Treatment continuity for people with serious mental illness is vital. Mental health treatment attrition is associated with increased rehospitalization (
1) and lower social and vocational functioning (
2). Continuous medical care is also essential for people with serious mental illness, given that general medical comorbidities are leading causes of premature mortality in this population (
3,
4). Notably, among veterans with serious mental illness, lower levels of outpatient general medical care are associated with higher rates of unforeseen mortality (
5) and lower survival rates (
6). Strategies that reengage patients lost to care and prevent treatment dropout may help reduce disparities in outcomes for this population.
Population management aims to mitigate health disparities by identifying and monitoring risk factors in vulnerable populations and proactively offering interventions, regardless of whether patients are currently seeking treatment (
7). Although population management has been widely used in general medical care, generally to target preventive interventions to patients at risk of rehospitalization (
8,
9), it has not yet been adopted in routine mental health practice. However, a recent Veterans Affairs (VA) quality improvement study (
10) demonstrated that population management could be implemented nationally to address the needs of veterans with serious mental illness. The study identified veterans with serious mental illness who had been lost to care for at least one year, conducted outreach to assess treatment needs, and invited the veterans to reengage in VA health care (
10). Treatment reengagement was associated with a sixfold decrease in mortality (
10), suggesting the utility of population management for improving outcomes for veterans with serious mental illness.
Identification of risk factors for adverse events in order to target prevention is an essential aspect of population management. Hence, the identification of specific risk factors for treatment dropout among people with serious mental illness is an important first step in preventing this adverse event. Prior studies identified patient characteristics associated with health care gaps among veterans with serious mental illness (
11); however little is known about health care utilization preceding loss to care. Identifying health care utilization factors associated with loss to care can help providers identify, in advance, patients who are at risk of dropout and inform the development of interventions designed to prevent dropout. Moreover, identification of health care utilization factors is more practical than identification of clinical symptom data, which can be difficult to ascertain from medical records.
This study aimed to identify health care utilization factors associated with loss to care among veterans with serious mental illness. We compared utilization by veterans with serious mental illness who subsequently dropped out of VA health care or who remained in care. Specifically, we examined whether utilization of inpatient, outpatient, general medical, and mental health services were uniquely associated with loss to care.
Methods
This study was part of a larger VA national initiative, SMI Re-Engage (
12), to reengage veterans with serious mental illness who were lost to care. In this study, we used a nested case control design to compare a cohort of veterans identified through SMI Re-Engage with a control cohort of veterans with serious mental illness who were not lost to care. This study was approved by the local medical center’s institutional review board. Informed consent was waived because data were from existing sources.
Employing previously described methods (
10,
12), we used VA administrative data from the National Psychosis Registry (NPR) (
13) to identify veterans who had been given a diagnosis of schizophrenia or bipolar disorder, had a VA visit in fiscal year (FY) 2008 or FY 2009 but no outpatient visits or inpatient hospitalizations of more than two days within the VA health care system in the following 12 months, and were alive throughout FY 2011. This procedure yielded 6,687 veterans in the lost-to-care cohort. To identify a control cohort, we used the NPR to randomly select 6,687 veterans with schizophrenia or bipolar disorder who received care during FY 2008 or FY 2009 and who were not subsequently lost to care. We excluded nonveterans and veterans with missing data.
We ascertained age, gender, race, marital status, history of homelessness, service status (Operation Enduring Freedom or Operation Iraqi Freedom [OEF/OIF]), diagnosis (schizophrenia or bipolar disorder), substance use diagnosis, service-connected disability status, and Charlson Comorbidity Index score (
14).
Variables included antipsychotic prescription receipt, number and length of general medical and mental health hospitalizations, and number of emergency, outpatient primary care, and mental health visits. For veterans lost to care, we obtained utilization data for the entire FY prior to the FY of dropout. Utilization data from FY 2007 were obtained for veterans who were lost to care in FY 2008 (46%), and data from FY 2008 were obtained for veterans who were lost to care in FY 2009 (54%). To control for possible historical trends in utilization, we obtained one FY of utilization data for the control cohort by randomly selecting 46% of the control cohort for analysis of data from FY 2007 and 54% for analysis of FY 2008 data.
Statistical analyses were performed with SAS, version 9.2. A conditional multivariate logistic regression model predicted odds of loss to care. Odds ratios (ORs) and 95% confidence intervals are presented.
Results
Among the 13,374 veterans, most were male, Caucasian, and middle-aged (
Table 1). During the FY for which data were obtained, the total sample had a mean of 14.4 outpatient mental health visits, 3.4 outpatient primary care visits, .3 mental health hospitalizations, and .2 general medical hospitalizations.
Veterans who were 65 years and older (OR=1.78) and Caucasian (OR=1.33), who served in OEF/OIF (OR=1.52), and who had a history of homelessness (OR=1.17) and a schizophrenia diagnosis (OR=1.16) were more likely than other veterans to be lost to care (
Table 1). Regarding health care utilization, having fewer outpatient mental health (OR=1.00) and general medical visits (OR=.96) was associated with higher odds of dropout, after adjustment for demographic and clinical factors. In contrast, increased number of mental health (OR=1.08) and general medical (OR=1.29) hospitalizations was associated with greater odds of loss to care.
Discussion
We identified health care utilization factors associated with subsequent loss to care among veterans with serious mental illness. Increased number of outpatient general medical and mental health visits was associated with lower odds of treatment dropout during the following year, whereas increased number of general medical and mental health hospitalizations was associated with higher odds of dropout. Population management strategies incorporating utilization monitoring might help prevent treatment dropout.
The association of dropout with general medical care was stronger than its association with mental health care. Continuity of outpatient general medical care is particularly important for persons with serious mental illness (
5,
6) because their risk of premature mortality is generally attributable to general medical comorbidities (
3). Our results indicated that continuous outpatient general medical care may also be an important factor in treatment retention.
Additionally, our findings that veterans with a history of homelessness or OEF/OIF service had higher odds of loss to care underscore the importance of current VA programs that facilitate treatment engagement among veterans who served in OEF/OIF or who are homeless, such as Make the Connection, Families at Ease, and the Homeless Veterans Initiative. Similarly, our findings illustrate the relevance of ongoing research regarding strategies for outreach to veterans with serious mental illness who are homeless in order to reengage them in care (
15).
Study limitations included lack of information regarding health care utilization at facilities not affiliated with the VA. Additionally, data were not available to ascertain potential reasons for loss to care, including veteran preference or ease of access to particular services. Moreover, the findings may not generalize to nonveterans.
Conclusions
Increased utilization of outpatient general medical care was associated with lower odds of subsequent loss to care among veterans with serious mental illness. Strategies addressing general medical care needs of this population through continuous outpatient treatment engagement may help prevent loss to care. Moreover, veterans with serious mental illness who served during OEF/OIF or who have a history of homelessness are particularly vulnerable to treatment dropout. VA efforts to engage homeless and OEF/OIF veterans should incorporate strategies to address the needs of veterans with serious mental illness. Future investigations should address whether population management’s ongoing risk monitoring and prevention improve treatment retention among veterans with serious mental illness and should ascertain veterans’ perspectives regarding treatment retention. Veterans’ perspectives about treatment retention may yield important information regarding potential reasons for loss to care among different subgroups of veterans, for example, veterans who served in OEF/OIF and those with a history of homelessness.
Acknowledgments and disclosures
Support for the study was provided by a Homeless Health Services Research Initiative (SDR-232) to Dr. Kilbourne from the Veterans Affairs (VA) Health Services Research and Development Service. Preparation of the manuscript was supported by an Advanced Psychology Fellowship award to Dr. Abraham from the VA Office of Academic Affiliations.
The authors report no competing interests.