Enhancing Treatment Reengagement for Veterans With Serious Mental Illness: Evaluating the Effectiveness of SMI Re-Engage

Care continuity for veterans with serious mental illness is a long-standing priority of the Veterans Health Administration (VHA) (1,2). Accordingly, from 2007 to 2009 the VHA Office of the Medical Inspector conducted a quality improvement demonstration project to facilitate treatment reengagement for individuals with prior VHA diagnoses of serious mental illness (specifically, schizophrenia or bipolar disorder) who had not received VHA care for at least 12 months (3,4). The project was developed in response to reports that 20% of VHA patients with serious mental illness had gaps in VHA care of at least 12 months over a five-year period (2,5); that patients with serious mental illness often have chronic general medical comorbidities necessitating treatment (6); and that among patients with schizophrenia, disengagement from VHA outpatient care was associated with increased risk of unforeseen mortality (defined as deaths among patients younger than age 80 with two or fewer days of inpatient hospitalizations in the year prior to death) (7).

Using the VHA National Psychosis Registry (NPR) (8), the project identified veterans with serious mental illness who had not received VHA care for at least 12 months and provided their contact information to clinicians at VHA medical centers nationwide. Clinicians then attempted to contact veterans who were appropriate for outreach (for example, veterans who were noninstitutionalized and whose current contact information was available) and facilitate their return to VHA care. Over a 20-month period, 71.6% of veterans returned to VHA care and the odds of mortality for those who returned to care were lower than for those who did not (4).

Favorable outcomes led VHA to issue a directive, Re-Engaging Veterans with Serious Mental Illness in Treatment, mandating that a reengagement program (SMI Re-Engage) (9) become a standard part of care. SMI Re-Engage was nationally implemented in 2012 at U.S. Department of Veterans Affairs (VA) medical centers and community-based outpatient clinics (10) and uses the NPR to identify veterans with prior VHA serious mental illness diagnoses who have not received VHA care for at least 12 months. When indicated (for example, veterans do not appear to be deceased, incarcerated, or otherwise institutionalized and have not returned to care on their own), clinicians conduct outreach and facilitate return to VHA care. The conceptual model for SMI Re-Engage (Figure 1) is that program implementation leads to contact with veterans, resulting in return to care, and reduced mortality risk. Prior research supports the first link; facility-level strategies to increase program implementation was shown to increase odds of contact (11). The analysis reported here evaluated the other links in the model.

As noted above, the demonstration project reported the percentage of veterans who returned to care within 20 months after their contact information was sent to VHA clinicians (4). However, the effectiveness of the outreach intervention was not assessed. The outcome (return to care) was reported for all veterans whom providers attempted to contact, regardless of whether they were actually contacted (4). Therefore, the demonstration project suggested that a population-based outreach intervention was feasible and that veterans would return to care; however, its intent-to-treat design left open the question of whether contact via the outreach intervention facilitates return to VHA care.

In addition, although the demonstration reported an association between return to care and a lower rate of mortality, it did not assess whether returning to care contributed to the observed reduction in mortality rates. Research showed that the mortality rate of veterans who returned to VHA care within the 20-month period (.3% deceased) was lower than the rate for those who did not return (3.9% deceased)—a difference that remained significant after the analysis controlled for covariates (4). However, the analyses did not account for temporal precedence—in other words, veterans may not have returned to care because they died before they could be contacted through the outreach intervention. Furthermore, analyses did not adjust for potential differences in individuals’ propensities to return to care based on demographic or clinical characteristics, for example. Consequently, the design left open the question of whether return to care via outreach intervention contributes to reduced mortality. By adjusting for individuals’ propensities to be contacted and return to care through the SMI Re-Engage program, we can more specifically assess the impact of the intervention on the intended outcomes of return to care and mortality, respectively.

This evaluation tested the hypotheses that SMI Re-Engage outreach contact facilitates return to VHA care and that return to VHA care via SMI Re-Engage reduces mortality risks. Because randomly assigning patients to conditions (contacted versus not contacted; returned to care versus not returned) to test intervention effectiveness was not feasible for an ongoing clinical program, we used propensity risk scoring to adjust for systematic effects of baseline factors on patients receiving the intervention (12,13).

Methods

Data Source and Patient Population

As part of SMI Re-Engage, 3,680 veterans with schizophrenia (ICD-9-CM codes 295.0–295.4 and 295.6–295.9) or bipolar disorder (ICD-9-CM codes 296.0, 296.1, and 296.4–296.8) who had no VHA care within at least 12 months (no VHA care was operationally defined by policy as no VHA outpatient visits or only an inpatient stay of two days or fewer [9]) were identified by using the VHA NPR (14). Contact information for these veterans was provided to the local recovery coordinator (LRC) at each VHA facility in May 2013 and November 2013, excluding decedents at the time of dissemination. (LRCs are mental health clinicians who promote recovery-oriented care [15]). Decedents were identified by using VHA records, including indicators from the Corporate Data Warehouse, Social Security Administration Death Master File, Beneficiary Identification Records Locator Subsystem, and VHA Vital Status file. After performing SMI Re-Engage outreach activities for a veteran, clinicians completed a program evaluation form that described outreach and reengagement efforts and results. Per VHA Handbook 1058.05, the study was operational in nature, did not constitute research, and did not require approval from an institutional review board.

To determine the causal effect of an intervention on an outcome by using propensity scoring, all members of the sample must have an opportunity to receive the intervention (12). Veterans included in the analyses testing the first hypothesis needed to have the opportunity to be contacted through SMI Re-Engage. Therefore, we excluded individuals who were inappropriate for outreach (specifically individuals who were incarcerated or otherwise institutionalized or deceased, who had already returned to VHA care independently, or who were ineligible for care or not a veteran) on the basis of clinicians’ reviews of medical records or administrative data. Of the 3,680 veterans whose information VHA clinicians received, 2,949 veterans (80%) were appropriate for outreach, and clinicians reported attempts to contact them. Listwise deletion removed four veterans with missing data, leaving 2,945 veterans in analyses for the first hypothesis. Analyses addressing the second hypothesis included all veterans who were successfully contacted via SMI Re-Engage (N=886) and thus had the opportunity to return to VHA care through the outreach intervention.

Measures for Analyses of Contact and Return to Care

Returned to care.

The dependent measure in the first analysis was time to return to care. Risk time for return to care began on the date veterans’ information was given to clinicians and ended 545 days later or on the date of return to care, whichever came first. Return to care was ascertained through VHA administrative data (Corporate Data Warehouse and National Patient Care Database) and operationalized as at least one VHA face-to-face outpatient encounter, inpatient hospitalization, or emergency department visit during the subsequent 545 days. Telephone utilization and chart review encounters were excluded to ensure that services performed as part of the outreach intervention were not inadvertently counted as returning to care.

Contact.

In the first analysis, the independent measure was a dichotomous variable that indicated whether a clinician successfully contacted the veteran. Contact was defined as the clinician’s response of “yes” to the question “Were you able to contact the veteran?” on the program evaluation form. Through SMI Re-Engage program training, clinicians were instructed to respond “yes” to this question when they had a telephone conversation with a veteran, correspondence via letters through traditional mail with a veteran (specifically, reaching a veteran via letter and the veteran’s responding via letter or telephone), or a telephone conversation with a current caregiver of a veteran (for example, a spouse responding to the outreach and providing information regarding the veteran’s current status because the veteran was unable to do so) and to respond “no” in all other circumstances (for example, no response to voice messages or letters or next of kin were reached but were not currently caregivers of the veteran).

Measures for Analyses of Return to Care and Mortality

Mortality.

In the second analysis, the dependent measure was time to date of death within 545 days of the date when a veteran’s name was provided to a clinician. Mortality was ascertained through VHA indicators listed above. Risk time began when veteran information was provided to clinicians and ended 545 days later or the date of death, whichever came first.

Returned to care.

In the second analysis, the independent measure was whether a veteran returned to care within the 545-day period (yes or no) and was operationalized as described above. Date of return to care was the date of the first in-person VHA inpatient, emergency department, or outpatient encounter.

Covariates: Demographic, Clinical, and Utilization Variables

Except where otherwise noted, data for covariates were obtained from administrative data based on the last fiscal year the veteran received VHA care. Demographic factors included gender (male or female) and age (≥65 or <65). Clinical factors included service-connected disability (yes or no) and psychiatric diagnosis (schizophrenia or bipolar disorder). Veterans who received both schizophrenia and bipolar diagnoses during the last fiscal year of VHA care were coded as having schizophrenia (14). The Charlson Comorbidity Index (16) accounted for the severity of medical comorbidity during the 12 months prior to leaving VHA care; index scores were categorized (0, 1, and ≥2). Higher scores indicate more severe medical comorbidity.

Utilization variables included the following within the 12 months prior to the veteran’s last VHA visit before loss to care: outpatient mental health visits (none, one to four, and more than four), outpatient primary care visits (none, one or two, and more than two), mental health hospitalizations (yes or no), medical hospitalizations (yes or no), and distance in miles from the veteran’s last known address to the nearest VHA facility (defined by the straight-line distance between the population centroid of the patient’s most recent zip code of residence and the nearest VA medical center or community-based outpatient clinic) (2). Length of time lost to care was operationalized as number of days between the veteran’s last VHA visit prior to the gap in utilization and the date the veteran’s information was provided to a clinician.

Also included as a covariate was the number of attempts to contact the veteran (number of telephone calls made and letters sent) as reported by the clinician on the program evaluation form (one, two or three, or four or more times).

Statistical Analyses

We conducted descriptive analyses of covariates in the full sample of veterans who clinicians attempted to contact through SMI Re-Engage and by return-to-care status and mortality status. We conducted bivariate analyses to assess unadjusted associations between contact and return to care and between return to care and mortality. [Further details about methods are presented in an online supplement to this article.] We conducted Cox proportional hazards analysis (survival analysis) (17) to assess the effect of contact on return to care. Individuals who did not return to care or died during the period before returning to care were censored. Covariates described above were included to isolate the effect of contact on return to care. The analysis included a weight representing the propensity of being contacted, which was calculated with a logistic regression model that used inverse probability of treatment weighting (12,13) [see online supplement for details regarding weight calculations (18–23)].

A second survival analysis (16) with propensity scoring (12,13) assessed the effect of return to care via SMI Re-Engage on mortality among veterans who were contacted through SMI Re-Engage (18–23) [see online supplement]. Individuals who did not die within the time frame were censored. Return to care was entered as a time-dependent covariate to account for the potential for length bias—that is, veterans would have more opportunity to return to care if they lived longer (24).

Post hoc analyses were conducted to further explore the association between contact and type of return-to-care visit among the 941 veterans who returned to care. A bivariate analysis was used to assess the unadjusted association between contact and type of return-to-care visit (inpatient stay or emergency department visit versus outpatient visit) [see online supplement]. Logistic regression weighted for the propensity to be contacted and including the covariates described above was used to assess the adjusted association between contact and type of return-to-care visit (18–23) [see online supplement].

Results

Sample

The mean±SD age of the 2,945 veterans who received outreach was 52.7±14.2. Most were men (90%), and most had a diagnosis of bipolar disorder (65%). Over one-third (36%) had a service-connected disability. Most veterans (76%) did not have any of the general medical comorbidities in the Charlson Comorbidity Index listed in their medical record during the last 12 months in which they received care. During the last 12 months in which the veterans received care, 16% had a psychiatric hospitalization, 7% had a general medical hospitalization, 45% had one to four outpatient mental health visits, and 46% had one or two outpatient primary care visits (Table 1).

Contact and Return to Care

Consistent with bivariate findings [see online supplement], the survival analysis assessing the effect of SMI Re-Engage contact on return to care indicated that being contacted increased the risk of returning to care in the 545-day period (hazard ratio [HR]=3.40, p<.001) (Table 1). Having a service-connected disability was associated with an increased risk of returning to care (HR=1.30, p<.001). Two covariates predicted a lower risk of returning to care: being age 65 or older (HR=.58, p<.001) and having more medical conditions (Charlson Comorbidity Index score ≥2 versus 0) (HR=.65, p<.01). An interaction (HR=.997, p<.001) between risk time and contact (yes or no) was included in the model to account for nonproportional hazards (24).

Return to Care and Mortality

As with the bivariate findings [see online supplement], using survival analysis we did not observe a significant effect on mortality of return to care via SMI Re-Engage (Table 2). Greater mortality risk was associated with being age 65 or older (HR=2.92, p<.05) and having a medical condition (Charlson Comorbidity Index score of 1 versus 0; HR=3.45, p<.05).

Contact and Type of Return-to-Care Visit

The post hoc bivariate cross-tabular analysis indicated a significant association between contact status and type of return-to-care visit among the 941 veterans who returned to care in the 545-day period [see online supplement]. Among veterans who returned to care, those who were contacted were more likely to return to outpatient care than veterans whom providers attempted to reach but could not contact (88% versus 74%) [see online supplement]. Veterans whom providers were unsuccessful in their efforts to contact were more likely to have an inpatient hospitalization or emergency department visit upon returning to care, compared with those who were contacted (26% versus 12%). The association of contact with type of return-to-care visit remained significant after adjustments for covariates and the propensity to be contacted (odds ratio [OR]=2.42, p<.001) (Table 3). Two variables were associated with lower odds of an outpatient return-to-care visit (versus an inpatient or emergency department visit): having a general medical hospitalization (OR=.43, p<.05) or a mental health hospitalization (OR=.42, p<.001) during the year prior to leaving VHA care.

Discussion

We assessed the effectiveness of SMI Re-Engage, a reengagement program for veterans with serious mental illness who had not received VHA care for at least 12 months. Findings indicate that contacting a veteran via outreach increased the risk of the veteran returning to VHA care within 18 months of when the veteran’s information was provided to a clinician. Returning to care via SMI Re-Engage did not appear to reduce mortality risk. Results of post hoc analyses suggest that outreach contact contributed to returning to outpatient care, specifically. Veterans whom providers were not able to contact (despite attempts to do so) had lower odds of returning to outpatient versus inpatient or emergency care compared with veterans who were contacted.

Disengagement from general medical and mental health care is associated with negative outcomes for individuals with serious mental illness (7,25). Outreach interventions are part of well-established mental health programs (for example, assertive community treatment) (26) for this population, yet outreach to patients with serious mental illness who have disengaged from care for a substantial period (for example, for more than 12 months) is not standard in mental health care. Although prior evaluations of similar programs—such as the demonstration project on which SMI Re-Engage is based and NYC Mental Health Care Monitoring Initiative, a non-VA outreach program—indicate that a substantial proportion of individuals with serious mental illness will return to care (3,4,27), the effectiveness of such outreach programs has not been assessed. This evaluation extends this literature by indicating that outreach facilitated return to care among patients with serious mental illness who were disengaged from care. This evaluation also established outreach contact via SMI Re-Engage as an effective component of VHA care continuity for veterans with serious mental illness. Efforts are under way to identify practices associated with successfully contacting veterans via the SMI Re-Engage program.

The finding that return to care via SMI Re-Engage was not associated with reduced mortality risk is inconsistent with the findings of the demonstration project on which SMI Re-Engage is based (4). We employed more robust statistical analyses, and we limited analyses to veterans who were contacted via SMI Re-Engage and who thus had the opportunity to return to care through the outreach program. In comparison, the demonstration project evaluated whether return to care, with or without being contacted via outreach, was associated with mortality. Assessment of more proximal outcomes associated with return to care, such as symptom reduction or improvement in quality of life, would strengthen future evaluations.

Consistent with the demonstration project (4), our study found that veterans age 65 or older and veterans with general medical comorbidities had a lower risk of return to care and a higher risk of mortality within 18 months. Further evaluation is needed to ascertain whether older veterans with serious mental illness and general medical comorbidities are receiving health care outside VHA, such as through community nursing homes or Medicare.

The study had some limitations. Our data did not include the date when clinicians contacted veterans via SMI Re-Engage. Therefore, it was not possible to use contact as a time-dependent covariate in the analysis predicting return to care. This issue has been addressed operationally to enhance future evaluations. Facility characteristics (for example, appointment availability) and provider skills (for example, building a strong alliance with veterans contacted, assessing needs, explaining care options, and navigating bureaucratic issues) were not assessed in this analysis. We were also unable to assess veterans’ use of non-VHA care and whether all veterans identified for outreach had a need for VHA services. Notably, individuals with serious mental illness have variable recovery trajectories, and all may not currently need mental health care (28). Nevertheless, general medical comorbidities experienced by this population (6) suggest that some level of engagement in primary care is likely needed (29). These factors, unaccounted for in this study, could have influenced whether a veteran returned to VHA care.

Conclusions

Clinician outreach via SMI Re-Engage facilitated care reengagement for veterans with serious mental illness who experience extended gaps in use of VHA services. SMI Re-Engage is a relatively straightforward, brief intervention carried out by a single provider and is consistent with population-based panel management (30) strategies. SMI Re-Engage is implemented by mental health providers within VHA, a large integrated health care organization with a national medical record system. Outside VHA, fragmentation of mental and general medical care systems may limit the ability of mental health providers to identify medical needs and facilitate continuity of general medical care (31). Identifying mechanisms for retention and reengagement in mental health and general medical care for people with serious mental illness outside VHA is critical for improving the health of this population. Within VHA, efforts to improve care continuity for veterans with serious mental illness include identifying those at risk of loss to care (32), developing strategies to improve care retention, and sustaining the SMI Re-Engage program.