Improving Patient Selection and Prioritization for Hospital at Home Through Predictive Modeling.

Area of Science:

• Healthcare Management
• Clinical Informatics
• Predictive Analytics

Background:

• Hospital at Home (HaH) programs provide acute care in patients' residences.
• Current patient selection relies on time-consuming manual chart reviews, creating bottlenecks.
• Identifying clinically and socially appropriate patients is crucial for HaH success.

Purpose of the Study:

• To develop and implement a predictive model to automate and optimize patient selection for Hospital at Home programs.
• To improve the efficiency and accuracy of identifying eligible patients for HaH care.
• To streamline the patient screening and enrollment process.

Main Methods:

• Development of a predictive model integrating clinical and social factors.
• Creation of a web application and data pipeline for eligibility scoring.
• Provider utilization of the model to prioritize chart reviews and screenings.

Main Results:

• The predictive model achieved an Area Under the Curve (AUC) of 0.77 during training and 0.75 in production testing.
• The algorithm identified inconsistencies in enrollment criteria, which evolved during the study.
• The system successfully streamlined patient identification for the Hospital at Home program.

Conclusions:

• A predictive model can significantly enhance the efficiency of patient selection for Hospital at Home care.
• The developed tool aids in prioritizing patient chart reviews and screenings.
• This approach addresses challenges in patient identification and enrollment for novel care models.