Predictive Modeling of Hospital Readmission of Schizophrenic Patients in a Spanish Region Combining Particle Swarm...

Area of Science:

• Medical Informatics
• Computational Psychiatry
• Health Services Research

Background:

• Hospital readmissions are key indicators of care quality and patient outcomes.
• High readmission rates increase healthcare costs and negatively impact patient quality of life.
• Predictive models for hospital readmissions can guide treatment selection and preventive strategies.

Purpose of the Study:

• To develop predictive models for the readmission risk of patients diagnosed with schizophrenia.
• To enhance predictive accuracy by combining the particle swarm optimization (PSO) algorithm with machine learning classification algorithms.

Main Methods:

• Utilized a database of 6089 readmission records for schizophrenia patients from 11 public hospitals in Spain (2005-2015).
• Employed machine learning classification algorithms integrated with the particle swarm optimization (PSO) algorithm.
• Evaluated model performance using metrics such as AUC, recall, accuracy, and F1-score.

Main Results:

• The Random Forest algorithm combined with PSO demonstrated superior performance.
• Achieved an Area Under the Curve (AUC) of 0.860.
• Reported a recall of 0.959, accuracy of 0.844, and F1-score of 0.907.

Conclusions:

• The developed predictive models offer a significant contribution to improving patient care for schizophrenia.
• These models can facilitate the implementation of targeted preventive measures.
• The study highlights the potential for reducing healthcare system costs through improved readmission risk prediction.