Since its introduction to the United States in 1988, laparoscopic cholecystectomy (LC) has replaced open cholecystectomy (OC) as the standard operation to treat symptomatic cholelithiasis. However, its popularity is not without consequence. The rate of major injuries to the common bile duct or biliary tract during LC has been reported to be between 0.11% and 1.4%. [1, 6, 10, 16]. By comparison, bile duct injury (BDI) after OC is reported to be between 0.1% and 0.3% [10, 12, 16]. Although BDI remains a relatively rare event, it has been associated with significant morbidity and mortality worldwide [1, 3, 17]. Health care costs in the United States for the management of major BDI are estimated at >$50,000 per patient, with an average inpatient hospital stay of 32 days [13]. Although clinical outcomes are generally expected to be good [17], the quality of life after BDI may be poor [3]. As such, ≤61% of BDI results in litigation, at a cost of $40–$100 million per year [9]. Major biliary injury is now one of the most common litigated surgical complications [4].

In this current investigation, our specific aim was to report the 10-year trend and associated mortality rate for BDI requiring operative repair in the United States during the years 1990 to 2000 as represented in the largest all-payer administrative database currently available in this country. From these data, we then attempted to identify the major factors associated with death in patients undergoing biliary reconstruction for the management of BDI after LC.

Materials and methods

Study population

For this retrospective cohort, the study population is drawn from the Nationwide Inpatient Sample (NIS) compiled by the Healthcare Cost and Utilization Project (HCUP, Agency for Healthcare Research and Quality, Rockville, MD, USA) for the years 1990–2000. This sample represents the all-payer inpatient experience of a 20% stratified probability sample of American nonmilitary, nonfederal hospitals for each year under consideration. The NIS is an administrative database. All patients undergoing LC at a given facility, whether as inpatients, transfers, “outpatient/same-day” or “23-h admissions,” are entered into the hospital database as inpatients for administrative purposes if they are to undergo operation or if they have been transferred from another facility. Each individual discharge abstract for this population of patients is statistically weighted to provide a national representation of diagnoses and procedure volume.

Data extraction

Because other investigators have questioned the accuracy of ICD-9 coding for injury occurring as a consequence of LC [16], we chose not to use this approach in our analysis. Instead, we identified patients with BDI by searching for biliary reconstruction codes and then excluding all patients who underwent reconstruction for biliary tract malignancy or for inflammatory or neoplastic pancreatic disease. Using the appropriate procedure codes as defined by the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM), in patients undergoing LC (51.23) or partial LC (51.24) were identified for each year under review and their data were extracted from the overall data set. Data for patients undergoing a reconstruction procedure (Appendix) were also extracted as an independent sample. From these two samples, the data for individuals who had undergone both LC and a biliary reconstruction procedure were then extracted for further analysis. Excluded reconstructions associated with biliary diagnoses or procedure codes unlikely to be a result of BDI are listed in the Appendix.

No specific ICD-9 code existed for LC in 1990; therefore, and the total volume is reported as OC for this year. A specific ICD-9 code for LC was introduced in October 1991. The LC volume for the first three quarters of 1991 was estimated as a constant proportion (as derived from the last-quarter volume) of all cholecystectomies performed. After these cases were used (n = 203,241) to project LC volume for 1991, they were excluded from further analysis because no biliary reconstruction procedures could be associated with LC without the specific ICD-9 codes for this procedure.

Data analysis

Data extraction for the primary (cholecystectomy) and secondary (reconstruction) procedures, as well as the calculation of national averages and statistical analysis, was performed using the Statistical Package for the Social Sciences (SPSS, Chicago, IL, USA). For univariate analysis, continuous variables were compared using the Mann-Whitney U test to compare medians, given the nonparametric distribution of the data. Categorical variables were compared using the chi-square test to compare proportions. Variables found to be significant (p < 0.05) under univariate analysis were subjected to a multivariate stepwise regression model to examine the relationship between these variables and the outcome of interest (death) for patients who had undergone a reconstruction procedure. Specific patient, institutional, and outcome variables were chosen based on our determination of their importance in comparing patients with and without BDI after LC.

Results

For the study period, a total of 73,427,129 weighted NIS inpatient abstracts were surveyed to derive the study cohort. After excluding the projected LC volume for the first 9 months of 1991, a total of 2,841,186 LCs were available for examination of subsequent biliary reconstruction procedures.

Univariate analysis comparing those patients who underwent biliary reconstruction after LC to those who did not is shown in Table 1. The two groups differed significantly in all variables examined except for ethnic group (p = 0.083). The percentage of LC has increased over the period 1991 to 2000 (Table 2). In 1992, 324,794 cholecystectomies were performed laparoscopically, representing 64.06% of all cholecystectomies performed during that year. For 2000, 324,783 LC were performed, representing 75.3% of all cholecystectomies. Over the study period, LC was associated with a low mortality rate (mean, 0.45%; range, 0.33–0.58%).

Table 1 Comparison of patient, institutional, and outcome characteristics between patients who had an uncomplicated LC and those who sustained BDI necessitating operative repair after LC, 1991–2000
Full size table

A total of 4,497 cases of biliary reconstruction after LC were reported between 1991 and 2000. Within this group of patients, the average rate of BDI requiring reconstruction was 0.15% for the years under study. Biliary reconstructions ranged from 0.25% in 1992 (2.5 per 1,000 LC) to 0.09% in 1999 (0.9 per 1,000 LC) (Fig. 1). For 2000, the most recent year for which data are available, biliary reconstruction was performed in 0.10% of all patients who underwent LC (1 per 1,000 LC). The average mortality rate for patients undergoing biliary reconstruction for the years 1991–2000 was 4.5%. The mortality rate was significantly different in this group as compared to those undergoing uncomplicated LC (p < 0.001) during the same time period (Table 1). There is considerable variation in the mortality rate over the years under review, but it appears to have stabilized after 1998 (Fig. 2).

Figure 1
figure 1

Number of annual reconstructions and percentage of total annual laparoscopic cholecystectomy requiring reconstruction 1991–2000.

Full size image
Figure 2
figure 2

Annual mortality rate in patients undergoing a biliary reconstruction procedure after laparoscopic cholecystectomy, 1991–2000.

Full size image

Univariate analysis of patients undergoing biliary reconstruction showed no significant difference between patients who died after this procedure and those who survived in terms of sex, income level, or the location and teaching status of the hospital where they received treatment (Table 3). There was a significant difference between these two groups of patients in comparing age, admission type and source, principal diagnosis, ethnic group and primary payer. In a multivariate logistic regression model (Table 4), age, African American ethnicity, admission type and source, and hospital location and teaching status were all found to be significantly associated with death after biliary reconstruction (p < 0.025).

Table 2 NIS-derived national volumes for laparoscopic cholecystomy and associated mortality for 1990–2000
Full size table
Table 3 Decedent analysis among 4,061 patients undergoing biliary reconstruction after laparoscopic cholecystectomty
Full size table
Table 4 Multivariate analysis of in-hospital mortality following biliary reconstruction after laparoscopic cholecystectomy
Full size table

Discussion

In this population-based study of >2.8 million LCs performed over a 10-year period, we found that the percentage of cholecystectomies performed laparoscopically has increased between 1991 and 2000, while the volume of all cholecystectomies has decreased due to a decrease in the proportion of open cases. The incidence of BDI, as measured by biliary reconstruction procedures after LC, has decreased by >50% over the same period (2.5 cases per 1,000 in 1992 to 1 case per 1,000 in 2000). Our data are in agreement with the lowest rates, but they are not consistent with the highest rates for BDI reported over the last 7 years [1, 6, 10, 16]. The most likely explanation for this finding is that minor BDI not requiring biliary reconstruction were excluded from this study, but they constitute a percentage of the injuries included in most other studies of this subject.

Further, our data seem to support the contention that the increase in BDI, after the introduction of LC, was a consequence of inexperience with the new laparoscopic technique [14]. This may be indicative of the socioeconomic pressures that caused this new technology to come into widespread use before mechanisms to ensure adequate training in LC were established. The emphasis on teaching this procedure within general surgery residency training programs has increased experience with LC, and the use of intraoperative methods to define biliary anatomy [2, 6] may have increased the safety of LC. These and other factors have contributed to a decrease in the incidence of BDI to these levels seen with OC [10, 12, 16].

The average operative mortality rate for patients undergoing of biliary reconstruction after BDI associated with LC has been reported to be 2.8% on follow-up [7]. Our finding of a population-based, procedure-related mortality rate of 4.5% is alarming. In comparison to individuals in whom LC is uncomplicated, patients undergoing biliary reconstruction for BDI tend to be older, male, to have been admitted urgently to urban, nonteaching hospitals, and to have a larger number of associated diagnoses (Table 1). Many of the characteristics examined in our comparison of patients who incurred BDI and those who did not incur BDI are statistically different, yet the small percentile differences in certain categories raise questions about the clinical relevance of these findings. This finding is probably a consequence of a large sample size uncovering a significant difference between groups even when the percentage differences appear minimal. Only three-quarters of patients suffering BDI are discharged to home, as compared to 94% of patients in whom LC is uncomplicated. Boerma et al. have reported that BDI patients have a poor quality of life up to 5 years after biliary reconstruction [3]. In addition, Flum and Cheadle, in examining Medicare claims data, found a 26% mortality rate within the 1st year after BDI in an older population of patients than is reflected by the NIS database [7]. Indeed, it is not surprising that patients undergoing biliary reconstruction were older (Table 1). It would appear that elderly patients are at particular risk of dying after BDI requiring operative intervention. To lessen the risk in this group, particular attention should be paid to preoperative preparation and intraoperative monitoring.

It is surprising to find that most of the patients in our cohort who incurred BDI requiring reconstruction were admitted from the clinic or office setting and not from the emergency department. Our data do indicate that >50% of these admissions were emergent admissions (Table 1). Although the NIS data set does not specify the reason for emergency admission, most (both “BDI” and “No BDI”) carried the diagnosis of “Cholelithiasis with cholecystitis,” suggesting that the admission was based on symptoms or signs of gallbladder disease.

In our decedent analysis among those undergoing biliary reconstruction, there was no significant difference in sex, income level, and hospital location or teaching status between those who survived and those who died. The most significant finding in this analysis was that ethnic group had a dramatic influence on death after BDI. African Americans had a mortality rate >10%, which is more than twice the mortality rate for whites based on our decedent analysis and over a sevenfold increase in the risk of dying after reconstruction for BDI as compared to whites based on multivariate analysis. There is evidence from other fields, particularly oncology [8], that the mortality rate among African Americans with cancer is significantly higher than that for white individuals. Our findings indicate that procedure-related death should become a focus of future investigation as it relates to ethnic variations in outcome.

There are some important limitations to our study that may have an impact on the findings. First, because the NIS database does not include inpatient information derived from US. Military or Veterans Affairs hospitals, LC cases from these institutions are not represented in this study. Ferreira et al. reported a volume of 7,492 LC performed within the Veterans Affairs health care system between 1992 and 1995 [5]. Although these authors did not report the incidence of BDI within their study population, one could speculate, based on an average BDI rate of 1.76 per 1,000 LC as derived from our data for 1992–95, that this would add only 13 additional cases to the volume of BDI during this period. Nonetheless, the contribution of these centers to the overall national LC and BDI experience remain largely an unknown variable. Likewise, because of LC data are available for the last quarter of 1991 only, prediction of the true incidence of in-hospital BDI for this year is impossible. The biliary reconstruction rate for 1991 (Fig. 1) represents only the procedures performed during the last quarter of that year; thus, it does not include BDI requiring reconstruction during the previous 9 months.

Second, we have designed our investigation around the occurrence of BDI requiring operative intervention for repair and have not addressed the occurrence of BDI that is managed successfully without the need for open operation. We have chosen to exclude minor BDI cases (principally, cystic duct leak treated by stenting of the common bile duct or stricture amenable to balloon dilatation) because these injuries generally have a lesser impact on the health of the patient and are usually managed nonoperatively. In addition, to avoid overprediction of BDI, we specifically excluded ICD-9-CM code 576.2 from consideration (Appendix) because this code might represent a primary biliary stricture and not a secondary stricture as a result of BDI. After analyzing our data, we found no significant change in injury or mortality rates due to exclusion of this code.

Third, it is impossible to control for potential ICD-9 coding errors at the hospital level that may have had an impact on our ability to accurately extract patient abstracts. By using an approach that looks at biliary reconstruction coding instead of injury coding, we believe that we have minimized the potential impact of erroneous coding on our results.

Finally, it has become increasingly clear that the best postoperative results for reconstruction after BDI are achieved in the hands of specialized surgeons. Robinson and Stewart, in two separate reports, reinforced this assumption by demonstrating a superior outcome for these patients after referral to a major medical center with extensive hepatobiliary experience [11, 15]. Consequently, it has become common practice to transfer patients to these (usually urban) teaching facilities for definitive surgery. In our current sample, we attempted to control for this double hospitalization by extracting the data for both LC and reconstruction patients as separate samples from the NIS data set. With this approach, patients transferred from another hospital to the center where the bile duct reconstruction was performed would be captured on the basis of diagnosis or procedure (LC) codes on admission, which would be recorded in the admission source variables within the NIS database. We found no significant difference in the mortality rate for patients undergoing reconstruction based on hospital location or teaching status (Table 3) in the univariate analysis; moreover, there was only a slightly increased contribution of this variable to the mortality rate in the multivariate analysis (Table 4). However, we did find that these patients had a significantly higher risk of dying after biliary reconstruction if their admission source was another hospital or health facility (OR 3.712; 95% CI, 1.78–10.781 and OR 2.8; 95% CI, 1.224–6.221, respectively) than patients admitted from the emergency department of the treating facility (OR 0.30; 95% CI, 0.197–0.474). Because the NIS data set does not directly control for case complexity or critical illness at the time of admission, it is difficult to prove a direct relationship between these variables and death. However, the low risk of dying in either a rural hospital (OR 0.45; 95% CI, 0.239–0.855) or urban nonteaching hospital (OR 0.60; 95% CI, 0.399–0.890) does suggest that patients at higher risk for dying were transferred from these facilities to urban teaching hospitals and were sicker on admission than patients admitted from the accepting facility itself.

It is still a common belief that the incidence of BDI requiring reconstruction continues to be higher after LC than after OC. We hope that this belief will now be modified in light of our findings from this large population-based national data set.