Patient cost-sharing and insurance

arrangements are associated with

hospital readmissions after abdominal
surgery: Implications for access
and quality health care
Bora Youn, MS,a Marina Soley-Bori, MA,b Rene Soria-Saucedo, MD, MPH, PhD,b
Colleen M. Ryan, MD,c,d,e Jeffrey C. Schneider, MD,d,f Alex B. Haynes, MD, MPH,c,d,g,h
Howard J. Cabral, PhD, MPH,a and Lewis E. Kazis, ScD,b Boston, MA

Background. Readmission rates after operative procedures are used increasingly as a measure of hospital
care quality. Patient access to care may influence readmission rates. The objective of this study was to
determine the relationship between patient cost-sharing, insurance arrangements, and the risk of
postoperative readmissions.
Methods. Using the MarketScan Research Database (n = 121,002), we examined privately insured,
nonelderly patients who underwent abdominal surgery in 2010. The main outcome measures were risk-
adjusted unplanned readmissions within 7 days and 30 days of discharge. Odds of readmissions were
compared with multivariable logistic regression models.
Results. In adjusted models, $1,284 increase in patient out-of-pocket payments during index admission (a
difference of one standard deviation) was associated with 19% decrease in the odds of 7-day readmission
(odds ratio [OR] 0.81, 95% confidence interval [CI] 0.78–0.85) and 17% decrease in the odds of 30-
day readmission (OR 0.83, 95% CI 0.81–0.86). Patients in the noncapitated point-of-service plans (OR
1.19, 95% CI 1.07–1.33), preferred provider organization plans (OR 1.11, 95% CI 1.03–1.19), and
high-deductible plans (OR 1.12, 95% CI 1.00–1.26) were more likely to be readmitted within 30 days
compared with patients in the capitated health maintenance organization and point-of-service plans.
Conclusion. Among privately insured, nonelderly patients, increased patient cost-sharing was associated
with lower odds of 7-day and 30-day readmission after abdominal surgery. Insurance arrangements also
were significantly associated with postoperative readmissions. Patient cost sharing and insurance
arrangements need consideration in the provision of equitable access for quality care. (Surgery
2016;159:919-29.)

From the Department of Biostatistics,a Boston University School of Public Health; Center for the Assessment of
Pharmaceutical Practices (CAPP), Department of Health Policy and Management,b Boston University School
of Public Health; Department of Surgery,c Massachusetts General Hospital; Harvard Medical Schoold;
Shriners Hospitals for Children–Bostone; Department of Physical Medicine and Rehabilitation,f Spaulding
Rehabilitation Hospital; Codman Center for Clinical Effectiveness in Surgery,g Massachusetts General Hos-
pital; and Ariadne Labs,h Harvard School of Public Health and Brigham and Women’s Hospital, Boston,
MA

THE HOSPITAL READMISSIONS REDUCTION PROGRAM, estab-
lished by the Patient Protection and Affordable
Care Act, penalizes hospitals with greater-than-
Accepted for publication September 10, 2015.
Reprint requests: Lewis E. Kazis, ScD, Professor, Center for the
Assessment of Pharmaceutical Practices (CAPP), Department of
Health Policy and Management, Boston University School of
Public Health; 715 Albany Street, Boston, MA. E-mail: lek@bu.
edu.
0039-6060/$ - see front matter
Ó 2016 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.surg.2015.09.007
expected unplanned readmission rates among
Medicare patients with acute myocardial infarc-
tion, congestive heart failure, pneumonia, chronic
obstructive pulmonary disease, total hip arthro-
plasty, and total knee arthroplasty. The Centers
for Medicare and Medicaid Services (CMS) plan
to expand applicable conditions of the Hospital
Readmissions Reduction Program to coronary ar-
tery bypass graft surgery.1
Many studies have examined the predictors of
readmissions in various contexts,2 yet many issues,
to follow, remain unresolved. First, operative

SURGERY 919
920 Youn et al
readmissions should be distinguished from medi-
cal readmissions. Although medical readmissions
often are driven by diminished status of underlying
medical conditions, operative readmissions are
often the result of postoperative complications.3-5
Second, early readmissions may be a better quality
of care indicator than later in time readmissions,
because they may capture potentially preventable
care transition issues. Surgery patients discharged
from the hospital may need extensive transitional
care, including home nursing, wound care, and
outpatient follow-ups.5 Readmissions 3–4 weeks af-
ter surgery are more likely to be related to under-
lying medical or socioeconomic conditions.5
Third, little is known about readmission rates
and contributing factors among young and
middle-aged populations. Nonelderly patients ac-
count for more than 60% of short-term hospital
discharges in the United States,6 but much of the
existing knowledge about predictors of readmis-
sions is derived from Medicare patients aged
65 years and over.7-10 Fourth, the associations be-
tween patient cost-sharing, insurance arrange-
ments, and postoperative readmissions are
unknown. Factors that affect hospital utilization,
such as patient-cost sharing, may influence the de-
cision to return to the hospital.11 The range of
covered services, treatment waiting time, primary
care physician follow-up, and availability of urgent
care clinic vary by insurance plan and also may
affect surgical readmission rates.12 Health care
providers paid by capitation may have incentives
to lessen readmission rates.13 Few studies, however,
have shown how these factors affect hospital read-
missions after operative care.
This study examines the factors associated with
increased risk of unplanned hospital readmissions
after abdominal surgery in a privately insured,
nonelderly population. Our objective is to deter-
mine whether increased patient out-of-pocket pay-
ments reduce the odds of readmission and
whether readmissions are associated with the
patient’s insurance arrangements. We evaluated
both 7-day and 30-day readmissions to provide a
better understanding of potentially preventable
early readmissions.
METHODS
Data. This observational study used the Market-
Scan Commercial Claims and Encounters Database
from January 2010 to December 2010. The data-
base includes health insurance claims from large
employers and health plans across the United
States that provide private health care coverage
to more than 56 million employees, their spouse,
Surgery
March 2016
and dependents.14 The MarketScan database was
linked to 2011 American Community Survey 5-
Year Estimate data to obtain information on the
median household income of patient’s residential
area (3-digit zip-code).
Study sample. Nonelderly patients who under-
went at least 1 major abdominal surgery in 2010
were included in the analysis. The following pro-
cedures were selected based on Diagnosis-Related
Group codes: (1) appendectomy; (2) hepatobiliary
and pancreatic procedures; (3) major small and
large bowel procedures; (4) cholecystectomy; (5)
laparoscopic cholestectomy; (6) operating room
procedures for obesity; (7) peritoneal adhesiolysis;
(8) rectal resection; (9) stomach, esophageal,
duodenal procedure; and (10) splenectomy
(Supplementary Table I). Only the first abdominal
surgery was considered if the patient underwent
multiple operations during the study period. We
excluded from the analysis patients who were
transferred to another short-term, acute care hos-
pital, died during the index hospitalization, were
discharged against medical advice, or were still
within the same hospital (days billed for leave of
absence or interim billing). Using the algorithm
from 2014 Hospital-Wide All-Cause Unplanned Re-
admission measures developed by the Yale New
Haven Health Services Corporation/Center for
Outcomes Research & Evaluation, we also
excluded planned readmissions.15
Variables. Primary study outcomes were any
unplanned readmission within 7 days and 30 days
of discharge for abdominal surgery. Because oper-
ative patients are likely to be readmitted for post-
operative complications, both 7-day and 30-day
periods were assessed separately as primary
outcomes.
Primary predictors were insurance arrange-
ments and patient cost-sharing. On the basis of
incentives to use a particular network of providers,
primary care physician assignment, and capita-
tion,16 insurance plans were categorized as follows:
(1) health maintenance organization and point-of-
service with capitation (HMO/POS with capita-
tion), (2) exclusive provider organization (EPO),
(3) noncapitated POS, (4) high deductible plans,
(5) noncapitated preferred provider organization
(PPO), and (6) comprehensive plans (Table I).
For enrollees in HMO/POS with capitation,
EPO, and POS plans, primary care physicians
arrange medical services, including hospitaliza-
tions and referrals to specialists. Enrollees in
PPO and POS plans have financial incentives to
use selected networks of health care providers; en-
rollees can seek care out of network but will incur
Surgery
Volume 159, Number 3
Table I. Characteristics of insurance plans*
HMO/POS
Plan type with capitation
Require primary care physician as Yes
a gate keeper
Specialty referrals Yes
Cover out of network care Varies
Partially or fully capitated Yes
*Adapted from information available on the Thomson Reuters MarketScan Database.
EPO, Exclusive Provider Organization; HMO/POS with capitation, Health Maintenance Organization and Point-of-Service with capitation; POS, noncapi-
tated Point-of-Service; PPO, Preferred Provider Organization.
greater out-of-pocket payments. High-deductible
plan enrollees can pay a lower annual premium
to obtain health insurance in exchange for greater
deductibles. In capitated insurance plans, health
care providers are paid a set amount for each en-
rollee assigned.17,18 HMO/POS with capitation
was used as the reference group, because lesser
rates of hospital use are observed in HMO
settings.19
We calculated patient out-of-pocket (OOP)
payments by summing the amount of deductible,
copayment, and coinsurance during the index
admission. The distribution of OOP payments
was skewed because of some extreme values. It
was top-coded to a maximum value of $6,411
corresponding to the 99% percentile in the
adjusted analyses.20-22 This value is similar to the
OOP maximum for any individual Marketplace
plan under Affordable Care Act ($6,600 in
2015).23 Median household income of patient’s
3-digit zip-code residential area were categorized
into tertiles ($14,947
$46,309, $46,322
$58,429,
and $58,506
$135,199). Adjusted models
controlled for age, sex, comorbidities (Elixhauser
comorbidity index),24 surgery type, duration of
stay at index admission, discharge status, rural
area, and US geographic region.
Statistical analyses. v2 tests, t tests, and Wilcoxon
rank-sum tests were used to examine the difference
between those who were readmitted within 7 days
and 30 days versus those who were not readmitted.
Two multivariable logistic regression analyses were
performed to assess the likelihood of being read-
mitted in 7 days and 30 days after discharge.
Odds ratios with 95% confidence intervals and
P-values were reported.
Model calibration was determined by the
Hosmer-Lemeshow test and inspection of observed
versus predicted values by decile of predicted 7-day
and 30-day readmission risk. Model discrimination
was assessed by c-statistics. The robustness of
Youn et al 921
POS without High
EPO capitation Deductible PPO Comprehensive
Yes Yes No No No
Yes Yes No No No
No Yes, with Varies Yes, with N/A
greater cost greater cost
No No No No No
insurance plan and OOP payments findings were
tested with Bonferroni and bootstrap techniques.25
Multicollinearity was assessed with the variance
inflation factor. We examined the presence of clus-
tering effects by insurance arrangements using a
random effect mixed logistic model. Missing values
of the study variables were judged not consequen-
tial as they accounted for less than 5% of the obser-
vations. All analyses were conducted using SAS
version 9.3 software (SAS Institute, Cary, NC).
The Boston University Institutional Review Board
approved this study.
RESULTS
Patient characteristics by readmission status.
Our study sample included 121,002 patients
(Table II). The 30-day unplanned readmission
rate was 6.8%. Nearly half of the 30-day readmis-
sions occurred within 7 days (47.8%). Readmitted
patients were on average older (age group 45–54
and 55–64 years), sicker (with more comorbid-
ities), had a longer index duration of stay, and
faced less cost-sharing during the index admission
compared with those not readmitted.
Patient characteristics by insurance plan. Patient
characteristics differed by insurance arrangements
(Table III). Patients in the high-deductible plan
paid the greatest OOP payments, followed by pa-
tients in the PPO plan (median payment $1,309
and $994, respectively). Patients in the HMO/
POS with capitation plan and comprehensive
plan paid the lowest OOP payments (median pay-
ment $150 and $49, respectively). Patient’s OOP
payments as a percentage of area household in-
come were obtained to examine patient’s financial
burden. Approximately 22% of patients in the
high-deductible plan and 15% of patients in the
PPO plan paid more than 5% of their median
household income as OOP payments during index
admission. We found considerable variation in
patient characteristics between those in the
922 Youn et al Surgery
March 2016

Table II. Patient characteristics by readmission status*

Readmitted Readmitted
All patients within 7 d P valuey within 30 d P valuez
Overall, n (%) 121,002 3,924 (3.2) 8,209 (6.8)
Age group, y, n (%)
0
17 10,904 252 (2.3) 506 (4.6)
18
34 22,577 602 (2.7) 1,194 (5.3)
35
44 23,526 651 (2.8) <.001 1,457 (6.2) <.001
45
54 31,389 1,132 (3.6) 2,346 (7.5)
55
64 32,606 1,287 (4.0) 2,706 (8.3)
Sex, n (%)
Male 49,617 1,691 (3.4) .007 3,499 (7.1) .002
Female 71,385 2,233 (3.1) 4,710 (6.6)
Insurance plan, n (%)
HMO/POS with capitation 16,831 549 (3.3) 1,126 (6.7)
EPO 3,483 95 (2.7) 216 (6.2)
POS without Capitation 9,025 322 (3.6) .18 662 (7.3) .09
High deductible 7,863 236 (3.0) 513 (6.5)
PPO without Capitation 77,952 2,532 (3.3) 5,295 (6.8)
Comprehensive 3,108 104 (3.4) 232 (7.5)
Out-of-pocket payments, median (IQR), $x 746 (1,693) 392 (1,283) <.001 407 (1,341) <.001
Median household income of residence, 54,991 (14,673) 54,658 (14,311) .14 54,772 (14,417) .16
mean (SD), $jj
Number of Elixhauser comorbidity conditions, 1.0 (1.2) 1.3 (1.3) <.001 1.4 (1.3) <.001
Mean (SD)
Surgery type, n (%)
Stomach, esophageal, duodenal procedure 8,618 394 (4.6) 832 (9.7)
Hepatobiliary and pancreatic procedure 3,718 273 (7.3) 635 (17.1)
Major small & large bowel procedure 25,799 1,337 (5.2) 2,786 (10.8)
Splenectomy 591 26 (4.4) 50 (8.5)
Rectal resection 713 48 (6.7) <.001 115 (16.1) <.001
Peritoneal adhesiolysis 5,474 240 (4.4) 504 (9.2)
Cholecystectomy 2,508 77 (3.1) 165 (6.6)
Operating room procedures for obesity 19,108 366 (1.9) 881 (4.6)
Appendectomy 29,039 534 (1.8) 1,036 (3.6)
Laparoscopic cholecystectomy 25,434 629 (2.5) 1,205 (4.7)
Duration of stay, median (IQR), dx 3 (4) 5 (7) <.001 5 (6) <.001
Discharge status, n (%)
Discharged to home 114,483 3,443 (3.0) <.001 7,441 (6.5) <.001
Transferred/other 1,000 268 (26.8) 368 (36.8)
Urban/rural, n (%)
Urban 102,090 3,333 (3.3) .33 6,934 (6.8) .81
Rural 18,892 591 (3.1) 1,274 (6.7)
Region, n (%)
Northeast 18,918 609 (3.2) 1,258 (6.7)
Midwest 29,127 991 (3.4) .12 2,098 (7.2) <.001
South 48,373 1,579 (3.3) 3,350 (6.9)
West 24,492 744 (3.0) 1,498 (6.1)
*All percentages are row percent. v2 tests for categorical variables, T-tests for median household income of residence and Elixhauser comorbidities, and
Wilcoxon rank-sum tests (2-sided) for out-of-pocket payments and duration of stay. Missing values were 2.26% for insurance plan, 0.03% for median
household income, 4.56% for discharge status, 0.08% for region, and 0.02% for urban/rural.
yCompared with those not readmitted within 7 d.
zCompared with those not readmitted within 30 d.
xDuring the index admission.
jjThree-digit zip code area.
EPO, Exclusive Provider Organization; HMO/POS with capitation, Health Maintenance Organization and Point-of-Service with capitation; IQR, interquartile
range; POS, noncapitated Point-of-Service; PPO, Preferred Provider Organization; SD, standard deviation.
Surgery Youn et al 923
Volume 159, Number 3

Table III. Patient characteristics by insurance arrangements*

Health maintenance Exclusive Preferred
organization and provider Noncapitated provider
point-of-service with organization point-of-service High deductible organization Comprehensive
capitation (n = 16,831) (n = 3,483) (n = 9,025) (n = 7,683) (n = 77,952) (n = 3,108)
Number of Elixhauser comorbidity conditions, n (%)
0 7,407 (44.0) 1,561 (44.8) 4,203 (46.6) 3,549 (45.1) 33,398 (42.8) 930 (29.9)
1 4,876 (29.0) 1,168 (33.5) 2,872 (31.8) 2,434 (31.0) 22,894 (29.4) 882 (28.4)
2 2,640 (15.7) 527 (15.1) 1,239 (13.7) 1,266 (16.1) 12,691 (16.3) 675 (21.7)
3+ 1,908 (11.3) 227 (6.5) 711 (7.9) 614 (7.8) 8,969 (11.5) 621 (20.0)
Age group, y, n (%)
0
17 1,678 (10.0) 327 (9.4) 730 (8.1) 749 (9.5) 7,051 (9.1) 80 (2.6)
18
34 3,086 (18.3) 751 (21.6) 1,638 (18.2) 1,500 (19.1) 14,842 (19.0) 222 (7.1)
35
44 3,449 (20.5) 797 (22.9) 1,645 (18.2) 1,640 (20.9) 15,262 (19.6) 196 (6.3)
45
54 4,453 (26.5) 951 (27.3) 2,236 (24.8) 2,035 (25.9) 20,407 (26.2) 629 (20.2)
55
64 4,165 (24.8) 657 (18.9) 2,776 (30.8) 1,939 (24.7) 20,390 (26.2) 1,981 (63.7)
Sex, n (%)
Male 6,589 (39.1) 1,474 (42.3) 3,531 (39.1) 3,163 (40.2) 32,531 (41.7) 1,214 (39.1)
Female 10,242 (60.9) 2,009 (57.7) 5,494 (60.9) 4,700 (59.8) 45,421 (58.3) 1,894 (60.9)
Out-of-pocket 150 (500) 309 (1,092) 385 (1,168) 1,309 (2,273) 994 (1,742) 49 (297)
payments (OOP),
median (IQR), $y
OOP per median household income, n (%)z
<2% 14,033 (83.4) 2,615 (75.1) 6,385 (70.8) 3,472 (44.2) 41,046 (52.7) 2,628 (84.6)
2%# and <5% 2,083 (12.4) 742 (21.3) 1,966 (21.8) 2,667 (33.9) 25,073 (32.2) 311 (10.0)
5%# and <10% 591 (3.5) 115 (3.3) 585 (6.5) 1,403 (17.8) 9,823 (12.6) 118 (3.8)
10%# 124 (0.7) 11 (0.3) 89 (1.0) 321 (4.1) 2,010 (2.6) 51 (1.6)
Duration of stay, 2 (4) 2 (3) 3 (4) 3 (4) 3 (4) 4 (5)
median (IQR), dy
*Percentage may not sum to 100 because of rounding.
yDuring the index admission.
zOut-of-pocket payments as a percentage of median household income of residence (3-digit zip code).
IQR, Interquartile range.

comprehensive plan compared with the other
plans. Patients in the comprehensive plan had
more comorbidities, were more likely to be in
the oldest age group, and were classified as
retirees.
Readmissions by OOP payments and insurance
plan. Table IV gives the adjusted odds of readmis-
sion within 7 days and 30 days after discharge. Re-
sults suggest a $1,284 increase in patient OOP
payments during index admission (a difference
of one standard deviation) is associated with 19%
decrease in the odds of 7 day readmission (odds
ratio [OR] 0.81, 95% confidence interval [CI]
0.78–0.85) and 17% decrease in the odds of
30 day readmission (OR 0.83, 95% CI 0.81–0.86).
We also examined OOP payments by deciles
(Fig). Compared with patients who paid no OOP
payments, those who faced greater OOP payments
had consistently lower odds of 7 days and 30 days
readmission.
For insurance arrangements, patients in the
POS plan without capitation arrangements were
more likely to be readmitted both within 7 days
(OR 1.21; 95% CI 1.04–1.40) and 30 days (OR
1.19; 95% CI 1.07–1.33) compared with patients in
the HMO/POS plans with capitation. Patients in
the PPO plan without capitation also were more
likely to be readmitted both within 7 days (OR
1.11; 95% CI 1.01–1.23) and 30 days (OR 1.11;
95% CI 1.03–1.19) again compared with patients
in the HMO/POS plans with capitation. Patients in
the high-deductible plan were more likely to be
readmitted within 30 days (OR 1.12; 95% CI 1.00–
1.26). The odds of readmissions also were signifi-
cantly associated with surgery type, Elixhauser
comorbidities, duration of stay, age group, and
discharge status.
Model performance and validation. Both multi-
variable models showed good calibration. The P-
values of the Hosmer and Lemeshow test for the
7-day model and 30-day model were all greater
than 0.05 (0.34 and 0.39, respectively). There
were no differences between observed and ex-
pected readmission rates in subgroups of the
924 Youn et al Surgery
March 2016

Table IV. Risk-adjusted odds of 7-day and 30-day readmission

Readmitted within Readmitted within
7 d, odds ratio (95% CI) 30 d, odds ratio (95% CI)
Insurance plan
HMO/POS with capitation 1 [Reference] 1 [Reference]
EPO 1.00 (0.79–1.25) 1.07 (0.92–1.26)
High deductible 1.10 (0.93–1.29) 1.12 (1.00–1.26)x
POS without capitation 1.21 (1.04–1.40)x 1.19 (1.07–1.33)z
PPO without capitation 1.11 (1.01–1.23)x 1.11 (1.03–1.19)z
Comprehensive 0.83 (0.66–1.04) 0.88 (0.75–1.03)
Out-of-pocket payments* 0.81 (0.78–0.85)z 0.83 (0.81–0.86)z
Elixhauser comorbidity conditions, no. 1.08 (1.05–1.11)z 1.11 (1.09–1.13)z
Duration of stayy 1.66 (1.58–1.74)z 1.88 (1.81–1.94)z
Discharge status
Discharged to home 1 [Reference] 1 [Reference]
Transferred/other 4.91 (4.20–5.75)z 2.95 (2.56–3.40)z
Surgery type
Laparoscopic cholecystectomy 1 [Reference] 1 [Reference]
Stomach, esophageal, duodenal procedure 1.30 (1.13–1.50)z 1.42 (1.28–1.57)z
Hepatobiliary and pancreatic procedure 1.53 (1.30–1.81)z 1.97 (1.75–2.21)z
Major small and large bowel procedure 1.30 (1.16–1.45)z 1.36 (1.26–1.48)z
Splenectomy 1.22 (0.79–1.88) 1.22 (0.89–1.68)
Rectal resection 1.72 (1.25–2.36)z 2.21 (1.77–2.75)z
Peritoneal adhesiolysis 1.30 (1.10–1.52)z 1.36 (1.21–1.53)z
Cholecystectomy 0.93 (0.72–1.19) 0.99 (0.83–1.18)
OR procedures for obesity 0.88 (0.77–1.01) 1.14 (1.04–1.25)z
Appendectomy 0.92 (0.81–1.05) 0.97 (0.89–1.07)
c-statistic 0.69 0.71
*For a $1,284 difference in the patient’s out-of-pocket payments during the index admission (a difference of one standard deviation).
yDuring the index admission, log transformed.
zSignificant at .01 level.
xSignificant at .05 level.
Model adjusted for age group, sex, rural area, median household income, and geographic region (not shown on table).
CI, Confidence interval; EPO, Exclusive Provider Organization; HMO/POS with capitation, Health Maintenance Organization and Point-of-Service with
capitation; OR, operating room; POS, noncapitated Point-of-Service; PPO, Preferred Provider Organization.

model population. The goodness-of-fit for both
models was credible (c-statistics = 0.69 for 7-day re-
admission model, 0.71 for 30-day readmission
model).
There was no evidence of either collinearity
between the predictor variables or clustering by
insurance arrangements. All variance inflation
factor values were below 10. Intracluster correla-
tions were very small and nonsignificant. The
results of Bonferroni and bootstrap inference
assessing the association between insurance
arrangements, OOP payments, and 7-day and 30-
day readmissions remained stable and consistent.
We also examined selected statistical interactions
between insurance arrangements and other pre-
dictors in the model (OOP payments, Elixhauser
comorbidities, duration of stay, discharge status,
and age group). There were no improvements to
the model fit by the addition of these interactions
(results available upon request).
Sensitivity analysis. We performed stratified
analyses by insurance plans and examined whether
the effects of OOP payments are the same across
insurance plans. Table V gives the adjusted odds of
readmission within 7 days and 30 days after
discharge per an increase of $1,284 in OOP pay-
ments. The influence of OOP payments on read-
missions varied by insurance arrangements.
Patients in high-deductible, POS, and PPO plans
had 13%, 35%, and 20% reduced odds of 7-day re-
admissions for a $1,284 increase in OOP payments,
respectively. Odds of 30-day readmissions also were
reduced by 14%, 29%, and 18%, respectively. On
the contrary, OOP payments were not significant
predictors of 7-day and 30-day readmissions for pa-
tients in HMO/POS with capitation, EPO, and
comprehensive plans.
We examined the robustness of the main model
stratified by median household income tertiles and
by the 4 most frequent operations: major small and
Surgery Youn et al 925
Volume 159, Number 3

Fig. Risk-adjusted odds ratio of 7-day readmission (A) and 30-day readmission (B) by out-of-pocket payment deciles
(reference group = no out-of-pocket payment). The range of out-of-pocket payments of each decile are: $0 < D1
#$200, $200 < D2 #$300, $300 < D3 #$554.34, $554.34 < D4 #$879.12, $879.12 < D5 #$1,150, $1,150 < D6
#$1,479.41, $1,479.41 < D7 #$1,856.39, $1,856.39 < D8 #$2,300, $2,300 < D9 #$3,039.46, $3,039.46 < D10. ORs
are adjusted for age group, sex, rural area, geographic region, insurance plan, median household income of residence,
Elixhauser comorbidity conditions, duration of stay, discharge status, and surgery type.

large bowel procedures and operating room
procedures for obesity, appendectomy, and laparo-
scopic cholecystectomy (Supplementary Tables
II–V). ORs in the stratified analyses were similar
to the main model described in Table IV, especially
for the OOP payments variable. The overall rela-
tionships with insurance arrangements were fairly
consistent with the main model, except for appen-
dectomy and the highest income tertile. The odds
of 7-day readmission for patients in the highest in-
come tertile across insurance arrangements were
similar.
We assessed the effect of adjusting both the
7-day and 30-day readmission models for the
number of hospitalizations in 2010 before the
index abdominal surgery. Low OOP payment at
the index operation may be a surrogate for high
use. Multiple previous admissions may lead to
reaching caps on deductibles and other OOP
spending. The relationship between OOP and
insurance plans with 7-day and 30-day readmission
remained stable after we controlled for the num-
ber of previous hospitalizations (results not shown
but are available on request). In addition, we
excluded patients in comprehensive plans from
the main model and examined whether the results
remain stable. Those in comprehensive plans
considerably differ in demographics and comor-
bidities from those in other plans. Again, the odds
of 7-day and 30-day readmissions were similar to
the main model (results not shown but are avail-
able upon request).
DISCUSSION
To our knowledge, this is the first study to
examine the associations between patient cost
sharing and insurance arrangements with hospital
readmission in a nonelderly, primarily middle class
926 Youn et al
Table V. Risk-adjusted odds of 7-day and 30-day readmission per $1,284 increase in patient out-of-pocket
payments, stratified by insurance plan
Readmitted within 7 days,
odds ratio (95% CI)
HMO/POS with capitation
EPO
High deductible
POS without capitation
PPO without capitation
Comprehensive
ySignificant at .05 level.
*Significant at .01 level.
Model adjusted for age group, sex, rural area, geographic region, median household income of residence, Elixhauser comorbidity conditions, duration of
stay, discharge status, and surgery type (not shown on table).
EPO, Exclusive Provider Organization; HMO/POS with capitation, Health Maintenance Organization and Point-of-Service with capitation; POS, noncapi-
tated Point-of-Service; PPO, Preferred Provider Organization.
working population. We found that increased pa-
tient cost sharing at the index admission consis-
tently reduces the odds of readmission at both
7 days and 30 days. Patients in POS plans and PPO
plans without capitation were more likely to be
readmitted within 7 days and 30 days compared
with patients in HMO/POS with capitation plans.
Patients in high-deductible plans also were more
likely to be readmitted within 30 days compared
with patients in HMO/POS with capitation plans.
Patients’ financial concerns and insurance plan
structures are well-known predictors of hospital
use,26 and they may affect the decision to return to
the hospital after surgery. Epstein et al8 have
demonstrated that hospital readmissions are
related to the general incentives to use hospital ser-
vices. Different organizational characteristics and
service arrangements, such as primary care physi-
cian assignment and home health services
coverage, may affect readmission rates. According
to an analysis of Medicare data, health care systems
that are more dependent on hospital-based care
have greater readmission rates.11 From the
perspective of health care providers, those paid
by capitation have strong incentives to prevent re-
admissions and reduce costs by providing outpa-
tient follow-up and case management.13 Lemieux
et al13 have reported that patients in capitated
Medicare Advantage plan have significantly lower
surgical readmission rates than those in traditional
Medicare fee-for-service plans.
Few studies have specifically examined the asso-
ciation between private insurance arrangements
and readmissions.27-31 Most of them used data
from specific geographic regions or compared
broad categories of managed care plans versus
fee-for-service plans. Allen et al27 compared 30-
days readmission rates between managed care
Surgery
March 2016
Readmitted within 30 days,
odds ratio (95% CI)
0.94 (0.82–1.08) 0.96 (0.87–1.06)
0.70 (0.47–1.03) 1.00 (0.81–1.24)
0.87 (0.77–0.98)* 0.86 (0.80–0.94)y
0.65 (0.53–0.79)y 0.71 (0.62–0.82)y
0.80 (0.77–0.84)y 0.82 (0.79–0.84)y
0.85 (0.60–1.21) 0.90 (0.72–1.12)
versus fee-for-service plans in heart failure patients.
Two other studies examined the differences in re-
admission rates between HMO versus non-HMO
plans.30,31 None of these studies found statistically
significant differences between the 2 insurance ar-
rangements. Although our findings differ from
previous research on private insurance plans, this
study is based on national data with greater granu-
larity of more specific definitions of insurance cat-
egories and includes a postoperative patient
population.
To date, no studies have examined the associa-
tion between patient cost-sharing and readmission.
However, our findings are consistent with existing
literature suggesting an inverse relationship be-
tween the amount of patient cost-sharing and the
use of inpatient services.19,26,32 Cost-sharing may
become a financial barrier to care when patients
face various health conditions.32,33 Although
most of the readmissions in our study were associ-
ated with acute postoperative complications,
greater cost sharing during index admission can
make patients more cost-sensitive, influencing
their subsequent use of inpatient services. In the
RAND Health Insurance Experiment, hospital ad-
missions were 29% greater in the free-care group
compared with the cost-sharing group.26 Siu
et al32 found that cost-sharing reduces both medi-
cally appropriate and inappropriate hospital ad-
missions. Cost-sharing also reduced the
probability of obtaining medical care for acute ill-
nesses such as infections.33 Our findings on capi-
tated insurance plans also are consistent with
those from the RAND health Insurance Experi-
ment. Hospital admissions were 40% less in
prepaid group practice settings (HMOs)
compared with unmanaged fee-for-service plans.34
Evidence from many other studies also supports
Surgery
Volume 159, Number 3
that patients in HMOs use inpatient medical ser-
vices less frequently than those in fee-for-service
basis plans.19,35-37
We also observed a lack of financial protection
and underinsurance in a number of patients.
Waters et al38 defined lack of financial protection
as annual OOP payments exceeding 10% of house-
hold income. We only obtained OOP payments
during initial admission, yet 4% of patients in the
high deductible plan and 2.6% in PPO plan spend
more than 10% of their median household income
of residence for their abdominal surgery. Consid-
ering the high rate of underinsurance among pri-
vately insured adults in 2010,39,40 it is likely that
greater proportions of patients experienced cata-
strophic health expenditures during 2010.
Our results help better understand early read-
missions after surgical care. In our sample, nearly
half (47.8%) of the 30-day readmissions occurred
within a week of discharge. Approximately 40% of
patients who were readmitted within 7 days had a
primary diagnosis of surgical complications at
readmission, whereas only 20% of patients who
were readmitted in the last week of the 30-day
readmission period had the same primary diag-
nosis (Supplementary Tables VI–VIII in the data).
Future studies should explore whether interven-
tions targeting early readmissions can reduce over-
all readmission rates.
Limitations of the study include the following.
First, different patient mix in insurance arrange-
ments (selection bias) may have influenced our
results. Although we controlled for various patient-
level characteristics, including patient demo-
graphics, comorbidities, and area household
income, patients in each plan may have differed
in unmeasured characteristics. High-deductible
plans generally attract healthier and more cost-
conscious enrollees. On the contrary, sicker pa-
tients more often are enrolled in more generous
plans, because they are more likely to incur high
health care costs. The amount of cost-sharing and
insurance plan types may be related to patient
resources, access requirements, and underlying
health status.41,42 Even after our adjustments, the
results may still be confounded by unmeasured
patient characteristics. For example, patient-level
variables characterizing disease severity or socio-
economic status were not available; however, our
insurance and cost-sharing results remained stable
across several sensitivity analyses as previously
described. In addition, patients in comprehensive
plans were considerably different from those in
other plans in this study. They had more comor-
bidities, were more likely to be in the oldest age
Youn et al 927
group, and characterized as retirees. Results from
this sample may thus not be generalizable to other
comprehensive plans.
Second, although patients are unlikely to be
aware of the exact amount of OOP payment at the
time of readmission, most are likely to have some
understanding of the OOP payment structure of
their health insurance, and accordingly, become
cost-sensitive on their use of health services. Ac-
cording to the Kaiser Foundation Health Insur-
ance Literacy Survey, most US adults who have
health insurance appear to have good knowledge
of how insurance works in general.43
Third, we examined the amount of cost-sharing
only at the index admission. The overall financial
burden on each patient may include insurance
premiums and other OOP payments throughout
the course of their illness. Fourth, to date, the
interaction between cost-sharing and insurance
plan type is not well-known.44 Although we per-
formed stratified analysis by insurance plan, distin-
guishing the combined and individual effects of
OOP payments and insurance arrangements is
difficult to discern and also not always obvious to
the consumer of care. Finally, we used an adminis-
trative claims database. We could not control for
hospital-level characteristics including process of
care provided and detailed clinical variables.
Further, we lacked information on mortality after
discharge which might have biased our results.
The biases, however, are likely towards the null hy-
pothesis with less data and more noise in the
models suggesting our significant results are con-
servative. Given that this is a retrospective observa-
tional claims analysis, we caution the reader not to
draw hard and fast conclusions from the results;
however, this work provides opportunities to
generate hypotheses for future studies.
In contrast, our study has several strengths.
First, we examined readmissions after abdominal
surgery in a nonelderly primarily middle class
working population. Nonelderly patients account
for more than 60% of short-term hospital dis-
charges in the United States.6 Yet, their readmis-
sions are underexplored compared with previous
studies that focus on Medicare and Medicaid
patients. Second, to our knowledge, this is the first
study to assess the association between cost-sharing
and hospital readmissions in a middle-class popula-
tion. Our results are based on a national sample
and have important implications for equitable ac-
cess to care with potential consequences for health
care quality. Low readmission rates may not always
indicate good quality of care but difficult access to
essential medical care. Forgone medical care
928 Youn et al
attributable to financial barriers is inequitable and
should be avoided. Last, we separately examined
early operative readmissions within 7 days that
are more likely to be related to processes of care
during the index admission from 30-day readmis-
sions that are relevant to the current CMS policy.
We also excluded planned readmissions according
to the CMS algorithm.15 In our sample, 6.13% of
30-day readmissions were planned.
Our results may assist hospitals in identifying
patients at high risk of readmissions and devel-
oping better transition of care services. Policy
makers within public and private organizations
also may consider the timing of readmissions
when evaluating postsurgical rates as future
quality measures. Differential weights can be
ascribed to early readmissions within a week
and later readmissions after 3–4 weeks.5 Finally,
insurance plans may arrange their care structure
to reduce postsurgical complications and read-
missions. Case managers may shepherd patients
after discharge and individualize outpatient post-
operative services including home IV antibiotics,
medical equipment and therapies in addition to
coordination of community services such as
home nursing visits.
Patient cost-sharing and insurance arrange-
ments are significant predictors of both 7-day
and 30-day hospital readmission after abdominal
surgery among a privately insured, nonelderly
population. Patient’s financial concerns and insur-
ance plan structures should be considered when
addressing readmissions as this may have implica-
tions for equitable access and quality of care
during the postoperative period.
We thank James A. Rothendler, MD (Boston Univer-
sity School of Public Health) for his insightful comments
on the manuscript, and Omid Ameli, MD, PhD, MPH
(Boston University School of Public Health), for assist-
ing with data linkage.
SUPPLEMENTARY DATA
Supplementary data related to this article can be found
online at http://dx.doi.org/10.1016/j.surg.2015.09.007.
REFERENCES
1. Centers for Medicare & Medicaid Services. CMS to Improve
Quality of Care during Hospital Inpatient Stays [Internet].
2014 [cited 2015 Apr 6]. Available from: http://www.cms.
gov/Newsroom/MediaReleaseDatabase/Fact-sheets/2014-
Fact-sheets-items/2014-08-04-2.html.
2. American Hospital Association. Examining the drivers of
readmissions and reducing unnecessary readmissions for
better patient care. TrendWatch Sept. 2011.
Surgery
March 2016
3. Weber SM, Greenberg CC. Medicare Hospital Readmission
Reduction Program: what is the effect on surgery? Surgery
2014;156:1066-8.
4. Kassin MT, Owen RM, Perez SD, Leeds I, Cox JC,
Schnier K, et al. Risk factors for 30-day hospital readmis-
sion among general surgery patients. J Am Coll Surg
2012;215:322-30.
5. Girotti ME, Shih T, Dimick JB. Health policy update:
rethinking hospital readmission as a surgical quality mea-
sure. JAMA Surg 2014;149:757-8.
6. Centers for Disease Control and Prevention. Hospital Utili-
zation (in non-Federal short-stay hospitals) [Internet].
[cited 2015 Apr 6]. Available from: http://www.cdc.gov/
nchs/fastats/hospital.htm.
7. Jencks SF, Williams MV, Coleman EA. Rehospitalizations
among patients in the Medicare fee-for-service program.
N Engl J Med 2009;360:1418-28.
8. Epstein AM, Jha AK, Orav EJ. The relationship between hos-
pital admission rates and rehospitalizations. N Engl J Med
2011;365:2287-95.
9. Joynt KE, Orav EJ, Jha AK. Thirty-day readmission rates for
Medicare beneficiaries by race and site of care. JAMA 2011;
305:675-81.
10. Tsai TC, Joynt KE, Orav EJ, Gawande AA, Jha AK. Variation
in surgical-readmission rates and quality of hospital care.
N Engl J Med 2013;369:1134-42.
11. Goodman DC. The Revolving Door: A Report on US Hospi-
tal Readmissions. Robert Wood Johnson Foundation; 2013.
Available from: http://www.rwjf.org/content/dam/farm/
reports/reports/2013/rwjf404178.
12. Brooke BS, Stone DH, Cronenwett JL, Nolan B,
DeMartino RR, MacKenzie TA, et al. Early primary care pro-
vider follow-up and readmission after high-risk surgery.
JAMA Surg 2014;149:821-8.
13. Lemieux J, Sennett C, Wang R, Mulligan T, Bumbaugh J.
Hospital readmission rates in Medicare Advantage plans.
Am J Manag Care 2012;18:96-104.
14. Hansen LG, Chang S. White paper: health research data for
the real world. The Thomson Reuters MarketScan data-
bases. Baltimore: Thomson Reuters; 2011.
15. Yale New Haven Health Services Corporation/Center for
Outcomes Research & Evaluation. 2014 Measures Updates
and Specifications Report: Hospital-Wide All-Cause Un-
planned Readmission - Version 3.0. [Internet]. 2015 [cited
2015 Apr 6]. Available from: http://www.cms.gov/
Medicare/Quality-Initiatives-Patient-Assessment-Instruments/
HospitalQualityInits/Measure-Methodology.html.
16. Coey D. Physician Incentives and Treatment Choices in
Heart Attack Management. [Internet]. Stanford Institute
for Economic Policy Research; 2013. Report No: SIEPR Dis-
cussion Paper No. 12–027. Available from: ftp://dlib.info/
opt/ReDIF/RePEc/sip/12-027.pdf.
17. U.S. Bureau of Labor Statistics - Glossary [Internet]. [cited
2015 Sep 4]. Available from: http://www.bls.gov/bls/
glossary.htm.
18. Thomson Reuters MarketScanÒ Database. Available from:
https://marketscan.truvenhealth.com/marketscanportal/.
19. Hulka BS, Wheat JR. Patterns of utilization. The patient
perspective. Med Care 1985;23:438-60.
20. Centers for Medicare & Medicaid Services. Detailed Meth-
odology for the Total Per Capita Cost Measure for Medicare
Fee-for-Service Beneficiaries [Internet]. 2013. Available
from: https://www.cms.gov/Medicare/Medicare-Fee-for-
Service-Payment/PhysicianFeedbackProgram/Downloads/
Detailed_Methods_Total_Per_Capita_Costs_2-12-13.pdf.
Surgery
Volume 159, Number 3
21. Ash AS, Ellis RP. Risk-adjusted payment and performance
assessment for primary care. Med Care 2012;50:643-53.
22. Yaisawarng S, Burgess JF. Performance-based budgeting in
the public sector: an illustration from the VA health care
system. Health Econ 2006;15:295-310.
23. HealthCare.gov. Glossary: Out-of-pocket maximum/limit
[Internet]. [cited 2015 Apr 6]. Available from: https://
www.healthcare.gov/glossary/out-of-pocket-maximum-limit/.
24. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity
measures for use with administrative data. Med Care 1998;
36:8-27.
25. Efron B, Tibshirani RJ. An introduction to the bootstrap.
(softcover reprint of the original 1st ed. 1993 edition).
New York: Chapman and Hall/CRC; 1994:456.
26. Manning WG, Newhouse JP, Duan N, Keeler EB,
Leibowitz A, Marquis MS. Health insurance and the de-
mand for medical care: evidence from a randomized exper-
iment. Am Econ Rev 1987;77:251-77.
27. Allen LA, Smoyer Tomic KE, Smith DM, Wilson KL,
Agodoa I. Rates and predictors of 30-day readmission among
commercially insured and Medicaid-enrolled patients hospi-
talized with systolic heart failure. Circ Heart Fail 2012;5:672-9.
28. Lavernia CJ, Villa JM, Iacobelli DA. Readmission rates in the
state of Florida: a reflection of quality? Clin Orthop 2013;
471:3856-62.
29. Baker CL, Zou KH, Su J. Risk assessment of readmissions
following an initial COPD-related hospitalization. Int J
Chron Obstruct Pulmon Dis 2013;8:551-9.
30. Encinosa WE, Hellinger FJ. The impact of medical errors
on ninety-day costs and outcomes: an examination of surgi-
cal patients. Health Serv Res 2008;43:2067-85.
31. Hoyer EH, Needham DM, Miller J, Deutschendorf A,
Friedman M, Brotman DJ. Functional status impairment is
associated with unplanned readmissions. Arch Phys Med
Rehabil 2013;94:1951-8.
32. Siu AL, Sonnenberg FA, Manning WG, Goldberg GA,
Bloomfield ES, Newhouse JP, et al. Inappropriate use of
hospitals in a randomized trial of health insurance plans.
N Engl J Med 1986;315:1259-66.
33. Lohr KN, Brook RH, Kamberg CJ, Goldberg GA,
Leibowitz A, Keesey J, et al. Use of medical care in the
Youn et al 929
Rand Health Insurance Experiment. Diagnosis- and
service-specific analyses in a randomized controlled trial.
Med Care 1986;24(9 Suppl):S1-87.
34. Manning WG, Leibowitz A, Goldberg GA, Rogers WH,
Newhouse JP. A controlled trial of the effect of a prepaid
group practice on use of services. N Engl J Med 1984;310:
1505-10.
35. Robinson JC. Decline in hospital utilization and cost infla-
tion under managed care in California. JAMA 1996;276:
1060-4.
36. Greenfield S, Nelson EC, Zubkoff M, Manning W, Rogers W,
Kravitz RL, et al. Variations in resource utilization among
medical specialties and systems of care. Results from the
medical outcomes study. JAMA 1992;267:1624-30.
37. Luft HS. How do health-maintenance organizations achieve
their ‘‘savings’’? N Engl J Med 1978;298:1336-43.
38. Waters HR, Anderson GF, Mays J. Measuring financial pro-
tection in health in the United States. Health Policy 2004;
69:339-49.
39. Schoen C, Doty MM, Robertson RH, Collins SR. Affordable
Care Act reforms could reduce the number of underin-
sured US adults by 70 percent. Health Aff (Millwood)
2011;30:1762-71.
40. Magge H, Cabral HJ, Kazis LE, Sommers BD. Prevalence
and predictors of underinsurance among low-income
adults. J Gen Intern Med 2013;28:1136-42.
41. Fronstin P, Collins SR. Early experience with high-
deductible and consumer-driven health plans: findings
from the EBRI/Commonwealth Fund Consumerism in
Health Care Survey. EBRI Issue Brief [Internet]. 2005 [cited
2015 Sep 4];(288). Available from: http://papers.ssrn.com/
sol3/papers.cfm?abstract_id=870483.
42. Cohen A, Siegelman P. Testing for adverse selection in in-
surance markets. J Risk Insur 2010;77:39-84.
43. The Henry J. Kaiser Foundation. Assessing Americans’ Fa-
miliarity With Health Insurance Terms and Concepts
[Internet]. [cited 2015 Apr 6]. Available from: http://kff.
org/health-reform/poll-finding/assessing-americans-famili
arity-with-health-insurance-terms-and-concepts/.
44. Swartz K. Cost-sharing: effects on spending and outcomes.
Synth Proj Res Synth Rep 2010;20:42-5.