Predictors of rehospitalization and suicide in depressed inpatients receiving repetitive transcranial magnetic stimulation: A real-world retrospective cohort study.
To identify clinical predictors of 6-month rehospitalization or suicide events following repetitive transcranial magnetic stimulation (rTMS) in hospitalized patients with depression.
This retrospective cohort study analyzed electronic health records (EHRs) from a tertiary psychiatric hospital in Shanghai, China. Inpatients with depression (ICD-10 codes F31.3-F31.5, F32 or F33) treated with rTMS during hospitalization were included. Missing data were addressed using multiple imputation by chained equations (MICE). Based on univariate analyses of baseline characteristics, further multivariable logistic regression, group least absolute shrinkage and selection operator (LASSO), and random forest analyses were used to identify predictors of rehospitalization or suicide events within 6 months post-discharge. Firth logistic regression was used for bipolar depression (BD) subgroup analyses due to the limited sample size.
A total of 275 inpatients were included, including 222 with unipolar depression (UD) and 53 with BD. Among them, 25.5% experienced rehospitalization or suicide-related events. Comorbid substance use disorders (SUDs), previous electroconvulsive therapy (ECT), and use of benzodiazepines were independently associated with higher odds of these outcomes in the overall cohort and the UD subgroup (all p < 0.05). Predictive models in UD showed moderate discrimination and calibration, whereas no significant predictors were identified in BD. These findings reflect observations from a single-center, retrospective cohort.
Clinically accessible factors were associated with poor long-term outcomes in hospitalized patients with UD receiving rTMS. These findings may help inform patient stratification and support future research aimed at improving risk assessment and personalization of rTMS treatment.
This retrospective cohort study analyzed electronic health records (EHRs) from a tertiary psychiatric hospital in Shanghai, China. Inpatients with depression (ICD-10 codes F31.3-F31.5, F32 or F33) treated with rTMS during hospitalization were included. Missing data were addressed using multiple imputation by chained equations (MICE). Based on univariate analyses of baseline characteristics, further multivariable logistic regression, group least absolute shrinkage and selection operator (LASSO), and random forest analyses were used to identify predictors of rehospitalization or suicide events within 6 months post-discharge. Firth logistic regression was used for bipolar depression (BD) subgroup analyses due to the limited sample size.
A total of 275 inpatients were included, including 222 with unipolar depression (UD) and 53 with BD. Among them, 25.5% experienced rehospitalization or suicide-related events. Comorbid substance use disorders (SUDs), previous electroconvulsive therapy (ECT), and use of benzodiazepines were independently associated with higher odds of these outcomes in the overall cohort and the UD subgroup (all p < 0.05). Predictive models in UD showed moderate discrimination and calibration, whereas no significant predictors were identified in BD. These findings reflect observations from a single-center, retrospective cohort.
Clinically accessible factors were associated with poor long-term outcomes in hospitalized patients with UD receiving rTMS. These findings may help inform patient stratification and support future research aimed at improving risk assessment and personalization of rTMS treatment.
Authors
Wang Wang, Yang Yang, Wu Wu, He He, Wang Wang, Yang Yang, Mou Mou, Chen Chen, Fang Fang
View on Pubmed