Effect of hospital volume and teaching status on outcomes of acute liver failure
Abstract
Acute liver failure (ALF) often requires multidisciplinary support. Higher hospital volumes have been associated with better outcomes for surgical procedures, but whether such a relationship exists for ALF has not been explored previously. In this study, our aim was to examine if hospital volume affects mortality from ALF. Using data from the Nationwide Inpatient Sample for the years 2001 to 2004, we identified cases by the presence of a primary discharge diagnosis of ALF (International Classification of Diseases, 9th revision, Clinical Modification code 570.x). Hospitals were divided into low-, medium-, and high-volume hospitals on the basis of 1 to 5, 6 to 20, and more than 20 annual ALF discharges. There were 17,361, 6756, and 1790 discharges with ALF from low-, medium-, and high-volume hospitals, respectively. There was no difference in adjusted mortality between low- and high-volume hospitals (odds ratio 0.94, 95% confidence interval 0.68-1.28). Teaching hospitals had a trend toward lower mortality among patients with hepatic encephalopathy (odds ratio 0.69, 95% confidence interval 0.47-1.01). High-volume centers had a higher rate of orthotopic liver transplantation (OLT) primarily because they were transplant centers, had better in-hospital post-OLT survival, and showed a trend toward a shorter time to OLT. In conclusion, patients with ALF receiving care at teaching hospitals and high-volume centers tend to be sicker. However, teaching hospitals and high-volume centers have equivalent in-hospital survival despite caring for this more severely ill cohort. Liver Transpl 14:1347–1356, 2008. © 2008 AASLD.
Hospital characteristics and volume have been shown to influence patient outcomes.1-17 It is well established that high-volume centers have better outcomes for surgical procedures,13, 14, 16, 18-21 but there are limited data on the impact of volume on outcomes for complex nonsurgical conditions. Acute liver failure (ALF) is associated with progressive multiorgan failure and a high mortality.22 In addition to complications such as renal failure requiring support with dialysis and respiratory or circulatory failure, these patients also develop complications typically associated with liver disease, including coagulopathy, encephalopathy, and gastrointestinal bleeding. Thus, patients hospitalized with ALF often require specialized care. Specialist gastroenterologists or hepatologists may be better trained to prevent, recognize early, and treat these complications. Also, the poor prognosis of many of these patients requires prompt recognition of indications for and consideration of liver transplantation.23
Given the multidisciplinary and specialized support needed for ALF, it is conceivable that such a framework, including the availability of a liver transplant team, is more likely to be available at teaching hospitals or at centers that deal with a high volume of ALF in comparison with those centers that see few cases each year. Previous multicenter studies on outcomes of ALF in the United States have been mainly from tertiary care centers24 with the inherent problem of referral bias. Whether hospital characteristics or volume influences outcomes of patients with ALF, beyond individual patient comorbidity and complications, has not been explored previously. Such information, if available, would help identify specific system-related factors that can improve outcomes and have practice implications regarding the need for appropriate referral to specialty centers.
We performed this observational study using a nationwide representative sample with the following aims: (1) to compare the frequency of complications occurring in patients with ALF by hospital teaching status and volume; (2) to identify prognostic factors in determining mortality in hospitalized ALF patients; and (3) to examine if hospital teaching status, transplant center status, or volume affects outcomes after ALF.
Abbreviations
ALF, acute liver failure; ARF, acute renal failure; CI, confidence interval; CVL, central venous line; GI, gastrointestinal; HCUP, Healthcare Cost and Utilization Project; ICD-9-CM, International Classification of Diseases, 9th revision, Clinical Modification; NIS, Nationwide Inpatient Sample; OLT, orthotopic liver transplantation; OR, odds ratio.
PATIENTS AND METHODS
Data Source
The Nationwide Inpatient Sample (NIS) is the largest all-payer inpatient database maintained by the Healthcare Cost and Utilization Project (HCUP) of the Agency for Healthcare Research and Quality.25 It consists of a 20% stratified sample of all discharges from community hospitals in participating states, with survey stratification based on hospital location, bed size, region, ownership, and teaching status. This database has been used to provide reliable estimates for various gastrointestinal and liver diseases26 and analysis of hospital volumes.1, 13-15, 27-29 We used data from the HCUP-NIS for the years 2001 to 2004. The HCUP-NIS 2004 includes discharge data from hospitals across 37 participating states and contains information on nearly 8 million hospital stays. Within the NIS, each hospital discharge constitutes an individual entry and is coded with one primary diagnosis for that hospitalization, which is determined to be the chief reason for the hospital stay. It also lists up to 14 secondary diagnoses and 15 procedures associated with the stay.
Definition of the Study Population
Our study population comprised patients with a primary discharge diagnosis of ALF (International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM) code 570.x), as has been employed previously. Acute or fulminant liver failure is defined in the literature as liver failure associated with coagulopathy (international normalized ratio > 1.5) and encephalopathy that occurs within 8 weeks of onset. Although the ICD-9-CM code for ALF has not been validated with the clinical definition, it is the only available code that can be identified from administrative data sets that captures the population with ALF. This code has been used in previous studies with the recognized limitation that it may also include a few patients with milder liver disease or subacute or chronic liver disease.30, 31
Definition of Variables
Information was extracted from the database about the patients' demographic information. Comorbidity was classified on the basis of the Deyo modification of Charlson's index, which has been used to provide reliable estimates of comorbidity from administrative data sets.32, 33 However, codes pertaining to liver disease within Charlson's index were excluded when the overall score was calculated. Various known prognostic factors in ALF22, 23 were identified by ICD-9-CM codes. These include acidosis (ICD-9-CM code 276.2), acute respiratory failure (ICD-9-CM code 518.81), acute renal failure (ARF) including dialysis support (ICD-9-CM code 584.5, 584.6, 584.7, 584.8, 584.9, 39.95, V45.1, V56.0, V56.1),34 hepatic encephalopathy (ICD-9-CM code 572.2),35 coagulopathy (ICD-9-CM codes 286.6, 286.7, 286.9),35 and hypovolemia (ICD-9-CM codes 276.5, 276.51, 276.52, 785.59). The ascertained comorbid or inciting liver diseases included acetaminophen overdose (ICD-9-CM code 965.4, E935.4, E850.4),31, 36 hepatitis A (ICD-9-CM code 070.0, 070.1), hepatitis B (ICD-9-CM code 070.2x, 070.3x, V02.61), hepatitis C (ICD-9-CM code 070.41, 070.44, 070.51, 070.54, V02.62), hepatocellular cancer (ICD-9-CM code 155.0), alcoholic liver disease or alcoholism (ICD-9-CM code 571.1-571.3, 571.4x, 571.5-571.9, 572.2, 572.3, 572.8, 303.0, 303.0, 303.9, 790.3), fatty liver disease (ICD-9-CM code 571.0, 571.8), abnormal liver function tests (ICD-9-CM code 790.4), and hemochromatosis (ICD-9-CM code 275.0).30 We repeated the analysis excluding patients who despite having a primary diagnosis of ALF also had competing codes that could represent chronic liver disease, as was done by other authors.30
Definition of Hospital Characteristics
The hospital teaching status in the NIS is obtained from the American Hospital Association's Annual Survey of Hospitals. A hospital was considered to be a teaching hospital if it had an American Medical Association–approved residency program, was a member of the Council of Teaching Hospitals, or had a ratio of full-time equivalent interns and residents to beds of 0.25 or higher.
Annual hospital volumes were calculated on the basis of the number of discharges with a primary diagnosis of ALF from that hospital during that particular year for each year of study. Hospitals were classified as low-, medium-, or high-volume hospitals if they had 1 to 5, 6 to 20, or more than 20 annual discharges with ALF. Our specified hospital volume breakdown categories were selected a priori so that low-volume hospitals accounted for approximately two-thirds of the discharges, whereas high-volume hospitals accounted for the top 10% of discharges by volume. We varied this cutoff threshold in our sensitivity analysis. Classification of a hospital as a transplant or nontransplant center is not directly available in the NIS. As an alternate strategy, we classified hospitals as transplant centers if they had any discharges with ICD-9-CM procedure codes for orthotopic liver transplantation (OLT) during the study period (2001-2004).
Definition of Outcomes
Our primary outcomes of interest were in-hospital mortality, length of stay, and total hospitalization charges. As secondary outcome measures, we identified the frequency of liver biopsy (ICD-9-CM code 50.11), central venous line placement (ICD-9-CM code 38.93), blood transfusions (ICD-9-CM code 99.03, 99.04, 99.07), gastrointestinal endoscopy (ICD-9-CM code 45.11, 45.13, 45.14, 45.16, 44.43),37, 38 OLT (ICD-9-CM code 50.5, 50.51, 50.59),35, 39 and mechanical ventilation beyond 96 hours (ICD-9-CM code 96.72).
Statistical Analysis
Appropriate SVY survey estimation commands using the strata weights provided in the NIS were employed for analysis with Stata Intercooled 9.2 (StataCorp, College Station, TX). The t test and the chi-square tests were used to test the difference in the means or proportions between 2 samples, whereas an analysis of variance was used for comparison between multiple groups. Univariate logistic regression was performed to identify predictors of mortality in ALF. Factors that were significant in this analysis at P < 0.10 were included in the final multivariate model. Independent predictors were considered significant at P < 0.05. Similarly, univariate and multivariate linear regressions were performed to identify predictors of continuous outcomes. The multivariate analysis was repeated after the inclusion of only those variables that were statistically significant in the final model.
Subgroup analyses were performed to examine if hospital teaching status or volume affected outcomes in subgroups of patients with ALF. Because patients may not randomly choose to seek care at either teaching or high-volume hospitals, they may differ in measured and unmeasured characteristics that could influence both their choice of hospital and prognosis. To minimize this effect, we performed a propensity score analysis by methods described previously.40, 41 Propensity scores were calculated to determine the likelihood of seeking care at a teaching hospital and high-volume center with age, gender, race, insurance status, location (rural or urban), median income for zip code of residence, and Charlson's comorbidity score. This score was divided into tertiles and was included as a variable in the multivariate models.
The study was approved by the institutional review board of the Medical College of Wisconsin.
RESULTS
Patient Characteristics by Teaching Status
There were 11,831 and 14,076 discharges with a primary diagnosis of ALF from teaching and nonteaching hospitals, respectively. Nonteaching hospitals had a higher proportion of older patients (age 66 or older: 28.1% versus 17.8%, P < 0.001), whereas teaching hospitals had a greater proportion of nonwhite patients (56% versus 46.6%, P < 0.001). There were more patients with Charlson's score > 3 in teaching hospitals.
Teaching hospitals had a larger proportion of cases with hepatitis B (4.2% versus 1.5%, P < 0.001) or acetaminophen overdose (6.3 versus 2.5%, P < 0.001). Certain complications of ALF were more common in teaching hospitals, including acidosis (8.6%), ARF (23.6%), coagulopathy (16.8%), and encephalopathy (17.8%).
Patient Characteristics by Hospital Volume
In all, there were 17,360 discharges from low-volume centers, 6,756 discharges from medium-volume centers, and 1,790 discharges from high-volume centers (Table 1). Patients in high-volume centers had greater comorbidity and more complications, including ARF, respiratory failure, coagulopathy, and encephalopathy. There were also higher proportions of patients with hepatitis B and acetaminophen overdose in high-volume centers (8.1% and 10.4%, respectively) compared to low-volume centers (1.8% and 3.0%, respectively, P < 0.05).
| Characteristic | Low Volume (n = 17,361) | Medium Volume (n = 6,756) | High Volume (n = 1,790) | P Value |
|---|---|---|---|---|
| Age group (years) | 0.001 | |||
| 19-35 | 3,013 (17.4) | 1,523 (22.5) | 497 (27.8) | |
| 36-50 | 5,515 (31.8) | 2,236 (33.1) | 684 (38.3) | |
| 51-65 | 4,189 (24.1) | 1,743 (25.8) | 446 (24.8) | |
| 66-80 | 3,199 (18.4) | 1,005 (14.9) | 142 (8.0) | |
| 81-90 | 1,445 (8.3) | 249 (3.7) | 21 (1.1) | |
| Sex | 0.503 | |||
| Male | 8,124 (46.8) | 2,989 (44.3) | 860 (48.0) | |
| Female | 9,235 (53.2) | 3,767 (55.7) | 930 (52.0) | |
| Race | 0.154 | |||
| White | 8,409 (48.4) | 3,479 (51.5) | 8,393 (46.9) | |
| Black | 1,578 (9.1) | 850 (12.6) | 163 (9.1) | |
| Hispanic | 1,267 (7.3) | 611 (9.1) | 174 (9.6) | |
| Other | 798 (4.6) | 3,057 (4.5) | 210 (11.8) | |
| Missing | 5,311 (30.6) | 1,510 (22.4) | 4,067 (22.7) | |
| Insurance | 0.000 | |||
| Private | 5,917 (34.1) | 1,790 (36.8) | 922 (51.5) | |
| Medicare | 5,787 (33.3) | 1,723 (25.5) | 300 (16.9) | |
| Medicaid | 2,938 (16.9) | 1,293 (19.1) | 339 (18.9) | |
| Self-pay | 2,717 (15.7) | 1,254 (18.5) | 228 (12.7) | |
| Modified Charlson's comorbidity index* | 0.000 | |||
| 0 | 7,476 (43.1) | 1,790 (40.9) | 922 (33.3) | |
| 1 | 3,575 (20.6) | 1,723 (20.0) | 300 (19.1) | |
| 2 | 1,751 (10.1) | 1,293 (7.6) | 339 (3.9) | |
| ≥3 | 4,557 (26.3) | 1,254 (31.5) | 228 (43.6) | |
| Hospital region | 0.020 | |||
| Northeast | 3,199 (18.4) | 1,702 (25.2) | 1,015 (56.8) | |
| Midwest | 3,888 (22.4) | 940 (13.9) | 207 (11.6) | |
| South | 6,891 (39.7) | 2,806 (41.5) | 440 (24.5) | |
| West | 3,378 (19.5) | 1,308 (19.4) | 127 (7.2) | |
| Hospital bed size | 0.000 | |||
| Small | 2,961 (17.1) | 337 (5.0) | 0 (0.0) | |
| Medium | 5,492 (31.6) | 1,036 (15.3) | 91 (5.0) | |
| Large | 8,904 (51.3) | 5,383 (79.7) | 1,699 (95.0) | |
| Hospital location | 0.000 | |||
| Rural | 4,676 (26.9) | 166 (2.5) | 0 (0.0) | |
| Urban | 12,683 (73.1) | 6,590 (97.5) | 1,790 (100.0) | |
| Teaching status | 0.000 | |||
| Nonteaching | 12,360 (71.2) | 1,716 (25.4) | 0 (0.0) | |
| Teaching | 5,000 (28.8) | 5,041 (74.6) | 1,790 (100.0) | |
| Transplant center | 0.000 | |||
| No | 17,110 (98.5) | 4,406 (65.2) | 0 (0.0) | |
| Yes | 251 (1.5) | 2,350 (34.8) | 1,790 (100.0) | |
| Acetaminophen overdose | 0.001 | |||
| No | 16,833 (97.0) | 6,365 (94.2) | 1,606 (89.6) | |
| Yes | 526 (3.0) | 391 (5.8) | 187 (10.4) | |
| Hepatitis B | 0.000 | |||
| No | 17,048 (98.2) | 6,492 (96.1) | 1,645 (91.9) | |
| Yes | 311 (1.8) | 275 (3.9) | 145 (8.1) | |
| Acute renal failure | 0.000 | |||
| No | 14,955 (86.2) | 5,246 (77.6) | 1,173 (65.6) | |
| Yes | 2,404 (13.8) | 1,510 (22.4) | 617 (34.4) | |
| Dialysis | 0.000 | |||
| No | 16,815 (96.9) | 6,173 (91.4) | 1,523 (85.1) | |
| Yes | 544 (3.1) | 583 (8.6) | 267 (14.9) | |
| Acute respiratory failure | 0.000 | |||
| No | 16,312 (94.0) | 6,051 (89.6) | 1,477 (82.5) | |
| Yes | 1,047 (6.0) | 705 (10.4) | 313 (17.5) | |
| Acidosis | 0.012 | |||
| No | 16,242 (93.6) | 6,165 (91.3) | 1,642 (91.8) | |
| Yes | 1,117 (6.4) | 591 (8.7) | 148 (8.2) | |
| Coagulopathy | 0.004 | |||
| No | 15,333 (88.3) | 5,458 (80.8) | 1,500 (83.8) | |
| Yes | 2,026 (11.7) | 1,298 (19.2) | 290 (16.2) | |
| Hepatic encephalopathy | 0.000 | |||
| No | 15,318 (88.2) | 5,593(82.8) | 1,285(71.7) | |
| Yes | 2,041(11.8) | 1,163(17.2) | 508(28.3) | |
| Hyponatremia | 0.035 | |||
| No | 15,758 (90.8) | 6,217 (92.0) | 1,697 (94.8) | |
| Yes | 1,601 (9.2) | 539 (8.0) | 93 (5.2) | |
| Hypovolemia/shock | 0.000 | |||
| No | 13,908 (80.1) | 5,557 (82.3) | 1,562 (87.3) | |
| Yes | 3,451 (19.9) | 1,197 (17.7) | 228 (12.7) | |
| In-hospital mortality (%) | 17.2 | 20.9 | 27.3 | 0.000 |
- NOTE: Numbers in parentheses represent percentages unless specified otherwise.
- * Excluding codes for mild and severe liver disease.
Predictors of Mortality in ALF
On univariate analysis, older age and greater comorbidity were associated with higher mortality, whereas female sex predicted lower mortality. Acidosis, ARF, respiratory failure, coagulopathy, encephalopathy, hyponatremia, and hypovolemia were all associated with worse outcomes. Patients with hepatitis B also had worse outcomes. On multivariate analysis, older age [odds ratio (OR) 5.40, 95% confidence interval (CI) 3.39-8.59] and greater comorbidity (OR 3.62, 95% CI 2.70-4.85) remained associated with higher mortality (Table 2). Uninsured patients had worse outcomes than patients with private insurance (OR 1.36, 95% CI 1.02-1.77). Respiratory failure, ARF, coagulopathy, and hypovolemia were associated with higher mortality. Patients discharged from hospitals in the Midwest or South had better outcomes than patients discharged from hospitals in the Northeast.
| Predictor | Odds Ratio | 95% Confidence Interval |
|---|---|---|
| Age group (years) | ||
| 19-35 | 1.0 | |
| 36-50 | 1.65 | 1.19-2.29 |
| 51-65 | 2.65 | 1.89-3.73 |
| 66-80 | 4.18 | 2.75-6.36 |
| 81-90 | 5.25 | 3.28-8.39 |
| Sex | ||
| Male | 1.0 | |
| Female | 0.77 | 0.64-0.92 |
| Insurance | ||
| Private | 1.0 | |
| Medicare | 1.16 | 0.90-1.51 |
| Medicaid | 1.54 | 1.16-2.04 |
| Self-pay | 1.36 | 1.02-1.77 |
| Modified Charlson's comorbidity index | ||
| 0 | 1.0 | |
| 1 | 1.35 | 1.02-1.77 |
| 2 | 2.16 | 1.57-2.97 |
| ≥3 | 3.77 | 2.79-5.09 |
| Hospital region | ||
| Northeast | 1.0 | |
| Midwest | 0.73 | 0.56-0.95 |
| South | 0.74 | 0.58-0.95 |
| West | 0.92 | 0.72-1.16 |
| Acidosis | ||
| No | 1.0 | |
| Yes | 2.20 | 1.63-2.96 |
| Hepatitis B | ||
| No | 1.0 | |
| Yes | 2.23 | 1.45-3.42 |
| Renal dysfunction | ||
| No | 1.0 | |
| Yes | 3.05 | 2.46-3.78 |
| Acute respiratory failure | ||
| No | 1.0 | |
| Yes | 5.39 | 4.05-7.17 |
| Coagulopathy | ||
| No | 1.0 | |
| Yes | 1.41 | 1.12-1.78 |
| Teaching status | ||
| Nonteaching | 1.0 | |
| Teaching | 0.98 | 0.79-1.22 |
| Hospital volume | ||
| 1-5 | 1.0 | |
| 5-20 | 0.97 | 0.76-1.25 |
| 21 or more | 0.94 | 0.68-1.28 |
Differences in Outcomes by Teaching Status
On univariate analysis, teaching hospitals had higher mortality than nonteaching hospitals (OR 1.31, 95% CI 1.11-1.54), but this difference was not significant after adjustments for age, comorbidity, and various prognostic factors (OR 0.97, 95% CI 0.79-1.20; Table 3). Among patients with ALF and an associated diagnosis of hepatitis B, teaching hospitals had a trend toward improved survival on the multivariate analysis (OR 0.30, 95% CI 0.06-1.66), whereas this was not seen in ALF with acetaminophen overdose or other etiologies. On subgroup analysis, patients with encephalopathy had a trend toward lower mortality in teaching hospitals (OR 0.69, 95% CI 0.47-1.01). There was no mortality benefit in teaching hospitals compared to nonteaching hospitals among patients with ARF or coagulopathy. However, there was a trend toward better survival among the 1,146 patients with both encephalopathy and coagulopathy (fulminant hepatic failure; OR 0.70, 95% CI 0.31-1.61). Patients at teaching hospitals had a longer adjusted hospital stay (0.83 days, 95% CI 0.33-1.33) and higher hospital charges ($5713, 95% CI $1950-9476).
| Subgroup | Number of Patients (n) | Teaching Hospital* | High-Volume Hospitals† |
|---|---|---|---|
| All patients (unadjusted) | 25,907 | 1.31 (1.11-1.54) | 1.81 (1.41-2.33) |
| All patients (adjusted) | 25,907 | 0.97 (0.79-1.20) | 0.98 (0.72-1.34) |
| Propensity score–adjusted model | 25,907 | 0.96 (0.78-1.18) | 0.98 (0.71-1.33) |
| Hepatitis B | 706 | 0.30 (0.05-1.65) | 2.23 (0.46-10.81) |
| Acute renal failure | 5,493 | 1.18 (0.86-1.62) | 0.98 (0.63-1.53) |
| Coagulopathy | 3,599 | 1.24 (0.83-1.86) | 1.91 (1.00-3.65) |
| Hepatic encephalopathy | 3,690 | 0.69 (0.47-1.01) | 0.88 (0.51-1.53) |
| Coagulopathy and encephalopathy | 1,146 | 0.70 (0.31-1.61) | 0.62 (0.19-2.05) |
| Required mechanical ventilation > 96 hours | 1,149 | 1.23 (0.53-2.88) | 0.60 (0.20-1.82) |
- * Compared to nonteaching hospitals.
- † Compared to low-volume hospitals.
Differences in Outcomes by Hospital Volume
Multivariate analysis did not reveal significantly different mortality in medium- (OR 1.00, 95% CI 0.79-1.28) or high-volume centers (OR 0.98, 95% CI 0.72-1.33) in comparison with low-volume centers in the overall study population or in any specific subgroup. Because patients transferred from other hospitals to high-volume centers are likely to be sicker with a higher mortality, we repeated the analysis after adjusting for transfer status (admitted on transfer from another short-term hospital) in the multivariate model. This analysis yielded a trend toward lower mortality in high-volume centers (OR 0.79, P = 0.17) but not in medium-volume centers (OR 0.98, P = 0.84). The annual hospital volume among the high-volume centers ranged from 21 to 44 discharges. We repeated the analysis for in-hospital mortality within this stratum alone. Within the subgroup of patients at high-volume hospitals, individual center volume was not predictive of in-hospital mortality. Patients at high-volume hospitals had a nearly 5-day longer adjusted hospital stay (4.5 days, 95% CI 1.5-7.4) and higher hospital charges ($51,240, 95% CI $18,179-84,301).
Analysis by Transplant Center Status
All hospitals with a high-volume of ALF admissions were transplant centers, whereas only 1% of ALF discharges from low-volume hospitals were from transplant centers (Table 1). Patients from transplant centers were more likely to have coagulopathy (21.4% versus 12.4%, P = 0.001), encephalopathy (26.0% versus 11.9%, P < 0.001), or renal failure (36.4% versus 18.4%, P < 0.001). On multivariate analysis after adjustments for age, comorbidity, and the presence of complications, discharges from transplant centers had a higher adjusted mortality (OR 1.62, 95% CI 1.23-2.13). Because of the high degree of colinearity between transplant center status and hospital volume, these variables were not included in the same multivariate model. Among transplant hospitals, there was a trend toward lower mortality at high-volume ALF centers (OR 0.53, 95% CI 0.25-1.12) but not medium-volume centers (OR 0.86). A greater proportion of patients at high-volume hospitals underwent OLT in comparison with patients at medium-volume hospitals, even after the analysis was restricted to transplant centers alone (19.7% versus 11.9%, P = 0.002). However, this difference was not statistically significant on multivariate analysis (OR 0.72, 95% CI 0.45-1.15).
Frequency of Diagnostic and Therapeutic Procedures in ALF
Patients in teaching hospitals were more likely to undergo a liver biopsy (12.2% versus 5.9%, P < 0.001) than patients in nonteaching hospitals (Table 4). They were also more likely to receive a blood transfusion or require prolonged mechanical ventilation or placement of a central venous line. This was also true in medium- and high-volume hospitals in comparison with low-volume hospitals. Importantly, medium- and high-volume hospitals had a higher proportion of patients who underwent OLT (20.1% versus 0.2%, P < 0.0001), better in-hospital survival post-OLT, and a trend toward a shorter time to transplantation (8.9 days versus 13.9 days) than low-volume hospitals.
| Procedure | Hospital Volume | Teaching Status | |||
|---|---|---|---|---|---|
| Low Volume | Medium Volume | High Volume | Teaching | Nonteaching | |
| Liver biopsy (%) | 6.9 | 12.3* | 14.5† | 12.2* | 5.9 |
| Mean time to biopsy (days) | 4.3 | 3.9 | 6.1 | 4.3 | 4.6 |
| OLT (%) | 0.2 | 4.1* | 20.1* | 5.6* | 0.07 |
| Mean time to transplantation (in days) | 13.9 | 7.4 | 8.9 | 8.6 | 9.4 |
| Died in hospital post-OLT (%) | 33.3 | 5.6* | 14.2* | 11.7* | 0 |
| CVL placement (%) | 9.4 | 14.9* | 20.3* | 15.2* | 8.5 |
| Mean time to CVL placement (in days) | 3.2 | 4.8 | 5.2† | 4.7† | 3.1 |
| Upper GI endoscopy (%) | 7.3 | 8.3† | 11.0 | 8.6 | 7.1 |
| Blood transfusion (%) | 12.4 | 17.1* | 19.7* | 16.1† | 12.5 |
| Mechanical ventilation > 96 hours (%) | 2.2 | 6.2* | 18.9* | 7.7* | 1.7 |
- NOTE: P values for medium- and high-volume hospitals are for comparison with low-volume hospitals; P values for teaching hospitals are for comparison with nonteaching hospitals.
- Abbreviations: CVL, central venous line; GI, gastrointestinal; OLT, orthotopic liver transplantation.
- * P < 0.001.
- † P < 0.05.
Analysis for Interactions
We analyzed for interaction between teaching status and hospital volume with hospital location or bed size. None of the interaction terms were statistically significant. We also found no interaction between teaching status and hospital volume.
Sensitivity and Subgroup Analyses
We performed various sensitivity analyses to examine the robustness of our results (Table 3). First, we excluded patients who, despite having a primary diagnosis of ALF, also had competing secondary discharge diagnoses that could represent chronic liver disease and not ALF (n = 3,350). These included patients with hepatitis C, fatty liver, abnormal liver function tests, hemochromatosis, hepatocellular carcinoma, or liver abscess. This did not change the OR estimates for teaching hospitals (1.06, 95% CI 0.85-1.33) or high-volume centers (0.87, 95% CI 0.60-1.26). To minimize incomplete adjustment for severity of illness, we also adjusted for blood transfusions, placement of a central venous line, mechanical ventilation beyond 96 hours, or need to undergo upper endoscopy or liver biopsy. This also did not significantly change the estimates for teaching hospitals (OR 0.98) or high-volume centers (0.93). Adjusting for propensity score tertiles did not further alter our estimates. Modifying the volume thresholds for high-volume centers to 15 annual discharges (OR 1.33, 95% CI 0.98-1.80) or 30 discharges (OR 0.91, 95% CI 0.64-1.32) did not significantly change the OR estimates.
DISCUSSION
The impact of hospital volume or teaching status on patient outcomes is an important health concern.17 Several studies have shown that higher hospital volume is associated with better outcomes for surgical procedures.1-9, 11, 20, 21 However, there has been limited examination of this relationship for complex medical illnesses such as ALF. ALF often requires specialist care to recognize complications early and institute appropriate management. This is more likely to be present in centers that deal with a higher volume of ALF. To our knowledge, ours is the first study using a nationwide representative sample to examine the impact of hospital teaching status and volume on outcomes of patients with ALF.
We identified older age and greater comorbidity to be risk factors for mortality in ALF. This correlates with the King's College and Clichy criteria, in which age is a negative prognostic factor in non–acetaminophen-induced ALF.22, 23, 42, 43 Acidosis, coagulopathy, and ARF were also associated with higher mortality in our study, as has been recognized previously.23, 42, 43 We also confirmed that hepatitis B infection is associated with higher mortality in comparison with other causes of ALF.44 Interestingly, we identified a geographic variation in patient outcomes, with patients in the Midwest or South having a lower adjusted mortality than discharges from hospitals in the Northeast. This persisted after adjustments for the presence of complications and comorbidity and may imply differences in management between various regions.
Patients with ALF who were treated at teaching hospitals and high-volume centers tended to be sicker than their counterparts treated at nonteaching hospitals or low-volume centers. However, after adjusting for age, comorbidity, and the other complications that have been shown to have prognostic significance, we did not find a difference in in-hospital mortality by teaching status or hospital volume. This differs from previous studies examining surgical outcomes or intensive care unit admissions.2-4, 6-9, 11, 18-21 Prior examination of outcomes for certain medical conditions from teaching hospitals revealed differences only in some, not all, conditions.17 There could be various explanations for our findings. It is possible that in ALF, the presence of complications such as coagulopathy or renal or respiratory failure may more critically influence mortality than hospital characteristics. This may differ from surgical procedures in which the surgeon's experience may play a dominant role in determining patient outcome. However, equivalent survival at teaching hospitals and high-volume centers, even though they care for a more severely ill cohort, may represent vindication for these hospitals despite higher treatment costs.17
We also identified a trend toward better survival in patients with both encephalopathy and coagulopathy in teaching hospitals and high-volume centers. Teaching hospitals also had a trend toward lower mortality in the subgroup of patients with encephalopathy alone (OR 0.68, 95% CI 0.47-1.01). We postulate several reasons for these findings. Supportive care for patients with encephalopathy, including provision of respiratory support for advanced encephalopathy, early recognition of cerebral edema, use of intracranial pressure monitoring, and institution of appropriate therapeutic interventions, may be more readily available in high-volume centers. There may also be differences in the availability of physician and nursing staff.15 Recent multicenter studies of ALF outcomes have been limited to large academic medical centers. Although such reports shed important light on the epidemiology and prognostic factors in ALF, they do not examine interinstitution variations among different levels of hospitals. It may be important to include low-volume centers and nonteaching hospitals in this cohort to prospectively examine the role of specialist physicians (for example, hepatologists and intensivists) and other hospital factors in outcomes of ALF.
Hepatitis B accounts for 7% to 19% of ALF cases in the United Kingdom and United States24, 44 and is associated with a high mortality. Teaching hospitals showed a trend toward improved survival in patients with hepatitis B (OR 0.30, P = 0.16), a difference that was not identified for any other etiology of ALF. This may be, at least in part, due to the higher rate of OLT in teaching hospitals, as OLT has been shown to improve survival to over 80% in patients with hepatitis B virus–related ALF.44
Uninsured or self-pay patients had significantly worse survival in our study than those patients with private insurance. They were also less likely to be transferred (7.4 versus 11.4%, P = 0.001) and undergo a liver biopsy (5.7% versus 10.7%, P < 0.001) or OLT (0.2% versus 4.7%, P < 0.001). The difference in mortality is similar to that found by Nguyen et al.35 among patients admitted for complications of portal hypertension and raises questions of disparities in care. Whether the difference in survival is due to the delayed presentation of uninsured patients, impaired access to health care, or reluctance in seeking care due to financial burden merits further prospective study.
Analysis by transplant center status revealed that patients with ALF receiving care at transplant hospitals were likely to be sicker with a higher frequency of complications such as coagulopathy, encephalopathy, or renal failure. This higher acuity of illness likely explains our finding of higher adjusted mortality in transplant hospitals compared to nontransplant hospitals. However, among transplant hospitals, there was a trend toward lower mortality for hospitals with a high volume of ALF (OR 0.53, 95% CI 0.25-1.12).
High-volume centers and teaching hospitals had a larger proportion of patients undergoing liver biopsy or requiring blood transfusion or prolonged mechanical ventilation, which reflected the severity of disease in these patients. Importantly, high-volume centers had a significantly larger proportion of patients who underwent OLT, although this difference was neutralized after stratification by transplant center status. They had a trend toward a shorter median time to transplantation and significantly better in-hospital survival post-OLT in comparison with low-volume centers. Although we did not have information on the volume of OLT performed at each hospital, it is likely that high-volume ALF centers are also high-volume centers for OLT. Our findings correlate with several recent studies that showed better outcomes in centers with a high-volume of OLT.45-47 However, our results further support the practice that patients with ALF who have a disease course suggesting a possible need for OLT should be transferred to high-volume centers.
Our study has several strengths. To our knowledge, it is the first study examining the differences in characteristics as well as outcomes of patients with ALF between different hospitals by teaching status and volume of ALF discharges. We were able to adjust for comorbidity and various important prognostic factors. Using a nationwide representative sample decreases some of the bias from single-center studies, especially from large teaching hospitals, as we have shown that patients from these centers tend to be more severely ill. We identified cases of ALF by using a previously employed definition using ICD-9-CM codes.30 In addition, separate analyses after the exclusion of patients with ICD-9-CM codes that could be confused with chronic liver disease did not change our estimates. We also controlled for the fact that certain patients may be more likely to seek care at teaching hospitals or high-volume centers by including propensity score tertiles in our multivariate models.
The major limitation of our study is the reliance on ICD-9-CM coding for a diagnosis of ALF. By performing our analysis after excluding patients with possible competing codes for chronic liver disease, we decreased the likelihood of inclusion of this subgroup in our study. Previous authors have reported that about 20% of the patients in their study using administrative codes for ALF were coded as chronic liver disease in the follow-up period.30 Moreover, the high in-hospital mortality of over 20% in our cohort suggests that we did capture a majority of patients with ALF and that only a few patients with mild liver disease were included in our study. We also repeated the analysis including patients with codes for both encephalopathy and coagulopathy, thus capturing the standard clinical definition of ALF. Also, the misclassification of cases as ALF is likely to be random between the different volume strata and thus unlikely to affect our estimates and results significantly. We controlled for disease severity using a number of different complications and interventions. However, as in most observational studies, there is the possibility of unmeasured confounders. In our analysis, patients treated at high-volume centers or teaching hospitals were sicker than their counterparts. It is likely that any such unmeasured confounder that adversely affects prognosis in ALF would also be more common in these centers, thus shifting our results toward better survival in these centers. Given the higher number of ALF discharges in our study in comparison with previous estimates,48 it is possible that some patients who do not fit the definition of ALF but have subacute liver failure are included in our analysis. However, we expect this to be similar across all hospitals and do not expect it to modify our results. We also attempted to minimize this by including only individuals with a primary diagnosis of ALF.
In conclusion, we demonstrate that patients with ALF receiving care at teaching hospitals and high-volume centers tend to be sicker. However, teaching hospitals and high-volume centers have equivalent in-hospital survival despite caring for this more severely ill cohort, thereby vindicating their higher costs. Uninsured patients have higher mortality compared to those with private insurance. High-volume centers have a greater rate of utilization of OLT primarily because they are transplant centers, show a trend toward shorter time to OLT, and have improved short-term in-hospital survival after OLT for ALF.
