|Home | About | Journals | Submit | Contact Us | Français|
To evaluate the impact of Medicaid managed care organizations (MCO) on health care access for adults with disabilities (AWDs).
Mandatory and voluntary enrollment data for AWDs in Medicaid MCOs in each county were merged with the Medical Expenditure Panel Survey and the Area Resource File for 1996–2004.
I use logit regression and two evaluation perspectives to compare access and preventive care for AWDs in Medicaid MCOs with FFS. From the state's perspective, I compare AWDs in counties with mandatory, voluntary, and no MCOs. From the enrollee's perspective, I compare AWDs who must enroll in an MCO or FFS to those who may choose between them.
Mandatory MCO enrollees are 24.9 percent more likely to wait >30 minutes to see a provider, 32 percent more likely to report a problem accessing a specialist, and 10 percent less likely to receive a flu shot within the past year. These differences persist from the state evaluation perspective.
States should not expect a dramatic change in health care access when they implement Medicaid MCOs to deliver care to the adult disabled population. However, continued attention to specialty care access is warranted for mandatory MCO enrollees.
After more than a decade of experimentation with Medicaid managed care (MMC) for adults with disabilities (AWD), there is little evidence about how this policy change influences beneficiaries' access to health care (Ireys, Thornton, and McKay 2002). Yet the health and quality of life of persons with disabilities is particularly sensitive to the accessibility of their health care (Iezzoni 2002; Lawthers et al. 2003; U.S. Department of Health and Human Services 2005; Iezzoni and O'Day 2006;). While the relative effects of MMC on care access for nondisabled adults have been well studied (Zuckerman, Brennan, and Yemane 2002; Garrett, Davidoff, and Yemane 2003; Garrett and Zuckerman 2005; Kaestner, Dubay, and Kenney 2005; Le Cook 2007;), scholars caution against generalizing from such research to a population with a substantially different health profile (Rowland et al. 1995; Sisk et al. 1996; Currie and Fahr 2005;). Recognizing this gap between research and practice, Medicaid programs and the research community are building an evidence base to inform decisions about how best to care for this population (Landon et al. 2004; California Department of Health Services 2005; Volpel, O'Brien, and Weiner 2005; Center for Health Care Strategies Inc. 2006;).
This study contributes to that effort by assessing health care access and preventive care use for AWDs in Medicaid managed care organizations (MCOs) relative to fee-for-service (FFS). I apply two evaluation strategies to a nationally representative sample in an effort to reconcile the extant population-specific findings (Lo Sasso and Freund 2000; Coughlin, Long, and Graves 2009;). First, I assess the effect of being enrolled in an MCO relative to FFS, an evaluation perspective that may be most relevant to beneficiaries, advocacy groups, and to the program staff who monitor health plan performance (Highsmith and Somers 2003). The second evaluation strategy adopts the state's perspective and assesses the effect of MCO implementation on the total eligible population, including beneficiaries who opt out of MCOs or choose FFS where it is an option. This perspective may be particularly relevant to policy makers within, and outside of, Medicaid programs because it captures the overall impact of this programmatic change, including any potential spillover effects (Currie and Fahr 2005).
In response to the population's disproportionate impact on the Medicaid budget, states have expanded MMC programs to include AWDs (United States General Accounting Office 1996; Congressional Budget Office 2006;). By 2004, MMC was available for AWDs in 66 percent of U.S. counties, up from 43 percent in 1996 (Burns 2008). While Medicaid regulations support a variety of health plan types, Medicaid MCOs have been a popular choice among states for their beneficiaries, with disabilities also growing in prevalence from 14 percent of counties in 1996 to 25 percent in 2004 (Burns 2008).
However, the plan characteristics that make Medicaid MCOs popular—defined provider networks, comprehensive services, and capitated financing—have also led to conflicting expectations about their effects on care access for AWDs (Tanenbaum and Hurley 1995; United States General Accounting Office 1996, 2004; Fox et al. 1997; Regenstein 2000; Bachman, Drainoni, and Tobias 2004). For example, capitated financing, combined with imperfect risk adjustment, may provide MCOs with an incentive to avoid the most costly patients, or to underserve enrollees (Kronick et al. 1996; United States General Accounting Office 1996, 2004; Meyers, Glover, and Master 1997). Alternatively, that same incentive may encourage MCOs to monitor health more aggressively, facilitate the use of services appropriate to patient needs, and avoid more serious and costly health issues later (Master et al. 1996). Finally, MCOs may have limited experience serving a disabled population and lack an adequate network of ancillary service and specialty care providers (Tanenbaum and Hurley 1995).
The relative effects of Medicaid MCOs on care access for nondisabled beneficiaries are mixed. Relative to FFS Medicaid, Medicaid MCO programs are associated with an equal or improved likelihood of having a usual source of care (USC), and an equal or a lower probability of emergency room (ER) use among adults (Coughlin and Long 2000; Garrett, Davidoff, and Yemane 2003; Garrett and Zuckerman 2005;), although this relationship varies depending on the length of Medicaid enrollment (Lo Sasso and Freund 2000). Additionally, Medicaid MCO enrollees report shorter travel times to the USC and shorter wait times once there to see their provider (Sisk et al. 1996; Coughlin and Long 2000;). Preventive care use outcomes vary by the recall period. Medicaid MCOs are associated with lower or no difference in the likelihood of receiving a Pap smear or breast exam in the past 12 months (Zuckerman, Brennan, and Yemane 2002; Garrett and Zuckerman 2005;), but a higher probability of receipt of a Pap smear in the past 2 years (Coughlin and Long 2000).
Relative to FFS, mandatory Medicaid MCOs are associated with diminished health care access for enrollees but either no change or improvement in access at the programmatic level, that is, for the total eligible AWD population (Lo Sasso and Freund 2000; Coughlin, Long, and Graves 2009;). These mixed findings may be a function of the evaluation strategies deployed. Lo Sasso and Freund (2000) assessed the relative effects of mandatory enrollment in a Medicaid MCO and found a higher probability of inpatient admissions for ambulatory care sensitive conditions and a higher rate of ER visits among MCO enrollees. Coughlin, Long, and Graves (2009) assessed the relative effects of living in a county with voluntary or mandatory Medicaid MCOs on Medicaid beneficiary outcomes. In this program approach, they observed a greater likelihood of having a USC for preventive health care among beneficiaries in MCO counties relative to those in FFS counties. Additionally, they found a greater likelihood of contact with a variety of providers relative to FFS county beneficiaries when the sample was restricted to urbanites.
Each perspective addresses distinct policy-relevant questions and could legitimately yield different findings. However, the enrollment evaluation was conducted in two California counties using administrative claims data from 1989 to 1992, while the program evaluation was national in scope, collapsed both mandatory and voluntary MCO programs into one category, and used survey data from 1997 to 2004. The different measures used, geographic variation in markets and Medicaid programs, and secular changes in MMC may all contribute to their mixed results. By holding these factors constant, this study offers a potential reconciliation of the historical findings while providing national estimates for each evaluation approach.
Four data sources are merged by the year and subject's county of residence. I pool data from the Household Component of the Medical Expenditure Panel Survey (MEPS) (1996–2004), a nationally representative survey of the U.S. civilian noninstitutionalized population. From the MEPS, Medicaid beneficiary enrollment status in FFS or an MCO is identified. To identify MMC county status and if enrollment is voluntary or mandatory, I use documents from the Centers for Medicare and Medicaid Systems (Centers for Medicare and Medicaid Services 2004, 2005) to build a dataset that describes the county Medicaid MCO status and enrollment mechanism in each U.S. county for the adult disabled population between 1996 and 2004 following a modified version of Garrett et al.'s data collection protocol (Garrett, Davidoff, and Yemane 2003).1 The Area Resource File and U.S. Census Small Area Income and Poverty Estimates contribute county-level characteristics (Health Resources and Services Administration 2005; U.S. Census Bureau 2007;).
From among adults aged 18–64 enrolled in Medicaid, I identify disabled beneficiaries as those who participate in the federal cash assistance program for persons with disabilities, the Supplemental Security Income (SSI) program. Medicare beneficiaries are excluded from this study because they are not uniformly subject to the same requirements within Medicaid MCOs as are Medicaid-only beneficiaries and may be excluded from mandatory MCO enrollment (United States General Accounting Office 1996). The total sample includes 2,438 person-years. Of these, I exclude 154 observations because they were not eligible for the access questions. An additional 105 observations are excluded due to missing data, most frequently observed in the MEPS Medicaid MCO enrollment variable.
I select access measures that are consistent with federal access requirements for Medicaid MCOs (United States Code of Federal Regulations 2006) and with measures used, or recommended for use, in state evaluations of Medicaid MCOs in disabled populations (Division of Health Care Quality Financing and Purchasing 2000; Minnesota Department of Health Services 2004; Kailes et al. 2005;).
The Medicaid program requires that each MCO enrollee have an ongoing source of primary care. Medicaid MCOs are also expected to consider the geographic location of providers and enrollees, considering distance, travel time, and the physical accessibility of provider locations when developing their provider networks. Additionally, they must provide timely access to care and have a mechanism in place to allow direct access to specialists for disabled beneficiaries (United States General Accounting Office 2004; United States Code of Federal Regulations 2006;). I assess if the beneficiary has a USC, the travel time to the USC, the wait time to see a provider there, and how much of a problem (if any) it is to see a specialist among those who report a need to see one.
Preventive care use measures are also assessed. AWDs are particularly susceptible to secondary conditions and co-morbidities that timely preventive care and screenings could prevent or mitigate (Lawthers et al. 2003; Diab and Johnston 2004; Kinne, Patrick, and Doyle 2004; U.S. Department of Health and Human Services 2005;). Managed care has long been associated with the provision of primary and secondary preventive health care (Miller and Luft 1994; Phillips et al. 2004; Berman, Armon, and Todd 2005; Rizzo 2005;). As such, Medicaid MCOs may represent an opportunity to increase delivery of these services. Moreover, several measures of preventive care use and screening are supported by evidence-based national guidelines (U.S. Department of Health and Human Services 2000; United States Preventive Services Task Force 2007;), thereby providing a relatively objective standard of care.
I construct the explanatory variable for the enrollment models by joining individual-reported Medicaid MCO enrollment status from the MEPS to the Medicaid MCO enrollment mechanism, mandatory or voluntary, present in the subject's county of residence as described in the county MMC dataset. For program effect models, the explanatory variable is simply the county's designation as FFS, mandatory MCO (MMCO) or voluntary MCO (VMCO).
In theory, individual and county-level data regarding Medicaid MCO status may not agree. For example, a working age SSI/Medicaid beneficiary may disenroll from MMCOs under a limited set of circumstances. Additionally, measurement error may contribute to inconsistency between the two data sources. Following Zuckerman, Brennan, and Yemane (2002), I assess the concordance in Medicaid MCO county and enrollment status for the enrollment models. As expected, I find that a small proportion of the unweighted sample, approximately 4.6 percent, reports enrollment in Medicaid FFS but lives in a mandatory Medicaid MCO county. I consider this group to be voluntary FFS (VFFS) enrollees given that states are required to permit disenrollment from MMCOs under limited circumstances. Surprisingly, however, 11.9 percent of the sample reports enrollment in a Medicaid MCO but lives in a county without any Medicaid MCO available. After further analysis, there is some evidence that these individuals may have incorrectly reported MCO enrollment status.2 I classify this group as FFS-only (FFSO) enrollees and conduct sensitivity analyses that exclude all discordant observations.
At the individual level, I control for predisposing, enabling, and need-based factors (Andersen and Aday 1978; Andersen et al. 1983;). These include age, sex, race/ethnicity, highest degree earned, residence in a metropolitan statistical area, annual income, marital status, family size, employment in the past 12 months, and self-reported physical and mental health and activity limitations (Mitchell, Khatutsky, and Swigonski 2001; Coughlin, Long, and Kendall 2002; Hill and Wooldridge 2002;).
I include lagged and concurrent county characteristics associated with MCO implementation, HMO market entry, and/or health care access. These variables include the percent of all residents below the federal poverty level, average per capita income, median household income, the population density (Moscovice, Casey, and Krein 1998; Slifkin et al. 1998; Duggan 2004;), HMO penetration rate (Slifkin et al. 1998; Duggan 2004; Currie and Fahr 2005;), the number of physicians per 10,000 residents, and the percent of households with an SSI beneficiary. Additionally, concurrent indicators for the presence of a Medicaid prepaid health plan (PHP) or primary care case management program (PCCM) in the county are included.3 PHPs provide a limited set of services that vary widely from transportation only to behavioral health care. Beneficiaries in FFS and MCOs can potentially be enrolled in a PHP as determined by the state Medicaid program.
Finally, I include state dummy variables to adjust for residual state-level characteristics that may influence the county's or individual's MCO status and the outcome such as Medicaid program characteristics or socioeconomic factors (Garrett, Davidoff, and Yemane 2003; Currie and Fahr 2005;).
Yi is a measure of health care access per person-year, X is a vector of individual characteristics associated with MCO enrollment and/or the outcome, GEO is a vector of county characteristics associated with MCO implementation and/or the outcome, and State and Year are a set of dummy variables to control for state characteristics and secular events that may confound the relationship between plan type and the outcome. In model (1), MMCO is an indicator of mandatory enrollment in a Medicaid MCO, VMCO is an indicator of voluntary enrollment in a Medicaid MCO, and VFFS is an indicator of voluntary enrollment in FFS either in a VMCO county where individuals may choose between an MCO and FFS or, less commonly, an MMCO county in which the beneficiary is enrolled in FFS care.4 In model (2), COUNTY_MMCO indicates that the county has mandatory MCO enrollment and COUNTY_VMCO indicates that the county has voluntary MCO enrollment. The reference group for both models is the same: beneficiaries who reside in counties where no Medicaid MCO is available and are thus enrolled in FFS (FFSO).
I identify access outcomes associated with each plan type from Medicaid plan enrollment/county program status variation within states, controlling for overall time trends, and conditional on observed personal and county characteristics. Several analytic challenges must be overcome to yield valid estimates. Unobserved individual characteristics associated with the outcome and the beneficiary's decision to enroll in an MCO may bias the enrollment estimates. To mitigate this potential source of bias, I disaggregate Medicaid MCO enrollment effects into those that are less prone to selection bias, mandatory FFS and MCO counties, from those where it may be a more consequential factor, voluntary MCO counties (Bindman et al. 2005; Kaestner, Dubay, and Kenney 2005;). I can then describe potential selection into MCOs by comparing VMCO enrollment to program results and examining descriptive statistics for those who choose MCOs and FFS within VMCO counties.
The second major challenge to the validity of my models is county-level omitted variables bias. County-specific time trends are one possible solution; however, even in large datasets it is often impractical to include them (Garrett, Davidoff, and Yemane 2003; Kaestner, Dubay, and Kenney 2005;). More commonly in the MMC literature, concurrent county control variables (Zuckerman, Brennan, and Yemane 2002; Garrett, Davidoff, and Yemane 2003; Kirby, Machlin, and Cohen 2003; Coughlin, Long, and Graves 2009;) or difference-in-difference models are used (Garrett and Zuckerman 2005; Kaestner, Dubay, and Kenney 2005; Le Cook 2007;). However, the eligibility requirements for SSI Medicaid beneficiaries are extraordinary—even relative to the nondisabled Medicaid beneficiaries—in terms of both income and level of disability. Thus, it is unclear that a relevant comparison group, whose health care access is equally sensitive to potential omitted variables, can be constructed to deploy difference-in-difference models. Alternatively, Currie and Fahr (2005) use instrumental variables (IV) and lagged geographic characteristics predictive of their main explanatory variable, state-level MMC enrollment. However, in contrast to this earlier work where IV methods address enrollment endogeneity, their potential role here is to address policy endogeneity, the state's decision to implement MCOs in a particular county. For the enrollment models, this requires an instrument that predicts the state's decision to implement MCOs in a county and is related to access only through the individual's subsequent MCO enrollment decision. It thus necessitates a strong correlation between the county's Medicaid MCO status and the individual enrollment decision. Such a strong correlation does not exist in the Voluntary MCO counties, limiting the feasibility of an IV analysis in this context. My strategy then to mitigate county-level omitted variables bias is to include lagged, concurrent, and time-invariant county variables that have been empirically or theoretically related to MCO implementation, HMO market entry, and/or health care access (Currie and Fahr 2005). Lagged county variables account for potential changes in county characteristics related to access that may also influence the state's implementation of Medicaid MCOs in the county (Baker 1997). Concurrent and time invariant county factors address additional geographic and market characteristics that may modify the outcome, independent of the beneficiary's plan type (Garrett, Davidoff, and Yemane 2003).
I compare the weighted descriptive statistics for the population across enrollment groups and program groups using student's t-tests and Pearson χ2 tests. Because observations are allocated to the plan types differently within the enrollment and program models (Table 1), descriptive statistics are provided for the sample subset according to individual enrollment status and then by county program status (heretofore enrollment and program samples). I use logit regression to estimate the models presented in Equations (1) and (2). Standard errors are estimated using a Huber variance estimator (Huber 1967), where observations are clustered by primary sampling unit to account for the complex survey design of the MEPS and within-person correlation over time.5 All analyses are weighted to reflect the civilian noninstitutionalized, adult disabled Medicaid population in the United States and conducted using Stata 9.0 (StataCorp 2005). Results are presented in terms of average marginal effects.
Within the enrollment effect models, I conduct sensitivity analyses to address potential enrollment misclassification. Following Zuckerman's response to potential misclassification bias (2002), I exclude all discordant observations.
In the full enrollment sample, the majority of observations are in FFS care, 51 percent in FFSO, 24 percent enrolled in VFFS, 15 percent enrolled in MMCO, and the remaining 10 percent enrolled in VMCO (Table 1). Overall, the majority of the observations are female, educated at less than a high school level, and report being in fair or poor physical health. Relative to FFSO enrollees, all other enrollment groups include a higher proportion of non-white beneficiaries. MMCO enrollees have a lower rate of employment in the past 12 months, and fewer VFFS enrollees report fair or poor physical health than FFSO enrollees. A higher percentage of each comparison group resides in urban areas with higher household, per capita income, and higher HMO penetration relative to FFSO (Table 2). Seventy-six percent of MMCO enrollees reside in counties with mandatory PHPs in operation compared with roughly 20 percent of FFS enrollees. In the program sample, VFFS enrollees are distributed according to their county's program status such that 19 percent of the sample observations are now allocated to MMCO and 29 percent to VMCO (Table 1). Beneficiaries in VMCO counties report a lower rate of fair or poor physical and mental health than the reference group (Table 2).
The process of obtaining care varies across several measures for MMCO enrollees (Table 3). They are 24.9 percent more likely to wait >30 minutes to see a provider at a USC appointment than FFSO enrollees. Additionally, relative to FFSO enrollees, MMCO enrollees are 32 percent more likely to report a problem accessing a specialist and 10.2 percent less likely to report receipt of a flu shot within the past 12 months. Most of these results persist in the program effect models. Beneficiaries in MMCO counties are 21.7 percent more likely to wait >30 minutes to see a provider at USC, 22.9 percent more likely to report problems obtaining specialty care, and 10.3 percent less likely to receive a flu shot in the past 12 months than beneficiaries in FFSO counties. They are also 6.5 percent more likely to have a USC than FFSO county beneficiaries.
Voluntary MCO enrollees are 6.8 percent more likely to report having a USC. This finding does not persist in the program models. Rather, program effect results indicate that beneficiaries living in VMCO counties are 20.3 percent more likely to report a problem accessing specialty care than FFSO county beneficiaries.
When I exclude discordant observations from the enrollment models, those individuals who report MCO enrollment but live in a FFSO county and those who report FFS enrollment but live in a MMCO county, the findings for MCO enrollees are unchanged with one exception (results not shown). MMCO enrollees are no longer less likely to report a flu shot, although the sign and effect size are consistent with the main results, suggesting that the decreased sample size may explain the loss of statistical significance.
Stakeholders to the health care reform debate for disabled beneficiaries voice many concerns and aspirations for Medicaid MCOs; yet there is little evidence about how these beneficiaries fare in Medicaid MCOs relative to FFS. In general, I find that the shift in care delivery from FFS to MCOs for Medicaid adult beneficiaries with disabilities is not associated with dramatic changes in health care access. However, from the perspective of both MCO enrollees and programs, the finding of decreased access to providers, particularly to specialists, merits further attention from both the research and policy communities.
Securing access to specialty care for beneficiaries with disabilities is a programmatic priority (Regenstein and Anthony 1998; Perry and Robertson 1999; United States General Accounting Office 2004). In this study, at least 24 percent of beneficiaries (Table 3) report problems getting access to specialty care with MMCO enrollees reporting the most difficulty. Specialty care access may be more limited within Medicaid MCOs than FFS because of provider payment arrangements, provider network composition, and/or utilization management strategies (Tanenbaum and Hurley 1995; United States General Accounting Office 1996; Perry and Robertson 1999). The single measure available does not identify which of these factors, or others, may explain the observed limitation in specialty care access among MMCO enrollees and beneficiaries more generally. Additionally, the effects of MCOs, FFS, and/or other models of care on the secondary and tertiary care needs and outcomes of this Medicaid population remain largely unknown. Given the population's substantial and complex medical care needs, future research should address the patient, physician, and health system factors that impede specialty care access and their implications for health and function.
It is somewhat surprising that MCO enrollees do not fare better than FFS enrollees across preventive services more given managed care's long association with health promotion. Growing evidence suggests that delayed or under use of preventive care and screening among AWDs is associated with a range of poor outcomes (Roetzheim and Chirikos 2002; Wilber et al. 2002;). Thus, while the complex care needs of disabled beneficiaries are an appropriate priority for Medicaid redesign efforts, the population's primary care needs should not be overlooked.
While both the enrollment and program evaluations suggest modest between-group differences, they tell somewhat different stories. MMCOs are associated with worse provider access than FFSO from either evaluation perspective. However, from the program perspective the reduced effect sizes for provider access variables and the increased likelihood of a USC provide a partial reconciliation of prior research in this population (Coughlin, Long, and Graves 2000; Lo Sasso and Freund 2000;).
The observed differences between enrollment and program perspectives beg the question of why they differ. In prior MMC research of AWDs, differences in study setting, sample, and data may have explained the observed outcome differences between evaluation perspectives. I hold these factors constant; thus, it is possible to consider what may contribute to the different findings. In particular, I examine the possibility of individual selection effects by comparing the individual characteristics (Table 2) of those who report enrollment in FFS and MCOs in both VMCO and MMCO counties. I find modest evidence of favorable selection in VMCO counties as MCO enrollees are less likely to report fair or poor physical health than FFS enrollees (results not shown). This finding may explain the increased problems obtaining specialty care for VMCO county beneficiaries as a whole observed in the program effect models compared with VMCO enrollees. In MMCO counties, states must allow MCO disenrollment under limited conditions. However, I find no significant differences in the characteristics of those who report MCO and FFS enrollment (results not shown), suggesting that FFS and MCO attributes may influence the different MMCO enrollment and program results.
Because the working age SSI population represents <2 percent of the U.S. population (United States Social Security Administration 2006), it is challenging to obtain a sample that is large and nationally representative. Thus, this study may be underpowered for some measures. Claims and encounter data are alternative data sources for MMC evaluation that may provide a larger sample (Lo Sasso and Freund 2000). However, encounter data are typically available at the state or plan level and may not generalize well beyond those entities. Recognizing these trade-offs, I opted to use nationally representative data to complement existing state-based research. Research that pools encounter and claims data from multiple states represents an important next step to evaluate care access for AWDs, particularly if it permits clinically meaningful subgroups.
The performance of particular MCO and FFS delivery systems may diverge from the average results presented here. To my knowledge, only one study has evaluated the effects of particular MCO features within the disabled Medicaid population, finding few differences in access across MCO characteristics (Hill and Wooldridge 2002). Finally, within MCO and FFS counties, Medicaid PHPs and PCCMs may operate (Table 2). Research in adult populations has shown few differences in access between Medicaid PCCMs and FFS (Zuckerman, Brennan, and Yemane 2002; Garrett, Davidoff, and Yemane 2003; Garrett and Zuckerman 2005; Coughlin, Long, and Graves 2009;); however, the effects of dual enrollment in PHPs and MCOs among AWDs are not well understood and require additional research.
Motivated by both cost and quality concerns, state Medicaid programs are actively experimenting with how best to deliver care to adult beneficiaries with disabilities, with an emphasis on implementing Medicaid MCOs (Wisconsin Division of Health Care Financing 2004; California Department of Health Services 2005; Kaiser Commission on Medicaid and the Uninsured 2005;). This study contributes to their increasing efforts to evaluate care system effects by examining health care access from the perspectives of two key stakeholders: the enrollee and the state. On average, states should not expect a dramatic change in health care access when they implement Medicaid MCOs to deliver care to the adult disabled population. However, continued attention to specialty care access is warranted from both the program and enrollee perspectives.
Joint Acknowledgment/Disclosure Statement: Support for this research is gratefully acknowledged from the NIMH (T32 MH18029-22), the AHRQ (T32 000083), a Merck Quantitative Sciences Fellowship in Health Economics, and the University of Wisconsin Institute for Research on Poverty. I am grateful to John Mullahy, Bobbi Wolfe, and Maureen Smith for advice, instruction, and comments throughout the development of this study, to Bowen Garrett for sharing his time, data, and the Urban Institute's Medicaid managed care data collection protocol, and to Ray Kuntz for expert and collegial assistance with long-distance data analysis. The research in this paper was conducted at the Agency for Health Care Research and Quality's CFACT Data Center, and the support of AHRQ is acknowledged. The results and conclusions in this paper are those of the author and do not indicate concurrence by AHRQ or the Department of Health and Human Services. No other disclosures.
Disclaimers: None Disclosures: None.
1Complete documentation is available from the author.
2Misclassification of MCO enrollment status may be due to measurement error in the county and/or individual's Medicaid MCO status. I compared county-level MMC plan status in my dataset to aggregate results from a dataset constructed using the same sources by Coughlin and Long and found the results were nearly identical (Coughlin and Long 2004). To ascertain potential sources of error in individuals' self-report of Medicaid MCO status, I reviewed the list of state-specific Medicaid MCO plans that MEPS survey staff provide to respondents as they are asked about their Medicaid plan. The names of Medicaid MCOs were on this list as well as health plans that are not Medicaid MCOs. Specifically, the names of Medicaid prepaid health plans (PHPs) and Medicaid primary care case management plans (PCCMs) were also included on this list. A beneficiary cannot be dually enrolled in a PCCM, and a MCO or FFS. But s/he may be dually enrolled in a PHP, and a MCO or FFS. I expected, and found, greater discordance between individually reported MCO status and county MCO status in counties where either a Medicaid PHP or PCCM was present.
3Primary care case management (PCCM) is a fee-for-service plan that provides comprehensive health care and case management of primary health care services. The PHP is a capitated plan that provides limited, or carved out, services. The scope of PHPs ranges widely from transportation only to behavioral health care.
4Beneficiaries may be enrolled in FFS within MMCO counties for the following reasons: (1) they have not yet been assigned to an MCO; (2) they have disenrolled from an MCO per a limited set of circumstances defined by federal regulations U.S. 42 CFR 438.56 such as medically determined need for a service not provided by the MCO, patient grievances, etc., or (3) they are excluded from MCO enrollment most typically because of Medicare eligibility or the need for nursing home–level care. Medicare eligibles are excluded from this study. Thus, the former reasons are more likely to explain FFS enrollment in MMCO counties in this sample.
5The presence of singleton PSUs within strata prevented an adjustment for strata in the estimation of standard errors. This approach tends to yield conservative standard errors (Wolter 1985).
Additional supporting information may be found in the online version of this article:
Appendix SA1: Author Matrix.
Table S1: Unadjusted Health Care Access by Plan Type: SSI/Medicaid Beneficiaries Ages 18–64, MEPS 1996–2004.
Table S2: Access Associated with Enrollment in Medicaid MCO Relative to Medicaid FFS: SSI/Medicaid Beneficiaries Ages 18–64, MEPS 1996–2004 (Average Marginal Effects %).
Table S3: Access Associated with County Program Status, Medicaid MCO Relative to Medicaid FFS: SSI/Medicaid Beneficiaries Ages 18–64, MEPS 1996–2004 (Average Marginal Effects %).
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.