Primary and Behavioral Health Care Integration Program

Short abstract

This article describes an extension of the RAND Corporation's evaluation of the Substance Abuse and Mental Health Services Administration's Primary and Behavioral Health Care Integration grants program.

Abstract

This article describes an extension of the RAND Corporation's evaluation of the Substance Abuse and Mental Health Services Administration's Primary and Behavioral Health Care Integration (PBHCI) grants program. PBHCI grants are designed to improve the overall wellness and physical health status of people with serious mental illness or co-occurring substance use disorders by supporting the integration of primary care and preventive PH services into community behavioral health centers where individuals already receive care. From 2010 to 2013, RAND conducted a program evaluation of PBHCI, describing the structure, process, and outcomes for the first three cohorts of grantee programs (awarded in 2009 and 2010). The current study extends previous work by investigating the impact of PBHCI on consumers' health care utilization, total costs of care to Medicaid, and quality of care in three states. The evidence suggests that PBHCI was successful in reducing frequent use of emergency room and inpatient services for physical health conditions, reducing costs of care, and improving follow-up after hospitalization for a mental illness. However, PBHCI evidence does not suggest that PBHCI had a consistent effect on quality of preventive care and health monitoring for chronic physical conditions. These findings can guide the design of future cohorts of PBHCI clinics to build on the strengths with respect to shifting emergency department and inpatient care to less costly and more effective settings and address the continuing challenge of integrating care between specialty behavioral health providers and general medical care providers.

This study describes an extension of the RAND Corporation's evaluation of the Substance Abuse and Mental Health Services Administration's (SAMHSA's) Primary and Behavioral Health Care Integration (PBHCI) grants program. PBHCI grants are designed to improve the overall wellness and physical health (PH) status of people with serious mental illness (SMI) or co-occurring substance use disorders by supporting the integration of primary care and preventive PH services into community behavioral health (BH) centers where individuals already receive care. From 2010 to 2013, RAND conducted a program evaluation of PBHCI, describing the structure, process, and outcomes for the first three cohorts of grantee programs (awarded in 2009 and 2010). That evaluation found wide variation in program structures, a range of implementation barriers, and some consumer-level improvements in PH outcomes (e.g., cholesterol, diabetes management). The current study extends previous work by investigating the impact of PBHCI on consumers' health care utilization, total costs of care to Medicaid, and quality of care in three states.

Background

Adults with SMI suffer disproportionately from PH conditions. Compared with their non-SMI peers, adults with SMI are at increased risk for a range of acute and chronic diseases, including diabetes, cardiovascular disease, respiratory disease, cancer, and infectious disease (Jones et al., 2004; McGinty et al., 2012; Parks et al., 2006; Substance Abuse and Mental Health Services Administration, 2012). Life expectancy estimates for adults with SMI range from eight to 30 years lower than for the general population (Chang et al., 2011; Colton and Manderscheid, 2006; Saha, Chant, and McGrath, 2007; Walker, McGee, and Druss, 2015). Co-occurring medical and BH conditions are also disproportionately costly for public payers of health care, primarily Medicaid and Medicare (Kasper, Watts, and Lyons, 2010; Melek, Norris, and Paulus, 2014). These disparities have been attributed to modifiable risk factors such as smoking, alcohol and substance use, poor nutrition, lack of exercise, obesity, and high-risk sexual behaviors (Parks et al., 2006); side effects of psychotropic medications (Newcomer, 2007); housing instability and low socioeconomic status (Katon, 2003); and limited access to quality medical care (Lawrence and Kisely, 2010).

Fragmentation between the general medical and BH sectors—in terms of clinical practice, administration, and financing—is widely considered to be a significant contributor to the poor overall health outcomes associated with SMI (Druss, 2007; Horvitz-Lennon, Kilbourne, and Pincus, 2006; Committee on Crossing the Quality Chasm: Adaptation to Mental Health and Addictive Disorders, Board on Health Care Services, Institute of Medicine, 2006; Pincus et al., 2007; President's New Freedom Commission on Mental Health, 2003). As such, initiatives that promote medical and BH integration are expected to address the triple aims of health care reform: improved care experiences, improved health outcomes, and reduced per-capita costs (Katon and Unützer, 2013).

Improvements in care experience and health outcomes are expected to result from increased access to primary care and preventive medical services (because of service co-location or facilitated referrals) and increased collaboration and learning across BH and PH care providers (Alakeson, Frank, and Katz, 2010). Reductions in health care costs for adults with SMI are expected to result through decreases in hospitalizations and emergency department (ED) visits for preventable health conditions and fewer inappropriate visits to EDs (e.g., for primary care needs) (Nolte and Pitchforth, 2014). In practice, however, the effects of integration on health care costs for adults with SMI may be more complex. Given high levels of previously unmet medical needs, integrated care programs for adults with SMI may lead to increased visits to primary and specialty medical care, which can increase the cost of care particularly for consumers who had little-to-no contact with PH care services before.

In the current study, we examined the impact of PBHCI-funded integrated care for adults with SMI on health care utilization, total costs of care, and quality of care received using Medicaid claims data. Medicaid claims data provide a valuable perspective because they reflect a wide scope of services that (Medicaid-enrolled) individuals receive, which is particularly important given that adults with SMI may be transient (receiving services across multiple locations and health systems) and are likely to receive services across multiple levels of care (i.e., hospital, crisis, emergency, outpatient).

The prior RAND evaluation of PBHCI did not have information on utilization and costs of health care outside of the PBHCI grantee clinics. It also did not have information on utilization, costs, and quality among consumers treated in non-PBHCI clinics to whom the PBHCI enrollees could be compared. The current study was designed to address these limitations and, specifically, to investigate the following research questions:

1. What was the impact of PBHCI on utilization of ED and inpatient (IP) services?

   One of the major motivations for improving the quality of primary care services for adults with SMI is to shift care away from unnecessary or preventable ED visits or IP hospitalizations. The claims data allowed us to examine utilization of ED and IP services and to distinguish utilization for PH conditions, where effects are anticipated, from utilization for BH conditions, which are not directly targeted by PBHCI.

2. What was the impact of PBHCI on costs of care to Medicaid?

   Improvements in care for PH conditions are likely to have complex cost implications for Medicaid. The claims data allowed us to examine the impact of PBHCI on the total costs of care per person and to break these costs down by the site of care to gain insight into how PBHCI affects each of these components of total costs of care.

3. What was the impact of PBHCI on the quality of health care for PH conditions for the people treated in PBHCI grantee clinics?

By improving primary care services, PBHCI was expected to improve care for PH conditions. Although the prior evaluation documented some of these improvements, the current study examined the impact of PBHCI on quality of care from a different (Medicaid) perspective, which included documentation of services provided outside of each PBHCI clinic. These measures reflect not only the care that was directly provided but the programs' success connecting patients with care from external medical providers. The measures include services that were not provided by the PBHCI clinics, such as screening exams for colorectal cancer and follow-up after discharge from a hospitalization for mental illness.

Methods

This study used Medicaid claims data to estimate the impact of PBHCI grants on utilization, costs of care, and quality, using a difference-in-differences model. This model compared change in the outcomes associated with introduction of the PBHCI program into the grantee clinics with change over the same time period in a set of comparison clinics from the same state that did not receive PBHCI grants. The study was organized as a series of three state-level case studies. States were selected based on a number of state-specific characteristics (e.g., data availability, number of PBHCI grantees).

A group of comparison clinics was selected from each state based on information in the claims data sets. Specifically, we examined four claims-based provider characteristics: pattern of utilization, proportion of claims with a primary diagnosis of BH condition, proportion of claims with a primary diagnosis of schizophrenia, and caseload size. For PBHCI and control clinics, all consumers with at least one visit to the clinic with a diagnosis of a SMI during a year were considered members of that clinic's caseload for that year and, thus, included in analyses.

Three types of outcomes were examined: measures of ED and IP utilization, costs of care, and quality indicators. Utilization measures included any ED or IP visits for BH or PH conditions and frequent ED or IP usage. Cost outcomes included both binary indicators of whether or not an individual used a type of service (e.g., an IP stay) and continuous measures of total costs of care among users of that service (e.g., the total cost for IP stays among individuals with an IP stay). Quality-of-care measures included appropriately receiving services for diabetes monitoring, flu vaccine, cancer screenings, outpatient PH care, and follow-up after hospital discharge.

Results

Utilization, cost, and preventive services were examined in a total of five cohorts of PBHCI clinics: two cohorts in State 1, two cohorts in State 2, and one cohort in State 1. Evidence of PBHCI effects on utilization of ED and IP services was mixed across cohorts, but two clear patterns emerged with respect to frequent use of these services. First, in all five cohorts, PBHCI was associated with a reduction relative to comparison clinics in the proportion of consumers having four or more ED or IP visits, and this reduction reached statistical significance in three of the five cohorts. Second, the reduction in frequent utilization was specific to utilization for PH conditions. In three of the five cohorts, PBHCI was associated with a reduction relative to comparison clinics in the proportion of consumers having four or more ED or IP visits with a primary diagnosis of a PH condition.

For each cohort of clinics, we examined the impact of PBHCI on total costs of care to Medicaid and on costs for specific types of services—outpatient care, ED visits, and IP stays. PBHCI was associated with a reduction relative to comparison clinics in the total costs of care per consumer in three of the five cohorts. The impact of PBHCI on total cost was not statistically significant in the remaining two cohorts. Reductions in cost for specific types of care varied across cohorts. Statistically significant reductions in cost for outpatient services were found in two cohorts: in cost per user of ED services for one cohort and in cost per used or IP services for another cohort. Countervailing increases were found for costs per user of IPs services in one cohort and in two cohorts. PBHCI was associated with higher likelihood of having ED-related costs in one cohort and lower likelihood of having ED-related costs in another.

Few of the quality-of-care measures for primary care services were impacted by PBHCI, either positively or negatively. There did not appear to be a pattern to the effects that were found. An exception was a pattern of negative effects of PBHCI on quality indicators for State 3—that is, PBHCI clinic consumers were less likely to have received appropriate services, such as diabetes screenings, than comparison clinic consumers. It is important to note that consumers (in PBHCI or comparison clinics) may indeed have received such services despite these services not being reflected in the claims data, especially if grant funds were used to cover these services.

Conclusion

The current study on the impact of PBHCI on utilization of ED and IP services, total costs of care, and quality of care received for Medicaid beneficiaries yielded mixed results. We did find some evidence that PBHCI can be successful in producing positive changes in consumer health care utilization patterns. In particular, there was evidence that, in some of the groups of clinics studied, PBHCI reduced frequent utilization of ED and IP services, increased ambulatory follow-up after an ED or IP visit, and reduced total per-person costs of care to Medicaid. While there was considerable variation in these effects across groups of clinics studied (across states and years awarded), there were no results in which PBHCI significantly increased total per-person costs of care to Medicaid. Although our findings regarding the impact of PBHCI on quality of care did not yield positive results, Medicaid claims data may not reflect all services provided to consumers. In particular, care assessed by quality measures such as appropriate diabetes screening may have been paid with grant funds and thus may not be reflected in claims.

Results of this study should be interpreted in the light of the following limitations. First, the study was conducted in three of the 32 states that hosted PBHCI clinics during this time period. Given the variability in the results, even across these three states, it is reasonable to infer that there is wider variability in PBHCI impacts across the country. While the results demonstrate that PBHCI can have positive impacts on utilization and costs, they do not allow us to draw conclusions regarding the overall impact of the program on a national basis. Second, this study was conducted using entire clinic caseloads, while only a subset of individuals were actually enrolled in the PBHCI program. The apparent impact of the program may have been reduced by this more inclusive sample. Third, the PBHCI program requirements were being revised across the cohorts studied here and were further revised for the cohorts that came after. Therefore, these results should be interpreted as reflections of the impact of the early phase of the program. Later cohorts, which followed requirements revised in light of these early experiences, may have had different results.

Our findings raise a number of questions regarding the mechanisms of change that could be further investigated for lessons regarding continuing improvement in care. For example, although PBHCI impacts on total costs of care were similar across cohorts, the pathways through which those outcomes were achieved appear to be different in each cohort. This heterogeneity, which may result from different program implementation strategies or from different pre-PBHCI systems, deserves further investigation.