.4; 95% CI, 0.20.8; P = 0.012). In contrast, telemetry
(adjusted odds ratio, 5.7; 95% CI, 1.129.6; P = 0.038) and
ICU arrests (adjusted odds ratio, 1.9; 95% CI, 1.15.1; P =
0.041) were significantly associated with improved survival
following PVT/VF arrests in comparison with general
in-patient arrests
Intraoperative (adjusted odds ratio, 1.6; 95% CI, 1.2
4.4; P = 0.041) and PACU (adjusted odds ratio, 1.1; 95%
CI, 1.14.6; P = 0.05) were significantly associated with
better neurological outcomes in comparison to the events
occurring in the general in-patient location (table 2). On
sensitivity analyses, after exclusion of patients with CPC
scores >1, location was not associated with neurological outcome, suggesting that patient factors likely play a large role
in neurological recovery from perioperative arrest.
Patient-level Associations
The following variables were each associated with lower
survival rates to discharge on univariate analyses (table

1): Older age, congestive heart failure during admission


(23.9% survival vs. 32.4% control survival; P = 0.008),
respiratory insufficiency (25.8% survival vs. 34.8% control
survival; P < 0.001), renal insufficiency (23.8% survival vs.
32.1% control survival; P = 0.028), hepatic insufficiency
(19.8% survival vs. 32.2% control survival; P = 0.009),
baseline depression in neurological status (19.7% survival
vs. 32.9% control survival; P < 0.001), acute nonstroke
neurological event (23.2% survival vs. 32.1% control survival; P = 0.028), septicemia (15.3% survival vs. 33.3%
control survival; P < 0.001), shock (21.0% survival vs.
37.5% control survival; P < 0.001), major trauma (15.3%
survival vs. 33.7% control survival; P < 0.001), metabolic
abnormality (18.6% survival vs. 33.6% control survival;
P < 0.001), and metastatic or hematologic malignancy
(26.0% survival vs. 32.4% control survival; P = 0.027).
While myocardial infarction at admission showed no relationship with survival, the survival rate was lower in those
cases in which the arrest was specifically attributed to the
myocardial infarction (23.7% survival vs. 32.2% control
survival; P = 0.021).
Arrests attributed to arrhythmia (35.5% survival vs.
26.1% control survival; P < 0.001), inadequate natural airway (49.3% survival vs. 31.1% control survival; P = 0.01),
and white race (32.9% survival vs. 27.1% control survival;
P < 0.001) had higher rates of survival.
After multivariable adjustment, several patient-level
characteristics and medical conditions were associated with
lower survival from perioperative arrests (table 2), including age (adjusted odds ratio per additional year, 0.98; 95%
CI, 0.960.99), congestive heart failure during current
admission (adjusted odds ratio, 0.6; 95% CI, 0.40.9),
shock (adjusted odds ratio, 0.5; 95% CI, 0.40.6), metabolic abnormality (adjusted odds ratio, 0.6; 95% CI, 0.4
0.9), metastatic or hematologic malignancy (adjusted odds
ratio, 0.6; 95% CI, 0.40.9), renal insufficiency (adjusted
odds ratio, 0.7; 95% CI, 0.50.9), sepsis (adjusted odds
ratio, 0.5; 95% CI, 0.30.7), active or evolving myocardial infarction (adjusted odds ratio, 0.6; 95% CI, 0.40.9),
and trauma (adjusted odds ratio, 0.3; 95% CI, 0.10.5).
Arrhythmic cause of arrest (adjusted odds ratio, 1.4; 95%
CI, 1.11.7) was associated with higher adjusted odds of
survival to discharge. Age was the only patient-level risk
factor associated with both reduced survival to discharge
and poor neurological outcome. Inadequate natural airway
and admission neurological status were the other patientlevel factors associated with neurological recovery (table 2).
Event-level Associations
Of the processof-care measures, time to epinephrine administration was not related to any of the outcome measures
(table 4). Time to defibrillation was related to all survival
end points except neurological outcomes in survivors. Time
to invasive airway placement was related to 24-h survival and
neurological outcomes, but not survival to discharge.

Anesthesiology 2013; 119:1322-39 1328 Ramachandran et al.

PERIOPERATIVE MEDICINE

Table 3. Summary of Study End Points and Adjusted Survival Rates by Primary Arrest Rhythm Type
PEA
End Point
ROSC
Survival to 24 h
Survival to
discharge
Good neurological
outcome

Asystole

PVT/VF

Frequency
(%)

AOR
(95% CI)

Frequency
(%)

AOR
(95% CI)

P
Value

Frequency
(%)

AOR
(95% CI)

P
Value

533/944
(56.5%)
395/944
(41.8%)
249/944
(26.3%)
144/234
(61.5%)

Reference

551/972
(56.7%)
436/972
(44.9%)
296/972
(30.5%)
178/272
(65.4%)

1.0 (0.81.2)

0.901

1.2 (11.6)

0.106

1.0 (0.91.3)

0.772

1.3 (1.01.7)

0.032

1.0 (0.81.2)

0.78

1.5 (1.22.0)

0.001

0.9 (0.61.4)

0.75

401/608
(65.9%)
320/608
(52.6%)
254/608
(41.8%)
151/233
(64.8%)

1.0 (0.71.6)

0.866

Reference
Reference
Reference

Adjusted odds ratios are adjusted for acute nonstroke neurological event, baseline depression in neurological status, congestive heart
failure during admission, hepatic insufficiency, hypotension, metastatic or hematologic malignancy, metabolic and electrolyte abnormality, renal insufficiency, respiratory failure, septicemia, active or evolving myocardial infarction, inadequate natural airway, invasive ventilation, arrhythmia, time of day, weekend, location of arrest, admission diagnosis, and age. Admission cerebral performance category score
was included in adjusted model for neurological outcome. Model comparisons for neurological outcomes were made between survivors
discharged with no major disability and those with a moderate degree of disability or worse. See appendices 24 for full model outputs.
Multiple imputations and bootstrapping were performed to uncover hidden estimates for covariates included in the model. Estimates
provided in this table are derived from postimputation adjusted analyses.
AOR = adjusted odds ratio; PEA = pulseless electrical activity; PVT/VF = pulseless ventricular tachycardia or ventricular fibrillation;
ROSC = return of spontaneous circulation.

After multivariable adjustment, the event-level variables associated with survival to discharge included
duration of event (adjusted odds ratio, 0.96; 95% CI,
0.950.96; P < 0.001), weekend arrests (adjusted odds

ratio, 0.5; 95% CI, 0.40.7), after-hours arrests (adjusted


odds ratio, 0.6; 95% CI, 0.50.8), and need for invasive
ventilation (adjusted odds ratio, 0.4; 95% CI, 0.30.6).
Duration of event (adjusted odds ratio, 0.98; 95% CI,

Table 4. Relationship of Event Location, Process-of-Care Measures, and Outcomes


Time to Epinephrine
Event location
General inpatient
OR
PACU
Telemetry
ICU
P value
Event survival
Survived event
Nonsurvivors
P value
24-h survival
Survived 24 h
Nonsurvivors
P value
Survival to discharge
Survived to discharge
Nonsurvivors
P value
Neurological outcome
Good neurological status
Neurological injury
P value

Time to Defibrillation

Time to Invasive Airway

4 (0, 7)
0 (0, 2)
1 (0, 3)
3 (1, 5)
0 (0, 3)
<0.001*

4 (1, 7)
1 (0, 3)
1 (0, 4)
2 (1, 3)
1 (0, 2)
0.06

7 (5, 10)
2 (0, 5)
3 (1, 6)
8 (3, 10)
8 (4, 10)
<0.001*

0 (0, 3)
1 (0, 3)
0.48

0 (0, 3)
1 (1, 5)
0.01*

4 (1, 7)
4 (1, 8)
0.25

0 (0, 3)
1 (0, 3)
0.28

0 (0, 3)
1 (1, 5)
0.01*

3 (1, 7)
5 (1, 8)
0.03*

0 (0, 3)
1 (0, 3)
0.56

0 (0, 3)
1 (1, 4)
0.01*

3 (1, 7)
4 (1, 8)
0.15

1 (0, 3)
1 (0, 3)
0.24

1 (0, 2)
1 (0, 5)
0.76

3 (1, 6)
5 (2, 7)
0.03*

Median (25th, 75th centiles) values are displayed. Good neurological outcome refers to Cerebral Performance Score of 1. Asymptotic
significances are displayed.
* The significance level is 0.05.
ICU = intensive care unit; OR = operating room; PACU = postanesthesia care unit.
Anesthesiology 2013; 119:1322-39 1329 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Fig. 2. Adjusted odds ratios of survival to hospital discharge (with 95% CIs) stratified by primary rhythm and hospital location.
General inpatient locations were considered the reference value for each rhythm type and location. Asystolic events occurring
in the operating room (OR) and postanesthesia care unit (PACU) were independently associated with survival to discharge compared to general inpatient arrests. Pulseless electrical activity (PEA) arrests were associated with worse survival to discharge
in intensive care unit (ICU) locations. Telemetry and ICU locations were independently associated with survival to discharge in
pulseless ventricular tachycardia (PVT)/ventricular fibrillation (VF) arrests. See appendices 57 for full model outputs. * P < 0.05
on adjusted analyses.

0.970.99) and need for invasive ventilation (adjusted


odds ratio, 0.6; 95% CI, 0.40.9) were both associated
with neurological outcomes on adjusted analyses.
Neurological Status at Admission and Outcomes
Admission CPC score of 1 was noted in 1,431 patients
(among 2,240 patients with recorded admission CPC
scores), of whom 566 survived to hospital discharge. Of
these survivors with admission CPC1, 445 patients (80%)
had CPC1 at discharge. Thus, 31.1% of patients who
were admitted to hospital with CPC1 were discharged in
a neurologically intact state following perioperative cardiac
arrests. The strongest patient-level risk factor associated
with neurologically intact survival was admission CPC
score of 1 (table 2; adjusted odds ratio, 21.7; 95% CI,
12.537.6).
The relationship between the number of preoperative
coexisting medical conditions and outcomes is described in
figure 3. Survival to discharge decreased with increasing neurologic disability at admission, from 39.6% (566/1,431) for
those with no major disability, 22.6% (115/509) for those
with moderate disability, 19.6% (36/184) for those with
severe disability, and 6% (7/116) for those in a coma or vegetative state (P < 0.001). In patients with good baseline neurological status, number of coexisting medical conditions was
associated with worse survival to discharge (P < 0.001) and
worse survival at 24h (P = 0.001) but not survival of event
(P = 0.42) or neurological outcome among those surviving to
hospital discharge (P = 0.64). In patients with preevent neurologic deficits, number of coexisting medical conditions was
not associated with any of the outcome measures (event survival, P = 0.76; 24-h survival, P = 0.19; survival to discharge,
P = 0.39; good neurological outcome, P = 0.77).

Discussion
In a large cohort of patients with perioperative cardiac arrests,
we found that one in three patients survived to hospital

discharge, representing a significantly higher survival rate


than previous studies on in-hospital arrests on general hospital floors.6,13,14 Survival after arrests with shockable rhythms
are significantly better than asystole and PEA arrests. Asystolic arrests in the OR and PACU were associated with better survival than asystolic arrests in other locations. Notably,
location-specific differences in process-of-care measures were
observed, suggesting variability in response times that may
contribute to survival and neurological outcomes for perioperative patients with cardiac arrest.
The overall survival-to-discharge rates in this study
are significantly higher than that in previous reports of
in-hospital arrests from within the GWTG-R database
(15.317%),2,6,13,14 but confirms previous reported rates on
survival from perioperative arrests.4 One putative reason for
this overall survival benefit seen in perioperative arrests is
because surgical causes for cardiac arrest are more likely to be
reversible. Alternatively, the immediate availability of physician led care in the OR and the PACU could improve survival from cardiac arrest by influencing the speed and quality
of response.2
Primary Arrest Rhythms
There were significantly higher survival rates across all primary rhythm types (asystole, 30.5%; PEA, 26.4%; and
PVT/VF, 41.8%) compared to that in previous in-hospital
GWTG-R data (asystole, 10%; PEA, 10%; VF, 34%; and
PVT, 35%).6 In particular, the high survival rates from asystolic perioperative arrests are a unique finding of the current
study, reflecting potentially reversible causes. In a previous
subanalysis of 24 patients with anesthesia-attributed arrests,
asystole was associated with a much higher survival rate
(80% survival to discharge),4 related to a predominance of
medication and airway causes. In addition, location-specific
differences in survival outcomes with asystole appear to
support the view that immediate availability of anesthesia
providers in the intraoperative and PACU locations may

Anesthesiology 2013; 119:1322-39 1330 Ramachandran et al.

PERIOPERATIVE MEDICINE

Fig. 3. Relationship of number of coexisting medical conditions and perioperative cardiac arrest outcomes. (A) Event survival,
(B) survival to 24h after event, (C) survival to discharge, (D) good neurological outcome (cerebral performance category [CPC]
1) at discharge. The X-axis shows number of coexisting medical conditions and Y-axis shows frequency (%) of outcome. Solid
lines represent patients with good neurological status (CPC 1) at admission and interrupted lines represent patients with baseline
neurologic deficits at admission (CPC 24). There is no relationship between number of comorbidities and neurologically intact
survival in patients without baseline neurologic deficits at admission.

contribute significantly to improved outcomes in comparison with other locations. This association may be important especially in the context of the modifiable risk such as
arrhythmic cause of asystolic arrest, which was observed to
have improved survival. Thus, while asystole may indicate
a dismal prognosis for out-of-hospital arrest, perioperative asystole in the perioperative setting should be treated
aggressively. Intact survival appears to depend less on the
presenting rhythm, and more on the etiology of arrest, timing of interventions, and quality of advanced cardiac life
support.
Location-specific Differences
Significant location-specific differences have been described
for time to invasive airway placement in overall in-hospital
arrest literature.14 The first 24h after surgery are associated with a heightened risk of respiratory failure needing
emergent airway intervention,13,15 suggesting that delays in

definitive airway management may be of greater importance


in perioperative arrest outcomes. In the current study, time
to invasive airway placement was related to 24-h survival and
neurological outcomes, but not survival to discharge. The
faster times to defibrillation, epinephrine administration,
and invasive airway placement seen in the OR and PACU
support the hypothesis that ready availability of skilled perioperative care directly impacts clinical response to cardiac
arrest. It has been previously shown by others that defibrillation within 2min of the arrest improves survival from PVT/
VF arrests (39.3 vs. 22.2% for delayed defibrillation).8 In
our study, only 20% of the arrests were PVT/VF and survivors had shorter time to defibrillation. However, as trauma
and shock were independently associated with worse survival
to discharge with PVT/VF arrests, intraoperative PVT/VF
arrests associated with these risk factors were less likely to be
responsive to defibrillation. Thus, the OR may be the one
location where speed of defibrillation offers little benefit for

Anesthesiology 2013; 119:1322-39 1331 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

the patient suffering from surgical exsanguination.10 This


may also explain the higher survival seen with PVT/VF
arrests in postoperative locations (telemetry, ICU) as they
are generally less likely to encounter major exsanguination
and more likely to have other causes of PVT/VF arrest.
Decisions on location of postoperative care may be
extremely important, and our study provides some insight
into the location-specific differences in outcomes of highrisk patients. The finding that neurological outcomes are
independently better in telemetry or ICU patients compared
to general in-patient areas supports the value of increased
intensity of monitoring in high-risk patients in the postoperative period. There is evolving evidence that patient
surveillance systems16 on general care units, but not just continuous pulse oximetry monitoring17 may reduce morbidity.
Our data suggest that the presence or immediate availability
of skilled care (seen with PACU, ICU, and, to a lesser extent,
telemetry beds) may be crucial in modifying survival from
early postoperative cardiac arrests. Additional supportive evidence for this is the significantly shorter time to epinephrine
administration and invasive airway placement seen in the
OR and PACU in comparison to general care locations.
Patient-level and Event-level Associations
Previously, Peberdy et al.7 found that survival outcomes were
substantially lower during the night hours and the weekend,
with the greatest effect-size noted for arrests in the OR and
postoperative care unit. Chan et al.8 also demonstrated that
survival rates from in-hospital cardiac arrest were lower during nights and weekends, even when adjusted for potentially
confounding patient-related, event-related, and hospital
characteristics. In the current study, we identified lower odds
of survival during night-time hours and weekends. It is possible that, in common with in-hospital arrests from other
locations, other unmeasured system factors contribute to
the observed time- and day-related differences in survival
outcomes. This finding illustrates the limitations of therapy
despite fairly intense one-on-one management with faculty
anesthesiologist and surgeons.
Measures of acuity of response in the current study,
namely, time to invasive airway placement and epinephrine
administration, were associated with better survival rates.
This finding suggests that anesthesiologists have the opportunity to modify several factors that can directly influence
outcomes from arrests. On the contrary, the lower survival
out of hours may reflect the lack of breadth of clinician
resources and limited choices for delaying surgery due to
greater severity of the primary surgical condition. These
findings confirms Sprungs data4 and mirror findings from
the National Confidential Enquiry into Patient Outcome
and Deaths.18 However, this finding needs to be viewed
in the context of emerging evidence that elective general
surgery is safe during late hours with careful provision for
appropriate perioperative care.19 Thus, it may be possible
to modify survival outcomes in high-risk emergent surgical

patients by preferentially scheduling such cases during more


routine hours, ensuring more appropriate perioperative provider expertise, or both when feasible.20
Baseline Neurological Status and Outcomes
Finally, there are significant controversies that exist around
suspension of do-not-resuscitate orders in patients presenting for surgery, as limited data exist on the relationship
between preoperative comorbidities and perioperative cardiac arrest survival.2123 In the current study, survival to
discharge fell steeply with increasing neurologic disability
at admission, from 39.6% for those with no major disability, 22.6% for those with moderate disability, 19.6% for
those with severe disability, and 6% for those in a coma or
vegetative state. Among patients with baseline neurologic
deficits, the number of coexisting medical diagnoses did
not change survival outcomes. In contrast, among patients
with normal baseline neurological status, the total number
of coexisting diagnoses was associated with worse survival
to discharge and worse survival at 24h, but the comorbid load bore no apparent relationship with neurological
outcome among those surviving to hospital discharge. In
other words, the risk of patient harm (facilitating survival,
but with new neurologic deficits) is not increased among
those patients with normal baseline neurological status,
even with an increasing comorbid disease burden. Therefore, the decision to suspend or retain do-not-resuscitate
orders in the perioperative period should consider baseline
neurological status to a greater degree than the total number of coexisting medical diagnoses. The policy of routinely
suspending do-not-resuscitate orders in the perioperative
period may not be appropriate for all surgical patients,
particularly those who are already in a coma or persistent
vegetative state.
Our study findings should be interpreted in the context
of the several limitations described in detail previously.10
First, the registry hospitals may not be representative of all
U.S. hospitals and therefore our findings may not be generalizable, although they represent the best summary evidence
available. Second, the GWTG-R does not collect certain
data that would have been informative for our analyses,
including preoperative medications, intraoperative medications, physiologic data, and postoperative interventions.
The variables designating cause of arrest were abstracted
from the medical record by GWTG-R research coordinators not involved in clinical care of the patient, and therefore, their accuracy has not been validated. The registry does
not capture events that do not elicit a hospital response or
the use of a crash cart, and thus cardiac procedural and
operative arrests may be treated as routine and may not be
part of this registry.
In conclusion, one in three patients survived to hospital discharge after perioperative cardiac arrests and two of
three survivors had neurologically intact survival. The relatively high rates of survival and good neurological outcomes

Anesthesiology 2013; 119:1322-39 1332 Ramachandran et al.

PERIOPERATIVE MEDICINE

following intraoperative and PACU arrests suggest that


perioperative factors and immediate availability of skilled
anesthesia care have direct influence on improved survival
outcomes.

References
1. Runciman WB, Morris RW, Watterson LM, Williamson JA, Paix
AD: Crisis management during anaesthesia: Cardiac arrest.
Qual Saf Health Care 2005; 14:e14
2. Kayser RG, Ornato JP, Peberdy MA; American Heart
Association National Registry of Cardiopulmonary
Resuscitation: Cardiac arrest in the Emergency Department:
A report from the National Registry of Cardiopulmonary
Resuscitation. Resuscitation 2008; 78:15160
3. Braz LG, Mdolo NS, do Nascimento P Jr, Bruschi BA, Castiglia
YM, Ganem EM, de Carvalho LR, Braz JR: Perioperative cardiac arrest: A study of 53,718 anaesthetics over 9 yr from a
Brazilian teaching hospital. Br J Anaesth 2006; 96:56975
4. Sprung J, Warner ME, Contreras MG, Schroeder DR, Beighley
CM, Wilson GA, Warner DO: Predictors of survival following cardiac arrest in patients undergoing noncardiac surgery: A study of 518,294 patients at a tertiary referral center.
Anesthesiology 2003; 99:25969
5. Newland MC, Ellis SJ, Lydiatt CA, Peters KR, Tinker JH,
Romberger DJ, Ullrich FA, Anderson JR: Anesthetic-related
cardiac arrest and its mortality: A report covering 72,959
anesthetics over 10 years from a US teaching hospital.
Anesthesiology 2002; 97:10815
6. Peberdy MA, Kaye W, Ornato JP, Larkin GL, Nadkarni V, Mancini
ME, Berg RA, Nichol G, Lane-Trultt T: Cardiopulmonary
resuscitation of adults in the hospital: A report of 14720 cardiac arrests from the National Registry of Cardiopulmonary
Resuscitation. Resuscitation 2003; 58:297308
7. Peberdy MA, Ornato JP, Larkin GL, Braithwaite RS, Kashner
TM, Carey SM, Meaney PA, Cen L, Nadkarni VM, Praestgaard
AH, Berg RA; National Registry of Cardiopulmonary
Resuscitation Investigators: Survival from in-hospital cardiac
arrest during nights and weekends. JAMA 2008; 299:78592
8. Chan PS, Krumholz HM, Nichol G, Nallamothu BK; American
Heart Association National Registry of Cardiopulmonary
Resuscitation Investigators: Delayed time to defibrillation
after in-hospital cardiac arrest. N Engl J Med 2008; 358:917
9. Jacobs I, Nadkarni V, Bahr J, Berg RA, Billi JE, Bossaert L,
Cassan P, Coovadia A, DEste K, Finn J, Halperin H, Handley
A, Herlitz J, Hickey R, Idris A, Kloeck W, Larkin GL, Mancini
ME, Mason P, Mears G, Monsieurs K, Montgomery W, Morley
P, Nichol G, Nolan J, Okada K, Perlman J, Shuster M, Steen
PA, Sterz F, Tibballs J, Timerman S, Truitt T, Zideman D;
International Liaison Committee on Resuscitation; American
Heart Association; European Resuscitation Council;
Australian Resuscitation Council; New Zealand Resuscitation
Council; Heart and Stroke Foundation of Canada;
InterAmerican Heart Foundation; Resuscitation Councils of
Southern Africa; ILCOR Task Force on Cardiac Arrest and
Cardiopulmonary Resuscitation Outcomes: Cardiac arrest
and cardiopulmonary resuscitation outcome reports: Update
and simplification of the Utstein templates for resuscitation registries: A statement for healthcare professionals

from a task force of the International Liaison Committee


on Resuscitation (American Heart Association, European
Resuscitation Council, Australian Resuscitation Council, New
Zealand Resuscitation Council, Heart and Stroke Foundation
of Canada, InterAmerican Heart Foundation, Resuscitation
Councils of Southern Africa). Circulation 2004; 110:338597
10. Mhyre JM, Ramachandran SK, Kheterpal S, Morris M,

Chan PS; American Heart Association National Registry for
Cardiopulmonary Resuscitation Investigators: Delayed time
to defibrillation after intraoperative and periprocedural cardiac arrest. Anesthesiology 2010; 113:78293
11. Raghunathan TE, Lepkowski JM, Van Hoewyk J, Solenberger
P: A Multivariate technique for multiply imputing missing
values using a sequence of regression models. Surv Methodol
2001; 27: 8595
12. Raghunathan TE, Reiter JP, Rubin DB: Multiple imputation
for statistical disclosure limitation. J Off Stat 2003; 19: 116
13. Taylor S, Kirton OC, Staff I, Kozol RA: Postoperative day one:
A high risk period for respiratory events. Am J Surg 2005;
190:7526
14. Wong ML, Carey S, Mader TJ, Wang HE, American Heart
Association National Registry of Cardiopulmonary
Resuscitation Investigators: Time to invasive airway placement and resuscitation outcomes after inhospital cardiopulmonary arrest. Resuscitation 2010; 81:1826
15. Ramachandran SK, Haider N, Saran KA, Mathis M, Kim J,
Morris M, OReilly M: Life-threatening critical respiratory
events: A retrospective study of postoperative patients found
unresponsive during analgesic therapy. J Clin Anesth 2011;
23:20713
16. Taenzer AH, Pyke JB, McGrath SP, Blike GT: Impact of
pulse oximetry surveillance on rescue events and intensive
care unit transfers: A before-and-after concurrence study.
Anesthesiology 2010; 112:2827
17. Pedersen T, Moller AM, Hovhannisyan K: Pulse oximetry for
perioperative monitoring. Cochrane Database Syst Rev 2009:
CD002013
18. Gray A: United Kingdom national confidential enquiry into
perioperative deaths. Minerva Anestesiol 2000; 66:28892
19. Sessler DI, Kurz A, Saager L, Dalton JE: Operation timing and
30-day mortality after elective general surgery. Anesth Analg
2011; 113:14238
20. Kelz RR, Freeman KM, Hosokawa PW, Asch DA, Spitz FR,
Moskowitz M, Henderson WG, Mitchell ME, Itani KM: Time
of day is associated with postoperative morbidity: An analysis of the national surgical quality improvement program
data. Ann Surg 2008; 247:54452
21. Obolensky L, Clark T, Matthew G, Mercer M: A patient and
relative centred evaluation of treatment escalation plans: A
replacement for the do-not-resuscitate process. J Med Ethics
2010; 36:51820
22. Waisel DB, Simon R, Truog RD, Baboolal H, Raemer DB:
Anesthesiologist management of perioperative do-not-resuscitate orders: A simulation-based experiment. Simul Healthc
2009; 4:706
23. Waisel D, Jackson S, Fine P: Should do-not-resuscitate orders
be suspended for surgical cases? Curr Opin Anaesthesiol
2003; 16:20913

Anesthesiology 2013; 119:1322-39 1333 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Appendix 1. Get With The GuidelinesResuscitation Investigators


Besides the author Paul Chan, M.D., M.Sc., members of the
Get With The Guidelines-Resuscitation Adult Task Force
include: Comilla Sasson, M.D., M.S., and Steven Bradley,
M.D., M.P.H., University of Colorado; Michael W. Donnino, M.D., Beth Israel Deaconess Medical Center; Dana P.

Edelson, M.D., M.S., University of Chicago; Robert T. Faillace, M.D., Sc.M., Geisinger Healthcare System; Romergryko
Geocadin, M.D., Johns Hopkins University School of Medicine; Raina Merchant, M.D. M.S.H.P., University of Pennsylvania; Vincent N. Mosesso, Jr., M.D., University of Pittsburgh
School of Medicine; and Joseph P. Ornato, M.D., and Mary
Ann Peberdy, M.D., Virginia Commonwealth University.

Appendix 2. General Estimating Equation Models to Predict Event Survival with and without Data Imputation for
Missing Data
Preimputation
Risk Factor

AOR

Acute nonstroke neurological event


0.80
Baseline depression in neurological status
0.65
Congestive heart failure during admission
0.80
Hepatic insufficiency
0.64
Hypotension/hypoperfusion
0.83
Metastatic or hematologic malignancy
0.94
Metabolic and electrolyte abnormality
1.17
Renal insufficiency
1.03
Respiratory failure
1.13
Septicemia
0.93
Active or evolving myocardial infarction
0.64
Inadequate natural airway
1.25
Arrhythmia
1.14
Day hours
1.91
Arrest rhythm (PEA reference value)
Asystole
1.07
PVT/VF (shockable rhythms)
1.37
White vs. nonwhite
1.05
Event location (general in-patient unitreference)
Operating room
1.10
PACU
2.10
Telemetry
1.56
ICU
1.49
Weekend
0.83
Invasive ventilation in place prearrest
0.64
Illness category (medical noncardiac reference value)
Medical cardiac
1.52
Surgical cardiac
1.44
Surgical noncardiac
1.73
Trauma
0.75
Age at hospital admission
0.99
Duration of event
0.98

Postimputation for Missing Data

95% CI

P Value

AOR

95% CI

P Value

0.511.26
0.460.93
0.571.12
0.411.00
0.681.01
0.681.30
0.891.55
0.831.28
0.881.43
0.651.31
0.460.90
0.702.22
0.941.39
1.472.48

0.335
0.018
0.189
0.051
0.068
0.714
0.263
0.79
0.337
0.665
0.009
0.446
0.175
<0.001

0.90
0.71
0.74
0.72
0.77
1.01
1.20
1.03
1.13
0.90
0.67
0.98
1.23
1.63

0.551.46
0.510.98
0.541.02
0.491.06
0.640.92
0.761.33
0.951.52
0.841.26
0.921.40
0.641.27
0.490.92
0.621.55
1.041.45
1.292.04

0.662
0.037
0.065
0.099
0.005
0.968
0.129
0.804
0.236
0.558
0.012
0.938
0.015
<0.001

0.851.35
1.021.83
0.831.32

0.572
0.036
0.675

0.99
1.23
0.97

0.811.20
0.961.59
0.801.19

0.901
0.106
0.8

0.751.61
1.343.29
0.743.26
0.942.36
0.641.06
0.381.08

0.642
0.001
0.239
0.088
0.133
0.095

1.28
2.13
1.51
1.46
0.84
0.69

0.881.86
1.423.19
0.772.97
0.942.27
0.681.04
0.500.94

0.2
0
0.233
0.094
0.117
0.019

0.942.46
0.972.15
1.292.33
0.491.14
0.991.00
0.970.98

0.091
0.071
<0.001
0.181
0.029
<0.001

1.57
1.49
1.45
0.72
0.99
0.98

1.032.39
1.052.10
1.131.86
0.461.15
0.991.00
0.970.98

0.035
0.025
0.003
0.171
0.003
<0.001

AOR = adjusted odds ratio; ICU = intensive care unit areas; PACU = postanesthesia care unit; PEA = pulseless electrical activity;
PVT/VF = pulseless ventricular tachycardia or ventricular fibrillation.

Anesthesiology 2013; 119:1322-39 1334 Ramachandran et al.

PERIOPERATIVE MEDICINE

Appendix 3. General Estimating Equation Models to Predict Survival to 24h Postarrest with and without Data
Imputation for Missing Data
Preimputation
Risk Factor

AOR

Acute nonstroke neurological event


0.92
Baseline depression in neurological status
0.72
Congestive heart failure during admission
0.99
Hepatic insufficiency
0.84
Hypotension/hypoperfusion
0.56
Metastatic or hematologic malignancy
0.81
Metabolic and electrolyte abnormality
0.75
Renal insufficiency
1.01
Respiratory failure
0.91
Septicemia
0.71
Active or evolving myocardial infarction
0.62
Inadequate natural airway
2.37
Arrhythmia
1.39
Day hours
1.77
White vs. nonwhite
1.08
Arrest rhythm (PEA reference value)
Asystole
1.08
PVT/VF (shockable rhythms)
1.35
Event location (general in-patient unitreference)
Operating room
1.18
PACU
1.76
Telemetry
0.92
ICU
0.82
Weekend
0.67
Invasive ventilation in place prearrest
0.42
Illness category (medical noncardiac reference value)
Medical cardiac
1.74
Surgical cardiac
1.84
Surgical noncardiac
1.71
Trauma
0.31
Age at hospital admission
0.99
Duration of event
0.96

Postimputation for Missing Data

95% CI

P Value

AOR

95% CI

P Value

0.621.37
0.521.00
0.721.37
0.511.37
0.460.69
0.621.07
0.551.02
0.811.27
0.721.15
0.491.03
0.430.89
1.274.43
1.131.71
1.292.45
0.861.37

0.696
0.048
0.969
0.478
<0.001
0.132
0.062
0.905
0.427
0.069
0.01
0.007
0.002
<0.001
0.497

1.00
0.75
0.96
0.88
0.55
0.90
0.70
0.98
1.02
0.66
0.55
1.57
1.32
1.56
1.02

0.671.49
0.541.02
0.711.29
0.571.36
0.450.67
0.701.17
0.530.93
0.801.22
0.821.26
0.460.94
0.370.81
1.002.48
1.091.59
1.192.05
0.821.27

0.995
0.07
0.776
0.561
<0.001
0.441
0.013
0.888
0.859
0.02
0.002
0.052
0.005
0.002
0.853

0.861.36
1.011.81

0.497
0.042

1.03
1.30

0.851.25
1.021.66

0.772
0.032

0.771.79
1.122.76
0.471.81
0.521.28
0.510.88
0.250.70

0.448
0.015
0.808
0.377
0.004
0.001

1.20
1.73
0.90
0.74
0.61
0.50

0.781.82
1.132.65
0.471.75
0.461.19
0.480.77
0.380.67

0.405
0.011
0.761
0.21
<0.001
<0.001

1.032.94
1.222.78
1.262.31
0.170.56
0.980.99
0.950.97

0.039
0.004
<0.001
<0.001
<0.001
<0.001

1.62
1.81
1.42
0.29
0.99
0.96

1.022.56
1.242.64
1.071.87
0.170.50
0.980.99
0.950.96

0.041
0.002
0.015
<0.001
<0.001
<0.001

AOR = adjusted odds ratio; ICU = intensive care unit areas; PACU = postanesthesia care unit; PEA = pulseless electrical activity;
PVT/VF = pulseless ventricular tachycardia or ventricular fibrillation.

Anesthesiology 2013; 119:1322-39 1335 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Appendix 4. General Estimating Equation Models to Predict Survival to Discharge in Patients with Pulseless
Ventricular Tachycardia or Ventricular Fibrillation Arrests with and without Data Imputation for Missing Data
Preimputation

Postimputation for Missing Data

Risk Factor

AOR

95% CI

P Value

AOR

95% CI

P Value

Acute nonstroke neurological event


Baseline depression in neurological status
Congestive heart failure during admission
Hepatic insufficiency
Hypotension/hypoperfusion
Metastatic or hematologic malignancy
Metabolic and electrolyte abnormality
Renal insufficiency
Respiratory failure
Septicemia
Active or evolving myocardial infarction
Inadequate natural airway
Arrhythmia
Day hours
Event location
Operating room
PACU
Telemetry
ICU
Weekend
Invasive ventilation in place prearrest
White vs. nonwhite
Illness category
Medical cardiac
Surgical cardiac
Surgical non cardiac
Trauma
Age at hospital admission
Duration of event

0.99
0.64
0.58
1.19
0.34
0.83
0.56
0.58
0.72
0.83
1.21
2.15
1.07
1.55

0.472.07
0.241.69
0.291.19
0.324.48
0.220.54
0.431.60
0.291.10
0.350.97
0.481.08
0.322.13
0.562.61
0.2816.52
0.661.76
0.832.90

0.981
0.365
0.137
0.792
<0.001
0.569
0.095
0.039
0.111
0.692
0.636
0.461
0.775
0.169

0.99
0.68
0.66
0.99
0.40
0.70
0.60
0.52
0.81
0.65
0.97
1.57
1.04
1.48

0.551.78
0.311.52
0.361.18
0.323.09
0.270.60
0.441.13
0.311.16
0.340.81
0.571.15
0.291.48
0.481.98
0.445.66
0.701.55
0.952.32

0.979
0.348
0.162
0.992
<0.001
0.15
0.132
0.004
0.242
0.308
0.944
0.49
0.84
0.084

2.32
2.51
5.44
3.51
0.73
0.64
1.32

0.826.56
0.857.41
0.6545.79
1.1210.95
0.381.40
0.241.68
0.742.36

0.112
0.096
0.119
0.031
0.345
0.362
0.346

1.75
2.05
5.72
1.95
0.63
0.69
1.05

0.813.79
0.954.39
1.1029.66
1.065.07
0.381.04
0.361.32
0.641.70

0.158
0.066
0.038
0.049
0.071
0.267
0.854

7.34
2.30
1.66
0.12
0.97
0.96

2.0626.13
0.905.88
0.763.64
0.030.48
0.960.99
0.940.97

0.002
0.083
0.204
0.002
0.001
<0.001

3.64
2.42
1.80
0.22
0.98
0.96

1.528.72
1.184.95
1.013.20
0.070.71
0.970.99
0.950.97

0.004
0.015
0.045
0.012
0.001
<0.001

AOR = adjusted odds ratio; ICU = intensive care unit areas; PACU = postanesthesia care unit.

Anesthesiology 2013; 119:1322-39

1336 Ramachandran et al.

PERIOPERATIVE MEDICINE

Appendix 5. General Estimating Equation Models to Predict Survival to Discharge in Patients with Asystolic Arrests
with and without Data Imputation for Missing Data
Preimputation

Postimputation for Missing Data

Risk Factor

AOR

95% CI

P Value

AOR

95% CI

P Value

Acute nonstroke neurological event


Baseline depression in neurological status
Congestive heart failure during admission
Hepatic insufficiency
Hypotension/hypoperfusion
Metastatic or hematologic malignancy
Metabolic and electrolyte abnormality
Renal insufficiency
Respiratory failure
Septicemia
Active or evolving myocardial infarction
Inadequate natural airway
Arrhythmia
Day hours
Event location
Operating room
PACU
Telemetry
ICU
Weekend
Invasive ventilation in place prearrest
White vs. nonwhite
Illness category
Medical cardiac
Surgical cardiac
Surgical noncardiac
Trauma
Age at hospital admission
Duration of event

0.70
0.79
0.50
0.62
0.31
0.33
1.18
0.77
0.87
0.31
0.57
3.36
1.65
2.31

0.281.71
0.381.64
0.251.00
0.241.56
0.200.47
0.170.62
0.512.71
0.491.24
0.571.34
0.120.77
0.231.43
0.7215.72
1.172.33
1.343.97

0.428
0.526
0.05
0.308
<0.001
0.001
0.696
0.284
0.534
0.012
0.229
0.124
0.004
0.003

0.79
0.95
0.58
0.73
0.34
0.48
0.74
0.84
1.01
0.39
0.57
3.02
1.60
2.22

0.381.63
0.551.63
0.321.05
0.331.62
0.240.50
0.290.78
0.361.53
0.571.24
0.711.44
0.190.80
0.231.40
0.929.85
1.182.18
1.373.59

0.519
0.847
0.072
0.435
<0.001
0.003
0.414
0.387
0.952
0.01
0.22
0.068
0.003
0.001

2.52
3.36
0.57
0.99
0.54
0.17
1.35

1.135.63
1.398.13
0.132.40
0.392.49
0.320.94
0.070.40
0.802.28

0.024
0.007
0.442
0.979
0.028
<0.001
0.264

1.38
2.04
0.59
0.60
0.49
0.41
1.18

1.022.58
1.024.09
0.142.39
0.261.38
0.300.79
0.250.67
0.791.75

0.047
0.044
0.456
0.228
0.003
<0.001
0.427

1.78
1.03
1.88
0.25
0.98
0.95

0.684.69
0.442.40
0.933.79
0.070.93
0.970.99
0.930.97

0.24
0.947
0.078
0.039
0.003
<0.001

1.44
1.62
1.78
0.46
0.98
0.95

0.613.41
0.803.29
1.033.08
0.171.26
0.970.99
0.930.96

0.411
0.183
0.038
0.13
<0.001
<0.001

AOR = adjusted odds ratio; ICU, intensive care unit areas; PACU, postanesthesia care unit.

Anesthesiology 2013; 119:1322-39 1337 Ramachandran et al.

Perioperative Cardiac Arrests and Outcomes

Appendix 6. General Estimating Equation Models to Predict Survival to Discharge in Patients with Pulseless Electrical
Activity Arrests with and without Data Imputation for Missing Data
Preimputation

Postimputation for Missing Data


P Value

AOR

95% CI

0.772.78
0.351.35
0.321.13
0.241.69
0.551.19
0.421.18
0.271.04
0.491.05
0.541.21
0.160.92
0.010.62
0.804.57
0.921.82
0.803.33

0.246
0.275
0.116
0.364
0.28
0.183
0.063
0.091
0.299
0.031
0.015
0.144
0.132
0.174

1.49
0.61
0.68
0.54
0.78
0.64
0.46
0.66
0.92
0.32
0.19
1.53
1.34
1.52

0.772.91
0.301.23
0.381.20
0.191.51
0.521.16
0.391.06
0.240.89
0.450.97
0.611.40
0.140.78
0.050.69
0.693.39
0.981.84
0.862.71

0.239
0.168
0.184
0.239
0.212
0.082
0.021
0.033
0.707
0.011
0.011
0.297
0.067
0.153

0.66
1.19
0.64
0.51
0.67
0.31
1.07

0.281.58
0.472.99
0.142.91
0.201.32
0.371.19
0.140.70
0.681.69

0.355
0.71
0.568
0.166
0.169
0.005
0.769

0.52
0.89
0.41
0.35
0.49
0.53
0.88

0.251.08
0.411.93
0.091.83
0.150.80
0.280.85
0.320.87
0.591.32

0.079
0.773
0.242
0.012
0.011
0.012
0.551

1.51
1.99
1.31
0.27
0.98
0.96

0.534.31
0.924.29
0.742.33
0.090.79
0.970.99
0.940.97

0.437
0.08
0.351
0.017
0.001
<0.001

1.55
1.64
1.29
0.20
0.98
0.95

0.643.79
0.793.38
0.762.19
0.070.57
0.970.99
0.940.97

0.334
0.182
0.345
0.002
<0.001
<0.001

Risk factor

AOR

95% CI

Acute nonstroke neurological event


Baseline depression in neurological status
Congestive heart failure during admission
Hepatic insufficiency
Hypotension/hypoperfusion
Metastatic or hematologic malignancy
Metabolic and electrolyte abnormality
Renal insufficiency
Respiratory failure
Septicemia
Active or evolving myocardial infarction
Inadequate natural airway
Arrhythmia
Day hours
Event location
Operating room
PACU
Telemetry
ICU
Weekend
Invasive ventilation in place prearrest
White vs. nonwhite
Illness category
Medical cardiac
Surgical cardiac
Surgical noncardiac
Trauma
Age at hospital admission
Duration of event

1.46
0.68
0.60
0.64
0.81
0.70
0.53
0.72
0.81
0.38
0.09
1.91
1.30
1.64

P Value

AOR = adjusted odds ratio; ICU, intensive care unit areas; PACU = postanesthesia care unit.

Anesthesiology 2013; 119:1322-39 1338 Ramachandran et al.

PERIOPERATIVE MEDICINE

Appendix 7. General Estimating Equation Models to Predict Good Neurological Outcome in Patients with Good
Neurological Status at Admission, before, and after Data Imputation for Missing Data
Preimputation
Risk factor

AOR

Acute nonstroke neurological event


0.76
Baseline depression in neurological status
0.69
Congestive heart failure during admission
1.07
Hepatic insufficiency
0.66
Hypotension/hypoperfusion
0.71
Metastatic or hematologic malignancy
0.75
Metabolic and electrolyte abnormality
1.55
Renal insufficiency
0.72
Respiratory failure
0.88
Septicemia
0.28
Active or evolving myocardial infarction
0.96
Inadequate natural airway
0.44
Arrhythmia
0.83
Day hours
1.70
White vs. nonwhite
1.63
Arrest rhythm (PEA reference value)
Asystole
0.83
PVT/VF (shockable rhythms)
1.04
Event location (general in-patient unitreference)
Operating room
5.72
PACU
5.17
Telemetry
2.46
ICU
6.19
Weekend
1.34
Invasive ventilation in place prearrest
0.03
Age at hospital admission
0.98
Duration of event
0.98
Illness category (medical noncardiac reference value)
Medical cardiac
1.23
Surgical cardiac
1.31
Surgical noncardiac
1.49
Trauma
1.16

Postimputation for Missing Data

95% CI

P Value

AOR

95% CI

P Value

0.212.75
0.242.04
0.383.04
0.123.54
0.411.23
0.371.51
0.603.96
0.351.47
0.511.54
0.080.94
0.392.36
0.171.13
0.501.37
0.704.14
0.883.02

0.671
0.504
0.897
0.627
0.219
0.417
0.363
0.366
0.658
0.04
0.929
0.087
0.461
0.241
0.124

1.00
1.02
1.09
0.48
0.90
0.82
1.03
0.73
0.91
0.43
1.31
0.52
0.89
1.09
1.23

0.273.78
0.254.13
0.412.94
0.181.27
0.591.39
0.471.44
0.512.07
0.401.33
0.571.46
0.141.35
0.553.15
0.231.18
0.581.35
0.482.52
0.722.11

0.995
0.981
0.857
0.138
0.642
0.487
0.933
0.311
0.704
0.146
0.545
0.118
0.572
0.831
0.451

0.431.59
0.551.98

0.572
0.906

1.09
1.07

0.681.77
0.661.74

0.718
0.773

1.7518.65
1.6216.45
0.5910.21
1.5225.25
0.583.10
0.000.29
0.970.99
0.971.00

0.004
0.005
0.215
0.011
0.498
0.002
0.005
0.029

1.66
1.88
0.80
1.61
1.20
0.60
0.98
0.98

0.713.86
0.804.47
0.213.00
0.515.10
0.632.28
0.331.08
0.960.99
0.960.99

0.242
0.15
0.742
0.419
0.578
0.086
0.001
0.001

0.374.15
0.433.99
0.593.75
0.255.47

0.733
0.63
0.395
0.853

1.77
1.50
1.44
0.99

0.714.43
0.723.09
0.454.64
0.981.00

0.223
0.276
0.536
0.086

AOR = adjusted odds ratio; ICU = intensive care unit areas; PACU = postanesthesia care unit; PEA = pulseless electrical activity; PVT/
VF = pulseless ventricular tachycardia or ventricular fibrillation.

Anesthesiology 2013; 119:1322-39 1339 Ramachandran et al.

Intraoperative Cardiac Arrests in Adults Undergoing


Noncardiac Surgery
Incidence, Risk Factors, and Survival Outcome
Sumeet Goswami, M.D., M.P.H.,* Joanne E. Brady, M.S., Desmond A. Jordan, M.D.,
Guohua Li, M.D., Dr.P.H.

This article has been selected for the Anesthesiology CME Program. Learning objectives
and disclosure and ordering information can be found in the CME section at the front
of this issue.

ABSTRACT
Background: Intraoperative cardiac arrest (ICA) is a rare but
potentially catastrophic event. There is a paucity of recent
epidemiological data on the incidence and risk factors for
ICA. The objective of this study was to assess the incidence,
risk factors, and survival outcome of ICAs in adults undergoing noncardiac surgery.
Methods: The authors analyzed prospectively collected data
for all noncardiac cases in the American College of Surgeons
National Surgical Quality Improvement Program database
from the years 2005 to 2007 (n = 362,767).
Results: The incidence of ICA was 7.22 per 10,000 surgeries.
After adjustment for American Society of Anesthesiologists
physical status and other covariates, the odds of ICA increased
progressively with the amount of transfusion (adjusted odds
ratios = 2.51, 7.59, 11.40, and 29.68 for those receiving 13,
46, 79, and 10 units of erythrocytes, respectively). Other
* Assistant Professor of Anesthesiology, Senior Staff Associate,
Associate Professor of Clinical Anesthesiology, Divisions of Cardiothoracic Anesthesiology & Critical Care Medicine, Department
of Anesthesiology, College of Physicians and Surgeons, Columbia
University, New York, New York. Professor of Epidemiology and
Director of the Center for Health Policy and Outcomes in Anesthesia and Critical Care, Department of Anesthesiology, College of
Physicians and Surgeons, and Department of Epidemiology, Mailman School of Public Health, Columbia University.
Submitted for publication November 15, 2011. Accepted for
publication July 13, 2012. Support was provided solely from institutional and/or departmental sources. American College of Surgeons
National Surgical Quality Improvement Program (ACS NSQIP) and
the hospitals participating in the ACS NSQIP are the source of the
data used herein; they have not verified and are not responsible for
the statistical validity of the data analysis or the conclusions derived
by the authors.
Address correspondence to Dr. Goswami: Department of Anesthesiology, College of Physicians & Surgeons of Columbia University, 630 West 168th Street, PH 505, New York, New York 10032.
sg767@columbia.edu. This article may be accessed for personal use
at no charge through the Journal Web site, www.anesthesiology.org.

What We Already Know about This Topic


The reported incidence of intraoperative cardiac arrest varies
widely, in large part due to paucity of epidemiological data
that allows quantitative assessment of associated risk factors
This study determined the incidence, risk factors, and survival
outcome of intraoperative cardiac arrests in adults undergoing
noncardiac surgery

What This Article Tells Us That Is New


Intraoperative cardiac arrest occurs at a rate of approximately
7 per 10,000 noncardiac surgeries, with a 30-day mortality
rate of 63%
The most important risk factor is the amount of intraoperative
erythrocyte transfusion

significant risk factors for ICA were emergency surgery (adjusted


odds ratio = 2.04, 95% CI = 1.452.86) and being functionally
dependent presurgery (adjusted odds ratio = 2.33, 95% CI =
1.693.22). Of the 262 patients with ICA, 116 (44.3%) died
within 24h, and 164 (62.6%) died within 30 days.
Conclusions: Intraoperative blood loss as indicated by the
amount of transfusion was the most important predictor of
ICA. The urgency of surgery and the preoperative composite
indicators of health such as American Society of Anesthesiologists status and functional status were other important
risk factors. The high case fatality suggests that primary prevention might be the key to reducing mortality from ICA.

NTRAOPERATIVE cardiac arrest (ICA) is a rare but


potentially catastrophic event that is associated with high
mortality. The reported incidence of ICA varies considerably
across studies.17 One principal reason for the lack of consistency could be that the incidence of ICA has decreased with

Copyright 2012, the American Society of Anesthesiologists, Inc. Lippincott


Williams & Wilkins. Anesthesiology 2012; 117:101826

Anesthesiology, V 117 No 5 1018

This article is accompanied by an Editorial View. Please see:


Muoz E III, Pan W: Sometimes you have to revisit the past to
understand the present. ANESTHESIOLOGY 2012; 117:94850

November 2012

PERIOPERATIVE MEDICINE

improved technology and clinical practices.8,9 Study periods


vary from 5 to 18 yr,4,1012 and thus the impact of changing technology and clinical practices may result in variation
in the incidence of ICA across individual studies. Another
reason is that most studies are based on data from single
institutions and consequently suffer from limited external
validity.1,3,6,1214 Furthermore, the incidence of ICA also varies in different countries during similar periods, based on the
quality and the availability of health care.7,14 To increase the
study sample size, many previous studies report the combined incidence of ICA in adult and pediatric patients or
cardiac and noncardiac procedures.1,3,7 Finally, some studies have reported the combined incidence of cardiac risk
in patients suffering from perioperative cardiac arrest and
myocardial infarction,15 outcomes with potentially different
risk factors and incidence. Thus the reported incidence of
ICA ranges from 1.1 to 34.6 cardiac arrests per 10,000 anesthetics.7,12 Although there is a wide variation in the reported
incidence, the case fatality of ICA has remained consistently
high at approximately 6080% since the 1950s.1,57,10,14 The
paucity of epidemiological data prevents us from accurately
estimating the incidence of ICA and quantitatively assessing
the associated risk factors. The risk profile of our patients
continues to change as we operate on patients who are older
and sicker.1618 With the changing risk profile of our patients,
it is important to identify modifiable risk factors and intervene by changing our clinical practices to further decrease
the morbidity and the mortality in the operating room. The
goal of this study was to assess the incidence of and risk factors for ICA in adults undergoing noncardiac surgery. The
results of this study may help cardiologists, primary care
physicians, anesthesiologists, and surgeons to improve the
risk stratification of patients and develop interventions to
lower the incidence of ICA in high-risk patients.

Materials and Methods


Data
Data from this study came from the American College of
Surgeons National Surgical Quality Improvement Program
(ACS NSQIP). The ACS NSQIP is a nationally validated,
risk-adjusted, outcomes-based program to measure and
improve the quality of surgical care among all participating
hospitals. A surgical clinical reviewer captures the data on
136 variables, including preoperative patient characteristics,
intraoperative variables, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major
surgical procedures using standardized protocols. Participating hospitals submit data to ACS NSQIP through either the
general and vascular surgery module or the multispecialty
American College of Surgeons National Surgical Quality Improvement Program - Program Overview. Available at: http://site.acsnsqip.org/about/. Accessed April 13, 2012.
# ACS-NSQIP: User Guide for the 2008 Participant Use Data File.
Available at: http://site.acsnsqip.org/wp-content/uploads/2012/03/
Final-user-guide-08-PUF.pdf. Accessed April 13, 2012.

module. Each of the two modules includes a high- and a


low-volume category. The hospitals participating in the
high-volume module must submit a minimum of 1,680
cases, whereas the hospitals participating in the low-volume
module must include a minimum of 900 cases.
There is a systematic sampling system called the 8-day
cycle for the hospitals not able to capture all their surgical
cases. To obtain a representative sample of operations, the
first 40 consecutive eligible operations are entered into the
NSQIP in each 8-day cycle, with each cycle starting on the
different day of the week. There are 46, 8-day cycles in 1
yr, and the program requires that data be submitted for 42
of those cycles.# The hospitals participating in the highvolume, general, and vascular surgery model capture the
first 40 consecutive cases meeting the inclusion/exclusion
criteria in the 8-day cycle for a total of 1,680 cases annually.
The hospitals participating in the low-volume, general, and
vascular surgery model are required to submit all general
and vascular cases that meet the inclusion/exclusion criteria
in the 8-day cycle. A minimum of 900 cases must be
submitted annually. The hospitals participating in the highvolume, multispecialty model must submit approximately
20% of each of the following subspecialties: general,
gynecologic, neurologic, orthopedic, otolaryngology,
plastic, cardiac, thoracic, urologic, and vascular. If 20%
of the hospital volume is less than 1,680 cases annually,
the hospital must submit a higher percentage to reach
a minimum of 1,680 cases. The hospitals participating
in the low-volume, multispecialty model must submit
the maximum number of cases that meet the inclusion/
exclusion criteria, with a minimum of 900 cases annually.
The ACS NSQIP includes all cases receiving general, spinal,
or epidural anesthesia. Carotid endarterectomy, inguinal
herniorrhaphy, parathyroidectomy, thyroidectomy, breast
lumpectomy, and endovascular abdominal aortic aneurysm
repair cases are included regardless of the anesthesia
modality. The following patients are not included in ACS
NSQIP: those receiving monitored anesthesia care; those
receiving peripheral nerve block or local anesthesia; trauma
cases, transplant cases, concurrent cases; patients less than 16
yr of age; and brain-dead patients categorized as American
Society of Anesthesiology (ASA) physical classification 6.
Hospitals with a 30-day follow-up rate of less than 80%
and whose surgical volume does not meet eligibility criteria
are also excluded.
To support the acquisition of high-quality data, ACS
NSQIP from its start has relied on robust clinical data with
rigorous definitions, collected by trained and audited personnel. The mechanisms for collection of high-quality data
include: presence of a dedicated surgical clinical reviewer
for data collection at each institution; initial and ongoing
reviewer training, examination and support; creation and
continual review of rigorous data definitions; audits for
data integrity and interrater reliability.19 The high quality of
ACS NSQIP data has been recognized by the Institute of

Anesthesiology 2012; 117:101826 1019

Goswami et al.

Intraoperative Cardiac Arrests in Noncardiac Surgery

Medicine and the Joint Commission. The interrater reliability audit is a tool to assess the quality of data collected at the
participating site; it evaluates the disagreement rates between
the surgical clinical reviewer and the site reviewer. Interrater reliability is calculated as a percentage using the number
of disagreements divided by the total number of variables
reviewed. The overall disagreement on variables has been
extremely low from the beginning (3.15% in 2005) and has
steadily fallen (1.56% in 2008).19

0.05 and the significance level of the Wald chi-square for an


effect to stay in the model was 0.05. The model goodness of
fit was assessed with the Hosmer and Lemeshow test.23 Using
the same process, a stratified analysis of high- and moderaterisk procedures was performed to assess risk factors for each
surgical procedure group. Statistical analysis was performed
using SAS version 9.2 (SAS Institute, Cary, NC).

Study Sample
The study cohort included all the noncardiac cases in the ACS
NSQIP database for the years 20052007 (n = 362,767). This
consists of data from 121 institutions in 2005 and 2006 and
from 183 institutions in 2007. ICA is defined as the absence
of cardiac rhythm or presence of chaotic cardiac rhythm
that requires the initiation of any component of basic and/
or advanced cardiac life support. Patients who have automatic
implantable cardioverter defibrillator that fire but have no
loss of consciousness are not included as ICA cases. Excluded
from this study were 1,130 patients who underwent cardiac
surgery, as indicated by the Current Procedural Terminology
codes.20 Surgical procedures were divided into two categories
of low/moderate risk and high risk, based on the incidence of
cardiac death and nonfatal myocardial infarction as discussed
in the guidelines published by American College of Cardiology/American Hospital Association, the European Society of
Cardiology, and the European Society of Anaesthesiology.21,22
Patients having vascular surgery were included in the high-risk
group and patients undergoing other procedures, such as endovascular repair of abdominal aortic aneurysm and carotid endarterectomy, were included in the low/moderate-risk group.

Of the 362,767 patients studied, 262 had an ICA, yielding an incidence of 7.22 per 10,000 surgeries. Of the 262
patients with ICA, 116 (44.3%) died within 24h, and 164
(62.6%) died within 30 days. The incidence of ICA varied
significantly with preoperative patient and clinical variables
(table 1). Specifically, the risk of ICA increased with age,
ASA physical status classification, impaired functional status, presence of comorbidities, and surgical-risk level (table
1). Emergency surgery and participation by resident surgeon
were also associated with a substantially increased risk of ICA
(table 1). The highest incidence of ICA, 814.25 per 10,000
surgeries, was found among patients receiving intraoperative
transfusion of 10 or more erythrocyte units (table 1).
Complications associated with tracheal intubation such
as airway (lip, tongue, pharyngeal, or laryngeal laceration)
injury or failure to intubate were not observed in any of
the patients suffering from ICA. Of the 262 patients, one
patient suffered from a tooth injury. None of the patients
who could not be intubated progressed to an ICA.
Multivariable logistic regression modeling revealed that
the risk of ICA increased progressively with the amount of
intraoperative transfusion; compared with patients with no
transfusion, the adjusted odds ratios were 2.51, 7.59, 11.40,
and 29.68 for those receiving 13, 46, 79, and 10 or more
erythrocyte units of transfusion, respectively (fig. 1; table 2).
Excluding the patients receiving more than 4 units of blood
72h before surgery (n = 1,573) did not change the results in
any meaningful way.
The odds of ICA also increased in a doseresponse fashion
with ASA physical status classification. Other risk factors for
ICA were emergency surgery (adjusted rate ratio = 2.04, 95%
CI = 1.45 2.86), and being functionally dependent presurgery
(adjusted rate ratio = 2.33, 95% CI = 1.693.22) (table 2).
The Hosmer-Lemeshow test indicated that the multivariable
logistic regression model (table 2) fitted the data adequately
(chi-square = 9.484, df = 8, P = 0.302). In patients undergoing
low- or moderate-risk procedures, the odds of ICA was lower
in patients with higher BMI. Compared with patients with
BMI less than 25, the adjusted odds ratios were 0.84 and 0.31
for patients with BMI between 3039 and 40 and higher,
respectively. Stratification analysis indicated that the amount of
intraoperative transfusion was the most important predictor of
ICA regardless of the risk level of surgical procedures. The odds
of ICA associated with emergency surgery and preoperative
functional status, however, appeared to be more pronounced

Statistical Analysis
The study protocol was reviewed and judged to be exempt
by the Columbia University Institutional Review Board.
The incidence, odds ratios, and 95% CI of ICA were calculated according to demographic and preoperative clinical characteristics, such as age, sex, race, body mass index
(BMI), and emergency case status and patient comorbidities,
including but not limited to dyspnea, chronic obstructive
pulmonary disease, heart disease, hypertension, impaired
sensorium, coma, cortical disease, spinal disease, renal failure
or on dialysis, diabetes, sepsis, and bleeding disorders. Preoperative and intraoperative transfusion histories were also
examined. A subset analysis was also performed by excluding the patients who received more than 4 units of blood
72h before surgery. Interactions between surgical procedure
risk and preoperative clinical characteristics on ICA were
assessed. Multivariable logistic regression was used to choose
predictors of ICA. Age, sex, race, procedure risk, renal disease, heart disease, cerebrovascular disease, and dyspnea were
forced into the model, and the remaining variables were
chosen using stepwise selection. The significance level of the
score chi-square for entering an effect into the model was
Anesthesiology 2012; 117:101826

Results

1020

Goswami et al.

PERIOPERATIVE MEDICINE

Table 1. Incidence and 95% CI of Intraoperative Cardiac Arrest per 10,000 Surgeries by Pre- and Intraoperative
Characteristics, American College of Surgeons National Surgical Quality Improvement Program, 20052007
Characteristic
Age, yr
1649
5069
70
Sex
Female
Male
Race
White
Black
Other/unknown
Body mass index, kg/m2
<25
25<30
30<40
40
Surgical-risk level
Low/ moderate risk
High risk (vascular)
Emergency surgery
No
Yes
Resident surgeon participation
No
Yes
ASA physical status classification
12
3
4
5
Functional health status before surgery
Independent
Partially or totally Dependent
Dyspnea
No
Moderate exertion
At rest
Severe chronic obstructive pulmonary disease
No
Yes
Current smoker within 1 yr
No
Yes
Heart disease||
No
Yes
Hypertension requiring medications
No
Yes

Anesthesiology 2012; 117:101826

No. of Surgeries* Number of Intraoperative


(n = 362,767)
Cardiac Arrests (n = 262)

Incidence per 10,000


Surgeries (95% CI)

138,804
143,127
80,834

34
98
130

2.45 (1.633.27)
6.85 (5.498.20)
16.08 (13.3218.85)

208,918
153,829

110
152

5.27 (4.286.25)
9.88 (8.3111.45)

256,798
35,339
70,630

183
27
52

7.13 (6.098.16)
7.64 (4.7610.25)
7.36 (5.369.36)

108,584
108,658
95,177
39,295

76
66
53
9

7.00 (5.438.57)
6.07 (4.617.54)
5.57 (4.077.07)
2.29 (0.793.79)

346,048
16,719

180
82

5.20 (4.445.96)
49.05 (38.4659.64)

315,352
47,415

97
165

3.08 (2.463.69)
34.80 (29.5040.10)

131,083
220,053

64
190

4.88 (3.696.08)
8.63 (7.419.86)

206,879
131,007
22,794
1,145

18
67
98
78

0.87 (0.461.27)
5.11 (3.896.34)
42.99 (34.5051.49)
681.22 (535.07827.38)

337,775
24,992

108
154

3.20 (2.593.80)
61.62 (51.9271.32)

321,006
36,254
5,507

175
40
47

5.45 (4.6410.96)
11.03 (7.6214.45)
85.35 (61.04109.65)

346,337
16,430

226
36

6.53 (5.677.38)
21.91 (14.7629.06)

286,386
76,381

198
64

6.91 (5.957.88)
8.38 (6.3310.43)

322,312
40,455

163
99

5.06 (4.285.83)
24.47 (19.6629.29)

202,194
160,573

94
168

4.65 (3.715.59)
10.46 (8.8812.04)
(continued)

1021

Goswami et al.

Intraoperative Cardiac Arrests in Noncardiac Surgery

Table 1. (Continued)
Characteristic

No. of Surgeries* Number of Intraoperative


(n = 362,767)
Cardiac Arrests (n = 262)

Impaired sensorium#
No
359,312
Yes
3,455
Coma**
No
362,452
Yes
315
Cerebrovascular disease
No
338,849
Yes
23,918
Spinal disorder
No
360,838
Yes
1,929
Renal disease
No
353,129
Yes
9,638
Diabetes mellitus with oral agents or insulin
No
311,481
Yes
51,286
Preoperative sepsis
No
328,804
SIRS||||/sepsis
30,212
Septic shock
3,751
Bleeding disorders
No
341,136
Yes
21,631
Transfusion of > 4 units of packed erythrocytes in 72h before surgery
No
361,194
Yes
1,573
No. of erythrocyte units transfused intraoperatively
0
343,054
13
14,937
46
3,103
79
694
10
786

Incidence per 10,000


Surgeries (95% CI)

220
42

6.12 (5.316.93)
121.56 (85.01158.12)

244
18

6.73 (5.897.58)
571.43 (313.70829.16)

219
43

6.46 (5.617.32)
17.98 (12.6123.35)

258
4

7.15 (6.288.02)
20.74 (0.4141.05)

221
41

6.26 (5.437.08)
42.54 (29.5455.54)

203
59

6.52 (5.627.41)
11.50 (8.5714.44)

139
58
65

4.23 (3.524.93)
19.20 (14.2624.13)
173.29 (131.51215.07)

202
60

5.92 (5.116.74)
27.74 (20.7334.75)

242
20

6.70 (5.867.54)
127.15 (71.72182.57)

98
42
38
16
64

2.86 (2.293.42)
28.12 (19.6336.61)
122.46 (83.74161.18)
230.55 (118.62342.48)
814.25 (622.641005.86)

*Data were missing for 2 patients on age, 20 patients on sex, 11,053 patients on body mass index, 11,631 patients on resident
surgeon participation, 942 patients on ASA physical status classification, and 193 patients on number of erythrocytes transfused.
Patients undergoing vascular surgery were included in the high-risk group, and all the other patients were included in the low/moderate-risk group. This was based on the guidelines published by American College of Cardiology.23Performed within 12h of admission or onset of symptomatology. Trauma cases were excluded from the data set. This variable focuses on the patients abilities to
perform activities of daily living. Activities of daily living are defined as the activities usually performed in the course of a normal day in
a persons life. Activities of daily living include: bathing, feeding, dressing, using the toilet, and being mobile. ||Congestive heart failure
in 30 days before the current surgery and/or myocardial infarction in 6 months before the current surgery and/or angina in 1 month
before the current surgery and/or previous percutaneous coronary intervention, and/or any previous cardiac surgery. #Mental status
changes and/or delirium in the context of the current illness. Patients with chronic or long-standing mental status changes were not
included. **Patient is unconscious or postures to painful stimuli or is unresponsive to all stimuli entering surgery.Hemiplegia and/or
history of transient ischemic attack and/or history of cerebrovascular accident.Paraplegia and/or quadriplegia. Acute renal failure
and/or on dialysis. Acute renal failure was defined as rapid, steadily increasing azotemia and a rising creatinine of > 3mg/ dl (265
mol/l) within 24h before surgery. ||||Systemic inflammatory response syndrome. The syndrome is recognized by the presence of 2
of the following within the same time frame: temperature > 38C or < 36C, heart rate > 90 beats/min, respiratory rate > 20 breaths/
min or partial pressure of carbon dioxide in arterial blood (PaCO2) < 32 mmHg, leukocyte count > 12,000 cell/mm3 (12109 cell/l), <
4,000 cell/mm3 (4109 cell/l) or > 10% immature (band) forms.
ASA = American Society of Anesthesiologists; SIRS = systemic inflammatory response syndrome.

Anesthesiology 2012; 117:101826

1022

Goswami et al.

PERIOPERATIVE MEDICINE

Table 2. (Continued)
Characteristic

Fig. 1. The adjusted odds ratios and 95% CI of intraoperative cardiac arrest according to the amount of intraoperative
erythrocyte transfusion in patients undergoing noncardiac
surgery, American College of Surgeons National Surgical
Quality Improvement Program, 20052007.

Table 2. Estimated Odds Ratios and 95% CI for


Intraoperative Cardiac Arrest by Pre- and Intraoperative
Characteristics, American College of Surgeons National
Surgical Quality Improvement Program, 20052007
Characteristic

Age, yr
1649
1.00
5069
1.09 (0.711.65)
70
1.17 (0.771.80)
Sex
Female
1.00
Male
1.00 (0.771.31)
Race
White
1.00
Black
0.99 (0.641.53)
Other/unknown
1.45 (1.052.00)
Surgical risk*
Low/ moderate risk
1.00
High risk (vascular)
1.11 (0.791.55)
Emergency surgery
No
1.00
Yes
2.04 (1.452.86)
ASA physical status classification
12
1.00
3
3.71 (2.136.47)
4
8.75 (4.7116.25)
5
30.27 (15.1160.62)
Functional health status before surgery
Independent
1.00
Partially or totally dependent
2.33 (1.693.22)
(continued)

Anesthesiology 2012; 117:101826

13
46
79
10

OR (95% CI)

OR (95% CI)

Heart disease
No
1.00
Yes
1.48 (1.111.96)
Coma||
No
1.00
Yes
1.91 (1.063.42)
Impaired sensorium#
No
1.00
Yes
1.12 (0.761.64)
Cerebrovascular disease**
No
1.00
Yes
1.09 (0.761.57)
Renal disease
No
1.00
Yes
1.37 (0.942.01)
Diabetes mellitus with oral agents or insulin
No
1.00
Yes
0.89 (0.651.22)
No. of erythrocyte units given intraoperatively
0
1.00
2.51 (1.693.71)
7.59 (4.9411.67)
11.40 (6.2220.88)
29.68 (18.6647.18)

*Patients undergoing vascular surgery were included in the


high-risk group, all the other patients were included in the low/moderate risk group. This was based on the guidelines published by American College of Cardiology.23 Performed within
12h of admission or onset of symptomatology. Trauma cases
were excluded from the data set. This variable focuses on the
patients abilities to perform activities of daily living. Activities of
daily living are defined as the activities usually performed in the
course of a normal day in a persons life. Activities of daily living
include: bathing, feeding, dressing, using the toilet, and being
mobile. Congestive heart failure in 30 days before the current
surgery and/or myocardial infarction in 6 months before the current surgery and/or angina in 1 month before the current surgery
and/or previous percutaneous coronary intervention and/or any
previous cardiac surgery. ||Patient is unconscious or postures to
painful stimuli or is unresponsive to all stimuli entering surgery.
#Mental status changes and/or delirium in the context of the
current illness. Patients with chronic or long-standing mental
status changes were not included. **Hemiplegia and/or history
of transient ischemic attack and/or history of cerebrovascular
accident. Acute renal failure and/or on dialysis. Acute renal
failure was defined as rapid, steadily increasing azotemia and
a rising creatinine of > 3mg/ dl (265 mol/l) within 24h before
surgery.
ASA = American Society of Anesthesiologists.

in high-risk surgical procedures than in low/moderate surgical


procedures.

Discussion
The amount of intraoperative erythrocyte transfusion, a
surrogate measure of intraoperative blood loss, appears

1023

Goswami et al.

Intraoperative Cardiac Arrests in Noncardiac Surgery

to be the most important predictor of ICA. The complications associated with intraoperative blood loss and
blood transfusions are well-known risk factors for cardiac
arrests. A review article by Zuercher et al.9 indicates that
the cardiovascular complications from hypovolemia from
blood loss and hyperkalemia from transfusion of stored
erythrocytes are the most common causes of ICA. A study
by Smith et al.24 indicates that acidosis, hypothermia, and
hypocalcemia in association with hyperkalemia are contributory factors in patients suffering from transfusionassociated ICA. The temporal relationship between ICA
and blood transfusion is not recorded in the ACS NSQIP
database, thus it is unclear if ICA was a complication of
rapid blood loss or a complication associated with rapid
blood transfusion. Previous preparation by the surgical
team to rapidly transfuse blood as needed, good communication between the surgeons and the anesthesiologists, and point-of-care electrolyte testing may decrease
the incidence of ICA associated with blood transfusion.
We also found that the composite indicators of preoperative patient health were important predictors of ICA. High
ASA physical status was found to be a significant predictor
of ICA even after adjusting for various comorbid conditions
such as heart disease, cerebrovascular disease, renal disease,
diabetes, and preoperative sepsis. Previous studies have demonstrated similar association between ICA and higher ASA
physical status.7,13 Braz et al.7 analyzed data from 53,718 consecutive anesthetics over a 9-yr period, 92% of the patients
experiencing cardiac arrests had an ASA classification of 3
or higher. Newland et al.13 analyzed data from 72,959 consecutive anesthetics over a 10-yr period; 96% of the patients
experiencing cardiac arrests had an ASA classification of 3 or
higher. Although controversy exists regarding the reliability
of ASA physical status classification,25,26 this study demonstrates its robustness in predicting ICA even after adjusting
for other comorbid conditions.
Functional capacity is another composite indicator of
patient health. Decreased functional capacity can be caused
by several factors such as advanced age, poor pulmonary
reserve, deconditioning, and cardiac disease.21 Poor functional capacity has been associated with increased incidence
of postoperative cardiac events.22 Preoperative functional
status is also an important part of preoperative cardiac risk
assessment as per guidelines published by both American
college of Cardiology and European Society of Cardiology.21,22 We found that preoperative functional status is
another global indicator of patient health, which is a significant predictor of ICA.
The urgency of surgery is a well-established preoperative risk factor for ICA.3,7 Newland et al.13 reported that
60% of all cardiac arrests occurred in patients of having
emergency surgery. Our study indicates that the risk of
ICA associated with urgency of surgery was more pronounced in high-risk surgical procedures. The heightened

Anesthesiology 2012; 117:101826

risk of ICA in emergency surgeries is likely due to multiple factors. It is difficult to evaluate and optimize a
patient before surgical emergencies such as ruptured
aortic aneurysms, perforated or ischemic viscus, ischemic limb, or trauma. Furthermore, patients that require
emergency high-risk (vascular) procedures will have an
even higher risk of cardiac complications as either the
patients are likely to have additional cardiac risk factors or the patients might experience substantial intraoperative fluctuation in volume status, heart rate, and
blood pressure.
Historically, respiratory events have played a major
role in anesthesia-related cardiac arrests.3,4,6,27 Although
the ACS NSQIP database did not record any direct information related to the ease of ventilation or tracheal intubation, it did record some surrogate indicators of difficult
tracheal intubation such as tooth, tongue, pharyngeal or
laryngeal injury. The lack of association in the current
study between ICA and the inability to perform tracheal
intubation, or presence of surrogate measures of difficult
tracheal intubation, most likely reflects the improvement
in monitoring and clinical practices. Some of the changes
in monitoring include widespread adoption of pulse
oximetry, capnography, disconnection alarms, and lowpressure alarms.9 Improvement in clinical practices such
as adoption of a standardized algorithm for management
of difficult airway might also have decreased the incidence
of airway-related cardiac events.28
An intriguing finding of this study was that patients with
higher BMI had lower incidence of ICA. Previous studies
have reported that higher BMI is an independent predictor of short-term survival after in-hospital cardiopulmonary
arrest29 and of long-term survival after ventricular fibrillation
in out-of-hospital cardiac arrests.30 Whether morbid obesity
confers any protection against the occurrence of ICAs warrants further investigation.
This study has several notable limitations. First, due to
data limitations, it is not clear whether blood transfusion
was in response to ICA or whether complications associated
with blood transfusion resulted in ICA. Second, due to data
limitations, it is unclear whether the patients were medically
optimized before the surgery. Thus, it is not clear whether
some of these risk factors are modifiable. Data limitations
also make it difficult to analyze the role of known factors such
as anaphylactic shock in causing ICA. Another limitation of
this study is that the ACS NSQIP participating hospitals are
not a nationally representative sample, making it impossible
to generalize the findings from this study to other hospitals. Participation by hospitals is voluntary, and it is possible
that the hospitals with poor outcomes may not participate
in the ACS NSQIP program. Furthermore, smaller community hospitals may not be able to afford participation in
the ACS NSQIP program. Finally, the data from the current
study cannot be extrapolated to patients receiving monitored

1024

Goswami et al.

PERIOPERATIVE MEDICINE

anesthesia care, patients receiving peripheral nerve block or


local anesthesia, trauma cases, transplant cases, cardiac cases,
and patients less than 16 yr of age.
In conclusion, this large observational study indicates
that ICA occurs at a rate of approximately 7 per 10,000 noncardiac surgeries, with a 30-day mortality rate of 63%. The
most important risk factor is the amount of intraoperative
erythrocyte transfusion. Other significant predictors of ICA
include the urgency of surgery and the composite indicators of patient health such as ASA physical status and the
functional status before surgery. The high case fatality suggests that primary prevention might be the key to reducing
mortality from ICA.

References
1. Ament R, Papper EM, Rovenstine EA: Cardiac arrest during
anesthesia; a review of cases. Ann Surg 1951; 134:2207
2. Stephenson HE Jr, Reid LC, Hinton JW: Some common
denominators in 1200 cases of cardiac arrest. Ann Surg 1953;
137:73144
3. Keenan RL, Boyan CP: Cardiac arrest due to anesthesia. A
study of incidence and causes. JAMA 1985; 253:23737
4. Olsson GL, Halln B: Cardiac arrest during anaesthesia. A computer-aided study in 250,543 anaesthetics. Acta
Anaesthesiol Scand 1988; 32:65364
5. Aubas S, Biboulet P, Daures JP, du Cailar J: [Incidence and
etiology of cardiac arrest occurring during the peroperative
period and in the recovery room. Apropos of 102,468 anesthesia cases]. Ann Fr Anesth Reanim 1991; 10:43642
6. Girardi LN, Barie PS: Improved survival after intraoperative cardiac arrest in noncardiac surgical patients. Arch Surg
1995; 130:158; discussion 19
7. Braz LG, Mdolo NS, do Nascimento P Jr, Bruschi BA, Castiglia
YM, Ganem EM, de Carvalho LR, Braz JR: Perioperative cardiac arrest: A study of 53,718 anaesthetics over 9 yr from a
Brazilian teaching hospital. Br J Anaesth 2006; 96:56975
8. Richard L, Keenan CPB: Decreasing frequency of anesthetic
cardiac arrests. J Clin Anesth 1991; 3: 3547
9. Zuercher M, Ummenhofer W: Cardiac arrest during anesthesia. Curr Opin Crit Care 2008; 14:26974
10. Otteni JC, Pottecher T, Tiret L, Hatton F, Desmonts JM:
[Cardiac arrest during anesthesia and the recovery period.
Data from the INSERM survey 1978-1982]. Ann Fr Anesth
Reanim 1986; 5:28794
11. Chopra V, Bovill JG, Spierdijk J: Accidents, near accidents
and complications during anaesthesia. A retrospective analysis of a 10-year period in a teaching hospital. Anaesthesia
1990; 45:36
12. An JX, Zhang LM, Sullivan EA, Guo QL, Williams JP:
Intraoperative cardiac arrest during anesthesia: A retrospective study of 218,274 anesthetics undergoing non-cardiac
surgery. Chin Med J 2011; 124:22732
13. Newland MC, Ellis SJ, Lydiatt CA, Peters KR, Tinker JH,
Romberger DJ, Ullrich FA, Anderson JR: Anesthetic-related
cardiac arrest and its mortality: A report covering 72,959
anesthetics over 10 years from a US teaching hospital.
ANESTHESIOLOGY 2002; 97:10815
14. Sprung J, Warner ME, Contreras MG, Schroeder DR, Beighley
CM, Wilson GA, Warner DO: Predictors of survival following cardiac arrest in patients undergoing noncardiac surgery: A study of 518,294 patients at a tertiary referral center.
ANESTHESIOLOGY 2003; 99:25969
15. Gupta PK, Gupta H, Sundaram A, Kaushik M, Fang X, Miller
WJ, Esterbrooks DJ, Hunter CB, Pipinos II, Johanning JM,
Lynch TG, Forse RA, Mohiuddin SM, Mooss AN: Development

Anesthesiology 2012; 117:101826

and validation of a risk calculator for prediction of cardiac


risk after surgery. Circulation 2011; 124:3817
16. Roques F, Nashef SA, Michel P, Gauducheau E, de Vincentiis
C, Baudet E, Cortina J, David M, Faichney A, Gabrielle F,
Gams E, Harjula A, Jones MT, Pintor PP, Salamon R, Thulin
L: Risk factors and outcome in European cardiac surgery:
Analysis of the EuroSCORE multinational database of 19030
patients. Eur J Cardiothorac Surg 1999; 15:81622; discussion
8223
17. Ferguson TB Jr, Hammill BG, Peterson ED, DeLong ER,
Grover FL; STS National Database Committee: A decade of
changerisk profiles and outcomes for isolated coronary
artery bypass grafting procedures, 1990-1999: A report from
the STS National Database Committee and the Duke Clinical
Research Institute. Society of Thoracic Surgeons. Ann Thorac
Surg 2002; 73:4809; discussion 48990
18. Biancari F, Kangasniemi OP, Aliasim Mahar M, Rasinaho E,
Satomaa A, Tiozzo V, Niemel M, Lepojrvi M: Changing
risk of patients undergoing coronary artery bypass surgery.
Interact Cardiovasc Thorac Surg 2009; 8:404
19. Shiloach M, Frencher SK Jr, Steeger JE, Rowell KS, Bartzokis
K, Tomeh MG, Richards KE, Ko CY, Hall BL: Toward robust
information: Data quality and inter-rater reliability in the
American College of Surgeons National Surgical Quality
Improvement Program. J Am Coll Surg 2010; 210:616
20. American Medical Association: Current Procedural Termi
nology CPT 2008 Professional Edition. Chicago, American
Medical Association, 2007; 14357
21. Fleisher LA, Beckman JA, Brown KA, Calkins H, Chaikof
E, Fleischmann KE, Freeman WK, Froehlich JB, Kasper EK,
Kersten JR, Riegel B, Robb JF, Smith SC Jr, Jacobs AK, Adams
CD, Anderson JL, Antman EM, Buller CE, Creager MA, Ettinger
SM, Faxon DP, Fuster V, Halperin JL, Hiratzka LF, Hunt SA,
Lytle BW, Nishimura R, Ornato JP, Page RL, Tarkington LG,
Yancy CW; American College of Cardiology/American Heart
Association Task Force on Practice Guidelines (Writing
Committee to Revise the 2002 Guidelines on Perioperative
Cardiovascular Evaluation for Noncardiac Surgery); American
Society of Echocardiography; American Society of Nuclear
Cardiology; Heart Rhythm Society; Society of Cardiovascular
Anesthesiologists; Society for Cardiovascular Angiography
and Interventions; Society for Vascular Medicine and
Biology; Society for Vascular Surgery: ACC/AHA 2007 guidelines on perioperative cardiovascular evaluation and care
for noncardiac surgery: A report of the American College
of Cardiology/American Heart Association Task Force on
Practice Guidelines (Writing Committee to Revise the 2002
Guidelines on Perioperative Cardiovascular Evaluation for
Noncardiac Surgery): developed in collaboration with the
American Society of Echocardiography, American Society
of Nuclear Cardiology, Heart Rhythm Society, Society of
Cardiovascular Anesthesiologists, Society for Cardiovascular
Angiography and Interventions, Society for Vascular Medicine
and Biology, and Society for Vascular Surgery. Circulation
2007; 116:e41899
22. Poldermans D, Bax JJ, Boersma E, De Hert S, Eeckhout
E, Fowkes G, Gorenek B, Hennerici MG, Iung B, Kelm
M, Kjeldsen KP, Kristensen SD, Lopez-Sendon J, Pelosi P,
Philippe F, Pierard L, Ponikowski P, Schmid JP, Sellevold
OF, Sicari R, Van den Berghe G, Vermassen F; Task Force
for Preoperative Cardiac Risk Assessment and Perioperative
Cardiac Management in Non-cardiac Surgery; European
Society of Cardiology (ESC): Guidelines for pre-operative
cardiac risk assessment and perioperative cardiac management in non-cardiac surgery. Eur Heart J 2009; 30:2769812
23. Hosmer DW, Lemeshow S. Applied Logistic Regression. New
York, Wiley, 2000
24. Smith HM, Farrow SJ, Ackerman JD, Stubbs JR, Sprung J:
Cardiac arrests associated with hyperkalemia during red
blood cell transfusion: a case series. Anesth Analg 2008;
106:106291

1025

Goswami et al.

Intraoperative Cardiac Arrests in Noncardiac Surgery

25. Mak PH, Campbell RC, Irwin MG; American Society of


Anesthesiologists: The ASA Physical Status Classification:
Inter-observer
consistency.
American
Society
of
Anesthesiologists. Anaesth Intensive Care 2002; 30:
63340
26. Haynes SR, Lawler PG: An assessment of the consistency
of ASA physical status classification allocation. Anaesthesia
1995; 50:1959
27. Keenan RL, Boyan CP: Decreasing frequency of anesthetic
cardiac arrests. J Clin Anesth 1991; 3:3547
28. American Society of Anesthesiologists Task Force on
Management of the Difficult Airway: Practice guidelines for

Anesthesiology 2012; 117:101826

management of the difficult airway: An updated report by


the American Society of Anesthesiologists Task Force on
Management of the Difficult Airway. ANESTHESIOLOGY 2003;
98: 126977.
29. Danciu SC, Klein L, Hosseini MM, Ibrahim L, Coyle BW,
Kehoe RF: A predictive model for survival after in-hospital cardiopulmonary arrest. Resuscitation 2004; 62:
3542
30. Bunch TJ, White RD, Lopez-Jimenez F, Thomas RJ: Association
of body weight with total mortality and with ICD shocks
among survivors of ventricular fibrillation in out-of-hospital
cardiac arrest. Resuscitation 2008; 77:3515

1026

Goswami et al.

Articles

Perioperative and anaesthetic-related mortality in


developed and developing countries: a systematic review
and meta-analysis
Daniel Bainbridge, Janet Martin, Miguel Arango, Davy Cheng, for the Evidence-based Peri-operative Clinical Outcomes Research (EPiCOR) Group

Summary
Background The magnitude of risk of death related to surgery and anaesthesia is not well understood. We aimed to
assess whether the risk of perioperative and anaesthetic-related mortality has decreased over the past ve decades and
whether rates of decline have been comparable in developed and developing countries.
Methods We did a systematic review to identify all studies published up to February, 2011, in any language, with a
sample size of over 3000 that reported perioperative mortality across a mixed surgical population who had undergone
general anaesthesia. Using standard forms, two authors independently identied studies for inclusion and extracted
information on rates of anaesthetic-related mortality, perioperative mortality, cardiac arrest, American Society of
Anesthesiologists (ASA) physical status, geographic location, human development index (HDI), and year. The primary
outcome was anaesthetic sole mortality. Secondary outcomes were anaesthetic contributory mortality, total perioperative
mortality, and cardiac arrest. Meta-regression was done to ascertain weighted event rates for the outcomes.
Findings 87 studies met the inclusion criteria, within which there were more than 214 million anaesthetic
administrations given to patients undergoing general anaesthesia for surgery. Mortality solely attributable to
anaesthesia declined over time, from 357 per million (95% CI 324394) before the 1970s to 52 per million (4264) in
the 1970s80s, and 34 per million (2939) in the 1990s2000s (p<000001). Total perioperative mortality decreased
over time, from 10 603 per million (95% CI 10 42310 784) before the 1970s, to 4533 per million (44054664) in the
1970s80s, and 1176 per million (11481205) in the 1990s2000s (p<00001). Meta-regression showed a signicant
relation between risk of perioperative and anaesthetic-related mortality and HDI (all p<000001). Baseline risk status
of patients who presented for surgery as shown by the ASA score increased over the decades (p<00001).

Lancet 2012; 380: 107581


See Comment page 1038
Department of Anesthesia and
Perioperative Medicine
(D Bainbridge MD,
J Martin PharmD, M Arango MD,
Prof D Cheng MD) and
Department of Clinical
Epidemiology and Biostatistics
(J Martin), University of
Western Ontario, London, ON,
Canada
Correspondence to:
Dr Daniel Bainbridge,
Department of Anesthesia and
Perioperative Medicine, London
Health Sciences Centre
University Hospital,
339 Windermere Road,
Room C3-106, London, ON,
Canada N6A 5A5
daniel.bainbridge@lhsc.on.ca

Interpretation Despite increasing patient baseline risk, perioperative mortality has declined signicantly over the past
50 years, with the greatest decline in developed countries. Global priority should be given to reducing total
perioperative and anaesthetic-related mortality by evidence-based best practice in developing countries.
Funding University of Western Ontario.

Background
Death is one of the most feared complications of surgery,
yet the magnitude of risk of death related to surgery and
anaesthesia is not well understood. Fortunately, mortality
rates in the perioperative period are low.1 However, this
rarity makes quantication of perioperative mortality
dicult in individual studies.
More than 230 million major surgical procedures are
undertaken annually worldwide.2 Since fewer than 4% of
the major surgical procedures worldwide are done in
developing countries,2,3 there has been a substantial
paucity of research to assess whether perioperativerelated and anaesthetic-related mortality are greater in
developing than in developed countries.3,4 The recent
introduction of the Safe Surgery Checklist has further
focused eorts on reducing surgical morbidity and
mortality in both developed and developing countries.5
Although there is a widespread opinion that surgery and
general anaesthesia have become safer over time, a
comprehensive, systematic global analysis of evidence on
this topic has not been done.
www.thelancet.com Vol 380 September 22, 2012

In response to a call for action on gaps in surgery safety


in low-resource settings,6 we aimed to synthesise the
available global data on anaesthetic and surgery-related
deaths in high-income versus low-income settings to
assess, through meta-regression, whether the risk of
perioperative and anaesthesia-related mortality has
decreased progressively over time and whether temporal
trends dier for developed versus developing countries.

Methods
Search strategy and selection criteria
We did a systematic review with meta-analysis and metaregression in accordance with recent methodological
guidelines.7,8 The research question, search strategy, inclusion criteria, and statistical analyses were prespecied.
We did a systematic search to identify all observational or
randomised studies that reported perioperative mortality,
anaesthetic-related mortality, or perioperative cardiac
arrest. Two investigators (DB and JM) systematically
searched Medline, Cochrane Central, Scopus, and
Current Contents databases from the date of inception
1075

Articles

See Online for appendix

until April 7, 2009. Routine additional searches of


Medline were done thereafter until Feb 3, 2011. Search
terms included variants of anesthetic, anaesthetic,
anesthesia, anaesthesia, surgery, operation, intraoperative, perioperative, perisurgical, postoperative, postsurgical, death, mortality, survival, and cardiac arrest,
using text words and Medical Subject Headings (MeSH)
terms (appendix). Tangential electronic exploration of
related articles (ie, using links to related references to
search for additional articles) and extensive hand
searches of bibliographies of relevant reviews and related
journals were also done.
Studies in any language were included if they reported
on a population of at least 3000 patients who underwent
general anaesthesia for surgery in the hospital setting.
The minimum sample size of 3000 was chosen to
reasonably estimate adverse events that occur at a rate of
one in 1000 or less in accordance with the rule of three
sample size approximation9 and also to ensure that small
studies would not skew the event rate estimates since
occurrence of death and cardiac arrest were expected to
be lower than one in 1000 in most settings. Because the
aim was to ascertain outcomes in allcomers (ie, all
reported patients who underwent surgery), studies
reporting exclusively on regional or local anaesthesia or
those done in a non-hospital setting were excluded.
Similarly, to ensure the studies represented a mixed
population rather than being confounded by any one
particular surgical or patient subtype, studies that focused
exclusively on specic age groups (eg, the elderly only or
children only) or that reported only one surgical subgroup
(eg, cardiac surgery only) were excluded. Studies that
reported an estimated rather than actual denominator or
that sampled only a portion of adverse events from a
larger population and studies that reported only corrected

mortality rates on the basis of morbidity scoring systems


without providing raw data were also excluded.
Although most studies were expected to be descriptive
cohort studies without a specic intervention or control
group, randomised or comparative studies were also
eligible. We planned to exclude the intervention arm
from randomised trials if the intervention was not
standard of care and had shown a dierential eect on
death in the intervention versus the control group;
however, this issue did not arise because there was only
one large randomised trial eligible for this analysis, and
both arms were eligible for inclusion.

Data extraction and outcome denitions


Using standard forms, two authors independently
identied studies for inclusion and extracted information
on outcomes, baseline American Society of Anesthesiologists (ASA) status, year of recruitment, country
of origin, and method of data collection. Disagreements
were resolved by consensus. The primary outcome was
anaesthetic sole mortality, dened as death deemed to be
attributable only to anaesthesia. Secondary outcomes
were anaesthetic contributory mortality, dened as death
deemed to be partially related to anaesthetic conduct;
total perioperative mortality, dened as death from any
cause; and cardiac arrest.
Country development status was assigned according to
the UN Human Development Index (HDI), which is an
index based on life expectancy, literacy, enrolment in
further education, and per-capita income.10 Since any
countrys HDI can change over time, the HDI relevant to
the country during the specic time period of the study
was assigned. If HDI data were not available for that year,
then the HDI at the closest date available was used.

Statistical analysis
2755 records identified through
electronic database searching

407 records identified through other sources


(hand searching of bibliographies)

3162 records identified


104 duplicates removed
3058 records screened by two
reviewers
2881 records excluded after review
of title and abstract because of
irrelevance
177 potentially relevant studies
identified, and full text
collected for review

87 records included in qualitative


analysis and regression
analysis

Figure 1: Flowchart of study identication

1076

90 studies excluded
2 patient subgroup analyses
22 had no denominator
66 had no data or were reviews,
editorials, or case reports

For every outcome in every study, event rates, dened as


the number of events per million anaesthetic procedures
and their corresponding 95% CIs were calculated. The
xed eect model (primary analysis) and the random
eects model (secondary analysis to account for
heterogeneity) were used to calculate weighted event
rates across all studies. Since the results and conclusions
were not sensitive to the model used, only data from the
xed eect model are presented here for simplicity. We
did meta-regression using the method of moments to
assess whether death rates changed signicantly over
time and by HDI. Since many studies reported data over
a time period of several years, the data were assigned as
close as possible to the year in which the patients were
recruited. When data were provided only in aggregate
time intervals, the data were assigned to the median year
of the relevant interval of recruitment. Prospective versus
retrospective data were collected for sensitivity analyses.
In addition to meta-regression, we subgrouped
aggregate event rates per million patients for each double
decade and for each HDI category to calculate natural
www.thelancet.com Vol 380 September 22, 2012

Articles

event rates for each time period and by low versus high
HDI (<08 vs 08, respectively). For subgroup analyses,
we did the test for interaction to assess whether eect
sizes diered signicantly across subgroups. Event rates
per million were also measured by baseline ASA physical
status subgroups.11
Statistical analyses were done using Comprehensive
Meta-Analysis version 2.2 (Englewood, NJ, USA, 2008).
Statistical signicance was dened as a two-sided p<001.
Heterogeneity was assessed using I, which indicates the
proportion of variability between studies that cannot be
attributable to chance alone.12 Values of I higher than
50% were deemed to suggest signicant heterogeneity
between studies.

Role of the funding source


The sponsor of the study had no role in study design,
data collection, data analysis, data interpretation, writing
of the report, or decision to publish the results. The
corresponding author had full access to all the data in the
study and had the nal responsibility for the decision to
submit for publication.

Results
3162 abstracts were reviewed and 177 potentially
relevant full text articles were retrieved (gure 1). Of
these, 87 studies met the inclusion criteria, within
which there were more than 214 million anaesthetic
administrations given to patients undergoing general
anaesthesia for surgery (appendix). The appendix lists
the characteristics and designs of the 87 studies. Most
studies reported events intraoperatively and within the
rst 2448 h postoperatively, whereas only four studies
reported 30-day anaesthesia mortality. Exclusion of the
latter trials did not substantially aect the mortality
estimates. Three trials had no HDI data available, one
because multiple countries were studied and two from
Taiwan in which HDI data were not collected. Sensitivity
analysis by prospective and retrospective data collection
did not materially change the results. As expected,
statistical heterogeneity (I >50%) was detected for all
event rates.
Risk of anaesthetic sole mortality decreased progressively over the decades, from 357 per million (95% CI
324394) before the 1970s, to 52 per million (4264) in
the 1970s80s, and 34 per million (2939) in the
1990s2000s (table 1; gure 2). In a weighted metaregression, this reduction over time was signicant
(p=0000001; gure 3). When temporal trends were
analysed for high-HDI and low-HDI country subgroups
separately, only the high-HDI countries had a decline
over time (p<000001).
Anaesthetic contributory mortality decreased over time,
from 650 per million (95% CI 610693) before the 1970s,
to 323 per million (290360) in the 1970s80s, and 143 per
million (129157) in the 1990s2000s (p<000001 across
subgroups; table 1). This reduction in mortality risk over
www.thelancet.com Vol 380 September 22, 2012

time was statistically signicant in a meta-regression


(p<000001; appendix). Analysis of temporal trends for
high-HDI and low-HDI countries separately showed that
only high-HDI countries experienced a decline in
Events

Weighted event rate per


million (95%CI)

p value for subgroup


interaction
By HDI

By decade

<000001

Anaesthetic sole mortality


Pre-1970s*

403/1 294 158

357 (324394)

High HDI

403/1 294 158

357 (324394)

NA

Low HDI

NR

NR

1970s80s

86/2 380 920

52 (4264)

High HDI

50/1 761 384

32 (2442)

Low HDI

36/619 536

101 (72140)

1990s2000s

186/8 990 012

34 (2939)

High HDI

151/8 610 720

25 (2130)

Low HDI

32/274 692

141 (100199)

<000001

<000001

Anaesthetic contributory mortality


Pre-1970s*

925/1 625 266

650 (610693)

High HDI

867/1 447 338

684 (642729)

Low HDI

58/177 928

326 (252422)

332/1 176 999

323 (290360)

High HDI

150/649 744

234 (200275)

Low HDI

180/475 127

432 (373500)

1990s2000s

395/5 950 293

143 (129157)

High HDI

275/5 641 048

85 (7596)

Low HDI

120/309 245

1970s80s

467 (391559)

<000001

<000001

<000001

<000001

Total perioperative mortality


Pre-1970s*

13 253/1 939 879

10 603 (10 42310 784)

High HDI

11 227/1 761 951

10 467 (10 27610 662)

Low HDI

2026/177 928

11 387 (10 90411 890)

4696/1 570 070

4533 (44054664)

High HDI

1723/934 781

1982 (18912078)

Low HDI

2973/635 289

7336 (70787604)

1990s2000s

6567/7 047 928

1176 (11481205)

High HDI

5981/6 738 683

1095 (10671123)

Low HDI

586/309 245

2445 (22552651)

Pre-1970s*

342/587 906

672 (605748)

High HDI

342/587 906

672 (605748)

NA

1970s80s

<000001

<000001

<000001

<00001

Cardiac arrest

Low HDI
1970s80s
High HDI

NR

NR

3045/2 480 112

1872 (18071939)

2799/2 361 710

1798 (17321866)

<000001

<00001

Low HDI

221/66 274

3642 (31934155)

1990s2000s

4299/6 475 012

719 (698741)

High HDI

3906/6 094 142

659 (638680)

Low HDI

345/276 270

2068 (18612298)

<000001

HDI=human development index. NR=not reported. NA=not available. *Includes studies primarily from the 1950s and
1960s; however, because there were few 1940s studies, they were also included here. Sensitivity analysis revealed no
substantial change in event rates when data before the 1950s were excluded from analysis. HDI subgroup data
excludes trials for which no HDI-specic data are available, and therefore addition across subgroups will not always be
equal to the overall results.

Table 1: Anaesthesia mortality and cardiac arrests by decade and by country human development
index status

1077

Articles

Anaeshetic sole mortality (deaths/106)

10 000

1000

100

10

1
1937

1947

1957

1967
1977
Study year

1987

1997

2007

Figure 2: Event rates for anaesthetic sole mortality by year


Every circle represents a study; the circle size is representative of the studys population size. The year represents
the median year if the study reported a range of years. Low-income and middle-income countries are shown in
light red (human development index [HDI]<08) and high-income countries in dark red (HDI080).

664
728

Log-odds event rate

793
857
922
986
1051
1115
1179
1244
1308
1939

1946

1953

1960

1967

1974
1981
Study year

1988

1995

2002

2009

Figure 3: Meta-regression for risk of anaesthetic sole mortality by year


Every circle represents a study; the circle size is representative of the weight of that study in the analysis. The
relation between mortality and year of study was signicant, with a signicant decline over the decades
(slope 0053, 95%CI 0058 to 0049; p=0000001).

mortality risk over time (slope 0042; p<00001),


whereas low-HDI countries experienced an increased
mortality risk over time (slope +0019; p=00001).
Total perioperative mortality decreased progressively
over the decades, from 10 603 per million
(95% CI 10 42310 784) before the 1970s, to 4533 per
million (44054664) in the 1970s80s, and 1176 per million
(11481205) in the 1990s2000s (p<00001 across subgroups; table 1). This reduction in all-cause perioperative
mortality was statistically signicant in meta-regression
over time (p<000001; gure 4). When analysed separately, both high-HDI countries (slope 0050; p<00001)
and low-HDI countries (slope 0042; p<00001)
experienced a signicant decline in perioperative mortality
1078

over time, although the magnitude of reduction was


greatest in high-HDI countries.
The incidence of cardiac arrest varied over the decades,
with an overall decline in incidence by time that was
statistically signicant (p=000001; appendix). The event
rate in the 1970s80s was 1872 per million (95% CI
18071939), compared with 719 per million (698741) in
the 1990s2000s (table 1). Although the studies from
before the 1970s provided some data for cardiac arrest,
the event rate was less reliable for this early subgroup
because of the small numbers within each study before
1970 compared with the larger and more numerous
studies from the later decades. For this reason, although
the event rate for cardiac arrests before the 1970s was
calculated to be 672 per million (95% CI 605748), which
is lower than in the later decade subgroups, the metaregression suggested an overall trend toward a reduction
over time because of the weight placed on later years
given the study sample sizes (p=0004).
When data from all countries across all decades were
assessed in a weighted meta-regression by HDI, the rate
of anaesthetic sole mortality decreased progressively with
increasing HDI (p<000001; appendix). Table 1 shows
that the aggregate event rate for anaesthetic sole mortality
was at least three times higher in low-HDI versus highHDI countries (101 per million [95% CI 72140] vs 32 per
million [2442] in the 1970s80s, and 141 per million
[100199] vs 25 per million [2130] in the 1990s2000s).
Low-HDI countries were insuciently represented
before the 1970s for sub-analysis to be done.
Meta-regression of anaesthetic contributory mortality
by HDI also showed a signicant reduction in anaesthetic
contributory mortality as HDI increased (p<00001;
appendix), which is also shown by the higher aggregate
event rates for each decade, with the exception of
before the 1970s, when low-HDI countries were poorly
represented within the subgroup analysis (table 1).
Meta-regression of total perioperative mortality by
increasing HDI showed a signicant reduction in risk
of perioperative death as HDI increased (p<000001;
appendix), and aggregate event rates by decade group
showed signicantly greater event rates in low-HDI
versus high-HDI countries in every decade group (all
p<00001; table 1).
Weighted meta-regression showed a signicant relation
between higher HDI and lower risk of cardiac arrest
(p<000001; appendix). The incidence of cardiac arrest
was at least twice the rate in low-HDI versus high-HDI
countries in both the 1970s80s and 1990s2000s (table 1).
Perioperative all-cause mortality increased with increasing ASA status at baseline, from a low of 557 per
million (95% CI 458678) for ASA grade 1 to a high of
273 534 per million (253 688294 320) for ASA grade 5
(table 2). Patients with ASA grade 2 at baseline had more
than double the mortality risk of patients with ASA
grade 1. Patients in each successive higher category of
ASA status at baseline had four to six times the mortality
www.thelancet.com Vol 380 September 22, 2012

Articles

Discussion
This comprehensive systematic review with metaregression quanties the global risk of anaestheticrelated and total perioperative mortality and shows the
rates of change in these risks by time, country HDI, and
baseline ASA status. The results show a clear reduction
in anaesthetic-related and perioperative mortality over
the past 50 years, despite the increasing baseline ASA
risk status of patients and patient complexity. Cardiac
arrests in the perioperative setting have also declined
over time. The rate of decline in anaesthetic-related
mortality and perioperative mortality over the decades
has been greatest and most consistent in countries with a
high HDI compared with those with a low HDI.
Despite the widespread emphasis on patient safety in
anaesthesia since the early 1980s, and with more
concentrated eorts toward patient safety since the
1990s, this aggregate analysis of all published worldwide
data suggests a steady improvement in anaestheticrelated and perioperative mortality even preceding these
eorts, particularly in high-HDI countries. This nding
might have been as a result of early advancement in
anaesthesia and surgical practices (training, certication,
drugs, and techniques); improved selection of patients
for surgery (better matching of patient risk to procedure
benet); increasing advances in aseptic techniques and
equipment sterilisation; increased use of antibiotics;
improvements in physiological monitoring, and uid
and blood management protocols; improved postoperative critical care; and improvements in teamwork
and education of the expanded perioperative health-care
team. Unfortunately, the studies did not provide
sucient details to test these hypotheses specically. The
continued reduction in anaesthetic and perioperative
mortality in recent years is encouraging and suggests
www.thelancet.com Vol 380 September 22, 2012

that the recent heightened eorts toward standardisation


and patient optimisation (eg, treatment of disorders such
as high cholesterol, hypertension, and angina before
surgery); higher levels of experience, improved surgical
technologies and techniques; and optimum timing of
antibiotics, timeouts, safe surgery checklists, and related
protocols might be translating collectively into continued
improvements in patient outcomes.1,5,1315 The results of
this global meta-analysis are particularly noteworthy
since anaesthetic-related and perioperative mortality are
suciently rare that individual studies are generally
insucient in size to detect measurable dierences, and
cannot rule out the possibility that important dierences
might exist because of any single intervention or
accumulation of eorts.13 Although these results lend
support to an overall improvement in patient care
300
370
440
Log-odds event rate

risk when compared with the next lower category of ASA


status. The same numeric patterns were noted for cardiac
arrest when examined by baseline ASA status. Incidence
of cardiac arrest ranged from 193 per million (95% CI
149249) for patients with ASA grade 1 up to 234 121 per
million (186 231289 939) in those with ASA grade 5.
When subgroup analysis was done to dichotomously
compare patients with ASA grade 13 with those with
ASA grade 45, the relative risk of all-cause perioperative
mortality in the high-ASA status group compared
with the low-ASA status group was 48 (95% CI 4253;
table 3). Similarly, the relative risk of cardiac arrest was
61 (5072). When meta-regression by ASA status was
done for death or cardiac arrest, there was a signicant
increasing relation between higher ASA at baseline and
death or cardiac arrest (p<000001).
Meta-regression by time period showed that the
baseline ASA status of patients increased signicantly
over the decades, conrming the widespread opinion that
the baseline risk and complexity of patients presenting
for surgery increased over the decades (p<00001).

510
580
650
720
790
860
930
1000
1929

1937

1945

1953

1964

1970
1978
Study year

1986

1994

2002

2010

Figure 4: Meta-regression for total perioperative mortality by year


Every circle represents a study; the circle size is representative of the weight of that study in the analysis. The
relation between mortality and year was signicant, with a signicant decline over the decades (slope 0053,
95% CI 0054 to 0052; p<000001).
Mortality event rate per
million (95% CI)

Cardiac arrest event rate


per million (95% CI)

Mortality or cardiac arrest


event rate per million (95% CI)

ASA grade 1

557 (458678)

193 (149249)

379 (324442)

ASA grade 2

1408 (12541582)

1112 (9441310)

1301 (11841431)

ASA grade 3

9369 (876110 018)

ASA grade 4

61 797 (58 41265 365)

ASA grade 5

5936 (52086764)
75 193 (68 14682 904)

8515 (80229039)
64 823 (61 73068 058)

273 534 (253 688294 320) 234 121 (186 231289 939) 268 963 (250 430288 340)

ASA=American Society of Anesthesiologists. See also the appendix for details of raw event rates.

Table 2: Mortality or cardiac arrest by American Society of Anesthesiologists status

Event rate per million (95%CI)


ASA grade 13

RR (95%CI) p for subgroup


interaction

ASA grade 45

Mortality

4793 (45335068)

93 268 (88 98697 735)

48 (4253)

<000001

Cardiac arrest

2114 (19252328)

86 830 (79 25595 054)

61 (5072)

<000001

Mortality or cardiac arrest

3887 (37044078)

91 865 (88 10895 766)

50 (4753)

<000001

ASA=American Society of Anesthesiologists. RR=relative risk of adverse outcome for patients with ASA grade 45 vs
ASA grade 13.

Table 3: Mortality or cardiac arrest by high versus low American Society of Anesthesiologists subgroup

1079

Articles

contributing to better patient outcomes at present,


subsequent updates of this dataset should be done over
time to provide a continued global measure of the patterns in anaesthetic-related and perioperative mortality
as an aggregate representation of the translation of the
aforementioned patient safety eorts into tangible
outcomes over time in the real-world setting, both in the
developed and developing world.
The rate of improvement in perioperative mortality
over time is signicantly related to HDI, and this might
be a result of the ability of wealthier countries to increase
health-care investment in the technologies, techniques,
and training necessary to improve patient safety.
Conversely, this nding might also be a result of a
dierence in case mix, with less developed countries
tending to have higher rates of surgery for traumatic
injury and advanced diseases and fewer elective surgical
procedures for chronic stable conditions or minor
diseases.3,4 Since the studies identied in this analysis
reported insucient details to allow greater exploration
of outcomes by patient characteristics, the latter
hypothesis remains unproven, but should be pursued in
future research.
Although anaesthetic mortality remains low compared
with trac fatalities or suicide,16 it still remains high
compared with death caused by air travel, which is a
commonly used yardstick to benchmark the risk of
anaesthesia.1720 Quantication of acceptable risk for
individuals is often dicult since a surgical procedure is
associated with potential benets that need to be weighed
against potential risks when deciding whether to undergo
surgery. Using updated quantied risks from this
analysis might help facilitate informed decision making,
in particular when placed in perspective against other
everyday risks, and might be especially important for
patients undergoing procedures for which benets might
be small or uncertain and need to be closely balanced
against risks.
The global burden of disease and disability-adjusted
life-years3 that could potentially be improved through
surgical intervention, if surgery was more accessible in
the developed world,4 highlights the importance of
ensuring that improved safety measures will accompany
all eorts to improve access to surgery in low-HDI
countries. The two to four times increased risk of
anaesthesia-related and perioperative mortality in lowHDI compared with high-HDI countries shows the gap
that has existed between health-care systems, and this
inequity remains despite recent eorts to improve global
perioperative safety regardless of country and social
status. Targeted initiatives involving collaboration
between lower and higher HDI countries should be
developed to reduce this gap between actual anaesthesiarelated and perioperative mortality in low-HDI countries
and achievable perioperative mortality in high-HDI
countries. Closing this gap is an important achievable
global health improvement and would probably exceed
1080

the payback on other potential improvements that


compete for our time and resources in the developed
worldfor example, new and expensive surgical
technologies, which generally oer smaller marginal
benets in the developed world than the lives that could
be saved by closing the global perioperative mortality
inequity gap in the developing world.
These results need to be interpreted in light of the
limitations of the data available. Since most studies failed
to report sucient risk factor data to allow for adequate
risk stratication, we used crude death rates, without
adjustment for specic comorbidities or type of surgical
procedure. Therefore, biases and confounders that may
dier over time might have aected the results,
particularly with respect to patient risk factors and type of
surgery. We aimed to minimise selection bias by
using large studies (>3000) of allcomers undergoing
anaesthesia for all decades, and measured trends over
time by HDI and by baseline ASA score to minimise the
eect of dierent surgical procedures and patient risk,
which are probably of greatest dierence in countries of
dierent HDI (and thus include patients of dierent ages
and with dierent comorbidities). We calculated rates of
change over time within high-HDI and low-HDI country
settings separately to minimise the eect of confounders
between countries. Ideally, global registries would
provide more complete data with more detailed patient
and procedural information to allow for more informative
analyses in the future with adjustment for confounding.
Nevertheless, the consistency in reductions over time for
all outcomes measured in this analysis, considered
together with the fact that baseline ASA risk status
signicantly increased over time, suggests that the
trends might be further improved if more specic risk
adjustment could be done.
Although substantial eort was made to ensure that all
relevant studies were included in this analysis, some
studies from the 1940s and 1950s could not be retrieved.
Also, other existing studies, particularly from the developing world, were probably not identied (eg, because
they were not available in English or were not published in
an indexed journal). Publication bias might have
contributed to inadequate power to detect mortality trends
from the developing world because few studies were
available from the developing world in the earlier decades.
Readers are encouraged to notify the authors if additional
relevant studies exist so that new data can be incorporated
into subsequent updates of this analysis. In particular,
studies in languages other than English and from the
developing world would further improve the estimates by
HDI status.
Future studies on perioperative mortality should aim to
be more thorough in the reporting of information,
including outcomes by surgical subtypes, by anaesthetic
subtype, and by patient risk groups. Furthermore, the
likely cause of perioperative death should be provided
more consistently. Denitions of anaesthetic-related
www.thelancet.com Vol 380 September 22, 2012

Articles

mortality and perioperative mortality remain disparate


across studies (eg, subjective judgment is often needed
to class mortality as related to anaesthesia, and even the
timeframe for perioperative death varies, such as death
within 24 h vs within 7 days after surgery), and calls for
standardisation of denitions seem to be mostly
unheeded so far.19,2123
Despite an increase in patient baseline risk, perioperative and anaesthetic-related mortality rates have
steadily declined over the past 50 years, and this might be
an indicator of the cumulative eect of eorts to improve
patient safety in the perioperative setting over the
decades. However, the decline was greatest and most
consistent in developed countries, and overall rates of
perioperative and anaesthetic-related mortality remain
two to three times higher in developing countries. Global
priority should be given to reducing total perioperative
and anaesthetic-related mortality with evidence-based
best practice in developing countries to reduce the
disparity in mortality compared with developed countries.
Contributors
DB extracted data, reviewed the results, and wrote the manuscript.
JM did the statistical analyses, reviewed the results, and wrote the
manuscript. MA extracted data, reviewed the results, and edited the
manuscript. DC reviewed the results and edited the manuscript.

6
7

10
11
12
13
14

15

16

17

Conicts of interest
We declare that we have no conicts of interest.

18

Acknowledgments
Funding support was provided by the Department of Anesthesia
and Perioperative Medicine, University of Western Ontario, London,
ON, Canada.

19

References
1
Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume
and operative mortality for high-risk surgery. N Engl J Med 2011;
364: 212837.
2
Weiser TG, Regenbogen SE, Thompson KD, et al. An estimation of
the global volume of surgery: a modelling strategy based on
available data. Lancet 2008; 372: 13944.
3
Chu KM, Ford N, Trelles M. Operative mortality in resource-limited
settings: the experience of Medecins Sans Frontieres in
13 countries. Arch Surg 2010; 145: 72125.
4
McIntyre T, Zenilman ME. Globalization of surgery: lets get
serious. Arch Surg 2010; 145: 71516.
5
Haynes AB, Weiser TG, Berry WR, et al. A surgical safety checklist
to reduce morbidity and mortality in a global population.
N Engl J Med 2009; 360: 49199.

www.thelancet.com Vol 380 September 22, 2012

20

21

22

23

Bickler SW, Spiegel DA. Global surgerydening a research


agenda. Lancet 2008; 372: 9092.
Liberati A, Altman DG, Tetzla J, et al. The PRISMA statement for
reporting systematic reviews and meta-analyses of studies that
evaluate healthcare interventions: explanation and elaboration. BMJ
2009; 339: b2700.
Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of
observational studies in epidemiology: a proposal for reporting.
Meta-analysis Of Observational Studies in Epidemiology (MOOSE)
group. JAMA 2000; 283: 200812.
Eypasch E, Lefering R, Kum CK, Troidl H. Probability of adverse
events that have not yet occurred: a statistical reminder. BMJ 1995;
311: 61920.
Human Development Report 201020th Anniversary Edition.
New York: United Nations Development Program, 2010.
Saklad M. Grading of patients for surgical procedures.
Anesthesiology 1941; 2: 28184.
Higgins JP, Thompson SG. Quantifying heterogeneity in a
meta-analysis. Stat Med 2002; 21: 153958.
Jacobs DO. Variation in hospital mortality associated with inpatient
surgeryan S.O.S. N Engl J Med 2009; 361: 1398400.
de Vries EN, Prins HA, Crolla RM, et al. Eect of a comprehensive
surgical safety system on patient outcomes. N Engl J Med 2010;
363: 192837.
Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in hospital
mortality associated with inpatient surgery. N Engl J Med 2009;
361: 136875.
Ikkersheim DE, Berg M. How reliable is your hospital? A qualitative
framework for analysing reliability levels. BMJ Qual Saf 2011;
published online March 21. DOI:10.1136/bmjqs.2010.045187.
Gaba DM. Anaesthesiology as a model for patient safety in health
care. BMJ 2000; 320: 78588.
Cooper JB, Gaba D. No myth: anesthesia is a model for addressing
patient safety. Anesthesiology 2002; 97: 133537.
Lagasse RS. Anesthesia safety: model or myth? A review of the
published literature and analysis of current original data.
Anesthesiology 2002; 97: 160917.
World Federation of Societies of Anaesthesiologists. International
Standards for a Safe Practice of Anaesthesia. 2008. http://
anaesthesiologists.org/en/safety/2008-international-standards-for-asafe-practice-of-anaesthesia.html (accessed June 10, 2011).
Deshpande JK. Cause and eect or conjecture? A call for consensus
on dening anesthesia-related mortality.. Anesth Analg 2011;
112: 125961.
Mellin-Olsen J, Staender S, Whitaker DK, Smith AF. The Helsinki
Declaration on Patient Safety in Anaesthesiology. Eur J Anaesthesiol
2010; 27: 59297.
Braz LG, Braz DG, Cruz DS, Fernandes LA, Modolo NS, Braz JR.
Mortality in anesthesia: a systematic review. Clinics (Sao Paulo)
2009; 64: 9991006.

1081