Service Use by Medicaid Recipients With Serious Mental Illness During an RCT of the Bridge Peer Health Navigator Intervention

There is considerable evidence that individuals with serious mental illnesses suffer from high rates of morbidity and that they die 10 to 20 years earlier than their peers without a mental illness (1–4). In a meta-analysis by Walker and colleagues (2) of studies of mortality among people with serious mental illnesses, most deaths were determined to be due to acute and chronic illnesses (67%), rather than to other causes, such as accidents or suicide. High rates of cardiovascular, metabolic, and pulmonary diseases (1, 5) are endemic to this population, and addressing the health of individuals with serious mental illnesses is paramount. Their higher rates of health issues are often attributed to poor health habits (such as physical inactivity, low-quality diet, smoking, and treatment nonadherence) (6, 7) or to the side effects of their psychiatric medications (1, 8). However, inadequate health care provision and poor service coordination to prevent and treat these conditions also affect their health status (1, 9, 10).

Improving the health of and health care for this population requires addressing issues on system, provider, and individual levels. Numerous strategies and national initiatives are employed to improve the general medical care and physical well-being of this population (11–22), yet few evaluations have included Latinos (23) or have measured increased skill development of self-efficacy or self-management behaviors (11, 24).

The Bridge Intervention

The purpose of this study was to use Medicaid record data to examine general medical service use by individuals participating in the Bridge intervention (25–27) and compare it with service use by a waitlist control group. Previously, in a randomized controlled trial with 151 participants (the waitlisted group received the intervention after 6 months), we compared outcomes self-reported by participants in the Bridge and by those in a control group (26). Compared with the waitlisted group (N=75), Bridge participants (N=76) reported increased access to and use of primary care health services; decreased preference for emergency or urgent care services or avoidance of these types of services, compared with primary care services; and better relationships with their primary care providers. They also reported improved detection of chronic medical disorders, reductions in pain, and increased confidence to self-manage their health care, compared with the waitlisted group (26, 28).

The Bridge model is guided by Gelberg and colleagues’ (29) behavioral model of health service use for vulnerable populations, which was adapted for those with serious mental illness (16). In brief, the Bridge is a manualized comprehensive health care engagement and self-management intervention with three phases over 6 months; it is designed to teach individuals the skills to access and manage their services effectively (25–27, 30). With use of behavioral strategies of modeling (“for them”), coaching (“with them”), and fading (“by them”), the goal is to promote clients’ skill building so that they can achieve maximum independence in managing their health care over time. Clients’ individual skills and readiness determine the length of each phase. Through one-on-one sessions with the peer health navigator in real-world settings, participants learn skills to manage their health and health care to improve their access to and use of health care to prevent, identify, or manage health conditions. The model is comprehensive because it connects people to preventive, routine care (primary care and dental), specialty care, pharmacy services, and emergency services. It is an engagement model because many persons may deprioritize their health when they are faced with other pressing issues, such as housing or psychiatric issues, and they may need assistance to engage with health care. The Bridge is predicated on ensuring that individuals are activated about and have the self-management skills to manage their health and health care. The core activities of the Bridge consist of four components: engagement, assessment, and planning; coordinated linkages; patient education; and cognitive-behavioral strategies.

The Bridge is a peer-delivered model. Peer providers are individuals who have lived experience with mental illness. They can draw upon their experiences to empathize with their clients while promoting recovery and wellness. Peers are a rapidly growing segment of the mental health workforce in the United States and part of several evidence-based health care interventions for individuals with serious mental illness (23, 31, 32).

In the study reported here, the intervention’s assessment was extended with an examination of how participation in the Bridge influenced the use of Medicaid-covered health care services. We hypothesized that participation in the intervention would lead to more appropriate use of health care services, defined as reduced emergency room care and increased use of outpatient and inpatient general medical services. We use the term “appropriate,” because avoidance of outpatient or hospital care or failure to identify health conditions may be a main driver of the higher morbidity and mortality rates for this population.

Methods

Setting and Sample

The study was conducted from March 2014 to September 2015 at a large community mental health agency in Southern California that provides intensive, field-based services to individuals with serious mental illness, known as full-service partnership (FSP) services or field-capable clinical services (FCCS). Agency staff recruited participants from their existing caseloads by using a screening form that focused on unaddressed general medical health issues or service needs (see Brekke et al. [28] for details). Of the 151 persons who participated in the study, we received Medicaid data for 145 (73 in the immediate treatment group and 72 in the waitlist group). One participant’s data were dropped because more than 85% of data were missing, for a final sample of 144 individuals. Inclusion criteria were age 18 or older, admittance to one of the programs at the study site, local residence for at least 3 months, English fluency, capacity to give informed consent, and diagnosed as having schizophrenia, schizoaffective disorder, bipolar disorder, or major depression. Exclusion criteria included being under conservatorship, unable to give informed consent, or currently hospitalized.

Sources of Data

Medicaid records of service use were analyzed for 144 participants from 6 months prior to the baseline assessment to 6 months after baseline. Self-report data were collected at baseline, 6 months, and 12 months. Peer navigators maintained contact logs of all their phone calls and in-person meetings with participants, as well as navigation-related data. All procedures were approved by the institutional review boards of the University of Southern California and the University of California, Los Angeles, and the Committee for the Protection of Human Subjects of California. All participants signed consent and HIPAA agreements related to participation in the study and for access to their Medicaid records.

Measures

Service use.

Medicaid records of services were categorized into three domains: emergency room, general medical outpatient, and general medical inpatient. Counts of the dates of service visits for each type were created in two blocks (6 months prior to baseline and baseline to 6 months). Outpatient services included visits to primary and specialty care services for diagnosis, treatment, and follow-up care, unless specifically coded as inpatient care by the state.

Intervention contacts.

Navigators recorded their in-person contacts and phone calls when they spoke directly to participants and identified the skills on which they worked with participants. Skills were designed to encourage self-management of health care and included communication with medical staff, following up with medical staff, following treatment plans, making appointments, engaging a buddy, picking up medication, working on diet and exercise, management of anxiety and stress, arranging for lab work, and arranging transportation.

Treatment as usual.

Everyone in the study received treatment-as-usual mental health services, and there was no attempt to alter those services, which consisted of field-based, outpatient services adapted to participants’ needs. Treatment teams typically offered some minimal support for general medical care, such as arranging appointments, giving reminders, and providing help with transportation. However, these activities were usually of lower priority, compared with other pressing issues, such as mental health functioning, housing, legal issues, and enrollment in benefits.

Demographic information.

Participants self-reported their age, gender, race-ethnicity, and income. Primary mental health diagnoses and treatment programs were drawn from mental health records.

Mental health and functional outcomes.

Participants’ clinical psychiatric status was measured with the 24-item Behavior and Symptom Identification Scale (BASIS-24) (33). The items of the BASIS-24 were designed to assess a client’s perception of difficulty resulting from psychiatric symptoms and problems in functional areas (daily living skills, work, and social situations) over the past week. Scoring is normed and standardized (0 to 4 weighted sum). We present the subscale scores and total score for descriptive purposes.

Medical diagnoses.

Participants were presented with a checklist of 10 chronic health diagnoses, (e.g., diabetes, high blood pressure, and heart disease), and they checked off any applicable diagnosis (“ever” at baseline or “in the last 6 months” at subsequent assessments).

Statistical Analyses

All participants who completed the 6-month follow-up assessment (treatment group, N=61; waitlist control group, N=62) were included in the analysis as intent to treat (all cases included regardless of their receipt of the intervention). We examined the intervention effects by using comparisons of service use from baseline to 6-month follow-up between the group participating in the Bridge intervention and the waitlist control group.

All data were inspected for normality, overdispersion (the mean being greater than the standard deviation), and zero inflation. Because the service use variables are counts of service dates, we examined the data for whether Poisson, negative binomial, or zero-inflated regression models were the best fit to the data by using the COUNTFIT program in Stata, version 14. In instances where zero-inflated models were preferable, we also compared the model fit with logistic models. Zero-inflated models are split into two parts. The first part predicts the dependent variable for nonzero values. Second, the inflated model examines predictors of the zero values for true zeros or error. Regression models comparing the count of services over 6 months (from baseline) were used with each outcome, and the models tested study group and controlled for service use in the 6 months prior. For interpretation, the incident rate ratios (IRRs) are reported instead of the coefficients and predictive margins to estimate the effects of the intervention when the effects of study group were examined.

Results

Sample Baseline Characteristics

Baseline demographic and clinical characteristics of the full sample and of the intervention and waitlist groups are presented in Table 1. At baseline, no significant differences between groups were noted for any demographic or clinical factor or for the number of chronic health conditions (26).

Intervention Character and Fidelity

The Bridge intervention is an individualized program, and thus the number of contacts and skills worked on vary according to each client’s need and engagement. The mean±SD number of in-person contacts during the 6-month intervention period for the intervention group was 4.88±4.76, 20% of which included the navigator’s accompanying a participant to the doctor; the mean number of phone calls during the period was 6.18±6.68 (range 0–50). The mean number of skills worked on was 7.88±14.94. The large standard deviations for these variables reflect the individualization of the intervention.

All navigators completed a 2-day initial training on the intervention and completed a peer provider certification course. A supervision team of the principal investigator (J.S.B.), a supervisor who is also a mental health peer, and a nurse provided coaching in the initial sessions with clients and weekly supervision and evaluated navigators for their skills development and fidelity to the intervention. Intervention fidelity was measured with a 20-item instrument developed in our pilot work that uses interviews, role playing, observations, and case records. All navigators were evaluated after 4 months and rated at the “good” level of the scale or higher.

Between-Group Comparisons of Service Use

Service use outcomes were medical care visits to emergency rooms, general outpatient settings, and general inpatient settings (Table 2). During the 6 months prior to the study period, 33% (N=24) of the waitlist group and 35% (N=25) of the intervention group used the emergency room. In the 6 months after baseline, 35% (N=50) of the total sample used the emergency room (31% [N=22] of the intervention group and 39% [N=28] of the waitlist group). Emergency room use was tested by using zero-inflated negative binomial regression. Across all participants, including those with and without emergency room use, the mean number of emergency room visits by the intervention group was less than half the mean number of visits by the waitlist group (adjusted means, 0.72±0.19 versus 1.59±0.42; contrast comparison χ²=7.99, df=1 and 143, p=0.005).

For inpatient visits in the 6 months since baseline, 7% (N=5) of the intervention group had an inpatient visit, compared with 13% (N=9) of the waitlist group. For outpatient services, the adjusted mean number of outpatient visits in the intervention group was 3.36±5.13, compared with 3.06±4.54 visits in the waitlist group. Inpatient visits and outpatient visits were not significantly different between groups, as assessed by logistic regression and negative binomial regression, respectively.

Discussion

The Bridge intervention is designed to activate individuals about their overall health and improve their use of general medical care in the hopes of lowering morbidity and mortality. According to the Medicaid records of study participants, the Bridge peer health navigator intervention significantly decreased emergency room visits by over 50%, compared with the waitlist control group. This shift in emergency room use was also reflected in the self-reported outcomes in the previous study, in which Bridge participants reported a decreased preference for emergency services, in favor of primary care settings (26). Given the need to address the high costs of emergency room use by this population, this is an important domain of service use for remediation.

Even though Bridge participants had less emergency room use than the waitlist group had, no between-group differences were found for inpatient service use. Given the low rates of inpatient care in our sample, future studies should explore this relationship with a larger sample. We also anticipated that Bridge participants would significantly increase their use of outpatient medical services. Although this outcome was affirmed by participant self-reports (26), the analysis of Medicaid records did not find a significant increase in use of outpatient medical services (although the two studies differed in the types of care included; the self-reported visits included eye and dental visits and excluded specialty care). In the paper that examined self-reported outcomes, the detection of chronic health conditions increased in the intervention group (26), and such detection was associated with use of general medical outpatient services (Spearman r=0.31, p=0.02), which could suggest that other factors, such as monitoring of new conditions, may have influenced rates of outpatient visits. Exploration of additional factors that may have influenced rates of service use should be explored in a larger randomized controlled trial. Our self-reported findings of increased detection of chronic conditions (26) in the intervention group but not in the waitlist group might reflect the need for health care services that might otherwise have been missed without the presence of the navigator. Cumulatively, these findings in conjunction with previous self-reports (26, 28) suggest that the Bridge intervention is a valuable model for improving health care utilization by individuals with serious mental illnesses. Future studies with larger samples are required to understand the mechanisms of change in the Bridge, but this is the first study to use data from Medicaid records to assess outcomes of a peer-led intervention (11, 24).

This study had some limitations. The analyses were limited to individuals who were enrolled in Medicaid and to a 12-month period. Participants could have received dental and vision services not covered by Medicaid (which were excluded from the analyses). However, all participants self-reported that all of their primary care, specialty care, hospitalization, and emergency care services that they used were covered by Medicaid, and their self-reports were corroborated by case managers.

Conclusions

Addressing health disparities among individuals with serious mental illnesses is a high priority, as is improving the quality of life of this population and reducing early mortality. Mental health peers delivered an intervention that had a positive impact on health service use and consumer-oriented outcomes.