State Medicaid Coverage, ESRD Incidence, and Access to Care

Abstract

The proportion of low-income nonelderly adults covered by Medicaid varies widely by state. We sought to determine whether broader state Medicaid coverage, defined as the proportion of each state’s low-income nonelderly adult population covered by Medicaid, associates with lower state-level incidence of ESRD and greater access to care. The main outcomes were incidence of ESRD and five indicators of access to care. We identified 408,535 adults aged 20–64 years, who developed ESRD between January 1, 2001, and December 31, 2008. Medicaid coverage among low-income nonelderly adults ranged from 12.2% to 66.0% (median 32.5%). For each additional 10% of the low-income nonelderly population covered by Medicaid, there was a 1.8% (95% confidence interval, 1.0% to 2.6%) decrease in ESRD incidence. Among nonelderly adults with ESRD, gaps in access to care between those with private insurance and those with Medicaid were narrower in states with broader coverage. For a 50-year-old white woman, the access gap to the kidney transplant waiting list between Medicaid and private insurance decreased by 7.7 percentage points in high (>45%) versus low (<25%) Medicaid coverage states. Similarly, the access gap to transplantation decreased by 4.0 percentage points and the access gap to peritoneal dialysis decreased by 3.8 percentage points in high Medicaid coverage states. In conclusion, states with broader Medicaid coverage had a lower incidence of ESRD and smaller insurance-related access gaps.

Medicaid is a state and federal partnership designed to provide health insurance for low-income Americans. States have flexibility within federal guidelines to design Medicaid programs, including income eligibility levels. A consequence of this flexibility is that states vary in the proportion of low-income individuals they cover.^1–3 There is limited information about the implications of these variations on access to care. Previous studies have included only a subset of states, several relied on self-report to ascertain access to care, and most focused on children’s health or primary care services.^1–4 Chronic disease care is a major source of rising health care expenditures, and access to care for uninsured individuals with a chronic disease has eroded over the last decade.⁵ Currently, low-income nonelderly adults who are pregnant, have a disability, or are parents of minors are eligible for Medicaid. Beginning in 2014, the Affordable Care Act will expand Medicaid coverage to adults with incomes<133% of the federal poverty level,⁶ although not all states are expected to participate in this expansion.

The care of patients approaching ESRD may be a useful model to study the potential effect of Medicaid expansion on chronic disease care. ESRD affects >350,000 nonelderly Americans at a cost of $10 billion per year.⁷ When prevention efforts fail, preparation for ESRD is recommended as a means of promoting cost-effective treatment strategies, including kidney transplantation, home dialysis therapies, and permanent hemodialysis vascular access.^8–10 Although all Americans can qualify for Medicare coverage after the onset of ESRD, those aged<65 years must rely on other sources of insurance or pay out of pocket to cover pre-ESRD care. Risk factors for progressing from CKD to ESRD cluster among poor and uninsured Americans.¹¹ As a consequence, those most likely to benefit from timely access to chronic disease care are often least likely to receive it, a dilemma recently underscored by an expert panel convened by the US Centers for Medicare and Medicaid Services.¹²

Using national data, we sought to determine the relation between the extent of state Medicaid coverage and access to care among nonelderly adults approaching ESRD and whether that relationship differed based on one’s insurance status. We hypothesized that states with broader Medicaid coverage would have a lower incidence of ESRD. We also hypothesized that broader state Medicaid coverage would narrow access gaps between privately insured adults with ESRD and those covered by Medicaid or the uninsured.

Results

State Demographic and Insurance Characteristics

The average Medicaid coverage among low-income nonelderly adults between 2000 and 2007 was 32.5% (range, 12.2%–66.0%) (Supplemental Figure 1). There was more variation among states (SD_among=10.4%) than within states over time (SD_within=3.4%). States with a lower extent of Medicaid coverage had a similar age, sex, race, employment, and obesity distribution compared with states with broader Medicaid coverage, but a higher percentage of low-income adults and a lower percentage of rural adults (Table 1).

Table 1.

State socioeconomic characteristics by extent of Medicaid coverage among adults aged 20–64 years

State Characteristic	Medicaid Coverage in 2000
	<25.0%	25.0%–33.3%	≥33.3%
States (n)	10	19	19
Nonelderly population in millions	2.4 (1.0–4.4)	2.6 (1.5–4.9)	2.4 (0.8–3.8)
Age of nonelderly adults (yr)	40.4 (39.8–40.6)	40.3 (40.1–40.8)	40.7 (40.5–41.2)
Age group (yr)
20–44	62.1	62.3	60.9
45–64	37.9	37.7	39.1
Men	50.3	49.7	49.5
Race
White	87.0	80.9	89.8
Black	7.2	14.6	7.4
Other	5.8	4.6	2.8
Uninsured	43.2	35.5	29.5
Low income	30.0	29.9	27.0
Below poverty level	11.6	11.3	10.1
Unemployed	3.5	3.7	3.4
Rural residence	21.9	29.1	31.6
Obese	18.7	20.8	20.1

Data are presented as a percentage or median (interquartile range) unless otherwise indicated.

State Medicaid Coverage, ESRD Incidence, and Access to Care

The median incidence rate of ESRD was 265.8 cases per million persons. A 10 percentage points higher level of state Medicaid coverage was associated with a 1.8% decrease in ESRD incidence (Table 2). The findings were consistent by age, but varied slightly by race and sex. A 10 percentage points higher level of state Medicaid coverage was associated with significantly higher levels of early nephrologist care, waitlisting within 1 year of ESRD, transplantation within 1 year of ESRD, and permanent vascular access, but not with higher levels of peritoneal dialysis utilization (Supplemental Table 1).

Table 2.

Estimated difference in ESRD incidence associated with a 10 percentage points higher level of state Medicaid coverage among low-income nonelderly adults

Characteristic	Median ESRD Incidence Rate per Million Population (Interquartile Range)	Estimated Change for Each 10 Percentage Points Higher Level of Medicaid Coverage (95% Confidence Interval)
Overall	265.8 (188.8, 299.7)	−1.8% (−2.6% to −1.0%)
Age (yr)
20–44	109.8 (86.5, 137.2)	−1.8% (−3.3% to −0.3%)
45–64	510.4 (346.4, 570.3)	−1.8% (−3.4% to −0.1%)
Race
White	177.6 (157.2, 199.3)	−2.2% (−4.8% to 0.4%)
Black	853.1 (647.2, 908.3)	−0.2% (−1.9% to 1.5%)
Other	292.2 (229.2, 431.1)	0.7% (−1.5% to 2.9%)
Sex
Men	304.0 (225.3, 348.9)	−0.7% (−1.9% to 1.0%)
Women	226.1 (159.4, 266.3)	−3.1% (−4.6% to −1.6%)

Models are adjusted for state age, sex, and race distribution, in addition to fixed effects for calendar year and state.

State Medicaid Coverage, Insurance Status, and Access to Care

Table 3 shows the clinical characteristics of the ESRD patient cohort according to insurance status. Patients with Medicaid were older than the uninsured and slightly younger than patients with private insurance. There was a higher proportion of women among those with Medicaid versus privately insured or uninsured patients. There was a higher proportion of nonwhite patients among those with Medicaid or no insurance compared with those with private insurance. Patients with Medicaid also had a higher prevalence of most comorbidities compared with the uninsured and privately insured patients.

Table 3.

Characteristics of patients aged 20–64 years who developed ESRD between 2001 and 2008, stratified by insurance status

Characteristic	All (N=310,331)	Insurance Status
		Uninsured (n=57,430)	Medicaid (n=117,825)	Private (n=135,076)
Age (yr)
20–44	28.5	37.7	28.4	24.6
44–64	71.5	62.3	71.6	75.4
Mean±SD	49.6±10.6	47.0±11.3	49.7±10.7	50.7±10.1
Men	55.7	63.1	48.2	59.0
Race
White	56.3	49.1	50.5	64.4
Black	37.5	44.3	43.0	29.8
Other	6.2	6.5	6.4	5.8
Body mass index (kg/m2)
<18.5	3.6	4.0	4.5	2.6
18.5–24.9	37.9	31.0	38.3	40.4
25.0–29.9	26.9	27.9	24.7	28.5
≥30	30.1	35.7	30.7	27.2
Hypertension	82.0	82.6	82.0	81.7
Diabetes	50.7	42.4	58.5	47.4
Heart failure	22.8	20.5	28.5	18.8
Ischemic heart disease	14.1	9.6	16.4	14.1
Peripheral vascular disease	9.3	5.8	12.2	8.3
Cerebrovascular disease	6.4	4.6	8.9	4.9
Chronic lung disease	4.8	2.9	6.9	3.8
Cancer	3.7	2.2	3.0	4.9
Tobacco use	8.1	9.9	9.9	5.6
Alcohol dependence	2.3	3.5	3.0	1.2
Drug dependence	2.5	4.6	3.8	0.5
Unable to ambulate	3.7	1.9	6.5	2.0
Year
2001	11.1	11.3	10.8	11.2
2002	11.3	11.2	11.3	11.4
2003	11.7	12.1	12.0	11.2
2004	12.1	12.9	12.5	11.4
2005	12.9	12.7	13.0	12.9
2006	13.7	13.0	13.5	14.2
2007	13.5	13.2	13.3	13.8
2008	13.7	13.5	13.6	13.8

All categorical data represent column percentages.

Patients with Medicaid or no insurance were less likely to receive early nephrologist care, to be waitlisted within 1 year of ESRD, to receive a transplant within 1 year of ESRD, to utilize peritoneal dialysis at ESRD onset, and to have a permanent vascular access at ESRD onset compared with those who were privately insured (Figure 1).

Figure 1.

Rates of access to care among privately insured, Medicaid, and uninsured nonelderly adults between 2001 and 2008. AVF, arteriovenous fistula.

Gaps in access to care were narrower in states with broader Medicaid coverage (Figure 2). Specifically, placement on the transplant waiting list, transplantation, and use of peritoneal dialysis were all higher among individuals with Medicaid residing in states with broader Medicaid coverage (P values for interaction<0.01). Likewise, among the uninsured, transplantation and placement of permanent vascular access were higher in states with broader Medicaid coverage (P values for interaction<0.01). Paradoxically, the use of peritoneal dialysis was lower among uninsured adults at higher levels of state Medicaid coverage (P values for interaction<0.01). There was no difference in early nephrologist care at higher levels of state Medicaid coverage among patients with Medicaid or among the uninsured (P values for interaction>0.2). Similar findings were seen in sensitivity analyses substituting preemptive waitlisting and transplantation for outcomes at 1 year after the onset of ESRD.

Figure 2.

Adjusted odds ratios for indicators of access to care among Medicaid and uninsured nonelderly adults (referent group is privately insured) at varying levels of state Medicaid coverage. Models are adjusted for age, sex, race, body mass index, hypertension, diabetes, heart failure, ischemic heart disease, peripheral vascular disease, cerebrovascular disease, chronic lung disease, cancer, tobacco use, alcohol dependence, drug dependence, and inability to ambulate, in addition to fixed effects for calendar year and state. Note: P value for interaction of state Medicaid coverage×Medicaid insurance<0.01 for all outcomes except early nephrology care (P=0.2) and permanent vascular access (P=0.5). P value for interaction of state Medicaid coverage×uninsured<0.01 for all outcomes except nephrology care (P=0.9) and waitlisting within 1 year of ESRD (P=0.5).

To understand what these differences might mean for patients, we estimated the probability of each access to care outcome for a hypothetical 50-year-old hypertensive white woman with Medicaid or no insurance compared with a similar privately insured patient in three states with low, intermediate, and high levels of Medicaid coverage (Figure 3, Supplemental Table 2). For this hypothetical adult with Medicaid, high state Medicaid coverage was associated with a 7.7 percentage points smaller access gap for placement on the kidney transplant waiting list, a 4.0 percentage points smaller access gap for transplantation, and a 3.8 percentage points smaller access gap for utilization of peritoneal dialysis, whereas the access gap was narrowed by <1.0 percentage points for other indicators. For an uninsured adult, high state Medicaid coverage was associated with a 4.0 percentage points smaller access gap for transplantation and a 2.7 percentage points smaller access gap for permanent vascular access.

Figure 3.

Difference in access to care, or access gap, between private insurance and Medicaid or uninsured in three states with low, intermediate, and high rates of Medicaid coverage. Access gaps are based on predicted probabilities of access to care for a 50-year-old hypertensive white woman with ESRD in 2005. The average levels of Medicaid coverage for high, medium, and low coverage states are 58.6%, 31.3%, and 18.7%, respectively. WL, waitlist within 1 year of ESRD; Tx, transplant within 1 year of ESRD; PD, peritoneal dialysis; Neph, nephrology care for 12 months before ESRD; AVFG, placement of arteriovenous fistula or graft before ESRD.

Discussion

This study has three key findings. First, broader Medicaid coverage among low-income nonelderly adults was associated with a lower incidence of ESRD. Second, gaps in access to pre-ESRD care between privately insured adults and those with Medicaid were smaller in states with broader Medicaid coverage. Third, broader Medicaid coverage was associated with some spillover benefits for uninsured adults with ESRD, but these were small and not consistent across indicators of access to care.

High-quality care for patients with CKD requires a health care system capable of delivering sustained and coordinated care. Components of this care include disease education, evidence-based treatment adapted for the individual, self-care resources, and a multidisciplinary network of providers working in coordination with the primary care provider as the disease progresses.^13–15 Variations in the incidence of ESRD and quality of care at the regional level^16–18 raise the possibility that limited access to care at earlier disease stages may underlie variations in outcomes. Although nearly all Americans can receive public health insurance after the onset of ESRD, an estimated 10% of adults with nondialysis-dependent CKD are uninsured.¹⁹ These individuals are less likely to receive interventions to slow progression of CKD, including treatment of hypertension or use of medications to block the renin-angiotensin system.^19,20 They may also be less likely to receive subspecialty care as CKD progresses. This could result in lower access to interventions such as kidney transplantation, home dialysis, and permanent hemodialysis vascular access, with the costs of failing to provide timely care shifted to Medicare. Removal of financial and regulatory barriers to care is seen as a key step to improve ESRD outcomes based on extrapolation from universal healthcare systems,¹² but evidence that this approach would be effective in the US healthcare system has been lacking.

The findings of this study suggest that Medicaid expansion may address several objectives of the Healthy People 2020 CKD Initiative, including a reduction in the burden of ESRD. Our findings are consistent with some recent reports demonstrating lower rates of adult mortality and delayed care in states expanding Medicaid coverage, and improvements in mental health among those newly enrolled in Medicaid,^21,22 although one study found that new Medicaid enrollees do not experience better control of diabetes and hypertension in the first 2 years after enrollment.²³ In addition, we found that gaps in access to pre-ESRD care between privately insured adults and those with Medicaid were tangibly smaller in states with broader Medicaid coverage. These effects were most pronounced for access to the kidney transplant waiting list, early transplantation, and use of peritoneal dialysis, in which access gaps were reduced by 19%–29%. This finding is consistent with earlier studies demonstrating that broader Medicaid coverage is associated with improved access to primary and preventive care services.^3,24

Barriers to health care access are commonly grouped into five categories: affordability, availability, accessibility, accommodation, and acceptability.²⁵ Broad Medicaid coverage of low-income adults may address several of these barriers to access. First, broad Medicaid coverage addresses the affordability of medical care by providing more low-income adults with health insurance. However, if this was the only benefit of broad Medicaid coverage, it would not explain why we observed smaller access gaps with broader coverage for patients enrolled in Medicaid at the start of ESRD. In addition to affordability, broad Medicaid coverage may address the availability of medical care if it encourages more physicians, and subspecialists in the case of ESRD, to accept Medicaid patients. Although we found no difference in the percentage of patients seen by a nephrologist before ESRD at the individual level, there may have been differences in the frequency of nephrology care or in referrals to surgeons and transplant physicians in states with broader coverage. Broader coverage may also address accessibility and accommodation barriers by reducing the amount of uncompensated care provided by safety-net institutions, thereby allowing them to provide more comprehensive services to low-income patients.²⁴

Conversely, it is possible that states might fund Medicaid expansions by reducing direct support to safety-net institutions. In this case, one might expect to see reduced access to care for uninsured individuals in states with broader Medicaid coverage. Reassuringly, we found a small positive, rather than negative, spillover effect on access to care for uninsured adults with ESRD, with the exception of access to peritoneal dialysis. The reasons for this are not entirely clear, but may relate to selection of appropriate patients out of the uninsured pool and into the Medicaid pool, different financial incentives to support peritoneal dialysis among Medicaid enrollees versus the uninsured, or cycling between Medicaid and uninsured states.²⁶

Access gaps for several outcomes, while attenuated at higher levels of Medicaid coverage, still remained large, indicating that insurance coverage is important but not sufficient for eliminating treatment disparities. Individuals with Medicaid may remain underinsured due to limited benefit packages and low physician payments relative to Medicare or privately insured patients. Cultural and communication barriers may also impede access to care. Poor awareness of CKD could contribute to episodic care, rather than sustained care.²⁷

Our study has several strengths. First, we included 48 states and captured nearly all nonelderly adults with ESRD; thus, our findings should be broadly generalizable. Second, we controlled for a large number of measured individual patient characteristics as well as unmeasured state characteristics by using state fixed effects in all models. Third, we assessed clinically relevant and validated measures of access to pre-ESRD care.

There are also several limitations. Ideally we would want to posit the effect of an individual state increasing its extent of Medicaid coverage. However, because of the fact that there was less variability within a state than between states, this is more challenging. Instead our results are most accurate for estimating the effect of an individual staying in the same state and switching insurance coverage or an individual moving to a new state. To infer the effect of an individual state increasing its degree of Medicaid coverage involves more extrapolation. We did not account for differences in Medicaid benefit plans or Medicaid spending. States with broader Medicaid coverage generally have more generous benefit plans (i.e., higher physician reimbursement, lower copays).²⁸ Spending variations reflect differences in the case-mix and the cost of health care that are difficult to entangle from health care utilization, and this is why we chose to focus on Medicaid coverage. We had insurance information only at the onset of ESRD, and this may not accurately reflect insurance status in the preceding years.

In conclusion, expansion of Medicaid among low-income nonelderly adults could support ESRD prevention efforts and reduce access gaps in pre-ESRD care, underscoring the key role Medicaid plays in delivering chronic disease care.

Concise Methods

Cohort

We identified all adults in the continental United States who developed ESRD between January 1, 2001, and December 31, 2008, and were between 20 and 64 years of age (n=408,535) using data from the US Renal Data System (USRDS), a national ESRD registry.⁷

Predictors: State Medicaid Coverage and Patient Insurance Status

The two primary predictors of interest were state Medicaid coverage, defined as the percentage of low-income nonelderly adults covered by Medicaid in each state, and the insurance status of patients who developed ESRD. We obtained health insurance coverage information for each state from data compiled by the Kaiser Commission on Medicaid and the Uninsured during 2000–2007.^29–34 The percentage of state Medicaid coverage was calculated for each state and each year as the number of low-income nonelderly adult Medicaid enrollees divided by the number of low-income nonelderly adults who are enrolled in Medicaid or uninsured, using the methods of Weissman et al.³ Because low-income, nonelderly adults with insurance from sources other than Medicaid have access to health care regardless of the state’s Medicaid eligibility rules, they were not included in the denominator for the state Medicaid coverage calculation. Low-income was defined as income<200% of the federal poverty level.

Insurance status at the onset of ESRD was ascertained from the USRDS Medical Evidence Report. Insurance status was categorized as Medicaid, private insurance, or uninsured. For the analyses of access to care, individuals were excluded if insurance status was solely Medicare, Veterans Affairs, other, or missing. Individuals with multiple types of insurance were included in the analysis and categorized according to the presence of Medicaid or private insurance.

Outcomes: ESRD Incidence and Access to Care

On the basis of key objectives of the Healthy People 2020 campaign for CKD,³⁵ we defined two main outcomes of interest: ESRD incidence and access to care. We measured five indicators of access to care: (1) nephrology care at least 12 months before ESRD, (2) placement on the kidney transplant waiting list within 1 year of ESRD, (3) kidney transplantation within 1 year of ESRD, (4) use of peritoneal dialysis (versus hemodialysis) at the onset of ESRD, and (5) presence of a permanent vascular access for hemodialysis (versus central venous catheter) at the onset of ESRD. In sensitivity analyses, we also considered placement on the transplant waiting list and transplantation before ESRD, rather than within 1 year of ESRD, as outcomes.

We determined waitlist and transplantation rates using information from the respective USRDS analytic files. We determined use of peritoneal dialysis (versus hemodialysis) among individuals who did not receive a preemptive kidney transplant. Early nephrology care and the presence of a permanent vascular access at the onset of ESRD were ascertained from the Medical Evidence Report starting in 2005. Compared with Medicare claims, the accuracy of these variables is 70% and 96%, respectively.^36,37 We defined permanent vascular access as the presence of an arteriovenous fistula or graft (versus catheter) at the onset of ESRD, and we evaluated this outcome only among patients who initiated hemodialysis (sample size for each outcome shown in Supplemental Figure 2).

Other Variables

We ascertained both state- and individual-level covariates. We obtained state age, sex, race, income, employment, and residence data from US Census files. We obtained state obesity data as a proxy for each state’s health status from the US Centers for Disease Control and Prevention Behavioral Risk Factor Surveillance System.³⁸ We ascertained demographic and clinical information for individuals with ESRD from the USRDS including age, race, body mass index, and comorbidities.

Statistical Analyses

We calculated the median ESRD incident rate for different demographic groups. We then used negative binomial regression to model differences in ESRD incidence based on state Medicaid coverage. The model was adjusted for state age, sex, and race distribution, and contained a fixed effect for calendar year and each state to account for correlation of individuals within states. We used a similar approach to model access to care at the state level.

We then examined whether the extent of state Medicaid coverage had different associations with each of indicators of access to care based on individual insurance status. From the primary cohort, we identified 310,331 individuals with Medicaid, private insurance, or no insurance coverage. The primary parameter of interest was the interaction between state Medicaid coverage and individual insurance status. We used multivariable logistic regression models adjusted for available individual-level covariates. We included a fixed effect for each state to control for the heterogeneity in state characteristics. An interaction term between state percent Medicaid coverage and individual insurance status was estimated for each model. Both state- and individual-level models incorporated a lag effect in which a given year of Medicaid coverage corresponded to outcomes in the following year.

Finally, we generated predicted probabilities for each of the outcomes based on a representative person to provide additional perspective on the relation between state Medicaid coverage and insurance status. Three states with similar state fixed effects and respectively low (<25%), intermediate (25%–33%), and high (>33%) Medicaid coverage were chosen and the hypothetical person’s insurance status was varied. From these probabilities, we estimated the access gap, the difference in access to care between individuals with private insurance and those with Medicaid or no insurance, at different levels of state Medicaid coverage. For each of the models, 95% confidence intervals were calculated using robust SEMs. Analyses were conducted in SAS 9.3 (SAS Institute, Cary, NC) and R 2.14 (The R Project for Statistical Computing, Vienna, Austria) software.

Disclosures

M.K.T. has served as a scientific advisor to Amgen and has received an honoraria from Satellite Healthcare. Y.N.H. has received research funding from Satellite Healthcare through the Norman S. Coplon Extramural Grant Program. W.C.W. has served as a scientific advisor to or on data safety monitoring boards for Amgen, Bayer, GlaxoSmithKline, Keryx, Medgenics, and Medtronic.