Race Is a Predictor of In-Hospital Mortality After Cholecystectomy, Especially in Those With Portal Hypertension

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2008;6: Race Is a Predictor of In-Hospital Mortality After Cholecystectomy, Especially in Those With Portal Hypertension GEOFFREY C. NGUYEN,*, THOMAS A. LAVEIST, DORRY L. SEGEV, and PAUL J. THULUVATH* *Division of Gastroenterology, Department of Medicine, and the Department of Surgery, Johns Hopkins University School of Medicine, Baltimore, Maryland; Mount Sinai Hospital Division of Gastroenterology, University of Toronto, Toronto, Ontario, Canada; and the Hopkins Center for Health Disparities Solutions, Bloomberg School of Public Health, Baltimore, Maryland Background & Aims: Cholecystectomy is the most frequently performed gastrointestinal surgery in the United States. In this study, we characterized racial disparities in in-hospital mortality after cholecystectomy among patients with and without decompensated cirrhosis. Methods: All patients who underwent cholecystectomy between 1998 and 2003 were queried from the Nationwide Inpatient Sample, the largest population-based and geographically representative all-payer database of hospital discharges in the United States. Crude mortality among races was determined for those with and without cirrhosis with portal hypertension and subsequently adjusted for demographic and clinical factors. Results: In-hospital mortality associated with cholecystectomy was higher in the portal hypertensive group compared with those without portal hypertension (10.8% vs 1.4%; P <.0001). African Americans had greater adjusted mortality risk than whites in both the nonportal hypertensive (odds ratio [OR], 1.48; 95% CI, ) and portal hypertensive (odds ratio [OR], 2.37; 95% CI, ) groups, although the mortality gap was more pronounced in the latter. For portal hypertensive patients, undergoing cholecystectomy at a liver transplant center was associated with dramatically lower mortality (OR, 0.41; 95% CI, ). Conclusions: In-patient mortality after cholecystectomy is 7.8-fold higher in patients with portal hypertension compared with those without portal hypertension. African Americans experienced higher mortality than whites after cholecystectomy, especially in the presence of portal hypertension. Cholecystectomy at a liver transplant center may offer survival benefit for patients with portal hypertension. Cholecystectomy is the most common gastrointestinal surgery in the United States, with more than 700,000 surgeries performed each year. 1 It is also among the most frequently performed procedures in patients with cirrhosis. The development of clinically apparent portal hypertension such as ascites, variceal hemorrhage, encephalopathy, or hepatorenal syndrome marks the transition from compensated to decompensated cirrhosis. 2 This subgroup of patients has up to an 11-fold higher risk of mortality after cholecystectomy compared with those without cirrhosis. 3 In 2002, the Institute of Medicine released a pivotal report describing racial disparities in quality of health care access to medical procedures that resulted in suboptimal outcomes, especially in African Americans. 4 A higher surgical mortality among African Americans has been well documented for cardiac, vascular, and gastrointestinal surgeries. 5 7 It is probable that some of these disparities arise from quality of in-hospital care. Because the optimization of perioperative and postoperative care is paramount for patients with sequelae of portal hypertension (PHTN) after abdominal surgery, we expect that racial disparities in postsurgical mortality would be even more pronounced in this population. Thus, our aims were to compare racial differences in in-hospital mortality after cholecystectomy in patients with and without PHTN. Patients and Methods Data Source All data were extracted from the Nationwide Inpatient Sample (NIS) between 1998 and It is the largest all-payer database of national hospital discharges, maintained as part of the Healthcare Cost and Utilization Project by the Agency for Healthcare Research and Quality. The NIS is a 20% stratified sample of nonfederal, acute-care hospitals in the United States and includes community and general hospitals and academic medical centers, but excludes long-term facilities. Each record in the NIS includes a unique identifier, demographic data (age, sex, and race), hospital transfer status, admission type (emergent, urgent, or elective), primary and secondary diagnoses (up to 15), primary and secondary procedures (up to 15), expected primary and secondary insurance payers, total hospital charges, length of stay, and hospital characteristics (region, urban vs rural location, bed-size, and teaching status). We used Clinical Modification of the International Classification of Diseases, 9th revision procedural codes to identify all patients in the NIS who had undergone either open (51.22) or laparoscopic (51.23) cholecystectomy between 1998 and Predictor and Outcome Variables Data on race and ethnicity were derived from hospital administrative data and categorized as follows: white, African American, Hispanic, Asian or Pacific Islander, or other. For Hispanic patients, ethnicity took precedence over race and was coded as Hispanic, whereas all other race categories were Abbreviations used in this paper: CI, confidence interval; NIS, Nationwide Inpatient Sample; OR, odds ratio; PHTN, portal hypertension by the AGA Institute /08/$34.00 doi: /j.cgh

2 October 2008 RACE AND IN HOSPITAL MORTALITY 1147 non-hispanic. International Classification of Diseases, 9th revision, Clinical Modification diagnostic codes were used to identify patients with PHTN as a result of cirrhosis with the following inclusion criteria: (1) cirrhosis (571.2, 571.5, 571.6) and (2) sequelae of PHTN (572.3) including ascites (789.5), hepatic encephalopathy (572.2), and variceal bleeding (456.0, 456.2). Case-mix adjustment was performed using the Deyo et al modification of the Charlson Index. 8,9 A liver transplant center was defined as a hospital or medical center that performed at least one liver transplant during the year of the index admission. In-hospital mortality was the primary outcome. Secondary outcomes included measures of morbidity including the requirement for postoperative parenteral nutrition; postoperative length of stay, defined as the time interval between cholecystectomy and either discharge or death; and nonroutine discharge, which was defined as discharge to another facility or requirement for home care. We also used International Classification of Diseases, 9th revision, Clinical Modification coding algorithms to identify postoperative complications, which then were categorized further as wound infections, urinary complications, gastrointestinal complications (including acute liver injury), pulmonary complications, cardiac complications, and intraoperative complications. Statistical Analysis Data were analyzed using the Stata 9.0 SE software package (Stata Corp LP, College Station, TX). All analyses were stratified by the presence of cirrhosis and PHTN. These analyses took into account the stratified 2-stage cluster design using Stata s SVY (survey data) commands incorporating individual discharge-level weights. Two-way chi-squared analyses were performed to compare categoric variables among different racial groups and unpaired t tests compared differences in means of continuous variables. For those with and without PHTN, crude mortality was calculated for demographic subgroups (race, sex, health insurance payer, median neighborhood income, geographic region, rural vs urban location, and teaching vs nonteaching hospital status). Multiple logistic regression analysis was used to determine the association between race and mortality in the absence of PHTN while adjusting for age, sex, primary health insurance carrier, Charlson comorbidity, calendar year, type of admission (elective vs nonelective), geographic region, and hospital characteristics. In addition to the earlierdescribed covariates, logistic regression models for the subgroup with PHTN also incorporated disease presentation (ascites, encephalopathy, variceal bleeding, and hepatorenal syndrome) and admission to a liver transplant center. Interaction between PHTN and race, sex, and type of health insurance Table 1. Demographics of Patients Undergoing Cholecystectomy Without PHTN White (N 302,990) African American (N 38,802) Hispanic (N 56,304) Asian (N 8471) Other (N 12,475) Age, y (SE) 57.1 (0.1) 48.5 (0.2) a 45.1 (0.4) a 56.3 (0.6) 48.0 (0.5) a Female 198,874 (66%) 30,127 (78%) a 41,721 (74%) a 5343 (63%) b 8757 (70%) a Health insurance Private 139,679 (46%) 15,997 (42%) a 21,056 (38%) a 3967 (47%) a 5174 (41%) a Medicare 124,254 (41%) 11,052 (29%) a 10,780 (19%) a 2469 (29%) a 2630 (21%) a Medicaid 18,927 (6%) 7577 (20%) a 14,087 (26%) a 1347 (16%) a 2314 (19%) a Self-pay 11,170 (4%) 2418 (6%) a 6488 (12%) a 415 (5%) a 1712 (14%) a Other 6749 (2%) 1230 (3%) a 3067 (6%) a 233 (3%) a 533 (4%) a Elective admittance 90,357 (33%) 9975 (27%) a 10,935 (25%) a 1829 (32%) 3621 (32%) Indication Acute cholecystitis 106,733 (35%) 12,693 (33%) a 23,131 (41%) a 2943 (35%) 4639 (37%) b Chronic cholecystitis 162,927 (54%) 20,638 (53%) a 27,504 (49%) a 4670 (55%) 6176 (50%) b Cholelithiasis 11,892 (4%) 2498 (6%) a 2887 (5%) a 276 (3%) 789 (7%) b Acalculous cholecystitis 9790 (3%) 1169 (3%) a 1700 (3%) a 259 (3%) 461 (4%) b Other 11,648 (4%) 1804 (5%) a 1082 (2%) a 323 (4%) 410 (3%) b Charlson Index 0.83 (0.01) 0.85 (0.01) 0.49 (0.02) a 0.87 (0.05) 0.57 (0.03) a Region Northeast 73,600 (25%) 7616 (21%) a 8713 (17%) a 1119 (14%) a 4209 (36%) a Midwest 49,756 (17%) 3635 (10%) a 725 (1%) a 236 (3%) a 1432 (12%) a South 125,883 (40%) 24,024 (60%) a 23,222 (40%) a 1069 (12%) a 4581 (35%) a West 53,751 (17%) 3527 (9%) a 23,644 (42%) a 6047 (71%) a 2253 (17%) a Location Rural 52,017 (18%) 3782 (10%) a 3318 (6%) a 949 (11%) 1374 (11%) b Urban 250,871 (82%) 34,985 (90%) a 52,986 (94%) a 7522 (89%) 11,099 (89%) b Hospital size Small/medium 123,112 (40%) 15,679 (40%) 24,239 (42%) 3230 (38%) 4405 (34%) Large 179,776 (60%) 23,088 (60%) 32,065 (58%) 5241 (62%) 8068 (66%) Teaching status Nonteaching 201,691 (66%) 18,583 (47%) a 32,765 (57%) b 5002 (57%) 6526 (50%) a Teaching 101,197 (34%) 20,184 (53%) a 23,539 (43%) b 3469 (43%) 5947 (50%) a a P.001 compared with whites. b P.01 compared with whites.

3 1148 NGUYEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10 Table 2. Characteristics of Patients Undergoing Cholecystectomy With PHTN White (N 1338) African American (N 158) Hispanic (N 309) Asian (N 68) Other (N 71) Age, y (SE) 59.9 (0.6) 53.9 (1.1) a 56.6 (1.1) b 56.7 (1.8) 56.6 (1.7) c Female 555 (41%) 67 (43%) 131 (42%) 29 (43%) 30 (42%) Health insurance Private 479 (36%) 40 (25%) a 69 (22%) a 24 (36%) b 17 (25%) b Medicare 633 (47%) 63 (39%) a 114 (37%) a 22 (32%) b 30 (43%) b Medicaid 143 (11%) 39 (26%) a 86 (28%) a 18 (26%) b 11 (16%) b Self-pay 38 (3%) 10 a 16 (5%) a 10 b 10 b Other 36 (3%) 10 a 23 (8%) a 10 b 10 b Charlson Index 4.0 (0.1) 3.7 (0.2) 3.9 (0.1) a 4.3 (0.2) 4.0 (0.2) a Indication for cholecystectomy Acute cholecystitis 303 (22%) 39 (24%) 79 (25%) c 14 (21%) c 13 (18%) Chronic cholecystitis 723 (54%) 87 (55%) 146 (47%) c 29 (43%) c 40 (58%) Cholelithiasis 60 (5%) c 10 c 10 Acalculous cholecystitis 47 (4%) (5%) c 10 c 10 Other 205 (15%) 22 (14%) 64 (21%) c 19 (28%) c 11 (14%) Ascites 754 (56%) 97 (62%) 177 (58%) 41 (60%) 44 (61%) Encephalopathy 148 (11%) 21 (13%) 40 (13%) Variceal bleeding 31 (2%) Hepatorenal syndrome 85 (6%) (8%) Region Northeast 255 (20%) 25 (17%) a 34 (12%) a 7 (11%) a 10 Midwest 200 (16%) 18 (11%) a 10 a 10 a 12 (17%) South 572 (41%) 97 (61%) a 107 (34%) a 10 (14%) a 28 (39%) West 311 (23%) 18 (11%) a 163 (52%) a 48 (70%) a 21 (28%) Location Rural 170 (13%) 16 (10%) 14 (5%) b 10 c 14 (21%) Urban 1168 (87%) 142 (90%) 295 (95%) b 68 (100%) c 57 (79%) Hospital size Small/medium 455 (33%) 57 (36%) 102 (33%) 20 (30%) 19 (25%) Large 883 (67%) 101 (64%) 207 (67%) 48 (70%) 52 (75%) Teaching status Nonteaching 708 (52%) 70 (44%) 135 (43%) c 19 (26%) b 37 (51%) Teaching 630 (48%) 88 (56%) 174 (57%) c 49 (74%) b 34 (49%) Liver transplant center 258 (20%) 19 (12%) c 84 (28%) 29 (43%) b 16 (23%) NOTE. Actual number of observations when 10 or fewer cannot be disclosed in compliance with the Agency for Healthcare Research and Quality policy. a P.001. b P.01. c P.05. were assessed in logistic regression models for mortality. Logistic model goodness-of-fit was assessed using the deviance goodness-of-fit statistic (ldev Stata program; Philip Ender, UCLA, Los Angeles, CA). The distribution of postoperative time of death (categorized as 24 h, h, and 72 h) was compared among races. The cumulative incidence of death after cholecystectomy was determined for racial groups using the Kaplan Meier method. Comparisons among racial groups also were made for measures of morbidity including postoperative length of stay, requirement for parenteral nutrition, and nonroutine discharge. Ethical Considerations The research protocol was approved by the Institutional Review Board of the Johns Hopkins Medical Institutions. Results The demographics of patients with and without PHTN, stratified by race, who underwent cholecystectomy are shown in Tables 1 and 2, respectively. The prevalence of complications of PHTN did not differ among races. Open cholecystectomy was performed more commonly in patients with PHTN. There were no differences in open versus laparoscopic cholecystectomy in patients with PHTN among races except among Asians, in whom open cholecystectomy was performed in 85% of patients compared with 52% to 64% in other groups. Only 21% of patients with PHTN (n 406) underwent cholecystectomy at a liver transplant center. Compared with whites, African Americans were less likely (20% vs 12%; P.03) whereas Asians were more likely (20% vs 43%; P.001) to be admitted to a transplant center for surgery. Crude Mortality Rates Among those without PHTN there were 7813 deaths associated with cholecystectomy, reflecting an overall mortality rate of 1.4%. Mortality was considerably higher in the PHTN group with 273 deaths and a rate of 10.8% (P.0001).

4 October 2008 RACE AND IN HOSPITAL MORTALITY 1149 Table 3. Crude Mortality After Cholecystectomy Stratified by Demographic Factors No PHTN PHTN Variable N Mortality rate N Mortality rate Race Non-Hispanic white (ref) % % African-American % a 30 19% a Hispanic % b 27 8% Asian/Pacific Islander % 10 7% Other % 10 10% Sex Males (ref) % % Females % b % Health insurance Private (ref) % % Medicare % b % c Medicaid % b % Self-pay % % Median neighborhood income Below national median (ref) % % Above national median % c % Region Northeast (ref) % % Midwest % c % South % a % West % % Location Rural (ref) % % Urban % b % Hospital type Teaching (ref) % % Nonteaching % b % NOTE. Actual number of observations when 10 or fewer cannot be disclosed in compliance with the Agency for Healthcare Research and Quality policy. Ref, reference group. a P.01 compared with the reference group. b P.001 compared with the reference group. c P.05 compared with the reference group. Table 3 shows crude in-hospital mortality rates stratified by demographic factors for patients with and without PHTN. Among the group without PHTN, African Americans had a small but statistically significant higher mortality rate than whites (1.7% vs 1.5%; P.01); the mortality in females in this group was half that of males (1.1% vs 2.2%; P.001). In contrast, African American patients with PHTN experienced a 2-fold higher mortality than their white counterparts (19% vs 11%; P.01) with similar mortality in males and females. Patients without PHTN who had laparoscopic cholecystectomy had lower mortality than those who had the open procedure (0.5% vs 3.8%; P.0001). This association also was mirrored in the PHTN group (6.2% vs 13.6%; P.0001). Multivariate Analysis After multivariate adjustment, women had lower death rates after cholecystectomy in the absence of PHTN (odds ratio [OR], 0.67; 95% confidence interval [CI], ), but this difference was absent in the presence of PHTN. Predictably, older age was a moderate predictor of in-hospital mortality both in the absence and presence of PHTN. African Americans without PHTN had 48% greater odds of death than whites (95% CI, ) after cholecystectomy (Table 4), whereas Hispanics had 20% lower odds of death compared with whites (P.001). Racial disparities in mortality were accentuated by the presence of PHTN, for which the odds of death were more than 2-fold higher in African Americans than whites (OR, 2.37; 95% CI, ). Compared with being privately insured, greater odds of death were observed for Medicare (OR, 1.45; 95% CI, ), Medicaid (OR, 1.80; 95% CI, ), and self-pay/uninsured patients (OR, 1.42; 95% CI, ) in the absence of PHTN. However, among patients with PHTN, there were no differences in mortality among health insurance carriers. For patients without PHTN, being admitted electively for cholecystectomy was associated with 50% lower odds of mortality than nonelective admissions. Among this same subgroup, patients undergoing cholecystectomy in urban hospitals also had higher mortality compared with rural medical centers (OR, 1.37; 95% CI, ). Importantly, patients with PHTN who underwent cholecystectomy at a liver transplant center experienced significantly lower mortality (OR, 0.41; 95% CI, ). Specific complications of PHTN were associated independently with higher mortality (Table 4). Hepatorenal syndrome

5 1150 NGUYEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10 Table 4. Adjusted Mortality Associated With Cholecystectomy No PHTN PHTN Variable Odds ratio (95% CI) P value Odds ratio (95% CI) P value Race Non-Hispanic white Ref Ref Ref African American 1.48 ( ) ( ).001 Hispanic 0.80 ( ) ( ).11 Asian/Pacific Islander 0.86 ( ) ( ).31 Other 0.97 ( ) ( ).57 Female 0.67 ( ) ( ).72 Age (per 10 years) 1.71 ( ) ( ).001 Health insurance Private Ref Ref Ref Medicare 1.45 ( ) ( ).65 Medicaid 1.80 ( ) ( ).89 Self-pay 1.42 ( ) ( ).08 Charlson Index (per point) 1.29 ( ) ( ).03 Income above vs below national median 0.93 ( ) ( ).37 Elective vs nonelective admission 0.49 ( ).001 Calendar year (per year) 0.98 ( ) ( ).15 Variceal bleeding ( ).25 Ascites ( ).001 Hepatic encephalopathy ( ).001 Hepatorenal syndrome ( ).001 Region Northeast Ref Ref Ref Midwest 1.05 ( ) ( ).82 South 1.21 ( ) ( ).74 West 1.09 ( ) ( ).30 Location Rural Ref Ref Ref Urban 1.37 ( ) ( ).39 Teaching vs nonteaching 1.20 ( ) ( ).62 Liver transplant center 0.41 ( ).001 Ref, reference group. (OR, 8.49; 95% CI, ) and ascites (OR, 3.33; 95% CI, ) imparted the highest mortality risks. The vast majority of deaths after cholecystectomy occurred more than 72 hours after surgery for patients with and without PHTN. The postoperative time of death did not differ by race (Table 5). The cumulative incidence of death after cholecystectomy for patients with PHTN is shown stratified by race in Figure 1. The gap in mortality curve between African Americans and whites and Hispanics (P.001) emerged after 3 days and continued to widen thereafter. Effect of Portal Hypertension on Mortality In multivariate analysis, the independent effect of PHTN on in-hospital death associated with cholecystectomy varied with race. The presence of PHTN increased mortality by 4-fold in whites after adjustment for age, comorbidity, geographic region, and hospital characteristics (OR, 4.23; 95% CI, ), while its effect in African Americans was more pronounced (OR, 6.45; 95% CI, ). The interaction term between PHTN and African American versus white race was statistically significant (P.009). This effect modification of PHTN was not observed for the other races, sex, or health insurance. Morbidity Measures Among patients undergoing cholecystectomy without PHTN, the postoperative length of stay was longer among African Americans (4.0 days) and Asian/Pacific Islanders (4.3 days) compared with whites (3.6 days) (P.001) (Table 5). The postoperative length of stay did not differ by race among the subgroup with PHTN (Table 5). Compared with whites, Hispanic patients without PHTN were considerably less likely to require parenteral nutrition (1.6% vs 0.9%; P.001) and nonroutine discharge (14% vs 6%; P.001). The frequency of all postoperative complications in patients with and without PHTN is listed in Table 6. Among those without PHTN, Hispanics experienced lower overall rates of complications than whites (10% vs 16.3%; P.01), and Asians had more wound infections than whites (6.2% vs 4%; P.01). Among those with portal hypertension, African Americans were more likely than whites to experience postoperative gastrointestinal complications, which included acute hepatic injury (12.4% vs 7.5%; P.03). Discussion In the largest nationally representative sample of hospital discharges, we showed racial disparities in mortality after

6 October 2008 RACE AND IN HOSPITAL MORTALITY 1151 Table 5. Other Unadjusted In-Hospital Outcomes After Cholecystectomy White (N 302,990) African American (N 38,802) Hispanic (N 56,304) Asian (N 8471) Other (N 12,475) Group without PHTN Cholecystectomy Open 83,235 (28%) 11,043 (28%) 10,971 (20%) a 2512 (30%) 2852 (23%) a Laparoscopic 219,755 (72%) 27,759 (72%) 45,333 (80%) a 5959 (70%) 9623 (77%) a Postoperative time of death 24 h 123 (3%) 22 (3%) 16 (4%) b h 487 (11%) 70 (11%) 40 (10%) b 72 h 3973 (87%) 571 (86%) 360 (86%) 96 (92%) 113 (92%) b Postoperative LOS, d a 2.8 a 4.1 a 3.1 b Parenteral nutrition 4704 (1.6%) 677 (1.8%) 484 (0.9%) a 174 (2.1%) b 198 (1.6%) Nonroutine discharge 42,602 (14%) 4855 (13%) a 3535 (6%) a 890 (11%) a 1007 (8%) a (N 1338) (N 158) (N 309) (N 68) (N 71) Group with PHTN Cholecystectomy Open 847 (64%) 99 (64%) 195 (63%) 58 (85%) a 38 (52%) b Laparoscopic 491 (36%) 59 (36%) 114 (37%) 10 a 33 (38%) b Postoperative time of death 24 h h h 135 (92%) 26 (87%) 23 (86%) Postoperative LOS, d Parenteral nutrition 73 (6%) 15 (10%) 14 (5%) Nonroutine discharge 520 (39%) 68 (43%) 101 (33%) b 28 (40%) 24 (38%) NOTE. Actual number of observations when 10 or fewer cannot be disclosed in compliance with the Agency for Healthcare Research and Quality policy. LOS, length of stay. a P.001. b P.05. cholecystectomy in the general US population between 1998 and The higher mortality observed in African Americans compared with whites was amplified even further by the presence of PHTN, a well-known risk factor for death and complications in biliary surgery. 10 These findings have potential public health implications because cholecystectomy is among the most commonly performed gastrointestinal surgeries in patients with and without PHTN. Figure 1. Kaplan Meier mortality curves after cholecystectomy among patients with complications of PHTN. The cumulative mortality of whites (solid line), African Americans (dashed line), and Hispanics (dotted line) are shown as a function of postoperative days. There was an insufficient number of Asian/Others to allow accurate estimates. There are several lines of evidence for racial disparities in hospital mortality after major surgeries such as coronary bypass and carotid endarterectomy. 11,12 Previous studies have suggested that the higher mortality among African Americans may be partly attributable to hospital volume and hospital demographics that may in turn reflect quality of care. 12 Racial disparities in access to health care are documented throughout the literature, and it is well established that higher surgical volume for technically difficult or specialized procedures is associated with better hospital outcomes Unlike cardiac or vascular procedures requiring extracorporeal bypass, cholecystectomy is a routine and widely accessible procedure performed by most general surgeons. However, we found a modest increase in mortality in African Americans compared with whites despite the former being on average younger with a greater female predominance, which is associated with favorable outcomes. The odds of mortality for African Americans after adjustment for demographic factors and comorbidity were 70% higher than that for whites. Because cholecystectomy in the presence of PHTN is more technically challenging and the postoperative course is more complicated, our findings of an even greater racial gap in mortality in this subgroup was expected. We also observed lower mortality rates among Hispanics and females in the PHTN group. The lower rate of hospital deaths among Hispanics is consistent with the Hispanic Paradox, a reproducible finding of lower all-cause and disease-specific mortality among Hispanics compared with whites in the United States, 16,17 despite lower rates of health care access. The under-

7 1152 NGUYEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10 Table 6. Postoperative Complications After Cholecystectomy White (N 302,990) African American (N 38,802) Hispanic (N 56,304) Asian (N 8471) Other (N 12,475) No PHTN Wound infection 4.0% 4.4% a 2.7% a 6.2% a 3.0% a Urinary complication 1.2% 0.6% a 0.5% a 1.1% 0.8 a Pulmonary complication 4.8% 4.6% 3.0% 3.4% 4.5% GI complication 5.5% 5.1% a 3.1% a 5.5% 3.7% a Cardiovascular complication 2.4% 2.1% a 1.2% a 2.2% 1.6% a Intraoperative complication 2.4% 2.5% 1.8% a 3.1% a 1.8% a Any postoperative complication 16.3% 15.2% a 10.0% a 17.5% 11.6% a (N 1338) (N 158) (N 309) (N 68) (N 71) PHTN Wound infection 11.3% 12.0% 12.7% 11.5% 6.3% Urinary complication 3.0% 1.9% 2.0% 2.8% 0 Pulmonary complication 15.0% 14.4% 16.0% 17.2% 10.4% GI complication 7.5% 12.4% b 7.5% 6.9% 3.0% Cardiovascular complication 5.4% 3.3% 8.6% 8.6% b 1.3% Intraoperative complication 6.9% 9.7% 7.9% 10.1% 10.2% Any postoperative complication 35.6% 40.7% 40.8% 42.2% 29.1% GI, gastrointestinal. a P.01 compared with white race. b P.05 compared with white race. lying mechanisms of these paradoxic observations remain unclear. The overall complication rates were lower in Hispanics than whites (10% vs 16.3%), and whether these observations could partly explain the lower mortality remains uncertain because we have no reliable information regarding the causes of death. Our findings of gender differences in mortality among those without PHTN support previous findings by Russell et al 18 showing a 2-fold higher rate of mortality among men after cholecystectomy. In the same study, men were found to have higher rates of conversion from laparoscopic to open cholecystectomy and disease severity. In a Swedish population-based study, Nilsson et al 19 found no gender differences in mortality, which may reflect differences in disease severity in the 2 populations. In our analysis, the gender difference in mortality dissipated in the presence of and adjustment for specific complications of PHTN. This observation suggests that within a subgroup with similar increased complication risks, females had no survival advantage over men. Another notable but not unexpected finding was the mortality differences among health insurance carriers. Among those without PHTN, the higher adjusted mortality risk among Medicaid recipients likely may represent lower access to quality health care, particularly for major surgeries Insurance-based disparities in mortality were not observed in patients with PHTN, suggesting that the health care benefits afforded by private insurance do not confer additional benefit over government-sponsored insurance programs in the context of severe liver disease. Importantly, racial differences in mortality were, however, independent of health care access related to insurance status. Our Kaplan Meier analysis of the PHTN subgroup showed mortality in African Americans diverging from that of whites at 4 to 5 days after surgery, which would coincide with the timeframe for development of decompensation from PHTN. Disparities in mortality during this late postoperative period may arise from differential admission to hospitals staffed by surgeons who regularly perform surgery on patients with PHTN (eg, liver transplant surgeons) or hepatologists who are experienced in managing complications of PHTN postoperatively. Interestingly, we found that undergoing cholecystectomy at a liver transplant center was associated with 59% lower odds of in-hospital death compared with nonliver transplant centers. In many transplant centers, liver transplant surgeons rather than general surgeons perform cholecystectomy. Thus, their surgical experience with this high-risk population may explain lower mortality in these centers. Furthermore, most hospitals that perform liver transplant also are staffed by hepatologists who are able to optimize postoperative care. Ko et al 24 previously showed that the management of complications of end-stage cirrhosis by gastroenterologists reduces length of stay with a strong trend toward lowering in-hospital mortality. If these findings were confirmed further in prospective studies, it would be reasonable to recommend that biliary surgeries in patients with PHTN be performed only in transplant centers. Another explanation for improved survival in transplant centers may be that these centers performed cholecystectomy in a selective manner. Because asymptomatic cholelithiasis and noninflammatory edema of the gallbladder wall are common radiologic findings in patients with PHTN, it is possible that transplant centers performed cholecystectomy only when there was overwhelming evidence to suggest that the patient s symptoms were related to gallbladder disease. This high threshold for performing surgery may have improved survival. 25 Nevertheless, when these surgeries were performed in transplant centers, mortality was significantly lower than at other centers. The main limitation of the NIS administrative data was that it did not allow us to directly measure and therefore account for the severity of biliary or liver disease, which may contribute to racial differences in mortality outcomes. Mor-

8 October 2008 RACE AND IN HOSPITAL MORTALITY 1153 tality in patients with liver disease is related directly to the severity of liver disease, 26 but NIS datasets did not have sufficient information to assess the severity of liver disease either by Childs Pugh or Model for End-stage Liver Disease scores. We were, however, able to account for other wellknown predictors of mortality for cholecystectomy including age, sex, comorbidity, and urgency of procedure. Although the delayed timing of death in our study population suggests postoperative rather than immediate perioperative complications were the causes of death, our data do not allow us to reliably assess the specific causes of mortality. Future studies that prospectively characterize the etiologies of cholecystectomy-associated death will offer significant insights into the underlying mechanisms of observed disparities. There are also potential limitations with the assignment of race in the NIS. Race data were derived from the hospital discharge abstract and there was no additional information on whether it was self-reported or assigned by others. Because validation of this race data was not possible, there may have been some misclassification of racial categories. However, a study by Kressin et al 27 has shown strong agreement between race from administrative data and self-reported race information, particularly for whites and African Americans ( 90% agreement). We would expect that potential errors in racial assignment would not be differential with respect to surgical outcomes. These nondifferential race misclassifications potentially may underestimate racial variations in utilization of procedures. 27,28 Another potential limitation of the NIS dataset was that there are no reliable ways to validate the diagnosis of portal hypertension. However, a previous study, using a different administrative dataset, has found good concordance with the diagnostic coding in patients with portal hypertension. 29 In summary, we have shown significantly higher mortality after cholecystectomy in African Americans compared with whites. We also found a very high mortality (10.8%) in patients with PHTN, and, interestingly, mortality was significantly lower in this group when surgery was performed in a liver transplant center. Because cholecystectomy is perhaps the most pervasive surgery of the gastrointestinal tract, and also because gallstones are very common in patients with cirrhosis, these findings will have significant public health impact. Based on our study, we suggest that cholecystectomy should be performed in patients with PHTN only when there is firm evidence to suggest that a patient s symptoms are owing to gallbladder disease. In addition, the surgery should be performed ideally in a liver transplant center or in centers with considerable experience in the management of complications of PHTN. Biological, social, and cultural factors may contribute to the higher mortality observed in African Americans. 30 The NIS is a powerful nationally representative database that identifies issues in health care delivery and utilization that warrant further exploration. However, we will require primary data collection from well-designed cohort studies to elucidate the roots of the observed disparities between African Americans and whites to direct clinical guidelines and interventions to rectify them. References 1. Ahrendt SA, Pitt HA. Calculous biliary disease. In: Townsend CM, Beauchamp RD, Evers BM, et al, eds. Sabiston textbook of surgery: the biological basis of modern surgical practice. 17th ed. Philadelphia: Saunders, 2007: D Amico G, Garcia-Tsao G, Pagliaro L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies. J Hepatol 2006;44: Thulstrup AM, Sorensen HT, Vilstrup H. Mortality after open cholecystectomy in patients with cirrhosis of the liver: a populationbased study in Denmark. Eur J Surg 2001;167: Smedley BD, Stith AY, Nelson AR, Institute of Medicine, eds. Unequal treatment: confronting racial and ethnic disparities in health care. Washington, DC: National Academies Press, 2005: Becker ER, Rahimi A. Disparities in race/ethnicity and gender in in-hospital mortality rates for coronary artery bypass surgery patients. J Natl Med Assoc 2006;98: Konety SH, Vaughan Sarrazin MS, Rosenthal GE. Patient and hospital differences underlying racial variation in outcomes after coronary artery bypass graft surgery. Circulation 2005;111: Benavidez OJ, Gauvreau K, Jenkins KJ. Racial and ethnic disparities in mortality following congenital heart surgery. Pediatr Cardiol 2006;27: Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43: Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45: Isozaki H, Okajima K, Morita S, et al. Surgery for cholelithiasis in cirrhotic patients. Surg Today 1993;23: Becker ER, Rahimi A. Disparities in race/ethnicity and gender in in-hospital mortality rates for coronary artery bypass surgery patients. J Natl Med Assoc 2006;98: Lucas FL, Stukel TA, Morris AM, et al. Race and surgical mortality in the United States. Ann Surg 2006;243: Birkmeyer JD, Stukel TA, Siewers AE, et al. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349: Begg CB, Cramer LD, Hoskins WJ, et al. Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998;280: Dimick JB, Cowan JA Jr, Colletti LM, et al. Hospital teaching status and outcomes of complex surgical procedures in the United States. Arch Surg 2004;139: Sorlie PD, Backlund E, Johnson NJ, et al. Mortality by Hispanic status in the United States. JAMA 1993;270: Liao Y, Cooper RS, Cao G, et al. Mortality patterns among adult Hispanics: findings from the NHIS, 1986 to Am J Public Health 1998;88: Russell JC, Walsh SJ, Reed-Fourquet L, et al. Symptomatic cholelithiasis: a different disease in men? Connecticut Laparoscopic Cholecystectomy Registry. Ann Surg 1998;227: Nilsson E, Fored CM, Granath F, et al. Cholecystectomy in Sweden : a nationwide study of mortality and preoperative admissions. Scand J Gastroenterol 2005;40: Rosenberg M, Cohen F. Medicaid and physician reimbursement. Pediatrics 2006;118: Wang EC, Choe MC, Meara JG, et al. Inequality of access to surgical specialty health care: why children with governmentfunded insurance have less access than those with private insurance in Southern California. Pediatrics 2004;114:e584 e The Medicaid Access Study Group. Access of Medicaid recipients to outpatient care. N Engl J Med 1994;330: Asplin BR, Rhodes KV, Levy H, et al. Insurance status and access to urgent ambulatory care follow-up appointments. JAMA 2005; 294:

9 1154 NGUYEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No Ko CW, Kelley K, Meyer KE. Physician specialty and the outcomes29. Dy SM, Cromwell DM, Thuluvath PJ, et al. Hospital experience and cost of admissions for end-stage liver disease. Am J Gastroenterol 2001;96: Care 2003;15: and outcome for esophageal variceal bleeding. Int J Qual Health 25. Iannuzzi C, Cozzolino G, Negro G. Elective cholecystectomy in30. Nguyen G, Thuluvath PJ. Racial disparity in liver disease: biological, cultural or socioeconomic factors. Hepatology 2008;47: selected cirrhotic patients. Acta Chir Belg 1993;93: Millwala F, Nguyen GC, Thuluvath PJ. Outcomes of patients undergoing non-hepatic surgery: risk assessment and manage ment. World J Gastroenterol 2007;3: Kressin NR, Chang BH, Hendricks A, et al. Agreement between Address requests for reprints to: Paul J. Thuluvath, MD, FRCP, administrative data and patients self-reports of race/ethnicity. Division of Gastroenterology, Department of Medicine, The Johns Hopkins Hospital, 1830 East Monument Street, Baltimore, Maryland Am J Public Health 2003;93: Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet 2002;359: This work was supported by an AGA Research Scholar pjthuluv@jhmi.edu; fax: (410) Award.