Local access to subspecialty care influences the chance of receiving a liver transplant

Authors


  • See Editorial on Page 353

  • A. Sidney Barritt IV contributed to the study concept and design, the drafting of the manuscript, the data analysis, and a critical revision of the manuscript. Stephen A. Telloni contributed to the drafting of the manuscript and the preparation of the data. Clarence W. Potter contributed to the preparation of the data. David A. Gerber contributed to a critical revision of the manuscript. Paul H. Hayashi contributed to the study concept and design, the drafting of the manuscript, the data analysis, and a critical revision of the manuscript.

  • This work was supported in part by the National Institutes of Health (1KL2-RR025746-03 and UL1-RR025747).

Address reprint requests to A. Sidney Barritt IV, M.D., M.S.C.R., Division of Gastroenterology and Hepatology, Department of Medicine, University of North Carolina, Campus Box 7584, Chapel Hill, NC 27599-7584. Telephone: 919-966-2516; FAX: 919-966-1700; E-mail: barritt@med.unc.edu

Abstract

Prior studies have examined the impact of demographic factors on liver transplant outcomes. These factors may be surrogate markers for access to medical care. We investigated physician density in referred patients' hospital service areas (HSAs) as a factor in patients' probability of receiving a liver transplant. We performed a retrospective review of patients referred for liver transplantation from 2002 through 2010. Data on physician density were obtained from the Dartmouth Atlas of Health Care. The primary outcome was the receipt of a liver transplant. A Cox hazard analysis was used to control for various demographic and medical covariates. Over the time period, 1485 adult patients were considered for liver transplantation. Factors that influenced the hazard of receiving a liver transplant were the Model for End-Stage Disease (MELD) score at referral {Hazard ratios (HR) per point = 1.11 [95% confidence interval (CI) = 1.09-1.14]}, a secondary diagnosis of hepatocellular carcinoma [HR = 2.72 (95% CI = 1.76-4.20)], blood group AB [HR = 2.98 (95% CI = 1.52-5.87) with blood group A as the referent], the type of insurance [HR for Medicare = 0.36 (95% CI = 0.14-0.89) with commercial insurance as the referent], and the number of gastroenterologists in an HSA [odds ratio with each additional gastroenterologist per 100,000 population = 1.12 (95% CI = 1.01-1.25)]. Age, race, sex, distance to the transplant center, and residence in a rural community did not influence the chance of receiving a liver transplant. In conclusion, the hazard of receiving a liver transplant are influenced by the diagnosis, MELD score, and insurance status; in addition, patients were 12% more likely to receive a transplant with each additional gastroenterologist per 100,000 population in their local HSA. Local access to gastroenterology subspecialty care is an important factor in receiving a liver transplant. Liver Transpl 19:377–382, 2013. © 2013 AASLD.

Abbreviations
AIH

autoimmune hepatitis

CI

confidence interval

CTP

Child-Turcotte-Pugh

DSA

donor service area

HCC

hepatocellular carcinoma

HR

hazard ratio

HSA

hospital service area

MELD

Model for End-Stage Liver Disease

NASH

nonalcoholic steatohepatitis

PBC

primary biliary cirrhosis

PSC

primary sclerosing cholangitis

SES

socioeconomic status.

Prior studies have examined the impact of travel times to transplant centers, ethnicity, and other demographic factors on liver transplant outcomes. There are significant variations in access to liver transplantation for certain minority populations, including Hispanic and Asian transplant candidates.[1, 2] These ethnic groups tend to be listed for transplantation later in their disease progression and thus have higher Model for End-Stage Disease (MELD) scores at the initial transplant listing and, consequently, worse outcomes.[3] Patients with Medicare and Medicaid also have higher MELD scores at the initial listing and lower survival rates after transplantation in comparison with patients with private insurance.[3, 4] This is not unique to liver transplant patients, however, because kidney transplant patients face similar sociodemographic disparities.[5] In addition, the geographic location of transplant candidates may be a potential barrier. Despite the implementation of the MELD system, significant geographic variation exists in terms of access to liver transplantation within the United States.[6, 7] For instance, patients living in rural areas have lower rates of listing and receiving solid organ transplants.[8] This is somewhat surprising because the distance that a patient lives from a liver transplant center does not adversely affect transplant outcomes.[9] Finally, sex is also a factor in liver transplant rates because women have significantly lower transplant rates than men, and this sex disparity has increased since the implementation of the MELD system.[10]

Each of the aforementioned factors addresses important questions about access to medical care for patients in need of liver transplantation. Although access to local subspecialty care likely affects tertiary referral patterns, we hypothesized that local access to subspecialty care also continues to affect patient outcomes once tertiary referrals are made. This topic has not been specifically addressed in the medical literature to date. In the present study, we investigate physician density within the home hospital service areas (HSAs) of referred patients as a factor in their probability of ultimately receiving a liver transplant.

PATIENTS AND METHODS

We performed a retrospective single-center cohort study. All adult patients (age ≥ 18 years) referred for a liver transplant evaluation at University of North Carolina hospitals between February 27, 2002 and December 31, 2010 were identified via the University of North Carolina's transplant database (TransChart, LLC, Dublin, OH). The primary outcome was the receipt of a liver transplant. The secondary outcome was listing for transplantation. Patients with fulminant hepatic failure and retransplants were excluded. Data on physician density were obtained via the Dartmouth Atlas of Health Care from publicly available files on selected hospital and physician capacity measures. Data for physician density were taken from the year 2006 to bisect the time frame during which our cohort was considered for liver transplantation. Data collection was performed after approval from the institutional review board of the University of North Carolina.

Variable Definitions

Physician density was defined as the number of physicians per capita in an HSA. HSAs are defined by the Dartmouth Atlas Project as local health care markets for hospital care. Each HSA consists of a collection of zip codes that makes up a local hospital's catchment area.

Rural counties are those with population densities less than 250 people per square mile. Distance calculations were performed via Google Maps on a ZIP code to ZIP code basis and are in presented miles.

Insurance status was defined as a patient's primary insurer. Those who had dual insurers (eg, disability insurance plus commercial insurance) were defined as having multiple coverage.

Statistical Analysis

Bivariate comparisons were made between those patients who underwent transplantation and those who did not. Analyses were performed with the Student t test and the chi-square test as appropriate. Data obtained from the bivariate analysis were used to build a Cox hazard model with an a priori strategy to include those variables with a P value < 0.10 or those variables deemed to be of clinical importance as potential confounders. Potential interaction terms, particularly those regarding the distance to transplant centers and physician density as well as urban residence and physician density, were assessed, and none were found to be significant. A full model including statistically and clinically important variables and potential confounders was constructed and then reduced by a backwards elimination strategy (via a change-in-estimate approach with a threshold of a 10% change in the beta coefficient) to create the reduced model. The reduced model did not improve precision, so the full model was used. Data analyses were performed with Stata 11 (StataCorp, College Station, TX).

RESULTS

Between February 27, 2002 and December 31, 2010, 1485 adult patients with nonfulminant liver disease were referred to University of North Carolina hospitals for a liver transplant evaluation (Table 1). Six hundred four (41%) were listed for liver transplantation, and 242 (16%) underwent transplantation. Sixty-four percent of the population was male, and 76% of the population was Caucasian. Viral hepatitis [hepatitis B and hepatitis C (including dual diagnoses of hepatitis C and alcohol)] was the most common etiology of liver disease (44%) and was followed by nonalcoholic steatohepatitis (NASH)/cryptogenic cirrhosis (19%) and alcoholic cirrhosis (17%). Seven percent of the population had a secondary diagnosis of hepatocellular carcinoma (HCC). Fifty-three percent of the population lived in rural areas according to their ZIP codes, and the mean distance to the University of North Carolina was 92 miles. Thirty-four percent of the population had commercial insurance (eg, Blue Cross or Aetna), and 40% of the population had multiple insurers. Seven percent had Medicare only, and 16% had Medicaid only. The number of gastroenterologists per HSA for our referred patient population ranged from 0.64 to 9.77 per 100,000 population with a median number of 3.9 (interquartile range = 3.2-4.4). The overall number of gastroenterologists per HSA in North Carolina ranged from 0.64 to 9.77 per 100,000 population with a median number of 3.2 (interquartile range = 2.1-4.4; Fig. 1).

Figure 1.

Distributions of gastroenterologists in the referral population and in North Carolina. The percentages of the 2 populations—the entire state of North Carolina and the patients referred for transplantation at the University of North Carolina—living in HSAs with various densities of gastroenterologists are shown. An HSA was roughly equivalent in size to a patient's local county. There was no difference between the distribution of gastroenterologists in the referred population and the distribution of gastroenterologists in North Carolina overall (P = 0.90).

Table 1. Characteristics of the Patients Referred for Liver Transplant Evaluations
Patient CharacteristicValue
  1. a

    DSA A covers eastern North Carolina and includes the University of North Carolina and 1 other transplant center that is a negligible distance from the University of North Carolina (<8 miles). DSA B covers the western portion of the state and includes 1 liver transplant center 140 miles from the University of North Carolina.

Referred patients (n)1485
Patients listed for transplantation (n)604
Total transplants (n)242
Mean age at referral (years)54 ± 10
Male [n/N (%)]891/1369 (64)
Caucasian [n/N (%)]941/1235 (76)
Indication for transplantation [n/N (%)]
Viral hepatitis644/1472 (44)
NASH/cryptogenic cirrhosis273/1472 (19)
Alcoholic cirrhosis250/1472 (17)
AIH/PBC/PSC120/1472 (8)
Other185/1472 (13)
Secondary diagnosis of HCC [n/N (%)]105/1472 (7)
ABO blood group [n/N (%)] 
A465/1160 (40)
AB32/1160 (3)
B122/1160 (11)
O541/1160 (47)
Mean MELD score at referral14 ± 6
Mean CTP score at referral7 ± 2
Mean distance to North Carolina transplant centers (miles)a
DSA A92 ± 76
DSA B124 ± 71
Rural residents [n/N (%)]742/1406 (53)
Insurance status [n/N (%)] 
Medicare92/1282 (7)
Medicaid207/1282 (16)
Military25/1282 (2)
Commercial440/1282 (34)
Multiple coverage510/1282 (40)
None8/1282 (1)

A bivariate analysis of those patients who underwent transplantation and those who did not (Table 2) showed that in comparison with women, a higher percentage of men referred for transplantation ultimately underwent transplantation (71% of transplant patients were male, whereas 62% of nontransplant patients were male; P = 0.01). Patients who received a transplant had a higher MELD score at referral (15.5 versus 13.6, P < 0.001) and a higher Child-Turcotte-Pugh (CTP) score at referral (7.8 versus 7.3, P < 0.001). The mean distance to the primary transplant center did not differ between those who underwent transplantation and those who did not. The numbers of acute care hospital beds, primary care physicians, and internists in a local HSA did not differ between those who underwent transplantation and those who did not. Patients who underwent transplantation had more gastroenterologists in their HSAs in comparison with those who did not undergo transplantation (4.01 versus 3.78 per 100,000 population, P = 0.008).

Table 2. Bivariate Analysis of Patients Who Underwent Transplantation and Patients Who Did Not
Patient CharacteristicTransplant Patients (n = 242)Nontransplant Patients (n = 1243)P Value
  1. a

    DSA A covers eastern North Carolina and includes the University of North Carolina and 1 other transplant center that is a negligible distance from the University of North Carolina (<8 miles). DSA B covers the western portion of the state and includes 1 liver transplant center 140 miles from the University of North Carolina.

Mean age at referral (years)54530.13
Male [n/N (%)]170/239 (71)721/1157 (62)0.01
Caucasian [n/N (%)]161/224 (72)780/1011 (77)0.10
Mean MELD score at referral15.513.6<0.001
CTP score at referral7.87.3<0.001
Etiology of liver disease [n/N (%)]
Viral hepatitis110/242 (45)534/1229 (43)0.001
Alcoholic liver disease22/242 (9)228/1229 (19)
NASH/cryptogenic cirrhosis62/242 (25)211/1229 (17)
AIH//PBC/PSC19/242 (8)101/1229 (8)
Other29/242 (12)155/1229 (13)
Secondary diagnosis of HCC [n/N (%)]41/242 (17)64/1214 (5)<0.001
ABO blood group [n/N (%)]   
A88/241 (37)377/919 (41)0.002
AB12/241 (5)20/919 (2)
B38/241 (16)84/919 (9)
O103/241 (43)438/919 (48)
Mean distance to North Carolina transplant centers (miles)a
DSA A87940.17
DSA B1271230.44
Rural residents [n/N (%)]124/240 (52)618/1166 (53)0.82
HSA acute care beds (per 100,000 population)2.302.290.77
HSA primary care doctors (per 100,000 population)68.868.00.43
HSA internists (per 100,000 population)26.125.30.07
HSA gastroenterologists (per 100,000 population)4.013.780.008
Insurance status [n/N (%)]
Medicare7/241 (3)85/1041 (8)<0.001
Medicaid16/241 (7)191/1041 (18) 
Military7/241 (3)18/1041 (2) 
Commercial82/241 (34)358/1041 (34) 
Multiple coverage129/241 (53)381/1041 (37) 
None0/241 (0)8/1041 (1) 

According to the Cox hazard model (Table 3), factors influencing the hazard of receiving a liver transplant were the MELD score at referral {hazard ratio (HR) per point = 1.11 [95% confidence interval (CI) = 1.09-1.14]}, a secondary diagnosis of HCC [HR = 2.72 (95% CI = 1.76-4.20)], blood group AB [HR = 2.98 (95% CI = 1.52-5.87) with blood group A as the referent], the type of insurance [HR for Medicare = 0.36 (95% CI = 0.14-0.89) with commercial insurance as the referent], and the number of gastroenterologists in an HSA [odds ratio with each additional gastroenterologist per 100,000 population = 1.12 (95% CI = 1.01-1.25)]. Age, race, sex, distance to transplant centers, and residence in a rural community did not influence the odds of receiving a liver transplant.

Table 3. Independent Factors Influencing the Risk of Receiving a Liver Transplant by Cox Proportional HRs
Referral CharacteristicHR for Receiving a TransplantP Value
  1. NOTE: The full model controlled for diagnosis, age, MELD score, sex, race, acute care beds per HSA, blood group, HCC, obesity, insurance status, distance to the transplant center, urban residence versus rural residence, and liver-kidney transplantation. Ninety-five percent CIs are shown in parentheses.

MELD score at referral1.11 (1.09–1.14)<0.001
Secondary diagnosis of HCC2.72 (1.76–4.20)<0.001
Blood group AB2.98 (1.52–5.87)0.002
Medicare only0.36 (0.14–0.89)0.03
Density of gastroenterologists per HSA1.12 (1.01–1.25)0.04

The liver diagnosis, MELD score, and insurance status were significantly associated with our secondary outcome: listing for liver transplantation (Table 4). In comparison with patients with a diagnosis of viral hepatitis, patients with alcoholic cirrhosis had an HR of 0.57 (95% CI = 0.42-0.78, P < 0.001) for being listed for transplantation. The MELD score gave an HR per point of 1.05 (95% CI = 1.03-1.07) for listing. The HRs for listing for Medicare and Medicaid patients were 0.43 (95% CI = 0.27-0.69, P < 0.001) and 0.28 (95% CI = 0.18-0.44, P < 0.001), respectively.

Table 4. Independent Factors Influencing the Odds of Being Listed for Liver Transplantation by Cox Proportional HRs
Referral CharacteristicHR for Being Listed for TransplantationP Value
  1. NOTE: The full model controlled for diagnosis, age, MELD score, sex, race, acute care beds per HSA, blood group, HCC, obesity, insurance status, distance to the transplant center, urban residence versus rural residence, and liver-kidney transplantation. Ninety-five percent CIs are shown in parentheses.

MELD score1.05 (1.03–1.07)<0.001
Alcoholic cirrhosis0.57 (0.42–0.78)<0.001
Medicare only0.43 (0.27–0.69)<0.001
Medicaid only0.28 (0.18–0.44)<0.001
Caucasian race0.76 (0.60–0.98)0.03

DISCUSSION

This study shows the novel finding that for patients referred for liver transplantation, the number of gastroenterologists in their home HSA independently increased the chance of receiving a liver transplant by 12% with each additional gastroenterologist per 100,000 population. Each MELD score point increased the chance of transplantation by 11%. Therefore, living in an HSA with 5 gastroenterologists per 100,000 population would increase the chance for transplantation by the same amount as having another 2 MELD points in comparison with living in an HSA with 3 gastroenterologists per 100,000 population (ie, a 24% increase in the chance for transplantation after referral). Although one may see this particular physician density as a surrogate for urban tertiary care, we did control for rural residence versus urban residence along with a variety of other factors, including the number of hospital beds, and the density of gastroenterologists remained significant; this implies a specific benefit from access to these subspecialists.

As expected, HCC increased the odds of receiving a transplant. Medicare insurance coverage was detrimental to a patient's chance of receiving a liver transplant but was better than no insurance coverage at all. Medicaid was numerically detrimental to a patient's chance for transplantation, but the finding did not meet statistical significance. These factors have been reported previously, but their persistence in our analyses attests to the validity of our main finding. Ethnicity, rural residence, and distance to the transplant center did not influence the chance of receiving a liver transplant at our center. Males were more likely to receive a transplant according to the bivariate analysis, but this finding did not persist in the multivariate analysis.

Access to local gastroenterologists improved the probability of transplantation for our patient population, most likely because many of our patients continued to receive medical care locally. Earlier recognition and referral by knowledgeable gastroenterologists do not explain the improved probability because we controlled for the MELD score at referral. Thus, it appears that our community gastroenterology colleagues provide a benefit after referral, most likely by keeping patients stable during the evaluation and while they are on the list.

Any discussion of access to care involves socioeconomic status (SES). At our center, we do not record household income during the transplant evaluation. The best assessment that we can make about SES is the insurance status: those of a higher SES are presumed to have some degree of commercial insurance, and those with a lower SES have Medicaid. This proxy has been used previously in the medical literature.[11]

As with many other centers, our referral base is geographically large: it measures more than 500 miles in diameter. Therefore, we are unable to see many of our patients at the clinic as frequently as a transplant practice in a high-density urban environment might. Our data suggest that having a local gastroenterologist is especially critical in these geographical circumstances. Standard-of-care guidelines regarding outpatient cirrhosis care (eg, spontaneous bacterial prophylaxis criteria[12]) may not be known to generalists. Such interventions will be delayed until the patient's next trip to the transplant center if a local gastroenterologist is not involved. Gastroenterologists are also more likely to pick up on subtle but important declines in a patient's status.

Having a local gastroenterologist is probably even more important when patients must be urgently admitted to their local hospital. Limited bed availability at a transplant center may delay inpatient transfers and leave care in the hands of local providers. In the last 10 years, several studies have shown the positive benefits of certain therapeutic maneuvers early in the hospitalization of liver failure patients. The administration of intravenous albumin on days 1 and 3 of spontaneous bacterial peritonitis care,[13] the immediate use of antibiotics in variceal bleeding[14] and at even slight signs of sepsis,[15] and the use of N-acetylcysteine for non-acetaminophen-induced acute liver failure[16] are just a few examples. Moreover, it is often the local gastroenterologist who will recognize the early need for urgent transfer and/or phone consultation with the transplant center, whereas the generalist may not.

In some community hospitals that refer to our center, there is only partial gastroenterology subspecialty coverage (eg, 3 of 4 weeks per month). Patients admitted to these hospitals on off weeks may have to forgo subspecialty care or await a transfer to a tertiary institution. A greater number of gastroenterologists in low-density areas (or improved strategies for uniform coverage) may lead to improved outcomes for patients awaiting transplantation. In addition to call coverage, improved care for liver failure patients may also come down to workload and time. Liver failure patients are some of the most labor-intensive patients in a gastroenterologist's practice and often require frequent clinic visits, hospitalizations, and phone calls. Increasing the number of such patients in a busy practice by just a few because of fewer gastroenterologists in town may affect one's ability to quickly see patients, address urgent issues, catch up on patient labs, and make thoughtful care decisions.

Our data also confirm that the hazard for listing for transplantation are largely influenced by the diagnosis, the disease severity, and the insurance carrier. In comparison with patients with a diagnosis of viral hepatitis, patients with alcoholic cirrhosis are approximately half as likely to be listed for liver transplantation. We suspect that this is due to the requirement for substance abuse relapse prevention counseling for patients with contemporary alcohol and drug abuse histories. Additionally, we find that substance abuse and mental health disorders are often comorbid with alcohol abuse, and this too is detrimental to the probability of being listed for liver transplantation.

Patients with only Medicare or Medicaid coverage are significantly less likely to be listed for transplantation than those who have commercial insurance alone. Many of these patients may have financial challenges that must be overcome before listing. Those patients with commercial insurance may have a better socioeconomic standing with work-related income and overall reduced personal out-of-pocket costs for health care. The deleterious impact of Medicaid insurance exceeds the positive impact of adding another gastroenterologist per HSA, so without insurance coverage reform, the impact of adding additional subspecialists to underserved areas may be moot.

The density of gastroenterologists was not associated with being listed for transplantation, likely because the decision to list for transplantation relies on factors not under the strict purview of the referring gastroenterologists. In other words, community gastroenterologists appropriately throw a wide net and refer more patients than are ultimately listed. The culling of these referrals is the transplant center's role. Therefore, one might not expect the listing of referred patients to be necessarily associated with gastroenterologist density.

The limitations to this study include its generalizability. Our results are germane to transplant programs with large rural catchments but are less likely to be as pertinent to transplant programs in larger urban centers with denser catchment areas. Additionally, this study cannot account for the patient who is not referred to our center. A 1997 population-based hospital discharge study in North Carolina found that age, sex, source of payment, type of liver disease, rural county of residence, and distance of residence from the transplant center were associated with rates of transplantation[17]; this study, however, did not consider subspecialty care. We did not account for multiple listings and patients who may have undergone transplantation elsewhere because we found this to be a small proportion of our referrals (<5%). Additionally, we used the MELD score at the time of referral as our only assessment of disease severity, and this single value may not capture variations in a patient's clinical course. Finally, the road from referral to transplantation is long with multiple variables to be considered, and our single-center study may have been underpowered to examine all variables fully.

In summary, the number of gastroenterologists in a local HSA has a significant impact on a patient's hazard of receiving a liver transplant. Our data are the first to highlight the important role that community gastroenterologists may play in the care of our transplant candidates. This benefit occurs after referral during the evaluation and wait-list times. In today's era of subspecialization within gastroenterology, it is important to acknowledge this need for retaining basic hepatology skills in the community. As more data accumulate on efficacious health care interventions for patients with cirrhosis, the instances when local gastroenterological care becomes critical for our pretransplant patients will likely increase. Having a skilled and well-educated community gastroenterology work force supporting a transplant center can be critical to many patients maneuvering through the transplant process.

Ancillary