Serving Persons With Severe Mental Illness in Primary Care–Based Medical Homes

The disproportionate burden of general medical comorbidities among people with severe mental illness suggests the need for more disease prevention and management. For instance, cardiovascular, pulmonary, and infectious diseases have been estimated to cause over half the premature deaths among people with schizophrenia (23). Primary care offers a potential intervention point for modifying health behaviors, such as counseling about smoking cessation or dietary changes.

Specialty mental health care visits are important for diagnosis and treatment adjustments. In addition, psychosocial therapies are almost exclusively delivered by specialty mental health care providers or in specialty treatment settings. However, primary care may substitute in part for some specialty mental health services, such as medication management and care coordination among mental health providers and services.

Medication adherence is notably low in the population with severe mental illness (24,25). Low medication adherence has been identified as a primary reason for hospitalization among individuals with schizophrenia (24,26).

Preventive health care use may also reduce mortality through timely identification of treatable diseases. For instance, lipid profiles can determine the risk of heart disease, one of the leading causes of death among persons with severe mental illness (27). Despite clinical consensus on the need for metabolic screening (28,29), screening rates for people taking antipsychotics remain low (30). Rates of cancer and resultant mortality have also been shown to be disproportionately high among people with mental illness (31,32). Screening for cancers has been shown to be lower in this population despite the greater risk (33,34).

Patient-centered medical homes have been gaining momentum as a model of primary care (35,36) and, beginning in 2008, have been recognized by the National Committee for Quality Assurance (NCQA). The medical homes approach is based on greater coordination of care among patients and providers, greater use of high-quality and safe medical technologies, a shifting of focus toward the full set of personal circumstances that may affect health status, greater access to care, and greater communication between patients and their provider team (37). Patient-centered medical homes may require transformation of Medicaid financing to incorporate the principles that define medical homes, which largely occur outside the fee-for-service structure (38).

Payment system reforms targeting persons with mental illness in insurance programs often focus on controlling mental health expenditures rather than on extracting efficiencies from general medical care. Reforms such as mental health carve-outs or prior authorization for psychotropic medications have demonstrated substantial reductions in expenditures on mental health treatments, often improving the efficiency of service provision. Mental health carve-outs, for example, have a demonstrated track record of reducing mental health expenditures by up to 40% without reducing the quality of care received (39).

These efficiency gains have generally been achieved through creating greater access to appropriate outpatient mental health care and offsetting reductions in psychiatric hospital utilization. Improvements in primary care for persons with mental illness have typically taken the form of integrated models of care, where a specialty behavioral health provider is colocated in primary care settings (40). This approach has been useful for less severe mental illnesses (1), but the evidence base for persons with severe mental illness has not been established. Given the disproportionately high level of chronic medical conditions experienced by persons with severe mental illness, it is likely that further efficiency gains can be made from transforming the delivery of medical care, although the degree to which this is true is unknown.

North Carolina’s Medicaid program has been a leader nationally in developing medical homes for patients with complex health problems (41–45). The process of phasing in the Medicaid medical homes model began in 1996 (41). Physicians organized themselves into nonprofit networks to support an enhanced version of primary care case management. Networks help practices coordinate care and implement disease management, case management, and other special emphasis programs for patients with chronic conditions (46). During our study period, both practices and networks received a $2.50 per-member per-month (PMPM) fee for serving as a medical home; rates have since increased. Providers that serve as medical homes receive enhanced training in team approaches in detecting and managing a number of chronic diseases, including diabetes and asthma. The medical homes program in North Carolina predates NCQA recognition, but North Carolina currently has one of the highest such patient-centered medical home recognition rates, with over 340 recognized practices statewide as of September 2012 (47).

We are unaware of any previous empirical tests of the effects of medical homes on care utilization for people with severe mental illness. However, previous research on other populations has found that medical homes improve use of preventive care (44) and increase quality (48). We therefore sought to answer two questions. First, would adult Medicaid enrollees with severe mental illness enrolled in medical homes use more primary care, use less care in emergency departments, and have greater medication adherence? Second, would changes in care from participation in a medical home be uniform across the three severe mental illnesses examined (schizophrenia, bipolar disorder, and major depressive disorder)?

We compared outcomes related to the use of health care services by persons with severe mental illness who were ever enrolled in medical homes with outcomes of similar persons with severe mental illness who were never enrolled in medical homes. During our study period (2004–2007), medical home enrollment was largely voluntary, raising the risk that selection effects might drive any observed differences in outcomes. To control for measurable factors that might confound focal estimates, we used 1:1 propensity score matching. Specifically, we ran logit models of ever enrolled or never enrolled in a medical home at the person level for each diagnostic group by using demographic characteristics and health condition indicators, including alcohol and drug use disorders (all ICD-9 codes indicating use, abuse, or dependence), other psychiatric and medical conditions, and a count of the total number of conditions (49). We selected propensity score–matched observations within one-quarter of a standard deviation caliper. Covariate balance in the matched samples was attained for all variables, as determined by a difference in covariates of less than one-half of a standard deviation (SD) (50). For most variables considered in this study, the difference was less than .2 SD.

We used generalized estimating equation (GEE) models of person-month–level outcomes from the matched sample. In its parameter estimates, GEE specifically accounts for the correlation of repeated observations of individuals. The gamma family was used for the adherence models, the Poisson for visit counts, and the binomial for binary variables on any service use and preventive care. Models with count and continuous outcomes used a log link to account for the right-skewed distribution of variables. Preventive care measures were run on person-level data by using generalized linear models with a binomial link.

Because of concerns about selection bias on unobservable differences by medical home status, we also ran our analyses using person-level fixed effects, which control for time-invariant differences between the two groups on measurable as well as unmeasurable factors through the inclusion of individual intercepts (51). Results were very similar to those reported here, indicating that unmeasured selection of time-invariant variables was not confounding these results.

We obtained Medicaid claims and enrollment data from the North Carolina Division of Medical Assistance through the Carolina Cost and Quality Initiative (www.shepscenter.unc.edu/ccqi) for 2004–2007, a period when there was growing enrollment in medical homes by Medicaid enrollees with severe mental illness. This period is relevant to other states only now ramping up their medical homes programs. Medicaid claims include all services and medications paid for by the Medicaid program. Adults age 18 or older with at least two outpatient visits or one inpatient stay with a diagnosis of bipolar disorder, schizophrenia, or major depression were identified in the claims data (N=272,149). Persons meeting criteria for bipolar disorder or schizophrenia were excluded from the sample with major depression. Persons dually enrolled in Medicare or who had Medicaid claims indicating nursing home use were also excluded because of the lack of data on pharmacy use after the implementation of Medicare Part D (N=89,110). Months with partial Medicaid coverage or with less than 90% of days enrolled were also excluded (N=3,601). Data were aggregated to the person-month level. The final matched sample size was 1,713,357 observations for 64,617 participants. The study was reviewed by the University of North Carolina Institutional Review Board.

Time-varying monthly outcomes included an indicator of any visit and the number of visits to primary care providers; any visit and the number of visits to specialty mental health providers; medication adherence for each target medication class; and emergency department visits. Medication adherence was calculated separately for antidepressants and antipsychotics according to the proportion of days covered (PDC), reflecting the proportion of days in a month for which the study participant had a filled prescription of any medication in each target class (52,53). Months without any dispensed medication were assigned an adherence value of 0 under the assumption of recommended continuous treatment. Because persons with bipolar disorder use multiple classes of medications depending on current symptoms, we omitted adherence results for that sample. Primary care visits were identified by using provider specialty codes of family and internal medicine in the claims data files, and specialty mental health visits were identified as claims for psychiatrists, psychologists, social workers, or counselors. We examined HEDIS indicators for preventive general medical care using procedure codes in the claims data files, including the receipt of cholesterol screening and cancer screening for age- and gender-appropriate populations according to the American Cancer Association guidelines—namely colorectal cancer screening for enrollees age 50 and older, breast cancer screening for women age 40 and older, and cervical cancer screening for women ages 21–65. Because of the infrequency of screening services, we modeled the receipt of each service at any time over the four-year study period, controlling for months of Medicaid enrollment and yielding one observation per person.

Medical homes enrollment was determined on a monthly basis regardless of the number of months enrolled.

In the matched samples (Table 1), we included 7,228 adult Medicaid enrollees with schizophrenia, 13,406 with bipolar disorder, and 45,000 with major depressive disorder. The average ages of persons in medical homes were similar across conditions, as were the number of comorbidities and the percentage who were Latino and of nonwhite, non–African-American race. The percentage of African Americans and the percentage of males both varied widely across the three conditions.

After controlling for baseline covariates, we found that persons with severe mental illness in medical homes had greater rates of use of both primary and specialty mental health care, compared with propensity matched controls (Table 2, column 3). The association between being enrolled in a medical home and having any primary care visits was similar across diagnostic groups, ranging from a 24 to 26 percentage point increase in the probability of one or more primary care visits per month (p<.01)—that is, a 57%−85% proportionate increase in the rates of primary care use compared with rates for persons not in medical homes. The effects of being enrolled in a medical home on the number of primary care visits were also similar across groups, amounting to about half (.54 to .61) a primary care visit per month. Enrollees in Medical homes had slightly increased use of specialty mental health care, representing less than a 1 percentage point increase, or a 1%−2% relative effect (p<.01). The increase in the number of specialty mental health visits ranged from about a fifth of a visit per month (.20) for people with major depression to more than half a visit per month (.62) for people with schizophrenia.

Medication adherence, as measured by the PDC, was relatively low. In the control group, patients with schizophrenia had medication for an average of 42% of days in a month over the study period, and those with depression had medication for 30% of days. Adherence among persons with schizophrenia in medical homes was 11 percentage points greater than adherence in the matched control group in propensity-adjusted results (p<.01), and adherence was 13 percentage points higher for those in medical homes with major depressive disorder (p<.01). These increases amount to a 26%−43% relative increase over adherence rates in the sample of patients not in medical homes. We also found lower rates of emergency department use in propensity-adjusted models for all three diagnostic groups enrolled in medical homes (p<.01), although the effects were modest, amounting to less than 1 percentage point decrease in use, or a 5% proportional effect (p<.01).

Even though medical home enrollees had greater use of outpatient services, medical home enrollment was associated with significantly more use of preventive services only for people with major depressive disorder (Table 3). For this group, enrollment in a medical home was associated with an almost 2 percentage point greater probability of having a lipid panel (p<.01), a 2.3 percentage point increase in the probability of being screened for colorectal cancer (p<.05), a 1.5 percentage point increase in the probability of being screened for breast cancer (p<.05), and a 1.2 percentage point increase in the probability of being screened for cervical cancer (p<.05). Although modest in absolute terms, given the low rates of screening in this population, these differences represent a 5%−12% relative increase, the highest being for breast cancer screening.

In this study, Medicaid enrollees with severe mental illness in a medical home had considerably greater use of primary care and higher medication adherence and modestly greater use of specialty mental health services than Medicaid enrollees who were not in a medical home, even after the analysis balanced the matched samples on an array of health comorbidities and sociodemographic characteristics. Evidence suggests that such enhanced health care use will decrease symptomatology (54,55) and the need for emergent care (56). It is somewhat surprising that the modest incentives provided by the Community Care of North Carolina program during our study period, amounting to $5 PMPM split equally between practices and networks, were associated with the relatively large effects on primary care use and medication adherence reported here.

Of interest, these access gains translated into increased use of preventive services only for persons with a major depressive disorder in medical homes. This difference across groups might reflect the greater integration of depression treatment in primary care and greater physician understanding and comfort in managing depression as opposed to psychotic conditions. This interpretation is also consistent with the finding of a proportionately greater number of specialty mental health visits for persons with schizophrenia and bipolar disorder compared with those with major depressive disorder.

A number of limitations should be considered in interpretation of these results. Medicaid claims data reflect only services paid through the Medicaid program and may not reflect all services received. Measures of severity of the target psychiatric conditions were not available in the claims data. Measures of medication adherence are based on filled claims only and may not reflect actual pill taking. If selection into the medical homes programs was not fully accounted for by the baseline demographic and medical comorbidity variables, then some degree of selection bias could have driven the results. However, results from fixed-effects models that controlled for time-invariant differences between those in medical homes and those not in medical homes yielded virtually identical results, indicating that unmeasured time-invariant selection was not likely to bias the reported results; time-varying differences could still pose a problem. Finally, North Carolina has been undergoing a series of mental health service reforms over the past decade (57). We do not believe these reforms disproportionately affected medical home enrollees, but if they did, then they could confound the results here attributed to medical homes.

Findings provide new evidence not previously available about how primary care patient–centered medical homes affect use of general medical care by persons with severe mental illness. Greater knowledge about the overall and conditional effects of medical homes on the quantity and quality of care will assist policy makers in their efforts to implement the ACA and to improve the efficiency and effectiveness of care for persons with mental illness in primary care settings.