CR Bingham, SMITREC & University of Michigan; FC Blow, SMITREC & University of Michigan

Objectives: Veterans with psychoses (VPsy) experience higher mortality rates than non-VPsys; nevertheless, relatively little research has examined this issue. Using data from the National Psychosis Registry (NPR), this study identified predictors of excess mortality among VPsy.

Methods: Sample data were obtained from the FY1999 NPR - a national registry of VPsy who received care in the VA during Fiscal Years (FY) 1988-99 - and included all veterans in the NPR who had died (as per BIRLs and the Veterans Benefits Administration) by the end of FY1999 (n = 135,405). The sample was 97.8% male, 74.5% Caucasian, 15.7% Black, and 3.7% Hispanic, (the remainder were from other racial groups), 41.3% currently married, 39.3% ever (but not currently) married, 18.0% never married, and 1.4% status unknown. Mental illness was represented as follows: schizophrenia, 30.2 %; bipolar disorder, 9.9%; other non-organic psychoses, 41.1%; paranoid states, 1.2%; and affective psychoses, 17.5%. The study sample was classified by gender, race, mental health diagnosis, and marital status, and age adjusted mortality rates were calculated overall, and within all combinations of gender, race, diagnosis, and marital status.

VPsy with excess mortality were identified using Configural Cluster Analysis (CCA). CCA utilizes multi-way contingency tables to identify "configurations" or cells with frequencies that are statistically significantly less than (anti-types) equal to, or greater than (types) the expected value. Based on the CCA, the sample was categorized into one of three categories: excess, representative, or scant mortality, and excess mortality risk was predicted using logistic regression analysis (LRA).

Results: The CCA identified 34 types (excess mortality n = 91,212). No antitypes were identified, resulting in only two groups of VPsy, excess mortality and representative mortality. Excess mortality was significantly more among veterans who were male (odds ratio (o.r.) = 2.105), African-American (o.r. = 2.78), Caucasian (o.r. = 1.54), had schizophrenia (o.r. = 6.52) or other non-organic psychosis (o.r. = 22.32), had were never married (o.r. = 1.98), and had been hospitalized for mental illness at least once (o.r., 5.02).

Results also indicated that the relation between utilization and high mortality risk was moderated by hospitalization for mental illness. High mortality rates for VPsy who had been hospitalized was predicted by more total years of VA health care, more intervening years without VA care (out-migration), fewer intervals of out-migration, and fewer consecutive years of out-migration immediately prior to death. High mortality risk for VPsy who had not been hospitalized for their mental illness was predicted by more total years of VA health care and fewer total years of out-migration.

Conclusions: High mortality risk is more common among VPsy with the most difficult to manage mental illnesses. Patterns of lost-to-care, overall, and lost-to-care immediately prior to death indicate that VPsy with high morality risk are more likely to receive intermittent VA care until their mortality risk becomes acute, at which point they are less likely to out-migrate.

Impact: Greater outreach efforts targeting high mortality risk VPsy and increases in accessibility of VA care may be needed to effectively serve this high risk population.