People with intellectual and developmental disabilities (IDD) are frequent users of health services. We examined how their service utilization of emergency department (ED), inpatient hospitalization, and primary care physicians changed as they transitioned from fee-for-service to Medicaid managed care (MMC). Our results showed that MMC reduced the utilization of all of these services. A substantial decrease in ED visits was associated with the reduction in visits due to mental/behavioral health conditions and conditions that could be nonemergent and manageable with the community-based health services. These findings suggest that health service utilization of people with IDD is related not only to their health needs, but also to the delivery model that provides their health services.
States have been transitioning Medicaid enrollees with disabilities from fee-for-service (FFS) to Medicaid managed care (MMC) models to reduce their health service utilization and associated state expenditures (Altman & Frist, 2015; Garfield et al., 2015; Lewin Group, 2009; Sparer, 2012). Persons with disabilities constitute a relatively small subset of Medicaid enrollees, but they utilize more acute health services, including high-cost hospital-based services, such as a visit to the emergency department (ED) and admission to inpatient hospital unit, than other enrollees (Garfield et al., 2015). Under MMC models, states contract with managed care organizations (MCOs) for the delivery of health services through their network of providers. The MCOs employ several strategies to increase access to community-based health services for enrollees including, for example, the assignment of a primary care physician (PCP) to each enrollee, referrals to specialists and other ancillary service providers in the network, medical homes, care coordination, extending the community provider's office hours, providing financial incentive to providers to meet quality standards, and promoting the use of alternative services for patients seeking ED services (Heller et al., 2015; Maeng et al., 2016; National Council on Disability, 2013; Summer & Hoadley, 2014). Through these strategies, enrollees under MMC are expected to have a stable and usual source of primary and specialty care in the community, which, in turn, is expected to lower the need for ED and inpatient hospitalization (Sparer, 2012).
Previous researchers reported inconsistent results regarding the impact MMC had in changing enrollee's health service utilization pattern. In one study, researchers found that MMC reduced the ED use of adult enrollees relative to those who were under FFS (Garrett & Zuckerman, 2005). They, however, did not find a corresponding increase in ambulatory service utilization. In another study, researchers found that MMC reduced the ER use for children, but not for adults (Garrett, Davidoff, & Yemane, 2003). Other researchers reported that the MMC effect in increasing physician visits was limited to those who lived in an urban area where many physicians practice (Coughlin, Long, & Graves, 2008). It appears that MMC impact may vary depending, at least, on the type of health services, target population, and the provider availability.
People with intellectual and developmental disabilities (IDD) use hospital-based health services more frequently than those without IDD (Blaskowitz, Hernandez, & Scott, 2019; Janicki et al., 2002; Lunsky et al., 2010; Morgan, Baxter, & Kerr, 2003). Although 20% of U.S. adults visit EDs every year (Capp, Rooks, Wiler, Zane, & Ginde, 2014; National Center for Health Statistics, 2015), the rate among people with IDD has ranged from 30% to 50% (Blaskowitz et al., 2019; Janicki et al., 2002; Venkat et al., 2009). The proportion admitted to an inpatient unit of a hospital has been reported as being between 16% to 18% of people with IDD (Blaskowitz et al., 2019; Janicki et al., 2002), compared with only 7% of adults in the general population (Adams & Benson, 2015). Complex health and behavioral needs of this population are often cited as a potential reason for their elevated utilization of these intensive services (Blaskowitz et al., 2019; Janicki et al., 2002; Lunsky et al., 2010; Morgan et al., 2003).
The purpose of the present study was threefold. First, we wanted to explore the extent and nature of hospital-based health services (i.e., emergency department and inpatient hospitalization) utilization and access to the community-based health services (i.e., outpatient visit to PCP) by people with IDD comprehensively. Past researchers focused on specific health conditions leading them to EDs (Chi, Masterson, & Wong, 2014; Lunsky et al., 2010; Lunsky, Tint, Robinson, Khodaverdian, & Jaskulski, 2011; Morgan et al., 2003), inpatient hospitalization (Ailey, Johnson, Fogg, & Friese, 2014; Balogh, Brownell, Ouellette-Kuntz, & Colantonio, 2010), and PCP utilization (Hall, Wood, Hou, & Zhang, 2007) separately. Our goal was to examine the utilization of hospital- and community-based health services simultaneously to better understand their health service utilization patterns.
Our second purpose was to examine the impact of MMC in changing the health service utilization of people with IDD. Health services for enrollees with IDD have already transferred to mandatory MMC in 10 states and either voluntarily or eligibility-based MMC in 22 states (Smith et al., 2016). Yet, studies on the impact of these changes on health service utilization of people with IDD are lacking. We hypothesized that Medicaid enrollees with IDD would reduce their utilization of ED and inpatient services and increase that of PCP under MMC relative to their utilization under FFS.
Our third purpose was to examine if MMC changed the underlying causes for hospital-based service utilization by people with IDD. A certain proportion of hospital-based service utilization by people with IDD was associated with health conditions that were nonemergent or treatable in the primary care setting (Balogh et al., 2010; Mann, Royer, McDermott, et al., 2015; Mann, Royer, Turk, et al., 2015). Reducing the overuse of high-cost hospital-based services for these conditions has been suggested as a means to contain state expenditures, one of the goals of MMC (Berwick, Nolan, & Whittington, 2008; McWilliams, Tapp, Barker, & Dulin, 2011). We were interested in if the promotion of the community-based health services through MMC has resulted in the reduction of avoidable use of hospital-based services. We hypothesized that Medicaid enrollees with IDD would reduce their utilization of ED and inpatient services due to conditions that were nonemergent or treatable in the primary care setting under MMC relative to their utilization under FFS.
Data for the present study were extracted from larger data sets obtained from the state Medicaid agency and the two MCOs for the evaluation of the integrated care program (ICP), part of Illinois's Medicaid reform initiative (State of Illinois, 2011). Under the ICP, Medicaid services for all “Seniors and Persons with Disabilities (SPD)” enrollees, excluding children under 18 years old and those who were dually eligible for Medicare, residing in the suburban counties around the city of Chicago, had their Medicaid services managed by one of two for-profit MCOs, selected via a competitive procurement process, starting in May 2011 (Illinois Department of Healthcare and Family Services, 2016). The state data set included demographics, administrative data, and FFS claims of all Medicaid enrollees who met the above ICP eligibility and resided either in the Chicago suburbs (i.e., treatment region) or the city of Chicago (i.e., comparison region) at least 1 month during the 9-month period from July 2010 to March 2011 and the 38-month period from May 2011 to June 2014. The MCO data set included claims of the enrollees in the Chicago suburbs who transitioned to MMC during the 24-month period between January 2012 and December 2013.
Enrollees who had at least one claim that included one or more of International Classification of Diseases–9th revision (ICD-9) codes (Buck, 2012) associated with IDD in the pre-ICP period were identified as subjects with IDD (see Yamaki, Wing, Mitchell, Owen, & Heller, 2018, for the list of ICD-9 codes used during the screening process). Among them, we included those who met the following criteria: aged 22 or older in July 2010 and enrolled in the Medicaid program continuously for 33 months before and after the implementation of ICP described later in the article to evaluate the persistent impact of MMC. We excluded enrollees who resided in intermediate care facilities for people with IDD (ICF/DDs) or nursing facilities to focus on the impact of MMC on community residents.
We examined the enrollee's health service utilization using their ED visits, inpatient hospital admissions, and outpatient visits to PCPs. First, we defined these variables by screening claims using the Healthcare Effectiveness Data and Information Set (HEDIS), a common performance measurement of health services used by the majority of health insurers in the United States (National Committee for Quality Assurance, 2016). HEDIS measures were adopted by the state of Illinois as standardized quality measures to evaluate the performance of the two chosen MCOs (Illinois Department of Healthcare and Family Services, 2012), and used in the state's project evaluating the ICP (Heller et al., 2014). We made some modifications of the HEDIS specification specific to the present study to include claims related to mental health conditions and to accommodate some limitations in our claims data. We aggregated the identified claims to create outcome variables that represent a frequency that a subject accesses each service per month.
Under HEDIS, any claims that met either one of the following two criteria were screened first as a potential ED visit: 1) “procedure” in the claim included one of five current procedural terminology (CPT) codes, where a code represents medical, surgical, and diagnostic procedure performed by the health service provider (i.e., 99281-99285); or 2) “procedure” included one of 5,775 CPT codes defined as the “ED Procedure Code Value Set” by HEDIS and “place of service” in the claim was “emergency room/free standing clinic.” Per the HEDIS specification, we then applied two additional HEDIS screens: 1) when there were multiple ED visits in a day, only the first one was counted; and 2) if the ED visit was followed by the same day inpatient hospital admission, it was not counted as an ED visit.
We omitted the following steps specified by HEDIS. The HEDIS definition uses the revenue code, a billing code for hospital services, as an additional screening criterion for an ED visit. We omitted it, as the code was not available in the FFS claims data from the state. Furthermore, the HEDIS definition excludes any ED visits if the claim meets the following four exclusion criteria: 1) “primary diagnosis” is mental and behavioral disorders, or “procedure” indicates 2) psychiatry services, 3) electroconvulsive therapy, or 4) treatment or detoxification of alcohol or drug abuse. We chose not to apply these exclusions because our preliminary analysis showed that a significant proportion of service utilization by our subjects was associated with these criteria. Removing ED visits associated with these conditions would undercount their ED utilization.
Inpatient hospital admission
Using the HEDIS definition for “inpatient utilization (IPU),” claims that met all of the following four inclusion criteria were screened initially: both (1) “admission date” and (2) “discharge date” were indicated, (3) “inpatient admission indicator” was “Yes,” and (4) “provider type” was “general hospital.” Then, the HEDIS definition calls for applying the IPU exclusion criteria, which remove (1) admissions related to mental health and behavior disorders; (2) infant delivery; and (3) surgery using either the discharge diagnosis or diagnostic related group (DRG) code, a standardized diagnosis classification system for reimbursement rates for hospitalization. We kept admissions related to mental and behavioral disorders due to the reasoning previously discussed and only excluded admissions related to infant delivery. We did not apply any of the IPU exclusion criteria defined by the DRG code because our claims data didn't include it.
Outpatient visits to PCPs
First, all claims that included one of 70 CPT codes specified in the HEDIS definition for “ambulatory outpatient visit (AMB)” were screened. Although the HEDIS definition of AMB further calls for the use of the revenue code and the aforementioned four ED exclusion criteria, we didn't apply those exclusions for the reasons previously mentioned. Among the claims screened for AMB, we only included claims of services provided by a physician whose self-reported specialty was either family practice, general practice, geriatrics, internal medicine, or pediatrics. We included pediatricians based on reports from caseworkers and nurses from pediatric specialty hospitals who support aging-out children. Through interviews conducted as part of the evaluation study on ICP (Heller et al., 2015), it was reported that some young adults with IDD had difficulty finding an adult provider and continued to see a pediatrician.
Causes for Hospital-Based Service Utilization
We identified the primary diagnosis for each hospital service utilization claim and screened them using the following three algorithms:
Clinical Classification Software (CCS)
CCS is a “grouper” that clusters diagnoses into a manageable number of clinically meaningful categories (Elixhauser, Steiner, & Palmer, 2014). We grouped the primary diagnosis using the CCS's single-level classification scheme.
Ambulatory care sensitive conditions (ACSCs)
ACSCs include 17 acute and chronic conditions that would not worsen or require intensive services if treated in a timely manner and managed properly in an outpatient setting (Billings, Anderson, & Newman, 1996). Assuming that a well-developed community-based health service would provide better access to health care for the patient and reduce preventable ED visits (Raven, Lowe, Maselli, & Hsia, 2013; Tang, Stein, Hsia, Maselli, & Gonzales, 2010) and unnecessary hospitalization, ACSCs have been used as an indicator for access to community-based health services (Billings et al., 1996; McCall, Harlow, & Dayhoff, 2001).
Nonemergent ED visit
The New York University algorithm assigns each health condition associated with an ED visit into four categories based on the probability of its urgency for medical care to identify preventable ED utilization (Billings, Parikh, & Mijanovich, 2000). More recent updates further collapse the visits into three mutually exclusive categories: emergent, nonemergent, and intermediate (Ballard et al., 2010). We examined the extent of nonemergent ED visits, visits due to a health condition which is likely to be either nonemergent or treatable by a PCP via an office visit.
We took into account the impact of the state's Save Medicaid Access and Resource Together (SMART) Act (Public Act 097-0689) that was implemented 7 months into the study's post-ICP period. The act either eliminated or restricted funding for some Medicaid health services, including, for example, group psychotherapy, chiropractic services, nonemergency dental care, podiatry services for nondiabetics, and eyeglasses (Heller et al., 2015; Illinois Department of Healthcare and Family Services, 2012). The act primarily affected subjects in the comparison group who remained under FFS, as MCOs maintained the available services at the pre-SMART Act levels in most cases (Heller et al., 2014).
To assess the impact of MMC, we employed an inverse propensity weighted difference-in-difference design, a quasi-experimental research design that simulates a pre- and post-test randomized control trial design in observational studies (Cook, 2015; Wing, Simon, & Bello-Gomez, 2018; Wong, Wing, Steiner, Wong, & Cook, 2013). Subjects who resided in the Chicago suburbs in July 2010 were labeled as the “treatment group.” Those who resided in the city of Chicago in July 2010 were labeled as the “comparison group.” Using the intention-to-treat approach (Gupta, 2011), we kept subjects in the initially identified group even though they may have moved between the two regions during the study period. Subject's health service utilization and primary condition associated with it, as previously described, served as our outcome variables. We followed the treatment group members in the Chicago suburbs and the inverse propensity weight (IPW) matched comparison group members in the city of Chicago across a 9-month preintervention period that extended from July 2010 to March 2011. Implementation of ICP in May 2011 in the Chicago suburbs marked the start of the intervention. The postintervention period was the 24 months from January 2012 to December 2013 due to the availability of MCO data. We excluded May 2011 to December 2011, the first 8 months of ICP implementation, from the postintervention period because of the gradual transition of enrollees to MMC and poor data quality during these months (Heller et al., 2015).
We employed two statistical techniques: propensity score matching and difference-in-differences (DID) regression analysis. We used the former to create a matched comparison group using the ICP-eligible enrollees in the city of Chicago who closely resembled the treatment group enrollees in the Chicago suburbs with respect to a large set of covariates measured prior to the implementation of ICP. Then, we used the latter to estimate the effect of MMC on our outcome variables to account for possible confounding and threats to validity that may have escaped the first stage-matching procedure.
Inverse propensity score weights (IPW)
We fit logistic regressions of membership in the suburban sample on a large set of pre-ICP covariates, including the subject's demographics, diagnoses related to disability, the long-term services and supports, the waiver enrollment, presenting health conditions, historical health service utilization, and state expenditures (see Yamaki et al., 2018, for the covariates used to create IPWs). The propensity scores were predicted values from the logistic regression model, and they represent the estimated probability that a comparison group member represents the treatment group member given these covariates (Pattanayak, Rubin, & Zell, 2011). The propensity scores were used to construct inverse propensity score weights (IPW) for each member of the city of Chicago sample. The IPW was set equal to 1 for each member of the suburban treatment group, and set equal to ps/(1-ps) for each member of the city of Chicago comparison group, where ps represents person's estimated propensity score (Wong et al., 2013). We examined the overlap of the IPW between the two groups after the initial creation, and deleted unmatched subjects from both groups (Caliendo & Kopeinig, 2008). Then, we reconstructed the propensity score with the same covariates, and IPW to better match subjects across the two groups. Conceptually, the IPW-matched comparison group served as a proxy for the treatment group members if they did not transition to MMC and remained under FFS in the DID analysis.
Panel data DID regression analysis has been used to estimate causal effects of health policies and programs in observational studies (Dimick & Ryan, 2014; Wing et al., 2018). In the present study, we estimated the effects of MMC by fitting panel data DID regression models to the IPW analytic sample. The effects were estimated in the form of the difference between the average change in outcomes from pre- and post-ICP period in the suburban treatment group subjects and the average change in outcomes from pre- and post-ICP period in the matched Chicago comparison group subjects. Table 1 describes this DID design in a mathematical formula. The DID model was used to account for possible confounding variables and threats to validity that may have escaped the first stage-matching procedure.
The following equation represents our DID regression model to estimate the effect of MMC on the subject's health service utilization:
The outcome variable, “Yit,” represented health service utilization for the subject “i” in month “t.” Independent variables, Treatmenti , PostICPt,, and PostSMARTt, were binary dummy variables. Treatmenti represented whether the subjecti was expected to transition to MMC. It was set to 0 for the comparison group member and 1 for the treatment group member. PostICPt noted if the observed month was before or after the start of ICP. It was set to 0 for the 9 months between July 2010 and March 2011, and set to 1 for the 24 months between January 2012 and December 2013. PostSMARTt referred to whether the observed month was before or after the implementation of the SMART Act. It was set to 0 for any months until June 2012, and 1 for months from July 2012 to December 2013. represented month fixed time effects that would impact both groups, and is an error term. Each coefficient represents a measure of the average difference in utilization between the two groups after accounting for baseline differences and time trends. The model was augmented to account for a change in policy, the implementation of the SMART Act in July 2012, that may have affected the treatment or comparison group in different ways. We used STATA ver. 15 (StataCorp, 2017) to conduct analyses with the level of significance set at .05. The first author's institutional review board approved this study protocol.
Table 2 compares covariates in the preintervention period between the treatment (n = 1,121) and the comparison (n = 1,102) subjects with IDD. Columns 1 and 2 represent unweighted within-group proportions and counts. White (41.5%) was the largest race group in the treatment group, but the majority of the comparison group was Black (54.6%). Almost one in every five subjects (18.7%) were Hispanic in the comparison group compared to one in every 10 (11.6%) in the treatment group. For both groups, the majority of subjects were younger than 40. Breakdowns by the IDD diagnoses across the two groups were similar except for moderate/severe level of ID diagnosis. The Developmental Disabilities (DD) waiver was the most frequently used Home and Community Based Services (HCBS) waiver program (67.5% for the treatment group; 49.0% for the comparison group), followed by the Persons With Disability waiver (12.2% and 20.5%, respectively). There were few differences in health conditions and utilization of various services between the two groups. Column 3 represents a profile of the comparison group after applying the inverse propensity score weight (IPW). The weighted profile of the comparison group was more similar to the treatment group than the unweighted one, suggesting the application of IPW successfully made the two groups similar to reduce the risk of biases in pre-existing group differences.
Overall Health Service Utilization
Descriptive analyses of ED visits indicated that 1,155 subjects, or 52% of all subjects including both the treatment and the comparison group, made 3,747 total ED visits during the study period. Among them, 24% visited the ED once, 11% visited twice, 14% visited between three and seven times, 3% visited eight or more times. About 33% of the subjects (n = 726) had an inpatient hospital admission. Sixteen percent were admitted once; 11% were admitted two or three times; and 6% were admitted four or more times. Eighty percent of the subjects (n = 1,773) had at least one outpatient visit to a PCP. Thirty-five percent visited a PCP one to five times; 27% visited a PCP six to 10 times; and 17% visited 11 times or more.
Causes for Hospital-Based Service Utilization
Table 3 summarizes the selected primary health conditions associated with utilization of hospital services. The most frequent condition associated with an ED visit was injuries (18.9%), followed by mental/behavioral health conditions (14.0%), and epilepsy (9.6%). One in every eight ED visits, or 12.5%, was associated with ambulatory care sensitive conditions (ACSCs). One-third, or 34.1%, were nonemergent ED visits. For the inpatient hospital admission, the most frequent conditions were mental/behavioral health conditions (37.9%), which included schizophrenia and other psychotic disorders (21.6%) and mood disorders (12.0%). Similar to ED visits, 11.3% of hospitalizations were associated with ACSCs.
Overall Impact of MMC
Figures 1 to 3 graphically show the monthly utilization of the three health services across the treatment and the IPW-matched comparison groups. The number of ED visits per month, shown in Figure 1, was similar across the two groups during the pre-ICP period. In the post-ICP period, the number for the treatment group was much lower than that of the IPW-matched comparison group across the months. In Figure 2, similarly, the number of inpatient admissions in the pre-ICP period was comparable across the two groups. In the post-ICP period, it was generally lower for the treatment group than the IPW-matched comparison group. Shown in Figure 3, the number of PCP visits each month was similar for the treatment and the IPW-matched comparison groups in both the pre- and post-ICP periods.
Table 4 summarizes the estimated impact of the ICP and the SMART Act on the health service utilization of the treatment group using coefficients of the DID regression model using the IPW-matched comparison group. In Column 1, the estimated average number of ED visits was 72.4/month (β0 + β1) for the treatment group during the pre-ICP period. Implementation of ICP, then, lowered the average number of visits by 41.4 (β2), p < 0.001, to 30.9/month (β0 + β1 + β2). The implementation of the SMART Act increased the visits by 5/month (β3) although it didn't reach statistical significance. Combining the impacts of both the ICP and the SMART Act, the net change of ED visits between the pre- and post-ICP periods for the treatment group was estimated to be −36.4/month (β2 + β3) F = 15.0, p < 0.001. In Column 2, similarly, the average number of inpatient admissions for the treatment group in the pre-ICP period was estimated as 37.7/month (β0 + β1). The ICP reduced it significantly by 20.8, p < 0.01, to 16.8/month (β0 + β1 + β2). The increase due to the implementation of the SMART Act, 12.1/month (β3), was insignificant. The net change between the pre- and post-ICP periods, −8.7/month (β2 + β3), was insignificant. In Column 3, the average number of PCP visits by the treatment group was 225.5/month (β0 + β1) in the pre-ICP period. ICP decreased the number significantly by 36.8 (β2), p < 0.05, to 188.6/month (β0 + β1 + β2). After the implementation of the SMART Act, the number increased significantly by 39.7 (β3), p < 0.01, to 228.3/month. The net change between the pre- and the post-ICP period, 2.8/month (β2 + β3), was negligible.
Impact of MMC on Avoidable Hospital Service Use
Table 5 summarizes the estimated impact of the ICP and the SMART Act on the selected primary health conditions associated with the hospital-based services utilization of the treatment group using coefficients of the DID regression model. The ICP was estimated to reduce the mean number of ED visits due to mental/behavioral health conditions by 9.5/month, p < 0.001, from 13.6/month in the pre-ICP period to 4.1/month in the post-ICP period. Similarly, ICP was estimated to reduce the number of nonemergent ED visits by 14.7/month, p < 0.01, from 25.2/month to 10.5/month. There was no observable ICP impact on ED visits due to injuries and ACSCs, or on inpatient hospital admission due to mental/behavioral health conditions and ACSCs. The SMART Act did not have a significant impact on hospital service utilization associated with any of these primary health conditions either.
In the present study, we examined the extent of health service utilization by people with IDD and how it changed as they transitioned from FFS Medicaid to MMC. To our knowledge, this is the first study examining utilization of multiple health services by people with IDD in relation to Medicaid health service delivery models. Overall, we found that people with IDD were heavy users of health services. Implementation of MMC decreased their utilization of both hospital- and community-based health services. A substantial decrease in ED visits was associated with decreases in visits due to mental/behavioral health conditions, and conditions that could be nonemergent and manageable with community-based health services. Our findings suggest that health service utilization of people with IDD is not only a function of their health needs/conditions, but also related to the health service delivery model.
Our findings on the utilization of the hospital-based services by people with IDD are generally consistent with previous findings. The proportion of our subjects who accessed the ED was similar to that reported previously (Blaskowitz et al., 2019; Venkat et al., 2009). Consistent with a previous study (Venkat et al., 2009), we found that injuries were the most frequent reason for ED visits among persons with IDD. The proportion of our subjects who were hospitalized was higher than that previously reported in other studies (Blaskowitz et al., 2019; Janicki et al., 2002).
Beyond ED visits specifically, our findings were consistent with other studies that suggested mental/behavioral health conditions and epilepsy were the leading causes of hospital-based service utilization for people with IDD (Ailey et al., 2014; Balogh et al., 2010; Lunsky et al., 2010). This signifies that these conditions are prevalent among people with IDD, difficult to manage due to their complexity, and may require access to comprehensive hospital-based services (Blaskowitz et al., 2019; Bowley & Kerr, 2000; Cooper, Smiley, Morrison, Williamson, & Allan, 2007; Lunsky et al., 2010; Morgan et al., 2003; National Association of State Directors of Developmental Disabilities Services and the Human Services Research Institute, 2013). Consistent with previous findings (Balogh et al., 2010; Mann, Royer, McDermott, et al., 2015; Mann, Royer, Turk, et al., 2015), we found that a notable proportion of the hospital-based service utilization by people with IDD was potentially avoidable. This may be an indication that some health conditions that are treatable at the primary care setting for the general population may require more comprehensive services at a hospital setting when presented by people with IDD (Mann, Royer, McDermott, et al., 2015). It is also possible that this finding indicates that people with IDD experience difficulty in accessing community-based services (Billings et al., 1996).
Additionally, there are at least two contextual factors that might have contributed to the increased use of hospital-based services by people with IDD. First, community-based IDD service providers may have policies and procedures requiring its staff members to seek ED services under certain circumstances for the safety of people for whom they provide care. In these situations, a staff member may feel compelled to seek ED services regardless of the presenting symptoms of the people. Second, the availability of mental health services for people with IDD in the community has been scarce in Illinois (Human Services Research Institute, 2008, 2012). Hence, community-based DD service providers may use the hospital-based services as an instant alternative for crisis intervention or a temporary residential placement (Lulinski Norris, 2014). Although the extent of these factors contributing to the hospital services utilization of our subjects is beyond the scope of the present study, it is critical to understand that the health service utilization of people with IDD is not only a function of their health needs, but also a function of the health and support service system surrounding them.
Our findings on the impact of MMC in reducing hospital-based health service utilization were not consistent with previous findings reporting little or no MMC impact on the utilization of the hospital-based services on enrollees with disabilities (Caswell & Long, 2015; Coughlin et al., 2008). These previous findings were, however, based on national-level population surveillance data that are likely to underrepresent people with IDD (Larson et al., 2001; Yamaki, 2005). The reduction in ED visits associated with mental/behavioral health conditions and nonemergent conditions found in the present study was encouraging. Access to the MCO's network of providers may have reduced the previously mentioned difficulty in accessing mental health services under FFS (Human Services Research Institute, 2008, 2012). Other MCO strategies implemented in the ICP (e.g., assignment of a PCP to each enrollee, and use of case management) might have also helped re-direct our subjects, who previously sought ED services for nonemergent conditions to community-based alternatives such as retail health and urgent care clinics (Weinick, Burns, & Mehrotra, 2010). The reduced utilization of high-cost hospital-based health services under MMC, however, did not produce the anticipated fiscal results of lowering the state Medicaid health service expenditures (Yamaki et al., 2018).
Although we expected to see increased utilization of community-based health services under MMC, we found that ICP reduced the number of PCP visits. Our finding might be reflective of the limited provider network MCOs offered during the initial phase of ICP. A report from the larger ICP evaluation study suggested that the number of physicians in the MCO network in July 2012 was still one-third of the number of those who registered for FFS Medicaid during the pre-ICP period, and that enrollees who transitioned to MMC reported difficulty in finding a new provider and a longer waiting period to see the provider during this period (Heller et al., 2015). Although the adequacy of the provider network has improved in the subsequent years, our findings could be attributable, at least partially, to the provider shortage during the initial ICP period. It is also plausible that the decreased PCP utilization could be impacted by our exclusion of nonphysician providers, such as nurses and physician assistants whom MCOs often use as lower-cost substitutes for PCPs (Coughlin et al., 2008), in our analysis. Future studies should consider including these alternative providers and other indicators (Capp et al., 2014; Cheung, Wiler, Lowe, & Ginde, 2012; Gindi, Blanck, & Cohen, 2016; Lowe et al., 2005; Wang, Tchopev, Kuntz-Melcavage, Hawkins, & Richardson, 2015) to better understand the extent of MMC impact on the access to community-based services for people with IDD. A definitive reason for the increase in PCP visits due to the SMART Act was unclear. We speculate that some of the optional services either reduced or dropped under the act might have been provided by PCPs, at least partially. Future studies are needed to address specific changes incurred by the act, as well as to explore how each of the two MCOs implemented these changes independently and their implications for community-based health service utilization among enrollees with IDD.
Interpretation of the present findings warrants at least two limitations. First, the present findings may have limited generalizability because this study was conducted in a pilot region of one state. The effects of MMC on health service utilization may vary with the type of program implemented, the number of providers in the given geographic area, and enrollee subgroups (Baker & Afendulis, 2005; Freund et al., 1989; Garrett & Zuckerman, 2005; Pollack, Wheeler, Cowan, & Freed, 2007; Verdier et al., 2009). Second, the scope of the present study excluded another important goal of MMC: reducing state Medicaid expenditures (Berwick et al., 2008; Garfield et al., 2015; Sommers, Kenney, & Epstein, 2014; Sparer, 2012; Yamaki et al., 2018). To better understand the full impact of MMC, the fiscal aspect needs to be examined together with changes in the service utilization.
State Medicaid programs, a fundamental source of acute health services for Americans with IDD, are in the midst of transformation. Although the present study demonstrated that MMC could reduce the utilization of ED and inpatient hospital services by enrollees with IDD, it is unclear whether the reduction is correlated with improved health status of each enrollee, another goal of MMC (Berwick et al., 2008). Monitoring changes in the delivery of state Medicaid programs and providing additional empirical data on the health implications these changes have for people with IDD are critical strategies to support their health and well-being in the community. Nonetheless, findings from the present study indicate that changes in the health service delivery system could reduce utilization, even for heavy users with complex health needs. This is an important implication for states considering or implementing Medicaid reform involving enrollees with IDD.