Health Outcomes from Home Hospitalization: Multisource Predictive Modeling.

Area of Science:

• Health Services Research
• Clinical Informatics
• Predictive Analytics

Background:

• Home hospitalization (HH) is a cost-effective alternative to traditional hospitalization for select patients.
• A 10-year analysis of a HH/early discharge (HH/ED) program showed high acceptance and health value generation.
• Health risk assessment was identified as a key area for improvement in HH/ED decision-making.

Purpose of the Study:

• To develop and evaluate predictive models for mortality and in-hospital readmission.
• To assess model performance at both the entry point and discharge from the HH/ED program.

Main Methods:

• Predictive modeling was conducted for mortality and in-hospital admission from January 2009 to December 2015.
• Multisource data included clinical, functional, and population health risk assessment variables.
• A random forest classifier was employed due to its superior performance.

Main Results:

• The random forest model achieved high predictive accuracy for mortality (AUROC 0.88-0.89) and moderate accuracy for readmission (AUROC 0.70-0.71).
• Models demonstrated strong performance at both entry and discharge points for the HH/ED program.
• Sensitivity and specificity values indicate robust predictive capabilities for patient outcomes.

Conclusions:

• The developed predictive models can inform clinical decision support systems for HH/ED program candidate selection.
• These models can guide healthcare professionals in managing transitions to community-based care post-discharge.
• Enhanced risk assessment through predictive modeling improves the overall value and effectiveness of home hospitalization programs.