Development and internal validation of a predictive model for prolonged intensive care unit stays in patients with psychotropic drug poisoning

Background

Some patients with psychotropic drug poisoning need intensive care unit (ICU) admission, but risk prediction models for prolonged ICU stays are lacking.

Objectives

Develop and evaluate a prediction model for prolonged ICU stays in patients with psychotropic drug poisoning.

Methods

The clinical data of patients with psychotropic drug poisoning were collected from the Medical Information Mart for Intensive Care (MIMIC)-Ⅳ 2.2 database. Patients were grouped by their ICU length of stay: non-prolonged (<2 days) and prolonged (≥2 days).

Variable selection methods included LASSO and logistic regression. The selected variables were used to construct the model, which was subsequently evaluated for discrimination, calibration, and clinical utility.

Results

The cohort included 413 patients with psychotropic drug poisoning, 49.4 % male, with a median age of 41 years. The variables stepwise selected for model construction through LASSO and logistic regression include sepsis, SAPS Ⅱ, heart rate, respiratory rate, and mechanical ventilation. The model showed good discrimination with an area under the receiver operating characteristic curve (AUC) of 0.785 (95 % CI: 0.736–0.833) and was validated well with bootstrap internal validation (AUC: 0.792, 95 % CI: 0.745–0.839). Calibration curves indicated good fit (χ2 = 4.148, P = 0.844), aligning observed and predicted rates of prolonged ICU stays. Decision curve analysis (DCA) showed positive net benefits across a threshold probability range of 0.07–0.85.

Conclusions

The model developed in this study may help predict the risk of prolonged ICU stays for patients with psychotropic drug poisoning.