Flum DR, Cheadle A, Prela C, Dellinger EP, Chan L. Bile Duct Injury During Cholecystectomy and Survival in Medicare Beneficiaries. JAMA. 2003;290(16):2168–2173. doi:10.1001/jama.290.16.2168
Author Affiliations: Departments of Surgery (Drs Flum and Dellinger), Health Services (Drs Flum and Cheadle), and Rehabilitation Medicine (Dr Chan), University of Washington, and the Division of Clinical Standards and Quality (Drs Chan and Prela), Centers for Medicare & Medicaid Services (CMS, formerly the Health Care Financing Administration), Region 10, Seattle, Washington. Dr Flum is formerly from the Robert Wood Johnson Clinical Scholars Program at the University of Washington.
Context Common bile duct (CBD) injury during cholecystectomy is a significant
source of patient morbidity, but its impact on survival is unclear.
Objective To demonstrate the relation between CBD injury and survival and to identify
the factors associated with improved survival among Medicare beneficiaries.
Design, Setting, and Patients Retrospective study using Medicare National Claims History Part B data
(January 1, 1992, through December 31, 1999) linked to death records and to
the American Medical Association's (AMA's) Physician Masterfile. Records with
a procedure code for cholecystectomy were reviewed and those with an additional
procedure code for repair of the CBD within 365 days were defined as having
a CBD injury.
Main Outcome Measure Survival after cholecystectomy, controlling for patient (sex, age, comorbidity
index, disease severity) and surgeon (procedure year, case order, surgeon
Results Of the 1 570 361 patients identified as having had a cholecystectomy
(62.9% women), 7911 patients (0.5%) had CBD injuries. The entire population
had a mean (SD) age of 71.4 (10.2) years. Thirty-three percent of all patients
died within the 9.2-year follow-up period (median survival, 5.6 years; interquartile
range, 3.2-7.4 years), with 55.2% of patients without and 19.5% with a CBD
injury remained alive. The adjusted hazard ratio (HR) for death during the
follow-up period was significantly higher (2.79; 95% confidence interval [CI];
2.71-2.88) for patients with a CBD injury than those without CBD injury. The
hazard significantly increased with advancing age and comorbidities and decreased
with the experience of the repairing surgeon. The adjusted hazard of death
during the follow-up period was 11% greater (HR, 1.11; 95% CI, 1.02-1.20)
if the repairing surgeon was the same as the injuring surgeon.
Conclusions The association between CBD injury during cholecystectomy and survival
among Medicare beneficiaries is stronger than suggested by previous reports.
Referring patients with CBD injuries to surgeons or institutions with greater
experience in CBD repair may represent a system-level opportunity to improve
Cholecystectomy is the most commonly performed elective abdominal surgical
procedure (∼750 000 each year) in the United States.1 Common
bile duct (CBD) injury occurs in 1 of 200 cholecystectomies, is a significant
source of patient morbidity after gallbladder surgery, and is the leading
source for medical malpractice claims against general surgeons.1 Repair
of a damaged CBD is a technically challenging undertaking that may best be
performed by experienced hepatobiliary surgeons. Published results of CBD
repair by such surgeons consistently demonstrate a low risk of operative mortality
and high rates of short-term and intermediate-term success without the need
for reoperation.2 Reviews of cases that result
in litigation, however, indicate a different clinical course for patients
with CBD injury.3 These patients have a high
rate of required reoperation, perioperative mortality, and significant deficits
in long-term quality of life and functional status. Determining the community-level
burden of CBD injury is challenging because both types of reports are susceptible
to bias. Reports by experienced surgeons may involve publication bias and
reports of cases that progress to litigation may involve selection bias.
The purpose of this study was to describe the impact of CBD injury among
Medicare beneficiaries who had undergone cholecystectomy. Specifically, we
sought to determine the population-level risk of death after CBD injury, to
describe the impact of patient and physician factors on survival, and to identify
any system-level interventions that might help improve outcome after CBD injury.
We used a retrospective cohort design, using Medicare National Claims
History (NCH) Part B data from January 1, 1992, through December 31, 1999,
that contained Current Procedural Terminology (CPT)
codes pertaining to either a cholecystectomy, a CBD repair, or biliary disease.
The method for determining CBD injury has been previously published.1 This study was exempted from University of Washington
Human Subjects Review.
The Medicare NCH Part B database, maintained by the Center for Medicare
& Medicaid Services (CMS), contains all the payment claims for the professional
component of services delivered to Medicare beneficiaries in either the inpatient
or outpatient setting. Each claim for services delivered to a beneficiary
can have 1 or more discrete billable service listed, along with information
about each service that includes the coded identifier of the physician providing
the care, the type of service performed, and the associated International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes. The Medicare Health Insurance Claim (HIC)
number identifies the patient for whom the services were delivered. Medicare's
Unique Physician Identification Number (UPIN) was used to match Part B data
to the American Medical Association (AMA) Physician Masterfile in a previously
described process.4 The AMA Physician Masterfile
includes data on physician specialty, board certification status during the
year of the procedure, surgeon age, race, sex, and year of graduation from
medical school. For this analysis, physicians whom the AMA database described
as general surgeons or surgical subspecialists were considered surgical specialists.
Two percent of Medicare providers could not be matched to the AMA data.
The date of death was obtained from the Enrollment Database, also maintained
by CMS, by matching on the beneficiary HIC number. The date of death in the
Enrollment Database is derived from the Social Security Administration's database,
Master Beneficiary Record. Of all records, 2.5% did not have complete Enrollment
Database data. Complete-case-only analysis was performed.
Patients were defined as likely having CBD injuries if they had a cholecystectomy
with intraoperative cholangiogram (CPT codes: 49311, 56341, 47563, 47605,
74300, and/or 74301) or without intraoperative cholangiogram (CPT codes: 49310,
56340, 56342, 47562, 47600, 47610, 47612, or 47620 without codes for intraoperative
cholangiogram) followed by reoperative repair of the CBD (CPT codes: 47701,
47720, 47721, 47740, 47760, 47765, or 47780) within the subsequent 365 days.
We excluded patients with diagnoses of hepatobiliary malignancies or choledochal
cysts (ICD-9 diagnostic codes: 155.1, 156, 156.1,
The Deyo modification5 of the Charlson
comorbidity index (0-3, with 3 indicating greatest comorbidity) was calculated
for each patient based on ICD-9 diagnostic codes
from all index records and all prior records that contained a diagnostic code
for biliary disease. Records with associated diagnosis codes of pancreatitis,
CBD stone, cholangitis, sepsis, and acute cholecystitis or procedure codes
including CBD exploration were considered complex.
The number of repair procedures (among patients covered by Medicare
after 1992) performed by each repairing surgeon before a given operation was
defined as repair case order and was considered a
marker of surgical experience at the time of the procedure. Repair case order
was also considered as a continuous variable.
The relationship of CBD injury and survival was assessed using multivariable
Cox proportional hazard regression analyses. The comparator group for the
relative hazard ratio (HR) of mortality after CBD injury included all patients
without CBD injury. Reported adjusted hazards for individual variables were
in comparison with the lower value of that specific variable. The median follow-up
for the entire cohort was 5.6 years, with 25% of patients followed up for
7.4 years or more. The longest follow-up was 10.6 years, and the longest common
(for both injured and noninjured patients) follow-up point was 9.2 years.
Survival was measured as the time from cholecystectomy until death or August
15, 2002. The proportional hazards assumption was confirmed by inspection
of Schoenfeld residuals and log-log plotting.
The cumulative incidence of death was estimated for the entire cohort
and then tested between groups of interest (those with and without injury)
and graphed as Kaplan-Meier plots. Unadjusted analyses were compared using
the log-rank test. Cox regression was used to estimate simultaneously the
effect of potential confounders on the association between patient and surgeon
characteristics and the hazard of death and to provide estimates of adjusted
survival at discrete time points. The analysis was repeated on the cohort
that was younger than 65 years to assess the impact of CBD injury on survival
in this younger subset of the population. Cox proportional hazards were estimated
using SAS (Version 8, SAS Institute, Cary, NC) and STATA (Version 7, STATA
Corp, College Station, Tex). A P value of <.01
was considered statistically significant.
A total of 1 570 361 patients—62.9% of whom were women
and whose mean (SD) age was 71.4 (10.2) years—had a cholecystectomy.
Of those 7911 (0.5%) were identified as having CBD injuries. Demographic characteristics
of patients with and without CBD injury have been previously decribed1 (Table 1).
Injured patients were slightly older, were more likely men, had a higher comorbidity
index, and were more likely to have had a cholecystectomy that was considered
complex. One third of all patients died within the follow-up period (median
survival 5.6 years, interquartile range, 3.2-7.4 years). By the last common
follow-up point (9.2 years), 55.2% of patients without CBD injury were alive
vs only 19.5% of patients who had a CBD injury. Most of the impact of CBD
injury appeared in the first 2 years (Figure
1) after the cholecystectomy. Within the first year after cholecystectomy,
the mortality rate (adjusted for age, sex, and comorbid illness) was 6.6%
in patients without CBD injury and 26.1% in those with injury. After approximately
3 years, the rates of death appeared to equalize. The unadjusted HR for death
was 4.09 times higher (95% confidence interval [CI], 4.06-4.11) and the adjusted
HR for death 2.8 times higher (HR 2.79; 95% CI, 2.71-2.88) for patients with
a CBD injury vs those without injury. The HR significantly increased with
advancing age, case complexity, and comorbidity index (Table 2). The impact of CBD injury on survival was also identified
in younger patients (Figure 1).
Of the 178 381 patients in the younger cohort (mean [SD] age, 54.8 [10.9]
years), the HR of death after CBD injury (adjusting for sex, age, and comorbidity
index) was 2.7 times higher than among those without CBD injury (95% CI, 2.4-3.0),
even though the absolute rate of death was lower in this younger cohort.
There were 1 458 821 patients without and 7719 patients with
CBD injury who had complete records available for survival analysis. Overall,
91.4% of these patients underwent only 1 operative repair and the odds of
a patient having multiple repair operations decreased by 30% with each successive
year compared with the year before it (odds ratio [OR], 0.70; 95% CI, 0.65-0.82).
Only 500 surgeons performed more than 10 repairs. The repairing surgeon was
the same as the injuring in approximately 75% of repairs. After adjusting
for other patient and surgeon variables, the hazard of death was 11% greater
when the repairing surgeon was the same as the surgeon who performed the cholecystectomy
(HR, 1.11; 95% CI, 1.02-1.20). Correspondingly, the level of experience of
the repairing surgeon was linked to survival. The median number of repairs
performed by repairing surgeons was 5 (interquartile range, 2-9) and increasing
surgical experience was associated with an increased likelihood of survival.
The adjusted hazard for death decreased 11% for every increase of 1 case order
in the repairing surgeon's level of experience (HR, 0.89; 95% CI, 0.82-0.98).
When considered at each point of the experience curve of the repairing surgeon,
with each repair the adjusted hazard of survival to at least 1 year increased
by approximately 2% (HR, 0.98; 95% CI, 0.97-1.00) and based on logistic regression
the predicted probability of death within the first year after repair was
derived for each level of experience.
Beyond experience, the characteristics of the repairing surgeon had
minimal effect on survival after repair. When controlling for all covariates
(Table 3), the features linked
to an increase in the hazard for death were patient age, comorbidity index,
and the same surgeon performing both the cholecystectomy and the repair. After
controlling for patient factors, those who had more than 1 repair operation
were not at significantly higher risk of death (HR, 1.05; 95% CI, 0.97-1.14).
Survival after cholecystectomy-related CBD injury was significantly
reduced in a nationwide cohort of nearly 1.6 million Medicare beneficiaries
undergoing cholecystectomy. Although this cohort represented approximately
one third of all patients undergoing cholecystectomy nationwide in any given
year, the age of this cohort was considerably older than the age of all patients
undergoing cholecystectomy and included younger patients with disability,
renal disease, or both. These patients were nearly 3 times more likely to
die within 10 years after cholecystectomy than noninjured patients, even after
controlling for age and comorbid illnesses. We also found that improved survival
was more likely when a different, more experienced surgeon performed the CBD
repair, but that 75% of repair procedures were in fact performed by the same
surgeon associated with the injury. Perhaps the most striking finding of this
study was the dramatic difference we found between published estimates of
mortality after CBD injury and the population-based outcomes. This may help
explain the apparent discrepancy between reported outcomes and those described
in cases of CBD injury that progress to litigation.
Since an acknowledged influence on the outcome of CBD repair is the
expertise of the repairing surgeon6 and the
number of attempted repairs before the definitive repair, it is possible that
surgeons with the best results are likely to review and publish their work.
In fact, a review of the outcomes of 40 published case series2 evaluating
561 (0.5) of 114 000 patients with CBD injuries indicated that repair
of the CBD resulted in failures in only 4.9% with no reported deaths. In the
largest case series to evaluate specifically the outcomes after repair of
the CBD, the mortality rate was 0% and the operative success 95%,7 with a very low rate of reoperation. Reports of repairs
of CBD strictures in the open era detailed a rigorous but quite successful
procedure with long-term satisfactory results in nearly 90% of those undergoing
repair.8- 13 Reported
short-term14 and long-term outcomes7,15- 28 in
the laparoscopic era are also quite good (Table 4) with 90% to 95% success and a low rate of reported mortality
(0%-14.2%). Of these case series, the average length of follow-up was 41.3
months and 17 (2.8%) of 602 patients with CBD injury died. It is difficult
to reconcile these excellent outcomes with the significant impairments in
quality of life identified after CBD.29,30
Several variables may be associated with adverse outcome following CBD
repair, including the degree of injury severity,31 associated
vascular injury,26 and comorbid conditions.
One of the most modifiable components of adverse outcome may relate to the
experience of the surgeon. In a study of 46 cases of CBD injury that progressed
to litigation, the primary surgeon's repair was successful only 27% of the
time vs a 79% success rate among surgeons at referral centers.32 In
that study only 47.8% of these cases were referred to the specialty center.
Stewart and Way33 reported that the rate of
successful primary repair to be only 17% when performed by the injuring surgeon
who is presumably less experienced at CBD repair.
A more recent evaluation of management trends, however, found that more
than 50% of procedures were performed by the primary surgeon27 and
our study further identified a 75% frequency of repair by the initial surgeon.
A report from the Connecticut laparoscopic surgery registry demonstrated that
a better rate of success was acheived among patients undergoing repairs at
the originating hospital than what has been previously noted.34 However,
recent surveys of medical malpractice cases involving CBD injuries reinforce
the importance of surgeon experience with repairs and note that on average
patients whose procedure was performed by less experienced surgeons had 2.2
repairs with much more varied outcomes.35
Population-based evaluations may be helpful in assessing a more generalizeable
rate of outcome after CBD injury than that identified in case series. Survival
after cholecystectomy in the population at large is good, with mortality rates
as low as 0.5% noted in a recent national survey.36 Mortality
after CBD injury, however, is harder to evaluate. In 2 recent analyses from
Switzerland, variable mortality rates of 0%37 and
9.4%38 were identified after CBD injury, with
little other than selection bias to explain this difference. Given that only
a fraction of patients are sent to referral centers and that the outcomes
of the repair are likely linked to the experience of the surgeon repairing
the injury, the community-based rate of mortality after CBD injury is likely
to be higher than previously reported.
The results of our study demonstrate the nationwide impact of CBD injury
on survival in Medicare beneficiaries and encompass a broad range of surgeon
types, experience level, and practice environments. In this way, we believe
these findings more closely reflect the experiences of the average patient
in the average community. Since this is, in fact, where CBD injuries occur,
we believe these estimates are more relevant when counseling patients. Furthermore,
evaluations in the community at large may better help identify system-level
opportunities for quality improvement.
As detailed in our previous work,1 there
were limitations in both the variables included in this database and the technique
used to select patients with CBD injury that may have affected the completeness
and accuracy of these estimates. Furthermore, in this analysis we describe
survival estimates after cholecystectomy and CBD injury in an older population
of patients, and it is unclear to what extent these findings may be generalizeable
to the community at large. The average age of patients undergoing cholecystectomy
in reported series of CBD injury (47 years)2 was
considerably lower than that identified in our study.
Although the absolute rate of death after CBD injury found in this study
may be limited to the population of patients older than 65 years, the relative
hazard of death within the follow-up period may have greater applicability
to the general community. Although most patients whose care is covered by
CMS are older than 65 years, those with chronic renal failure and those who
are medically disabled in all age groups are also included. To better evaluate
the relevance of our finding in a younger segment of the cohort, we performed
a subset analysis on the more than 175 000 patients younger than 65 years
(mean age 54.8 years) and found that the HR for death after CBD injury was
2.7 times higher in this younger population compared with the younger patients
without a CBD injury. This nearly identical rate suggests that the association
between CBD injury and survival is apparent across a broad selection of the
population undergoing cholecystectomy.
Unfortunately, cause of death was unavailable in this analysis. There
are other components of the data set that are limiting. For example, the case
order of the repair is a proxy for case order in that it is derived for each
surgeon based only on repairs performed in patients covered by Medicare after
1991. It therefore likely undercounts the experience of surgeons performing
repairs in patients not covered by Medicare or before 1992. Given that we
found a relationship between surgeon inexperience and outcome this may be
considered a conservative bias.
Furthermore, given that patients with CBD injury benefit from multimodality
care (eg, care provided by gastroenterologists, interventional radiologists,
diagnostic radiologists) and that surgeons with greater experience with CBD
repair probably also perform procedures at referral centers that specialize
in this multimodality care. On some level, surgeon experience may act as a
proxy for better hospital resources. Although in this analysis, we could not
distinguish the effect of hospital resources on patient outcome from surgeon
experience, we believe this potential confounding strengthens the recommendation
for patient referral to high-volume centers after CBD injury. Finally, this
is a case-only analysis, and records that did not have surgeon-matching information
were excluded (about 2% of all patients). The effect of these missing data
on the analysis is unclear, but there is no reason to suspect that missing
data were related to the exposure of interest.
In conclusion, this is the first study, to our knowledge, that demonstrates
the population-based impact of CBD injury on survival. We found that among
Medicare beneficiaries, after adjusting for all covariates, patients having
a major CBD injury were nearly 3 times more likely to die during the first
few years after cholecystectomy compared with patients without CBD injury.
Improved survival was noted when the repair was performed by a different surgeon
than the one who performed the cholecystectomy. Similarly, increased experience
of the repairing surgeon was linked to better outcomes. Based on this nationwide
analysis, we suggest that patients with CBD injury should be referred to a
surgeon and to an institution with greater experience in CBD repair. This
may represent a system-level opportunity to improve outcome after CBD injury