Integrating Primary Care Into Community Mental Health Centers: Impact on Utilization and Costs of Health Care

PBHCI and propensity-matched comparison groups were not significantly different on any of the variables used for matching (Table 1). PBHCI and comparison groups also did not differ on the number of days in the postperiod (PBHCI, mean=973±337, range 203–1,410, and comparison, mean=956±332, range 232–1,415 [clinic 1]; PBHCI, mean=845±371, range 104–1,420, and comparison, mean=834±391, range 141–1,420 [clinic 2]).

PBHCI clients at both sites had a higher average number of chronic illnesses than clients in the comparison groups, as measured by the Chronic Disability Payment System (CDPS) (13) (Table 2). PBHCI clients were also more likely than comparison clients to have diagnoses in the high psychiatric subcategory, denoting a more serious psychiatric condition (57% and 22%, respectively, p<.001 [clinic 1]; 30% and 13%, respectively, p<.001 [clinic 2]).

Table 3 presents a summary of unadjusted pre- and postperiod values for each of the dependent variables for each clinic. Table 4 presents results of the DID analyses comparing utilization and costs for outpatient medical care, inpatient hospital admissions, and emergency department visits for PBHCI and comparison groups at each clinic. Results of the DID analyses showed significant increases in the proportion of PBHCI clients using outpatient general medical services in the period following PBHCI enrollment compared with the comparison groups at both clinic 1 (p<.003) and clinic 2 (p<.001). Specifically, at clinic 1 the percentage of clients using outpatient medical services in the pre- and postperiods increased from 80% (N=298) to 92% (N=343) among PBHCI clients but changed little in the comparison group (from 61% [N=228] to 60% [N=224]). At clinic 2, the percentage of clients using outpatient medical services increased from 39% (N=152) to 76% (N=296) among PBHCI clients but changed little in the comparison group (from 28% [N=109] to 31% [N=121]).

At clinic 1, the intervention was also associated with a reduction in the proportion of clients with an inpatient hospital admission (from 18% [N=67] to 12% [N=45], p<.04), compared with an increase for the comparison group (from 15% [N=56] to 17% [N=63]).

At clinic 1, the intervention was also associated with a reduction in the proportion of clients with an inpatient hospital admission (from 18% [N=67] to 12% [N=45], p<.04), compared with an increase for the comparison group (from 15% [N=56] to 17% [N=63]). In addition, at clinic 1, DID analyses revealed trends for a decrease in the number of PMPM inpatient admissions (p<.08) and in PMPM inpatient costs (an estimated average PMPM reduction of $217.68, p<.06). At clinic 2, the percentage of clients with an inpatient hospital admission in the preperiod and postperiod did not differ for either the PBHCI clients (21% and 22%, respectively) or the comparison group (19% and 20%, respectively).

Although PBHCI participants showed a decrease in emergency department use and costs at both sites, these decreases were not significantly different from those observed in the respective comparison groups. Between 38% and 55% of PBHCIa and comparison clients used emergency department services in the pre- and postperiods.

PBHCI is the largest demonstration of integrated care for adults with serious mental illness in the United States and serves as a showcase of a novel model in which primary care is introduced in mental health settings where clients with mental illnesses already receive services. Although previous evaluations of PBHCI have described clinical outcomes, none to our knowledge have reported on the model’s impact on services and costs—information that payers need to assess the risks and potential payoffs of investing in providing integrated care services.

A primary goal of the SAMHSA PBHCI program is to increase clients’ use of primary care services. Results indicated that both programs studied here were successful in meeting this goal, even at clinic 1, which offered integrated care before the grant. This observation suggests that boosting funding even for programs with existing integrated services may have benefits. As such, identifying optimal funding levels for programs with different histories and levels of integration may be an important topic for future research.

The PBHCI program is based on the assumption that integrating behavioral and general medical care for persons with serious mental illness provides critical preventive services and acute care for emergent issues in a timely manner, thus reducing the use of inpatient hospitalization and emergency department services and associated costs. In this study, we saw evidence of reductions in inpatient hospital admissions among PBHCI clients at clinic 1, the clinic with existing integrated care services. These reductions were associated with a trend for an average cost reduction of $217.68 PMPM coupled with a trend for an average increase in outpatient medical costs of $50.56 PMPM, suggesting that PBHCI was a good investment at clinic 1. Such savings did not occur at clinic 2, which was new at providing integrated care. It is possible that the average follow-up period at clinic 2 (2.3 years, compared with 2.7 years at clinic 1) was not long enough to capture changes in use of inpatient hospital services, given that the clinic was identifying serious medical problems, often for the first time, in a population that had not been receiving general medical care. In fact, medical staff at clinic 2 argued that it may have been necessary for the program to increase inpatient hospital services for the first few years in order to serve pent-up need for treatment of serious medical conditions that had been ignored or undetected. Only after the program had addressed such needs, they contended, would it be reasonable to expect long-term reductions in use of inpatient services.

Clinic 1, on the other hand, had been providing integrated services long before PBHCI was introduced, so there was probably less pent-up need or demand for treatment of serious medical conditions, making it more possible to reflect the longer-term impact of an integrated care model on inpatient hospital use.

It is also possible that the relatively high proportion of clients with a history of homelessness at clinic 2 may have served to dilute the impact of the PBHCI intervention at that site, given the disruptive nature of homelessness on follow-through with taking prescribed medications and other medical recommendations. Given that almost two-thirds of clients at clinic 2 were homeless (compared with one-third in clinic 1), a more intense intervention that included the provision of housing may have been necessary to affect inpatient hospital use (14). We suggest these as important questions for future research. In any case, it is clear that the proportional increase in utilization of outpatient treatment between the pre- and postperiod, which was relatively small in clinic 1 (from 80% to 92%) and relatively large in clinic 2 (from 39% to 76%) was not, in itself, a good predictor of change in inpatient hospital utilization and costs. It is likely that other factors, such as history of integrated care or client characteristics, may also play an important role in determining the total cost of care.

We saw little evidence of change in emergency department use at either site. In fact, 40% to 52% of clients continued to use the emergency department in the year after enrolling in PBHCI. Similar findings have been reported in studies of homeless veterans (15,16) and persons with severe or substantial problem drug use (the majority of whom were homeless, mentally ill, or both) (17), even when access to primary care was available. One explanation for this pattern is that emergency department use may reflect unmet psychosocial needs of the homeless population (16). Supporting this explanation is the finding that providing services that specifically target homelessness has been successful in reducing emergency department use (14,18,19). Thus, to have an impact on emergency department use for the high-risk, vulnerable, and frequently homeless clients studied here, it may be necessary to supplement PBHCI-type services with integrated psychosocial services beyond usual care that specifically target homelessness. We recommend future research to test this idea with populations like the one studied here.

Important strengths of this study included the fact that this was the first utilization and cost study of PBHCI—information that is key for sustainability of individual programs. Data used in this study were objective and systematic, coming from health care claims, and were subjected to rigorous analysis that employed DID analyses with propensity-score matching. Because of its size and breadth, the PBHCI initiative provides a unique opportunity to learn about the range of benefits and challenges of integrating primary care into mental health centers. It also offers the opportunity to examine this important intervention with a naturalistic, ecologically valid design. Finally, this study is unique in that it examined potential effects of mature versus de novo integrated clinic services.

The study also had a number of limitations. First, although PBHCI and comparison groups appeared well balanced on variables used to create the propensity score match, differences between the two groups on two variables suggest potential sources of selection bias. PBHCI clients were more likely to receive outpatient medical services at baseline compared with the comparison group, and the percentage of clients with the most serious mental health problems was higher among PBHCI clients compared with the comparison group. Second, it is likely that more CDPS categories were reported for clinic 1 clients because of the greater frequency of medical contacts among clinic 1 versus clinic 2 clients and not necessarily because they had more medical problems. Third, the comparison group may have experienced spillover effects associated with being served contemporaneously at the same sites offering PBHCI.

Fourth, general medical services could have been undercounted, given that we had access to only service records at the single medical center with which the clinics were associated, and clients could have received services at other clinics and medical centers. Fifth, some PBHCI clients had a limited treatment period, but this outcome reflects real-life results with a seriously disabled safety-net population. Sixth, it was not possible to determine whether emergency department visits or inpatient hospital admissions were due to general medical or psychiatric problems. Seventh, we cannot provide data on the cost of the PBHCI program, given that such estimates were beyond the scope of this study. Finally, we lack detailed information on how community mental health centers implemented the project.

Findings suggest that investments in primary and behavioral health care integration can improve access to primary care for persons with serious mental illness, regardless of whether the clinic has a history of integration. Such investments also appeared to curb hospitalizations in the more established program, with a trend for potential savings of $217.68 PMPM, more than four times greater than the average increase in outpatient medical costs ($50.56 PMPM) associated with such investments, suggesting that PBHCI was a good investment at this clinic. Considering the medical complexities associated with serious mental illness and the historic lack of access to primary medical care among individuals with serious mental illness, such findings are promising and argue for expansion of integrated services for this population. They also have the potential to inform Medicaid and local hospital systems about the risks and payoffs of investing in integrated care for persons with serious mental illness and to create the opportunity for these payers to fund such services after federal PBHCI funding ends.

The authors thank Christina Clayton, L.I.C.S.W., C.D.P., and Nancy Sugg, M.D., M.P.H., for their vital contributions to this study.