Predicting the risk of suicide attempt in a depressed population: Development and assessment of an efficient predictive nomogram.

The purpose of this study was to develop and validate a user-friendly suicide attempt risk nomogram in depression, supporting timely interventions by clinicians. We collected clinical data of 273 depressed patients from January 2020 to January 2021. Suicide attempt was assessed conducting the Mini International Neuropsychiatric Interview. First, optimized features were filtrated through the least absolute shrinkage and selection operator regression analysis. Subsequently, we selected variables with nonzero coefficients and entered them into multiple logistic regression model and nomogram function to construct a visual predicting suicide attempt model. Additionally, the C-index, calibration plot and decision curve analysis, were applied to assess discrimination, calibration, and clinical practicability. Finally, the bootstrapping validation was applied to assess internal validation. Finally, eleven clinical features are screened out in the prediction nomogram. The model presented tiptop calibration and pleasant discrimination with a C-index of 0.853. A towering C-index value, up to 0.799, could also be attained in the interval validation analysis. In addition, decision curve analysis exhibited that our predictive model is clinically effective when the threshold is no less than 1%. These results demonstrate this predictive model was helpful for clinicians assessing the inpatient's suicide attempt recently and implementing individualized treatment strategies.