Abstract TMP92: Prevalence and Predictors of Recent Suicidal Ideation among US Stroke Survivors

Abstract

BACKGROUND: Although approximately one third of individuals develop depression after stroke, little is known about the prevalence of suicidal ideation (SI). In this study, we assessed the prevalence and predictors of SI among stroke survivors. METHODS: We assessed prevalence of SI among adults (≥20 years) with self-reported history of stroke who participated in National Health and Nutrition Examination Surveys - a series of cross-sectional surveys of a nationally representative sample of the US population - from 2005 through 2010. SI in the past 2 weeks was determined from item 9 of the Patient Health Questionnaire 9 (PHQ-9). Sociodemographic factors, health behaviors, and comorbid conditions were examined as potential correlates. Random forest models were used to identify potential predictors of SI. Variables were sorted in the order of importance according to mean decrease in GINI index. Final model selection was made using a stepwise backwards procedure. Results were compared to persons with history of myocardial infarction (MI), diabetes and cancer. RESULTS: Between 2005 and 2010, an estimated 6,235,528 adults in the US reported a history of stroke. The prevalence of SI among stroke survivors was 7.8% (n=400,640), a higher prevalence compared to individuals with history of MI (6.2%), diabetes (5.2%), and cancer (4.1%). The random forest model revealed that the following variables predicted SI among stroke survivors with 74% sensitivity and 81% specificity (overall accuracy of 80%) in decreasing order of importance: depression score, age, BMI, education level, income to poverty ratio, female sex, and marital status. The first 5 predictors were also among the top 6 most important predictors of SI among persons with history of cancer, MI and diabetes. The best threshold of the PHQ-8 depression score for distinguishing stroke survivors with SI was PHQ-8 ≥ 7. CONCLUSION: One in twelve adult stroke survivors in the US endorsed recent SI, a prevalence that exceeded that of several other major conditions including cancer. Key sociodemographic variables may identify stroke survivors at highest risk. These results suggest there may be a need to incorporate screening for SI into the routine assessment of stroke patients.