Recently we reported that race is an independent predictor of short- and long-term survival after liver transplantation.1 In our study, African-Americans and Asian Americans had poor outcomes compared to white Americans and Hispanic Americans. In transplant recipients, adherence to the immunosuppressive regimen and regular outpatient follow-up are essential for optimal outcome.2–9 It has been suggested that lower socioeconomic status (SES) may be one of the predictors of noncompliance with medical treatment.6, 10–12 It is possible that SES may have an effect on the outcome of adult liver transplant recipients since there is no universal coverage of posttransplant care and essential medications by federal or state government. It has been suggested that racial disparities in the outcome of liver transplant recipients, in addition to immunological differences, may well be related to differences in SES.13 Many studies have found a relationship between SES, defined by income or insurance, and survival.14–19 The present study was done to determine the effect of SES on posttransplant survival in adult liver recipients. We used neighborhood income (median and average income level as determined by zip code) and education as surrogate markers of SES. We also assessed the impact of insurance (payer) status on survival because insurance status may also be related to SES. Posttransplant survival was determined after adjusting for known independent predictors of survival in adult liver recipients.
Poor socioeconomic status (SES) may be associated with lower survival after liver transplantation. In a previous study, we showed that African-American race was an independent predictor of poor survival, and one of the major criticisms of our study was that we had not adjusted the survival for SES as a confounding variable. The objective of the present study was to determine the posttransplant outcome of adult liver transplant recipients based on neighborhood income, education, and insurance using the United Network for Organ Sharing (UNOS) database from 1987 to 2001. Patients (n = 29,481) were divided into 5 groups based on median income as determined by zip code: <$30,000, $30,001–$40,000, $40,001–$50,000, $50,001–$60,000, and >$60,000). Patients (n = 14,814) were divided into 4 groups based on level of education: higher than bachelor's degree; college attendance or technical school; high school education (grades 9-12); less than high school education. Insurance payer status (n = 23,440) was divided into Medicaid, Medicare, government agency, HMO/PPO, and private. Cox regression analysis was used to adjust the survival for other known independent predictors such as age, race, UNOS status, diagnosis, and creatinine. Results showed that neighborhood income had no effect on graft or patient survival either in the entire cohort or within different racial groups. Education had only marginal influence on the outcome; survival was lower in those with a high school education than in those with graduate education. Patients with Medicaid and Medicare had lower survival when compared to those with private insurance. African-Americans had a lower 5-year survival when compared to white Americans after adjusting for SES and other confounding variables. In conclusion, neighborhood income does not influence the outcome of liver transplantation. Education had minimal influence, but patients with Medicare and Medicaid had lower survival compared to those with private insurance. (Liver Transpl 2004;10:235–243.)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Patients and Methods
For this study, we used data from the United Network for Organ Sharing (UNOS) that included 50,616 patients who underwent orthotopic liver transplantation (OLT) in the United States between 1988 and 2001. We collected the following information on recipients and donors: age, gender, race, height and weight (of both recipient and donor) at time of transplantation, serum creatinine at time of transplantation, etiology of liver disease, ABO blood-type matching status, UNOS listing status, cold ischemia time, and 1-month, 1-year, 2-year and 5-year graft and patient survival. Body mass index (BMI) of both recipients and donors was calculated at the time of transplantation by dividing the weight (kg) by the square of height (m; kg/m2).
Median income, highest level of education achieved by the individual recipient, and insurance (payer) status were used as surrogate markers of SES. Because information on individual recipient's income was not available from the UNOS database, we used the zip code to determine neighborhood income, and this was used as a surrogate marker of income. The median income represented the median income of the neighborhood in which the patient resided at the time of registration for liver transplantation. Neighborhood income was obtained by incorporating the information from another database containing the median income for all zip codes in the United States (CACI Marketing Systems, Arlington, VA).
Complete data on median income determined by zip code, education, and payer status of the recipients at the time of registration for liver transplantation were available for 29,481, 23,440, and 14,814 patients, respectively (Fig. 1). Patients without zip code information (n = 4,985) were excluded from the zip-code income analysis. Similarly, 19,652 patients were excluded from education level, and 11,206 were excluded from payer status because of inadequate information. When we compared the demographics (income, education, and insurance status) of patients who were excluded from the study with those who were included, there were no clinically significant differences between the 2 groups (data not shown).
For the purpose of analysis, median income level by zip code was arbitrarily categorized into 5 groups: <$30,000, $30,001-$40,000, $40,001-$50,000, $50,001-$60,000 and >$60,000. Education level was divided into 4 groups: bachelor's degree or higher, college attendance or technical school, high school education (grades 9-12), and less than high school education. Insurance payer status was divided into 5 groups: Medicaid, Medicare, government agency, HMO/PPO, and private. The definition of UNOS listing status underwent change during the study period, so the old system did not correspond to the current listing system except for Status 1. This was taken into consideration when dividing UNOS listing status into 3 categories: UNOS Status 1 (fulminant hepatic failure), UNOS Status 2A (patients in the intensive care unit or who require intensive monitoring), and others. We classified the cause of liver disease as alcoholic liver disease, hepatitis C, hepatitis B, cryptogenic cirrhosis, primary biliary cirrhosis, and others. Patients with a diagnosis of non-A, non-B hepatitis were included as hepatitis C, patients with combined hepatitis B and C were included as hepatitis B, and patients with combined alcoholic liver disease and “postnecrotic” cirrhosis (possible viral hepatitis) were included as alcoholic liver disease. ABO matching was divided into 3 groups: “matched” (donor and recipient of same blood type); “compatible” (donor blood type O and recipient blood type other than O, or, recipient blood type AB and donor blood type either A or B); and “mismatched” (recipient blood type A, B, or O and donor blood type other than O or different from that of the donor). For the purpose of excluding outliers in our data, recipient BMI of <15 kg/m2 or >55 kg/m2, and creatinine value of 0 mg/dL, were considered missing. We did this on the assumption that these data may have beenentered incorrectly.
Continuous variables were compared by ANOVA and categorical variables by chi-square test. One month, 1-year, 2-year, and 5-year survival after OLT was determined by Kaplan-Meier survival analysis and log-rank test was used to analyze the differences in survival by Kaplan-Meier analysis. Cox regression analysis was used to adjust the survival for confounding variables that were found to be significant by univariate analysis. The following variables were included in the Cox regression analysis: recipient age, race, BMI, serum creatinine, UNOS listing status, cause of liver disease, and cold ischemia time; donor age; andthe SES parameters of median income by zip code, education level, and payer status. We selected these variables based on our previous experience with the same database.1 For categorical variables, white race, primary biliary cirrhosis (PBC) (diagnosis), ABO “matched,” UNOS status “others”, highest income level (>$60,000), highest education level (bachelor's degree or higher), and private insurance were used as indicator (reference) variables in Cox regression analysis because these factors are known to be associated with better outcome. For all analyses, a 2-tailed P value of .05 or less was considered significant. Statistical analyses were performed using SPSS version 10.0.6 (SPSS, Inc., Chicago, IL).
Patient characteristics differed according to median neighborhood income (Table 1), education (data not shown) and insurance (payer) status (data not shown). Fig. 2 shows the relationship between median neighborhood income, education level, and payer status when stratified by race. African-Americans patients and Hispanic Americans lived in a relatively lower income neighborhood compared to white Americans and Asian Americans. There was a higher proportion of Hispanic Americans with less than high school education than other races. When patients were stratified by insurance status, there were higher proportions of African-Americans and Hispanic Americans with Medicaid. In contrast, the proportion of white Americans in the Medicaid group was lower than in other groups.
|<$30,000 (n = 4,060)||$30,001–40,000 (n = 9,047)||$40,001–50,000 (n = 7,482)||$50,001–60,000 (n = 4,412)||>$60,000 (n = 4,480)||P value|
|Age (yr)||48.1 ± 1.2||49.3 ± 10.9||49.4 ± 10.8||50.6 ± 10.8||50.6 ± 10.6||<.01|
|BMI (kg/m2)||27.3 ± 5.8||27.3 ± 5.7||27.1 ± 5.5||26.9 ± 5.4||26.5 ± 5.2||<.01|
|Creatinine (mg/dL)||1.26 ± 1.05||1.28 ± 1.10||1.30 ± 1.11||1.25 ± 1.03||1.25 ± 1.0||.15|
|Cold Ischemia Time (h)||9.8 ± 5.1||9.7 ± 5.0||9.5 ± 4.9||9.6 ± 5.6||9.6 ± 4.9||.01|
|Donor age (yr)||35.4 ± 16.9||35.6 ± 16.9||35.4 ± 16.8||35.6 ± 16.8||36.0 ± 17.1||.48|
Survival by Neighborhood Income
When patients were stratified into groups by median neighborhood income, Kaplan-Meier survival analysis showed similar 5-year patient survival: <$30,000 – (74.2%), $30,001-$40,000 (73.8%), $40,001-$50,000 (75.7%), $50,001-60,000 – (76.7%), >$60,000 –(75.7%); P = .58—and 5-year graft survival—<$30,000 –(66.0%), $30,001-$40,000 –(66.5%), $40,001-$50,000 – (66.0%), $50,001-$60,000 –(66.6%), >$60,000 (66.4%); P = .91).
Cox regression analysis for patient and graft survivals, expressed as hazard ratio (HR), 95% CI, and P value, is shown in Table 2. Cox regression plots for patient and graft survival, after adjusting for confounding variables, are shown in Figure 3A. Again, this confirmed that there was no significant trend in graft and patient survival based on neighborhood income (Figure 3A). In addition, there were no differences in patient survival when patients were stratified by race and neighborhood income (Figure 3B).
|Patient||1 Year||P||2 Year||P||5 Year||P|
|<$30,000 (n = 1,521)||1.14 (0.91–1.42)||0.3||1.19 (0.97–1.45)||0.9||1.11 (0.93–1.33)||0.3|
|$30,001–40,000 (n = 3,412)||1.23 (1.02–1.48)||0.03||1.25 (1.06–1.49)||0.01||1.16 (1.00–1.35)||0.05|
|$40,001–50,000 (n = 2,958)||1.08 (0.69–1.32)||0.4||1.11 (0.93–1.32)||0.3||1.04 (0.69–1.22)||0.6|
|$50,001–60,000 (n = 1,632)||1.10 (0.89–1.37)||0.4||1.15 (0.95–1.40)||0.2||1.07 (0.69–1.27)||0.5|
|>$60,000 (n = 1,667)||Reference variable [indicator]|
|Medicaid (n = 1,523)||1.10 (0.92–1.31)||0.3||1.14 (0.97–1.33)||0.1||1.10 (0.96–1.27)||0.2|
|Medicare (n = 1,537)||1.13 (0.95–1.33)||0.13||1.18 (1.02–1.37)||0.02||1.21 (1.06–1.38)||0.01|
|Government (n = 399)||1.12 (0.82–1.52)||0.5||1.04 (0.78–1.39)||0.8||0.98 (0.75–1.28)||0.9|
|HMO/PPO (n = 738)||0.80 (0.60–1.08)||0.1||0.89 (0.68–1.17)||0.4||0.93 (0.72–1.21)||0.6|
|Private (n = 6,993)||Reference variable [indicator]|
|<Grade 9 (n = 576)||0.98 (0.74–1.31)||0.9||0.92 (0.71–1.20)||0.5||1.04 (0.83–1.31)||0.7|
|Grades 9–12 (n = 5,333)||1.11 (0.95–1.30)||0.2||1.16 (1.01–1.33)||0.04||1.19 (1.04–1.35)||0.01|
|College or technical school (n = 2,878)||0.92 (0.77–1.10)||0.3||0.99 (0.84–1.16)||0.9||0.99 (0.86–1.15)||0.9|
|Bachelor's degree/s (n = 2,403)||Reference variable [indicator]|
|<$30,000 (n = 1,521)||1.06 (0.66–1.27)||0.5||1.01 (0.91–1.27)||0.4||1.02 (0.88–1.18)||0.8|
|$30,001–40,000 (n = 3,412)||1.12 (0.97–1.31)||0.1||1.13 (0.98–1.30)||0.08||1.07 (0.95–1.20)||0.3|
|$40,001–50,000 (n = 2,958)||1.10 (0.94–1.28)||0.2||1.16 (0.96–1.27)||0.2||1.05 (0.93–1.19)||0.5|
|$50,001–60,000 (n = 1,632)||1.16 (0.98–1.37)||0.09||1.17 (1.00–1.37)||0.05||1.08 (0.94–1.25)||0.3|
|>$60,000 (n = 1,667)||Reference variable [indicator]|
|Medicaid (n = 1,523)||1.16 (1.01–1.33)||0.08||1.18 (1.04–1.34)||0.01||1.14 (0.96–1.28)||0.03|
|Medicare (n = 1,537)||1.17 (1.02–1.34)||0.02||1.18 (1.04–1.37)||0.01||1.19 (1.06–1.33)||0.002|
|Government (n = 399)||1.12 (0.87–1.43)||0.4||1.07 (0.84–1.35)||0.6||0.98 (0.79–1.23)||0.9|
|HMO/PPO (n = 738)||0.84 (0.67–1.05)||0.1||0.88 (0.71–1.09)||0.3||0.93 (0.73–1.11)||0.3|
|Private (n = 6,993)||Reference variable [indicator]|
|<Grade 9 (n = 576)||0.95 (0.76–1.20)||0.7||0.93 (0.75–1.15)||0.5||1.0 (0.82–1.22)||0.98|
|Grade 9–12 (n = 5,333)||1.02 (0.90–1.15)||0.8||1.06 (0.95–1.19)||0.3||1.09 (0.98–1.21)||0.1|
|College or technical school (n = 2,678)||1.01 (1.01–1.01)||0.5||1.01 (0.89–1.19)||0.9||1.01 (0.90–1.13)||0.9|
|>Bachelor's degree (n = 2,403)||Reference variable [indicator]|
Survival by Education Level
There was a positive trend in 5-year patient survival—higher than bachelor's degree (77.9%), college attendance or technical school– (77.6%), high school education (grades 9-12; 73.1%), less than high school education (70.8%); P < .01—and 5-year graft survival—higher than bachelor's degree (67.9%), college attendance or technical school –(69.2%), high school education (grades 9-12;65.3%), less than high school education –(64.9%); P = .02)—in those with better education.
On Cox regression analysis, high school (grades 9-12) education (Table 2) was a predictor of poor 2-year patient survival (HR 1.16; 95% CI, 1.01-1.33; P = .04] and poor 5-year patient survival (HR 1.19; 95% CI, 1.04-1.35; P = .01). Cox regression plots showed that the impact of education on patient survival was only marginal; however, graft survival was similar in all 4 groups (Figure 4).
Survival by Insurance (Payer) Status
Patients with HMO/PPO and Medicare coverage showed lower 5-year patient survival—Medicaid –(74.5%), Medicare –(67.3%,) government agency –(78.6%), HMO/PPO –(54.0%), private insurance –(76.6%); P < .01—and 5-year graft survival—Medicaid –(64.3%), Medicare –(61.1%), government agency –(68.6%), HMO/PPO –(62.0%), private insurance –(67.8%); P < .01— when compared to patients with private insurance.
When adjusted for other variables (Cox regression), only Medicare was associated with poor patient survival at 2 years (HR 1.18; 95% CI, 1.02-1.37; P = .02] and 5 years (HR 1.21; 95% CI, 1.06-1.38 P = .01). However, both Medicare and Medicaid were predictors of poor graft survival at 1 year, 2 –years, and 5 years (Table 2). Cox regression plots based on insurance status are shown in Figure 5.
Cox regression analysis showed that African-American race was associated with a low 5-year graft survival (HR 1.3; 95% CI, 1.1-1.5; P = .0003) and 5-year patient survival (HR 1.4; 95% CI, 1.2-1.6; P = .0002) when compared to white Americans after adjusting for age, sex, diagnosis, UNOS status, ABO matching, BMI, creatinine, neighborhood income, payer status, and education.
In this study, we showed that neighborhood income had no effect and education had minimal effect on posttransplant survival of adult liver transplant recipients in the United States. Insurance had moderate effect; Medicare and Medicaid were associated with relatively lower survival. However, it was reassuring to note that patients with HMO/PPO had a similar posttransplant outcome to patients with private insurances that offered the most comprehensive coverage. It is important to note that even though HMO/PPO had lowest patient and graft survival on the Kaplan-Meier survival analysis, survival was similar to that of private insurance when adjusted for confounding variables. This observation also underscores the importance of adjusting survival or outcome for other confounding factors, especially when studying a heterogeneous patient population.
There are many ways to determine SES in addition to those used in this analysis. We used zip code as a surrogate marker to determine median income. Income defined in this way does not in fact represent the true income of the individual, but it is a good reflection of the neighborhood income. It has been suggested that correlation between aggregate income obtained from zip codes (“imputed income”) and individual income may be poor, and a larger sample size is often required to show any statistical differences in survival.20–24 We believe that our sample size was sufficient to show existingdifferences in survival.18 Although individual income would have been ideal, it is hard to find large data sets with that information. Because posttransplant survival is dependent on many factors, it is important to have a very large sample size in order to adjust survival for many confounding factors, hence the use of UNOS data. Many previous outcome studies have shown a relationship between zip-code income and health outcome.14–19 It is likely that, in the United States, health care access and quality of health care delivery are dependent on neighborhood income rather than on individual income. As a result, we have not shown a difference in survival using zip codes as a surrogate marker of neighborhood income. To our knowledge, no previous studies determined posttransplant survival based on zip codes. It is possible that most transplant recipients are managed in transplant centers located in mostly urban, tertiary care centers; this may explain why neighborhood income had no impact on survival, a finding unlike that found in many other medical conditions.
One of the weaknesses of our study was missing data on education in many patients. Despite these missing data, we believe that we had a large enough sample size to determine the influence of education on post-transplant survival. In our study, patients with high school education (grades 9-12) had relatively poor patient survival. This is hard to explain because patients with education below grade 9 had an outcome similar to those with better education (higher than a bachelor's degree). Since there was no real trend, we believe that education had only minimal impact on survival. Combined results from analysis of zip-code income and education indicate that SES does not significantly influence posttransplant outcome. This again was a very reassuring observation.
An interesting observation was poor outcome in patients with Medicare and Medicaid as insurance. Medicare did not provide coverage for nonimmunosuppressive medications. Similarly, Medicaid coverage is highly variable from state to state. Whether inability to obtain medication was a factor in poor survival can only be determined by a prospective study. Another possibility, provocative as it may appear, is that there may be differences in health care access and delivery in recipients with Medicare and Medicaid because of poor insurance reimbursement. The lower survival in Medicare recipients is not related to older age, since survival was adjusted for age and other age-related comorbidity, such as serum creatinine. Although Medicaid may be a surrogate marker for lower income, Medicare is not usually provided on the basis of income. We believe that our observations merit further examination in prospective studies.
In other medical conditions, it has been shown that Medicaid patients are less likely to receive optimal care and more likely to have a poor outcome for both common medical conditions (e.g., asthma, myocardial infarction, and trauma) and rare medical conditions (e.g., cystic fibrosis) compared to privately insured patients.25–30 A large study involving 537,283 patients with acute myocardial infarction showed that Medicaid recipients and African were less likely to be transferred to a tertiary care center after admission.28 With regard to Medicare recipients, it has been shown in a cohort of patients with previous myocardial infarction that patients with supplemental insurance for prescription drug coverage were 6 times more likely to use 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) than patients without supplemental insurance.30 In a critical review, Adams et al. suggested that Medicare recipients are more likely to use essential medication if they have insurance coverage for drugs.31 On the basis of published studies and the present study, it is reasonable to suggest that liver transplant recipients who do not have adequate insurance benefits for drug coverage should be provided all essential medications by the state or federal government. This might reduce the survival differences among different groups. However, it is important to emphasize that we do not have any hard evidence to suggest that poor outcome in Medicare patients was related to the inability to obtain medications.
The stimulus for our current study was the observation made in one of our previous studies that African-American race was an independent predictor of poor survival after liver transplantation. In the present study, we examined whether neighborhood income influenced survival in different races and we found that, as in the entire cohor, it was not a predictive factor. When income, education, and insurance were included in multivariate analysis, race was still a strong independent predictor of survival. It is therefore important to note that survival difference in African-Americans may be due to other hitherto unrecognized factors that need to be studied prospectively.
The authors thank UNOS for providing the data.