Vous êtes sur la page 1sur 8

ORIGINAL ARTICLES

Comparison of Nurse, System and Quality Patient Care


Outcomes in 8-Hour and 12-Hour Shifts
Patricia W. Stone, PhD, RN,* Yunling Du, PhD, Rhabia Cowell, MS,* Norma Amsterdam, MA, RN,
Thomas A. Helfrich, JD, Robert W. Linn, JD, Amy Gladstein, JD, Mary Walsh, MSN, RN,
and Lorraine A. Mojica, BSN, RNC**

labor-management bargaining group to make an evidence-based


Background: Many nurses desire 12-hour shifts. However, there
decision. We encourage others to conduct similar studies.
are concerns about implementation.
Objective: We sought to compare the effects of 8- and 12-hour Key Words: nurse staffing, hospital administration, shift length,
shifts on nurse, system, and quality patient care outcomes. patient safety
Methods: We used a cross-sectional design with data collected from (Med Care 2006;44: 1099 1106)
multiple sources in 20032004, including a nurse survey and ad-
ministrative and patient records. We studied hospital nurses and
patients in general adult wards, with outcomes including burnout,
job satisfaction, scheduling satisfaction, preferences, intention to
stay, and employee safety. System outcomes included recruitment
and turnover, staffing, absenteeism, and related costs. A variety of T he hospital environment is a 24-hour, 7-day a week
service industry which, by necessity, translates into shift
work for hospital employees. Shift work has been considered
quality patient care outcomes were measured from the 3 different
types of data. detrimental to workers health, increasing stress-related to
Results: Thirteen New York City hospitals participated; 805 sur- worklife conflict1 and disruptions to circadian rhythms.2 In
veys were examined from 99 nursing units (response rate 42%). healthcare professionals, increased job stress has been found
Compared with nurses working 8-hour shifts, those working 12-hour to be correlated with high levels of burnout3 and increased
shifts were on average more satisfied with their jobs, experienced incidence of disease.4,5 The majority of nursing personnel
less emotional exhaustion, 10 times more likely to be satisfied with (the hospitals largest labor force) are women, which increases
schedules, 2 times as likely to perceive 12-hour schedules as the likelihood of difficulty balancing work and personal
important, and 58% less likely to report missing shifts; units with responsibilities.6 Because 12-hour shifts offer compressed
12-hour shifts had lower vacancy rates and weeks to fill the position work weeks (eg, three 12-hour shifts vs. five 8-hour shifts)
(all P values 0.05). There were no differences in patient many nurses perceive these schedules to be desirable even
7

outcomes.
without salary differentials. However, some managers and
administrators have been concerned about implementation.8
Conclusions: Nurses working 12-hour shifts were more satisfied. 9
There were no differences in quality outcomes. Flexibility and Smith and colleagues conducted a comprehensive sys-
tematic review of published evidence on 8- and 12-hour shifts
choice in shift length are important elements in a positive nurse
and concluded that longer shifts increased employee fatigue
work environment. This study represents an innovative attempt by a
but also increased job performance, which was typically
assessed using a survey instrument designed to measure
nurses reasoning ability for critical thinking. They postulated
From the *Columbia University School of Nursing, New York City, New the reason for these findings may be that employees working
York; Department of Clinical Biostatistics, Albert Einstein College, a compressed work week are more motivated, therefore
New York City, New York; 1199SEIU National Health & Human
Services Union, New York City, New York; League of Voluntary reducing possible detrimental effects. Results from 6 addi-
Hospitals and Homes: Linn & Green Consulting, Inc: New York City, tional studies directly comparing hospital nurses working 8-
New York; Beth Israel Hospital, New York City, New York; and and 12-hour shifts have been published since Smiths review.
**SVCMC Saint Johns Hospital Queens, Elmhurst, New York. Three research teams surveyed nurses and found they pre-
Supported by the 1199/SEIU League RN Planning and Placement Fund. ferred 12-hour and self-reported wellbeing and performance
This study was conducted as a result of an agreement between the Registered
Nurse Division of 1199 Service Employees International Union (SEIU), were not adversely affected by working longer shifts.10 12
New Yorks Health and Human Service Union (RN Division of 1199), Researchers in the United Kingdom observed nurses working
and the League of Voluntary Hospitals and Homes of New York 12- and 8-hour shifts and concluded those working the longer
(LVHH). shifts had less effective total performance score as measured
Reprints: Patricia W. Stone, Columbia University School of Nursing, 617 West
168th Street, New York, NY 10032. E-mail: ps2024@columbia.edu.
by an observer rating 5 different domains of nursing care (eg,
Copyright 2006 by Lippincott Williams & Wilkins care management).13 However, all of these studies were
ISSN: 0025-7079/06/4412-1099 limited to a single setting with the largest sample size being

Medical Care Volume 44, Number 12, December 2006 1099


Stone et al Medical Care Volume 44, Number 12, December 2006

FIGURE 1. A diagram of the concep-


tual framework guiding study. Dot-
ted lines and arrows represent the
potential confounding variables and
their relationship with the indepen-
dent and dependent variables.
240 nurses. Hoffman and Scott (2003) conducted a survey of nurses self-selected their place of employment based on
randomly selected Michigan licensed nurses (50% response qualifications and seniority within the hospital.
rate, n 252) and found nurses working 12-hour shifts were
more stressed than those working 8-hour shifts and there were
no differences in job satisfaction or salary. The nurses work- METHODS
ing 12-hour shifts were younger and had less work experi- Setting
ence, which may account for the increased stress.10 Con-
versely, in Poland, researchers found that nurses working Fourteen hospitals located throughout New York City
were eligible to participate. A cross-sectional research design
12-hour shifts compared with those working 8-hour shifts had
lower physical workload (measured in kilocalories per hour) was developed and nurse, system, and patient data were
collected from adult general inpatient units. The appendix
and better mood states as measured by fatigue, hostility, and
lists hospital characteristics such as number of beds and how
friendliness.14 In this study, the nurses working 12-hour shifts
also were younger. Based upon this previous research, the the 12-hour shifts were implemented (eg, number of annual
shifts per full-time equivalent). Institutional review board
effect of various shift length on nurse health, wellbeing, and
approval was obtained at all participating hospitals and the
productivity is uncertain.
Currently, many hospitals have implemented 12-hour primary investigators home institution.
shifts for nurses in critical care units.15 However, these longer
shifts often are not widely available. Some nurses seem to Conceptual Framework
prefer 12-hour shifts; however, there are legitimate concerns The study was guided by Donabedians theory of
about the safety of the employee and the patient in an healthcare quality, which includes the broad concepts of
extended work hour environment. This is especially true with structure, process and outcomes.18 Previously, investigators
an aging nursing workforce,16 which may not be able to as expanded the quality model to include worker outcomes.19 As
easily cope with the physical demands of extended shifts. illustrated in Figure 1, we further expanded the conceptual
Furthermore, with financial pressures put on hospitals and model to include system outcomes. In the model, the structure
concerns about patient safety, understanding economic con- of care represents setting specific characteristics (ie, hospital,
sequences of human resource policy decisions and how best nursing unit type, and average organizational climate per
to improve the nurse work environment is important.17 unit) and the nurse characteristics (demographics). These
The availability of 12-hour shifts became a difficult structural variables may confound the relationship between
issue between a union representing nurses and a multiem- our independent variable (shift length) and the outcomes of
ployer bargaining group. Because of the lack of clear evi- interest. For example, the age of the nurse may influence job
dence to guide decision-making regarding optimal shift satisfaction. The independent variable of interest in this study
length, this study was conducted as a result of their negotiated is the process of nurse shift length (ie, 8- vs. 12-hour shifts).
agreement. The aims of the study were to compare the effects Outcomes include those related to the nurse, system and
of 8- and 12-hour shifts on nursing, system and quality patient. Because of the contentious nature of the issue and the
outcomes, while controlling for confounding variables. Dur- conflicting results found in previous research, no a priori
ing the time the study was taking place, based on adminis- directional hypotheses were postulated. Data collection meth-
trative policies, preferences, supply and demand of nurses, ods are described followed by the operational definitions and
different nursing units offered different shift lengths and psychometric properties of the measures.

1100 2006 Lippincott Williams & Wilkins


Medical Care Volume 44, Number 12, December 2006 8-Hour Versus 12-Hour Shifts

Data Collection Nursing Outcomes


Nurse Survey Data for nurse outcomes came from the survey; variables
Direct care nurses were eligible to be surveyed during included burnout, job satisfaction, scheduling satisfaction,
the Fall of 2004. The survey (available from the correspond- scheduling preferences, intention to stay, employee health and
ing author upon request) contained well-developed measures safety. The Maslach Burnout Inventory (MBI) was the measure
found in the literature as well as some items developed of burnout and includes 3 subscales: emotional exhaustion (9
specifically for this study. To determine overall clarity, the items, 0.91), depersonalization (5 items, 0.77), and
survey was piloted in 30 noneligible nurse respondents prior personal accomplishment (8 items, 0.89).22 Job satisfaction
to distribution. was measured using the Job Enjoyment Scale,23 which asks
Using an anonymous self-report questionnaire format, respondents to rate job satisfaction on a 5-point Likert-type scale
which was coded to the hospital where it was distributed, the (7 items, 0.88) with scores ranging from 7 to 35.
respondent was asked to indicate the unit in which they Scheduling satisfaction, preferences, and intention to
primarily worked. Surveys were distributed on nursing units stay in current position were 1-item self-report measures. In
in 2 waves. To protect against duplicate responses, nurses previous research of physicians, self-report intention to stay
were instructed to respond 1 time. Furthermore, respondents had a sensitivity of 73.3%24 and is a frequently used in
demographics (ie, age, gender, race, ethnicity, education, and research assessing the nursing workforce.25 Employee health
years of experience) were examined; a survey distributed in and safety was measured by absenteeism and injuries. It is
the second wave that directly matched demographics from the well documented that employee reporting of occupational
first wave was considered a duplicate and discarded. injury using institutional mechanisms is low26,27 and self-
report of employee occupational health outcomes using sur-
28
vey methods has been validated. Therefore, we directly
Administrative Data asked respondents their history of occupational injury and
A variety of administrative data were collected for a absences in the last 4 months. This recall period was chosen
12-month study period (July 1, 2003 to June 30, 2004), because these are somewhat rare events, but the longer the
including (1) human resource files (ie, payroll files, recruit- period, the more likely a respondent is to underestimate the
ment and turnover data), (2) nursing management files (ie, occurrence.29
monthly patient census and patient incident reports), and (3)
patient discharge abstract data. All administrative data were System Outcomes
submitted to the research team electronically, compiled into a System outcomes included measures of recruitment and
database and aggregated at the level of the nursing unit. turnover, staffing and related costs, as well as absenteeism
Incident report data, which were used in this study as 1 and related costs. Recruitment measures included (1) average
method to measure patient outcomes, are formal systems vacancy rate for the year and (2) average number of weeks to
developed to collect and analyze information about unusual fill a position. Turnover was defined as the number of nurses
events.20 Risk adjusted patient safety indicators were devel- leaving their position per full-time equivalents per unit.
oped from 17 data elements from the patient discharge Payroll and census data were used to calculate: (1) 3 staffing
abstract (eg, age, sex, admission type, diagnosis related variables (ie, total care hours per patient days TCHPPD ,
group, major diagnostic category, the unit the patient was registered nursing hours per patient days RNHPPD and
discharged from, admit and discharge date). skillmix ratio of RNHPPD to TCHPPD ); (2) overall and
registered nurse specific staffing costs per patient day; and 3)
Operational Definitions and Psychometric absenteeism rates and costs per patient day.30 Staffing costs
Properties of Variables included all payroll expenditures such as regular time, differ-
entials, sick time, overtime, and agency personnel costs. To
Structural Variables account for fringe rates, nonagency payroll cost data were
Nursing demographics and organizational climate mea- multiplied by the mean factor representing these particular
sures were collected in the nurse survey. Organizational costs for this group (1.36).
climate was measured using a 30-item modified version of the
Perception of Nurse Work Environment (PNWE) scale.21 The Quality of Patient Care Outcomes
PNWE has 7 dimensions (eg, Nurse/Physician Collaboration Previous researchers have found patient adverse events
and Support for Professional Practice), in which respondents to be under reported using administrative mechanisms.31
are asked to indicate the extent the factor is present in their Therefore, nursing sensitive quality of patient care outcomes
work environment using a 5-point Likert scale. Internal were measured using 3 different types of data including
consistency of the composite scale, which was used in the 12-months of administrative incident reports, 12-months of
analysis, was excellent ( 0.95). Indicator variables for the patient discharge abstracts, and nurses perceptions of quality
hospital and unit type (medical/surgical or general specialty obtained from the survey.
wards such as oncology or orthopedics) were developed. Unit level measures of quality from incident report data
included (1) medication events per patient bed, (2) patient
Independent Variable falls per patient bed (defined as any unplanned descent to the
Each nursing unit was identified based on the type of floor), and (3) decubitus ulcer prevalence per patient bed.
staffing in place (12- or 8-hour shifts). These patient outcomes were chosen because similar mea-

2006 Lippincott Williams & Wilkins 1101


Stone et al Medical Care Volume 44, Number 12, December 2006

sures to the latter 2 have been endorsed by the National accurately check for this potential bias. To understand how
Quality Forum as nursing sensitive outcomes30; and, medi- the respondents may be similar and or different from the
cation errors have been found to be related to nurse staffing.27 population of interest, the demographics of those who re-
However, although most hospitals monitored these incidents, sponded to the study were compared with demographic data
definitions were not always similar. In all instances, original based on each hospitals Healthcare Employee Pension files
data were examined and, if needed, recoded to meet study for the same time period. The Pension data included basic
definitions. For example, 1 hospital separated witnessed and demographic data by personnel category at the hospital level of
unwitnessed falls and these categories were collapsed. If analysis. In addition, the response rate for each unit was calcu-
usable data were not available the unit was dropped from that lated; relationships between response rate and scheduling type as
analysis. well as nurse outcomes were examined (see below).
In a comprehensive review of literature conducted by Multivariate general estimating equations (GEE) were
the developers of the Patient Safety Indicators (PSI), 4 of 20 developed for each outcome variable.37 The GEE method was
measures were categorized as staffing sensitive (decubitus chosen as the primary data analytic technique because re-
ulcer, failure to rescue, postoperative pulmonary emboli or sponses may be correlated (ie, not independent). The GEE
deep vein thrombosis, and postoperative respiratory fail- inference is based on the sandwich estimators of standard
ure).32 Therefore, these 4 PSIs were chosen for analysis and errors. Although other methods, such as hierarchical linear
developed from the electronic patient discharge data using the modeling, increasingly are being used to handle correlated
publicly available software. All data were aggregated to the continuous data, the GEE method has been shown to be
unit the patient was discharged from. The software allowed robust when outcomes are categorical.37 To evaluate the
for risk adjustment to control for differing patient popula- effects of 8- versus 12-hour shifts on nursing outcomes and
tions by adjusting for age, sex, diagnostic-related groups, and nurses perceptions of quality, responses from all useable
comorbidity distribution of data compared with national surveys were included, the nurse was the unit of analysis,
norms. Furthermore, patient populations were standardized structural variables were entered into the models as covariates
by computing the event rate based on the number of patients and analyses were clustered by the unit in which the nurse
that are at risk for the event. Because postoperative compli- was employed. The effect of entering unit-level response rate
cations were so rare in this population, they were dropped and as a covariate in these models was examined. For the quality
failure to rescue (ie, deaths per 1000 patients having devel- of patient care outcomes (except nurse perceptions) the sam-
oped specified complications of care during hospitalizations) ple was limited to medical/surgical nursing units, the unit was
and decubitus ulcers (ie, cases of decubitus ulcers per 1000 the level of analysis, and the analyses were clustered by the
discharges with a length of stay greater than 4 days) were hospital. Differences in adjusted means were determined by
analyzed. the beta coefficient for continuous data. Logistic regressions
Three questions on the nurse survey were directed at and odds ratios were computed for dichotomous outcomes.
obtaining the nurses perception of quality of patient care and Statistical significance was examined with P values and 95%
included statements about quality (1) on the last shift worked, confidence intervals.
(2) in general, and (3) changes over the last year. The
responses for the first 2 items included categorical choices
ranging from excellent to poor. Responses to changes over RESULTS
the last year included improved, stayed the same, and dete- Thirteen of the 14 eligible hospitals (93%) enrolled in
riorated. Single-item questions assessing the clinicians per- the study. The median bed size was 300 (range, 212780). All
ception of quality of care provided have been used in a participating institutions were affiliated with an academic
number of studies.33,34 institution, nonprofit status and considered nonspecialty hos-
pitals. A total of 99 units (8-hour shifts, n 63; 12-hour
Data Analysis shifts, n 36) were analyzed.
Distributions and descriptive statistics of all variables Fifteen nurse surveys were discarded as duplicates; this
were reviewed. If a continuous variable was not normally left a total of 805 useable surveys with an overall response
distributed, it was either transformed using logarithmic tech- rate of 42% (range, 2270%). The majority (76%) of the
niques (ie, job satisfaction) or dichotomized at the median (ie, responses came from the first wave of distribution (n 612).
depersonalization, personal accomplishment, satisfied with There were no differences in the respondents demographic
current scheduling). Missing value analyses were conduct- profile compared with the Pension data; and response rate did
ed.35 Because of the relatively low amount of missing data not differ between 12- and 8-hour units (data not shown). The
( 5%) for responses in the Likert type scales, a mean majority of the nurses (n 504, 63%) worked 8-hour
substitution approach was taken imputing the missing score schedules. Comparisons of nursing demographics by sched-
with individual mean response for each scale.36 There was no ule and unit type are reported in Table 1. The only significant
attempt to impute missing nurse demographic data or items difference (P 0.05) in nurses demographics between shift
related to nurses perceptions of quality. length was related to race, with a greater proportion of those
Response bias is always a potential limitation in sur- working 8-hour shifts being black.
veys. The usual way of checking for this bias is comparing Table 2 displays the summary statistics and results of
those that responded to those who didnt on key variables. the multivariate regressions with nursing outcomes as the
Because the survey was anonymous, there was no way to dependent variables. Compared with nurses working 8-hour

1102 2006 Lippincott Williams & Wilkins


Medical Care Volume 44, Number 12, December 2006 8-Hour Versus 12-Hour Shifts

TABLE 1. Descriptive Statistics of Respondents Unit and from 12 to 14 units). Units staffed by nurses working 12-hour
Scheduling Type shifts had lower vacancy rates and weeks to fill the position
(P 0.05). Turnover, staffing, absenteeism, and related costs
12-Hr 8-Hr were not statistically significantly different between units.
(n 301), (n 504),
Variables n (%) n (%) P
Table 4 displays the quality patient care outcomes
results. There were no differences in any of the quality of
General nursing ward type 0.001
patient care variables. The incident report data came from 9
Specialty unit 167 (55.5) 349 (69.2)
(70%) participating hospitals and 13 (92%) hospitals pro-
Medical/surgical unit 134 (44.5) 155 (30.8)
vided the patient discharge abstracts (n 68,034). The mean
Nursing demographics
numerators and denominators across nursing units for the
Gender 0.111
PSIs were decubitus ulcer 32 of 1053 and failure to rescue 11
Female 279 (93.3) 452 (90.6)
of 149.
Male 20 (6.7) 47 (9.4)
Education 0.136
Diploma 25 (8.5) 44 (9.0) DISCUSSION
Associate degree 101 (34.2) 181 (37.2) Controlling for structural variables related to nurse
Baccalaureate degree 157 (53.2) 226 (46.4) demographics and employment settings, most nurses working
Masters degree 12 (4.1) 36 (7.45) 12-hour shifts in this sample reported higher job satisfaction
Race 0.002 and less emotional exhaustion, they preferred and were more
American. Indian 3 (1.1) 3 (0.6) satisfied with their schedules, and missed less shifts. The
Asian 104 (36.7) 124 (26.1) finding that nurses working 12-hour shifts report missing
Hawaiian, Pacific 7 (2.5) 7 (1.5) fewer shifts is not surprising, since they are scheduled to
Islander
work less shifts. Nurses working 12-hour shifts reporting
Black 104 (36.7) 243 (51.2)
that these shifts were important may indicate ability for
White 65 (23.0) 98 (20.6)
self-selection to units that provide these schedules. It is
Ethnicity 0.559
important to note, a minority preferred conventional 8-hour
Hispanic 9 (3.4) 20 (4.5)
shifts. These results support calls for flexibility and choice in
Non-Hispanic 256 (96.6) 426 (95.5)
shift length as important elements in positive nurse work
Mean (SD) Mean (SD) environments.38
Age 42.45 (10.82) 43.75 (9.70) 0.108
We did not find statistically significant differences
Years in healthcare 17.26 (10.12) 17.50 (10.33) 0.752
between groups in actual absenteeism or staffing costs. How-
Years in hospital 11.29 (7.87) 11.68 (8.38) 0.545
ever, small sample sizes due to lack of useable data were an
Years on unit 7.95 (6.68) 8.26 (7.00) 0.568
issue and may have prevented detection of important trends
that exist. Further larger scale research is needed to help
Specialty unit was defined as oncology, orthopedics, or detox. 2 analyses con- understand the extent and characteristics of absenteeism as
ducted for categorical data, t tests conducted for continuous data. Sample sizes of nurse
demographics vary because of missing data. Sample sizes are gender (n 798), well as other staffing related expenses, such as use of tem-
education (n 782), race (n 758), ethnicity (n 711), age (n 683), years in porary staff.39 In the meantime, we encourage policymakers
healthcare (n 783), years in hospital (n 704), and years on unit (n 663).
to analyze data from their own institutions to inform their
human resource management decisions. Although the best
shifts, those working 12-hour shifts were, on average, more data would come from a randomized controlled trial, that is
satisfied with their jobs, experienced less emotional exhaus- probably not feasible. Analyzing longitudinal data pre and
tion, were more than 10 times more satisfied with schedules, 2 post changes would be the next best choice.39
times as likely to perceive 12-hour schedules as important, and In 8-hour units, vacancy rates and weeks to fill posi-
58% less likely to report missing shifts (all P values 0.05). tions were significantly higher than the 12-hour units. Expen-
Although most respondents thought 12-hour shifts were impor- ditures related to vacancy have been found to be the single
tant, approximately 15% of the respondents working 12-hour largest component in estimating costs of nurse turnover,
shifts and 35% of those working 8-hour shifts did not. which ranges from $62,000 to $67,000 per nurse depending
Greater unit response rate was significantly related to on the service line.40 Vacancies may be related to the pref-
younger age (P 0.009), positive organizational climate (P erences of the nurses and internal transfers to units offering
0.001), lower emotional exhaustion (P 0.040), lower de- 12-hour shifts; however, we did not have data to test this
personalization (P 0.001), and higher personal achieve- hypothesis.
ment (P 0.001); however, results from the multivariate A strength of this study is the multiple ways quality of
regressions were similar to Table 2 when the unit response patient care was measured. We did not have sufficient power
rate was entered as a covariate (data not shown). to detect differences of incident reports or the PSIs; however,
The results of the system outcomes analyses are dis- the point estimates of all quality measures were similar by
played in Table 3. Four hospitals (29%) provided the recruit- shift length and do not provide any consistent evidence that
ment and turnover data, and 5 (36%) hospitals provided there were differences by shift length. To our knowledge, this
payroll data, which limited sample sizes (12-hour units is the first study that has examined administrative data, incident
ranged from 14 to 17 units, 8-hour units sample sizes ranged report data and nursing perceptions of quality in 1 study. Con-

2006 Lippincott Williams & Wilkins 1103


Stone et al Medical Care Volume 44, Number 12, December 2006

TABLE 2. Comparison of Nursing Outcomes by Shift Length


12-Hr Mean (SD), 8-Hr Mean (SD),
Variables n 301 n 504 Beta (95% CI) P
Job satisfaction 21.2 (4.3) 20.3 (4.3) 1.05 (0.13, 1.96) 0.025
Emotional exhaustion 21.4 (12.4) 25.9 (12.7) 5.86 ( 8.62, 3.10) 0.001

n (%) n (%) OR (95% CI)


Depersonalization 124 (41.2) 239 (47.4) 0.72 (0.471.09) 0.121
Personal accomplishment 140 (46.5) 223 (44.3) 1.11 (0.781.57) 0.56
Satisfied with current scheduling 232 (77.1) 260 (51.6) 10.44 (5.7418.99) 0.001
12-hr shifts are important 257 (85.4) 326 (64.5) 2.37 (1.413.98) 0.001
No intention to stay 83 (28.7) 152 (31.3) 0.83 (0.501.37) 0.47
Missed shifts 101 (33.6) 276 (54.8) 0.42 (0.290.60) 0.001
Injury 175 (58.1) 285 (56.6) 1.12 (0.791.61) 0.52
Results are from Generalized Estimating Equations models with responses clustered by nursing unit and structural variables entered into all to
control for potential confounding.
CI indicates confidence interval.

TABLE 3. Comparison of System Outcomes by Shift Lengths


Variables 12-Hr, Mean (SD) 8-Hr, Mean (SD) Beta (95% CI) P
Recruitment and turnover n 17 n 12
RN vacancy 5.0 (3.3) 13.5 (7.9) 10.00 ( 12.20, 7.90) 0.04
Weeks to fill RN position 5.3 (3.6) 10 (2.6) 4.39 ( 7.71, 1.08) 0.001
RN turnover 0.28 (0.30) 0.05 (0.05) 0.03 ( 0.10, 0.16) 0.63
Staffing and costs n 14 n 14
TC HPPD 6.00 (2.96) 7.96 (2.50) 1.74 ( 3.84, 0.36) 0.36
RN HPPD 3.21 (1.51) 3.25 (4.16) 0.05 ( 1.17, 1.08) 0.94
Skillmix 0.50 (0.09) 0.49 (0.07) 0.02 ( 0.05, 0.08) 0.61
Total staffing cost 285 (126) 353 (107) 41.2 ( 49.4, 131.6) 0.37
RN staffing cost 188 (77) 237 (78) 9.08 ( 77.4, 59.3) 0.79
Absenteeism and costs n 14 n 14
RN sick hour per patient day 0.224 (0.10) 0.313 (0.13) 0.07 ( 0.02, 0.01) 0.59
RN sick wage per patient day 5.82 (2.61) 8.00 (3.64) 1.69 ( 3.77, 0.38) 0.49
Recruitment and turnover data came from 4 hospitals. Staffing, staffing costs, and absenteeism data came from 5 hospitals. All 99 nursing units
were eligible to provide data. Results are from Generalized Estimating Equations models with data clustered by hospital.
RN indicates registered nurse; TC, total care; HPPD, hours per patient day; OR, odds ratio; CI, confidence interval.

tinued research is needed to determine the most accurate and sample from the participating hospitals. Response rates were
feasible measurement of the quality of nursing care. higher in units with better nursing outcomes. However, the
There are a number of potential limitations to this response rate did not affect the results of the multivariate
study. Because our analysis was in 1 geographical region, models.
which is known for long commute times, these results may Most of the previous literature examining effects of
not be generalizable to other locales.41 Future research inves- shift length in hospitals is limited to 1 study site. Studying
tigating the staffing preferences of nurses and recruitment and multiple hospitals is a definite improvement. However, this
retention issues across locales would better inform these was a convenience sample of limited size and the hospitals
issues. The PSIs were the only variables for which patient were part of multiemployer bargaining group with 1 union
underlying risk was controlled for. To limit this potential bias located in New York City. Although these hospitals are likely
for other patient outcomes, we limited our analysis to similar to be representative of many urban environments, they may
types of units (ie, medical/surgical units). Whenever survey- not reflect hospitals in rural regions or nonunion environ-
ing is conducted, response bias must be considered. Our ments. Consistent with concentrations of minority popula-
survey response rate is favorable to the 38% to 53% response tions in urban regions, these hospitals employee a higher
rates recently found in other anonymous multisite surveys of proportion of minorities than are found in nationally, which
hospital-based nursing personnel.42 The lack of differences may limit the generalizability.43
between the demographics of the Pension data and the survey One of the problems encountered was that many of the
respondents gave us confidence that we had a representative participating hospitals were not equipped to efficiently pro-

1104 2006 Lippincott Williams & Wilkins


Medical Care Volume 44, Number 12, December 2006 8-Hour Versus 12-Hour Shifts

TABLE 4. Comparison of Quality Patient Care Outcomes by Shift Length


Summary Statistics
Variables 12-Hour, Mean (SD) 8-Hour, Mean (SD) Beta (95% CI) P
Incident reports n 15 n 11
Medication events 0.05 (0.02) 0.05 (0.02) 0.001 ( 0.027, 0.028) 0.95
Falls 0.10 (0.03) 0.09 (0.03) 0.006 ( 0.020, 0.032) 0.64
Decubitus ulcers 0.09 (0.10) 0.09 (0.03) 0.005 ( 0.016, 0.025) 0.66
Risk-adjusted patient safety indicators n 13 n 21
Decubitus ulcer 0.03 (0.02) 0.03 (0.04) 0.002 ( 0.029, 0.026) 0.61
Failure to rescue 0.06 (0.03) 0.08 (0.06) 0.006 ( 0.040, 0.027) 0.94

n (%) n (%) OR (95% CI)


Nurse perceptions of quality n 289 n 491
Quality of nurse care last shift
Excellent/good 229 (79%) 358 (73%) 1.35 (0.795, 2.293) 0.27
Overall quality of nursing unit
Excellent/good 220 (76%) 352 (72%) 1.03 (0.625, 1.699) 0.90
Overall quality of nursing unit/past year
Improved 93 (32%) 113 (23%) 1.26 (0.843, 1.916) 0.25
The unit of analysis for incident reports and risk-adjusted patient safety indicators is medical surgical nursing units. The target sample size for these
outcomes was 34 units. The incident report data came from 9 hospitals and the patient discharge abstracts for use in the patient safety indicators came from
13 hospitals. For incident report and patient safety indicators, data were clustered by hospital in all Generalized Estimating Equations models. For the nurse
perceptions of quality, the unit of analysis of is the nurse respondents, data came from all 14 hospitals and in the GEE models data are clustered by the nursing unit.
OR indicates odds ratio; CI, confidence interval.

vide needed administrative data elements, such as aggregate of results. Data from this study are being used to inform
payroll and unit-based recruitment and turnover data. This negotiations for individual participating hospitals. The largest
limited the power for the system outcome analyses and may hospital in this study has converted multiple medical/surgical
have caused some selection bias. Lack of investment in units to 12-hour scheduling since the results have been
management support to help evaluate best practices has been reported. We encourage others to conduct similar rigorous
noted elsewhere.44 There are many new electronic and web- studies with the ultimate goal of implementing evidence-
based technology applications becoming available to assist based management practices to improve nurse work environ-
administration, nurse managers and clinical nurses more ment while improving quality of care.
effectively and efficiently provide quality care. As informat-
ics infrastructure rapidly increases in hospitals, the burden of
REFERENCES
participating in multisite research and reviewing data to make
1. Frone MR. Work stress and alcohol use. Alcohol Res Health. 1999;23:
evidence-based management decisions may be reduced.45 284 291.
We controlled for organizational climate and other 2. Klerman EB. Clinical aspects of human circadian rhythms. J Biol
potential confounding variables to determine the independent Rhythms. 2005;20:375386.
effect of shift length on the outcomes of interest. However, 3. Aiken LH, Clarke SP, Sloane DM, et al. Hospital nurse staffing and
patient mortality, nurse burnout, and job dissatisfaction. JAMA. 2002;
shift length may positively affect organizational climate. 288:19871993.
Further research examining the relationship between these 4. Mozurkewich EL, Luke B, Avni M, et al. Working conditions and
variables is warranted. adverse pregnancy outcome: a meta-analysis. Obstet Gynecol. 2000;95:
This study represents an innovative attempt by a labor- 623 635.
5. Karasek RA, Theorell T, Schwartz JE, et al. Job characteristics in
management bargaining group to make an evidence-based relation to the prevalence of myocardial infarction in the US Health
decision. Leadership from labor and management were com- Examination Survey (HES) and the Health and Nutrition Examination
mitted to participating in an objective study conducted by a Survey (HANES). Am J Public Health. 1988;78:910 918.
nonpartisan outside research team. In this context, conduct- 6. Messing K. Womens occupational health: a critical review and discus-
sion of current issues. Women Health. 1997;25:39 68.
ing this study offered unique challenges and opportunities 7. Josten EJ, Ng ATJ, Thierry H. The effects of extended workdays on
related to design and analysis. For example, the population of fatigue, health, performance and satisfaction in nursing. J Adv Nurs.
interest was a limited sample size; and, the choice of depen- 2003;44:643 652.
dent variables was based not only on previous research, but 8. Ugrovics A, Wright J. 12-hour shifts: does fatigue undermine ICU nursing
judgments? Nurs Manage. 1990;21:64A, 64D, 64F-64A, 64D, 64G.
also important concerns from both constituencies. Impor- 9. Smith L, Folkard S, Tucker P, et al. Work shift duration: a review
tantly, the knowledge gained from this study had the unique comparing eight hour and 12 hour shift systems. Occup Environ Med.
opportunity of being implemented into practice quickly, 1998;55:217229.
which increased the interest of all stakeholders (researchers, 10. Hoffman AJ, Scott LD. Role stress and career satisfaction among
registered nurses by work shift patterns. J Nurs Adm. 2003;33:337342.
labor and management) to ensure appropriate interpretation

2006 Lippincott Williams & Wilkins 1105


Stone et al Medical Care Volume 44, Number 12, December 2006

11. Gillespie A, Curzio J. A comparison of a 12-hour and eight-hour shift 28. Somville PR, Nieuwenhuyse AV, Seidel L, et al. Validation of a
system. Nurs Times. 1996;92:36 39. self-administered questionnaire for assessing exposure to back pain
12. Campolo M, Pugh J, Thompson L, et al. Pioneering the 12-hour shift in mechanical risk factors. Int Arch Occup Environ Health. 2005;110.
Australiaimplementation and limitations. Aust Crit Care. 1998;11:112115. 29. Jenkins P, Earle-Richardson G, Slingerland DT, et al. Time dependent
13. Fitzpatrick JM, While AE, Roberts JD. Shift work and its impact upon memory decay. Am J Ind Med. 2002;41:98 101.
nurse performance: current knowledge and research issues. J Adv Nurs. 30. National Quality Forum. National Voluntary Consensus Standards for
1999;29:18 27. Nursing-Sensitive Care: An Initial Performance Measure Set. A Conse-
14. Makowiec-Dabrowska T, Krawczyk-Adamus P, Sprusinska E, et al. Can nus Report. Washington, DC: National Quality Forum; 2004.
nurses be employed in 12-hour shift systems? Int J Occup Safety Ergon. 31. Mattke S, Needleman J, Buerhaus P, et al. Evaluating the role of patient
2000;6:393 403. sample definitions for quality indicators sensitive to nurse staffing
15. Rogers AE, Hwang WT, Scott LD, et al. The working hours of hospital patterns. Med Care. 2004;42(2 Suppl):II21II33.
staff nurses and patient safety. Health Aff (Millwood). 2004;23:202212. 32. Agency for Healthcare Research and Quality. AHRQ Quality Indicators:
16. General Accounting Office. Nursing Workforce: Emerging Nurse Short- Guide to Patient Safety Indicators. 03-R203. Rockville, MD: Agency for
ages Due to Multiple Factors. GAO-01-944. Washington, DC: United Healthcare Research and Quality; 2003.
States General Accounting Office; 2001. 33. Reschovsky J, Reed M, Blumenthal D, et al. Physicians assessments of
17. Institute of Medicine. Keeping Patients Safe: Transforming the Work their ability to provide high-quality care in a changing health care
Environment of Nurses. Washington, DC: National Academy Press; system. Med Care. 2001;39:254 269.
2004. 34. Sochalski J. Is more better? The relationship between nurse staffing and
18. Donabedian A. The quality of care. How can it be assessed? JAMA. the quality of nursing care in hospitals. Med Care. 2004;42(2 Suppl):
1988;260:17431748. II67II73.
19. Stone PW, Harrison MI, Feldman P, et al. Organizational climate of staff 35. SPSS MIssing Value Analysis 7.5. Chicago, IL: MaryAnn Hill/SPSS
working conditions and patient safety: an integrative model in advances Inc.; 2002.
in patient safety. In Henriksen K, Battles JB, Marks ES, et al., eds. 36. Downey RG, King C. Missing data in Likert ratings: a comparison of
Concepts and Methodology (Volume 2, pp. 467 481). AHRQ Publica- replacement methods. J Gen Psychol. 1998;125:175191.
tion No: 05-0021-2. Rockville, MD: Agency for Healthcare Research 37. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear
and Quality; Feb 2005. models. Biometrika. 1986;73:1322.
20. Jensen JR. Incident Reports-Part 1 and FDA Alert. Hospital Risk 38. Sullivan C, Reading S. Nursing shortages: lets be flexible. Collegian.
Manage. 1997;1. 2002;9:24 28.
21. Choi J, Bakken S, Larson E, et al. Perceived Nursing Work Environ- 39. Bartel A. Measuring the employers return on investments in training:
ment. Nurs Res. 2004;53:370 378. evidence from the literature. Ind Rel. 2000;39:502524.
22. Beckstead JW. Confirmatory factor analysis of the Maslach Burnout
40. Jones CB. The costs of nurse turnover, part 2: application of the nursing
Inventory among Florida nurses. Int J Nurs Stud. 2002;39:785792.
turnover cost calculation methodology. J Nurs Adm. 2005;35:41 49.
23. Taunton RL, Bott MJ, Koehn ML, et al. The NDNQI-Adapted Index of 41. Nanji A. Its a long ride from Queens. Newsday City Edition. 4-14-2005.
work satisfaction. J Nurs Meas. 2004;12:101122. 42. Asch DA, Jedrziewski MK, Christakis NA. Response rates to mail surveys
24. Rittenhouse DR, Mertz E, Keane D, et al. No exit: an evaluation of published in medical journals. J Clin Epidemiol. 1997;50:1129 1136.
measures of physician attrition. Health Serv Res. 2004;39:15711588. 43. Health Resources and Services Administration. The National Sample
25. Hayes LJ, OBrien-Pallas L, Duffield C, et al. Nurse turnover: a Survey of Registered Nurses 2000. Rockville, MD: Bureau of Health
literature review. Int J Nurs Stud. 2006;43:237263. Professions, Health Resources and Services Administration; 2000.
26. Sohn S, Eagan J, Sepkowitz KA. Safety-engineered device implemen- 44. Kovner AR, Elton JJ, Billings J. Evidence-based management. Front
tation: does it introduce bias in percutaneous injury reporting? Infect Health Serv Manage. 2000;16:324.
Control Hosp Epidemiol. 2004;25:543547. 45. Bakken S, Cimino JJ, Hripcsak G. Promoting patient safety and enabling
27. Blegen MA, Vaughn T, Pepper G, et al. Patient and staff safety: evidence-based practice through informatics. Med Care. 2004;42(2
voluntary reporting. Am J Med Qual. 2004;19:6774. Suppl):II49 II56.

APPENDIX

APPENDIX. Characteristics of Each Hospital


Hospital
1 2 3 4 5 6 7 8 9 10 11 12 13
Hospital size
Beds 212 780 530 387 235 200 222 300 461 250 332 358 279
No. medical/surgical units
8-hr 4 2 2 1 7 2 1
12-hr 1 4 3 3 3 1
No. adult general specialty units
8-hr 5 11 5 9 4 3 1 5 1
12-hr 1 4 1 2 1 6 2 1 3
No. eligible nurses for survey 181 388 221 171 139 68 123 150 134 41 150 117 52
Survey response rate (%) 31 33 52 70 48 25 33 26 40 22 62 48 58
Annual shifts by schedule type
8-hr 260 260 260 260 260 260 208 260 260
12-hr 169* 169 169 169 156 164 163 169 169 156 169
*One 12-hr unit scheduled nurses only 164 shifts.

1106 2006 Lippincott Williams & Wilkins

Vous aimerez peut-être aussi