How-To Guides

Over 24 years of prospective follow-up, 5.9% of patients with borderline personality disorder died by suicide. In the personality-disordered comparison group, 1.4% died by suicide. The between-group difference in time-to-suicide was not statistically significant in this sample (HR = 4.42, P = .150).

Yes. In this cohort, non-suicide death was more common than suicide among patients with borderline personality disorder. Over 24 years, 14.0% of borderline patients died from causes other than suicide, compared with 5.9% who died by suicide. In the comparison group, 5.5% died from non-suicide causes.

The most common non-suicide causes of death were cardiovascular causes. Across both diagnostic groups, cardiovascular deaths such as myocardial infarction occurred most often (n = 11), followed by substance-related complications such as liver cirrhosis (n = 5), cancer (n = 4), and accidents (n = 4).

The only significant baseline predictor of completed suicide in patients with borderline personality disorder was a greater number of prior psychiatric hospitalizations. The hazard ratio was 1.62 (P = .037). None of the other examined baseline predictors reached statistical significance, so no multivariate suicide model was tested.

In multivariate analysis, male sex and a greater number of prior psychiatric hospitalizations significantly predicted premature non-suicide death in patients with borderline personality disorder. Male sex was associated with a hazard ratio of 3.56 (P = .003), and more prior psychiatric hospitalizations with a hazard ratio of 2.93 (P < .001).

In bivariate analyses, several other baseline factors were also significant: lower socioeconomic status, government disability, history of drug use disorder, greater number of psychiatric medications, and obese-range body mass index (BMI ≥ 30).

Most borderline patients who died had never achieved recovery before death. Among patients with borderline personality disorder who died by suicide, 87.5% had never achieved recovery (χ2₁ = 16.11, P < .0001; n = 16). Among those who died from non-suicide causes, 87.5% were also never recovered (χ2₁ = 16.11, P < .0001; n = 24).

Recovery was defined as at least one 2-year follow-up period during which the patient was in remission from borderline personality disorder, had at least one emotionally sustaining relationship, and was functioning consistently and competently in full-time work or school. Being a houseperson also counted toward the work-functioning criterion.

This was a 24-year prospective follow-up study of 290 adult inpatients with borderline personality disorder and 72 personality-disordered comparison subjects. Participants were recruited during inpatient admission at McLean Hospital between 1992 and 1995 and were reassessed every 2 years. Because deaths were tracked prospectively and analyzed with survival methods, the study could estimate the timing of suicide and non-suicide mortality more accurately than retrospective chart reviews or follow-back studies.

In this sample, patients with borderline personality disorder had higher observed rates of both suicide and non-suicide death, but the between-group differences were not statistically significant. For suicide, the hazard ratio for time-to-suicide was 4.42 (P = .150). For premature death from other causes, the hazard ratio was 2.93 (P = .144).

The main limitations were the rarity of the outcomes, the limited set of predictors studied, and uncertain generalizability. Suicide and other deaths were relatively uncommon, which limited statistical power to detect between-group differences or predictor effects. The study also excluded patients with bipolar I disorder at baseline, and most participants were initially psychiatric inpatients who received ongoing outpatient psychotherapy and/or psychopharmacologic treatment, so the findings may not generalize to non-inpatient or untreated personality-disordered populations.