Researchers test new model with 2 existing models utilizing baseline data for suicidal thoughts and actions.
Researchers believe a new method could accurately predict the odds a patient will attempt suicide in the weeks following discharge of a psychiatric hospital for suicidal thoughts.
A team, led by Shirley B. Wang, Department of Psychology, Harvard University, examined whether modeling dynamic changes in real-time suicidal thoughts during psychiatric hospitalization might improve the predictions of post-discharge suicide attempts compared to using only baseline data or using the mean level of real-time suicidal thoughts during hospitalization.
After a discharge from a psychiatry hospital, patients are likely at the highest risk for suicide attempts for several weeks. However, real-time monitoring of suicidal thoughts using a smartphone application could be more indicative of the short-term risk than a single cross-sectional assessment.
To be included in the study, each individual must have been hospitalized for suicidal thoughts and/or behaviors.
In the prognostic study, the investigators examined 83 adults from the inpatient psychiatric unit at Massachusetts General Hospital. Each patient completed ecological momentary assessments surveys of suicidal thinking 4-6 times per day during hospitalization, as well as brief follow-up surveys assessing suicide attempts at 2 and 4 weeks following discharge.
Each individual in the study also completed at least 3 real-time monitoring surveys.
The mean age was 38.4 years old.
The researchers sought primary outcomes of suicide attempts in the month after discharge.
Comparing the Models
The mean cross-validated area under the curve (AUC) for elastic net models revealed predictive accuracy was fair for the model including baseline data (AUC, 0.71; first-third quartile, 0.55-0.88).
It was also good for the model using the mean level of real-time suicidal thoughts during hospitalization (AUC, 0.81; first-third quartile, 0.67-0.91).
Finally, it was deemed the best for the model using dynamic changes in real-time suicidal thoughts during hospitalization (AUC, 0.89; first-third quartile, 0.81-0.97).
This pattern of results held of other classification metrics, such as accuracy, positive predictive value, and Brier score, as well as when using different cross-validation procedures.
In addition, features assessing rapid fluctuations in suicidal thinking emerged as the strongest predictors of post-hospital suicide attempts, while a final set of models incorporating percentage missingness improved both the mean (mean AUC, 0.93; first-third quartile, 0.90-1.00) and dynamic feature (mean AUC, 0.93; first-third quartile, 0.88-1.00) models.
“In this study, collecting real-time data about suicidal thinking during the course of hospitalization significantly improved short-term prediction of posthospitalization suicide attempts,” the authors wrote. “Models including dynamic changes in suicidal thinking over time yielded the best prediction; features that captured rapid changes in suicidal thoughts were particularly strong predictors. Survey noncompletion also emerged as an important predictor of posthospitalization suicide attempts.”
Suicide is currently a leading cause of death, with more than 800,000 global suicides annually, 45,000 of which occur in the US.
The study, “A Pilot Study Using Frequent Inpatient Assessments of Suicidal Thinking to Predict Short-Term Post-discharge Suicidal Behavior,” was published online in JAMA Network Open.