Loading...
Predicting suicide is difficult at best. Always alert to clinical warning signs such as ideation, plans, and intent, clinicians are also particularly concerned with modifiable risk factors. Investigators analyzed electronic health record (EHR) data on nearly 2 million patients from two Boston hospitals over a 15-year period and identified 16,588 case subjects meeting criteria for suicidal behavior (based on ICD-9 codes, self-injury, and injury of questionable intent). Patients were followed for an average of 5.3 years.
Data-mining yielded a multifactorial model capable of predicting subsequent suicidal behavior with 90% specificity and 44%–46% sensitivity, identifying nearly half of all suicides and suicidal behaviors an average of 3 to 4 ye…