1008 — Predictors of Suicides in Patient Charts among Patients Receiving Depression Treatment
Kim HM, Smith EG, Ganoczy D, Walters H, Stano C, Ilgen M, Bohnert A, and Valenstein M, COE - Ann Arbor;
To identify factors noted in patient charts during one year proximal to completed suicides that are predictive of suicide among patients receiving depression treatment.
Case control design from a Veterans Health Administration depression treatment cohort from April 1, 1999 to September 30, 2004. The suicide sample was randomly selected from patients who died by suicide, stratified by geographic region, gender, and year of cohort entry. The control sample was randomly selected from the set of patients alive on the date of suicide death (index date) from the same strata as the suicide patient. Only those with at least one outpatient visit in the prior year were included in the analysis (N = 636). Variables were abstracted from all charts belonging to each patient in the year prior to the index date and consisted of the following: (1) clinical symptoms and diagnosis variables, (2) behavioral measures of interest, including suicide risk-related and substance abuse variables, (3) mental health treatments received, and (4) life events or stressors. Logistic regression was used to assess associations between the various factors and suicide.
Suicidality and substance abuse-related variables were the strongest independent predictors of suicide. In particular, the adjusted odds ratio (OR) of suicide was 5.3 (95% CI = 1.3-21.3) for history of suicide attempt, 2.9 (1.7-5.0) for suicidal ideation without an attempt, 3.0 (1.7-5.3) for alcohol abuse, and 7.2 (2.5-20.2) for prescription drug misuse. Of the stressful life events, difficulty with access to health care (OR = 2.6; 95% CI = 1.2-5.8) was a significant predictor of suicide. Receipt of VA substance abuse treatment was protective (OR = .30; 95% CI = .11-.81).
Many of the behavioral measures noted in charts are significant predictors of suicide even after controlling for affective symptom variables noted in charts as well as mental health diagnosis and health care utilization variables from administrative data.
This study highlights the importance of (1) incorporating effective assessments of suicide risks, including stressful life events, into routine assessments of patients treated for depression and (2) adjusting for behavioral measures in comparative research evaluating suicide risks or evaluating predictors of suicide risks.