Using Real-Time Monitoring of Suicidal Thinking to Predict Suicidal Behavior after Hospital Discharge

It is well documented that  the weeks following discharge from a psychiatric hospitalization constitute a period of high risk for suicide attempt.  This risk of post-discharge suicide is extraordinarily high among patients hospitalized with suicidal thoughts or behaviors —  nearly 200 times the global suicide rate.  While there is consensus that the post-discharge period is a time of heightened risk for suicidality, we have yet to determine the most effective means of monitoring suicidality and diminishing risk for suicide duing this vulnerable time. 

A new study from Shirley Wang, AM and colleagues in the Nock Lab looks at the possibility of employing real-time monitoring of suicidal thoughts (with a smartphone) as a means of assessing short-term risk for suicidality in patients after discharge.  More frequent real-time assessments, or what is called ecological momentary assessment (EMA), in individuals with suicidal thoughts have revealed that suicidal ideation varies dramatically over the course of most days.  Furthermore, prior retrospective studies have indicated that suicide attempts are often preceded by rapid fluctiuations in the intensity of suicidal thinking.

In this pilot study, 83 adults recruited from the inpatient psychiatric unit at Massachusetts General Hospital were monitored, receiving 4 to 6 semi-random smartphone prompts per day while hospitalized, and were instructed to complete surveys of suicidal thinking.  After discharge, they completed brief follow-up surveys assessing suicide attempts at 2 and 4 weeks post-discharge.   Among the 83 participants (mean [SD] age, 38.4 [13.6] years; 51.8% male; 83.1% White), nine (10.8%) made a suicide attempt during the month after discharge.  

Assessments reflecting rapid fluctuations in suicidal thinking and intention emerged as the strongest predictors of suicide attempt after discharge, and using EMA data greatly improved the ability to predict posthospital suicide attempts.  Including data on missingness — or nonresponse to prompts/questions about suicidal thinking — greatly increased predictive accuracy, a finding consistent with previous studies demonstrating that failing to respond to questions about suicidal thinking is a particularly strong predictor of the transition from suicidal thoughts to suicide attempt.  

The model based on data reflecting dynamic changes in real-time suicidal thoughts during hospitalization performed the best.  This dynamic feature model predicted post-disharge suicide attempts with good accuracy, with an area under curve or AUC of 0.89, which is high compared to the AUC values for other prediction models of suicidal behaviors. (An AUC of 0.89 indicates that this model identified 89% of individuals who would go on to attempt suicide.)

Most models predicting suicide attempt have reported extremely low positive predictive values (PPVs < .01), meaning that 99 out of every 100 individuals predicted to attempt suicide by these models will not attempt suicide. In comparison, the PPV for the dynamic feature model was much better at 0.39. 

This study indicates that collecting dynamic, real-time data on suicidality from patients during hospitalization can significantly improve the prediction of post-discharge suicide attempts. This is a pilot study, and studies including larger and more diverse samples are warranted to replicate these findings.  But this is exciting. This is a relatively simple and inexpensive means of assessing risk for suicidality, and one could easily imagine that extending EMA beyond the hospitalization may help to improve its utility in preventing suicide.

Deming CA, Harris JA, Castro-Ramirez F, Glenn JJ, Cha CB, Millner AJ, Nock MK.  Inconsistencies in self-reports of suicidal ideation and attempts across assessment methods.  Psychol Assess. 2021 Mar;33(3):218-229. 

Kleiman  EM, Turner  BJ, Fedor  S, Beale  EE, Huffman  JC, Nock  MK.  Examination of real-time fluctuations in suicidal ideation and its risk factors: results from two ecological momentary assessment studies.   J Abnorm Psychol. 2017;126(6):726-738. 

Nock  MK, Prinstein  MJ, Sterba  SK.  Revealing the form and function of self-injurious thoughts and behaviors: a real-time ecological assessment study among adolescents and young adults.   J Abnorm Psychol. 2009;118(4):816-827.

Wang SB, Coppersmith DDL, Kleiman EM, Bentley KH, Millner AJ, Fortgang R, Mair P, Dempsey W, Huffman JC, Nock MK.  A Pilot Study Using Frequent Inpatient Assessments of Suicidal Thinking to Predict Short-Term Postdischarge Suicidal Behavior.  JAMA Netw Open. 2021 Mar 1;4(3):e210591. Free article.

Shirley Wang is a Ph.D. student in the Clinical Science program at Harvard University and a member of the Nock Lab. Shirley’s research examines similarities and shared mechanisms underlying eating disorders, self-injury, and suicide, as well as the applications of data-driven methods (e.g., machine learning, network analysis) to conceptualize and predict these behaviors.