We investigated whether posttraumatic stress disorder (PTSD) was predictor of suicidal behavior even when adjusting for comorbid borderline personality disorder (BPD) and other salient risk factors. To study this, we randomly selected 308 patients admitted to a psychiatric hospital because of suicide risk. Baseline interviews were performed within the first days of the stay. Information concerning the number of self-harm admissions to general hospitals over the subsequent 6 months was retrieved through linkage with the regional hospital registers. A censored regression analysis of hospital admissions for self-harm indicated significant associations with both PTSD (β = .21, p.001) and BPD (β = .27, p.001). A structural model comprising two latent BPD factors, dysregulation and relationship problems, as well as PTSD and several other variables, demonstrated that PTSD was an important predictor of the number of self-harm admissions to general hospitals(B = 1.52, p.01). Dysregulation predicted self-harm directly (B = 0.28, p.05), and also through PTSD [corrected]. These results suggested that PTSD and related dysregulation problems could be important treatment targets for a reduction in the risk of severe self-harm in high-risk psychiatric patients.