Modelling admission lengths within psychiatric intensive care units.

Area of Science:

• Psychiatric Intensive Care
• Health Services Research

Background:

• Predicting length of stay in psychiatric intensive care units (PICUs) is crucial for resource allocation and patient flow.
• Previous attempts to identify predictors of PICU admission length have yielded limited success.

Purpose of the Study:

• To evaluate discharge destination as a predictor variable for length of stay in PICUs.
• To compare the predictive power of discharge destination with other potential variables.

Main Methods:

• A retrospective study analyzed 368 admissions across four PICUs.
• Patients were categorized into three types based on discharge destination.
• Bayesian models were employed to analyze the data.

Main Results:

• Patient grouping by discharge destination showed the highest intraclass correlation (0.37) in predicting length of stay.
• Patients transferred to more secure care experienced longer PICU admissions.
• The optimal model incorporated patient type (discharge destination) and unit, with an interaction effect.

Conclusions:

• Clinical pathway-based patient typing, specifically discharge destination, is a more effective predictor of PICU admission length than clinical diagnosis or specialized assessment tools.
• Bayesian modeling of admission lengths can be a valuable tool for service planning and patient monitoring in PICUs.