Repeated Survey of Suicidal Ideation Among Inpatients at a Psychiatric Unit Predicts Short-Term Posthospitalization Suicide Attempt

The researchers sought to determine whether prediction of suicide attempts after psychiatric hospitalization can be improved using frequent assessments of the level and variability of a patient’s suicidal thoughts.

Suicidal thinking and not completing surveys about suicidal ideation during hospitalization were strong predictors of posthospitalization suicide attempts. These findings, from a prognostic study of inpatients in a psychiatric unit, were published in JAMA Network Open.

Inpatients (N=104) at Massachusetts General Hospital psychiatric unit were recruited between 2016 and 2018 for this study. All patients had been admitted to the hospital due to suicidal thought or suicide risk. Surveys comprised a Self-injurious Thoughts and Behaviors Interview (SITBI) and ecological momentary assessment (EMA) of changes over time for suicide risk.

A total of 83 (79.8%) patients completed ³3 EMA surveys with a median of 18.5 (range, 3-120) surveys per patient. These patients were aged mean 38.43 (standard deviation [SD], 13.64) years, 83.1% were White, 42.2% were women, 98.8% reported a lifetime presence of suicidal thoughts, 75.6% had a previous attempt, 95.2% had a mood disorder, 39.8% a personality disorder, 36.1% an anxiety disorder, and 36.1% a substance use disorder.

After hospital discharge, 10.8% (n=9) attempted suicide. The attempts were identified by self-report (n=6) or medical record review (n=3).

A dynamic feature model predicted posthospitalization suicide attempt with an area under the receiver operating characteristic curve (AUC) of 0.89 (interquartile range [IQR], 0.81-0.97), accuracy of 0.85 (IQR, 0.81-0.88), sensitivity of 0.57 (IQR, 0.25-1.00), and specificity of 0.88 (IQR, 0.83-0.93).

The model was dominated by the variable probability of acute change (PAC) with an importance ~100%; followed by PAC resist, PAC intent, SD of desire, and maximum resist, which had <20% importance.

Including EMA non-completion to the model improved the prediction to an AUC of 0.93 (IQR, 0.90-1.00).

Limitations include that this study was based on few patients with little ethnic diversity treated at a single center. These findings may not be generalizable to other patient populations in differing care settings.

These data indicated that collecting dynamic information about suicide ideation among patients hospitalized for suicide risk may improve prediction of posthospitalization attempts and assist in making informed clinical decisions.

Disclosure: Multiple authors declared affiliations with industry. Please refer to the original article for a full list of disclosures.


Wang SB, Coppersmith DDL, Kleiman EM, et al. A pilot study using frequent inpatient assessments of suicidal thinking to predict short-term postdischarge suicidal behavior. JAMA Netw Open. 2021;4(3):e210591. doi:10.1001/jamanetworkopen.2021.0591.