A Framework for Predicting Impactability of Digital Care Management Using Machine Learning Methods.

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Digital Health Interventions

Background:

• Digital care management programs offer potential cost savings and quality improvements in healthcare.
• Current methods for targeting patients, often relying on "risk scores," are insufficient for identifying those most likely to benefit from interventions.
• A need exists for precise methods to identify patient impactability for digital health solutions.

Purpose of the Study:

• To develop and evaluate a framework using machine learning (ML) to define patient impactability for digital health interventions.
• To identify patients who are most likely to benefit from a digital care management program.
• To enable more efficient and effective allocation of healthcare resources.

Main Methods:

• Utilized anonymized insurance claims data combined with inferred sociodemographic data from a commercially insured population.
• Developed a cost prediction model to quantify differences in healthcare expenditures before and after intervention.
• Trained and compared various ML models (random forest, logistic regression) to classify patient impactability, optimizing parameters via grid search.

Main Results:

• The ML models achieved an overall sensitivity of 0.77 and specificity of 0.65 in categorizing patients as impactable or not impactable.
• The framework successfully defined impactability for a digital health intervention using ML methods.
• Demonstrated the ability to accurately predict which patients would benefit most from the digital care management platform.

Conclusions:

• Machine learning provides a viable method for defining patient impactability in digital health interventions.
• This framework facilitates efficient resource allocation by targeting patients with the highest likelihood of benefiting.
• The approach is generalizable to various interventions and supports closed-loop feedback systems for continuous healthcare improvement.