Enhancing pharmacist intervention targeting based on patient clustering with unsupervised machine learning.

Area of Science:

• Diabetes Management
• Health Services Research
• Machine Learning in Healthcare

Background:

• Adherence to the American Diabetes Association (ADA) Standards of Medical Care is suboptimal.
• Effective identification of patients for diabetes interventions is crucial for improving health outcomes.

Purpose of the Study:

• To assist pharmacists in identifying patients for diabetes control interventions.
• To apply unsupervised machine learning for patient stratification in diabetes care.

Main Methods:

• Analysis of the 2021 Medical Expenditure Panel Survey data.
• Utilized k-mode cluster analysis on patient features including preventive care adherence and demographics.
• Included adherence to HbA1c tests, foot exams, cholesterol tests, eye exams, and flu vaccination.

Main Results:

• Included 1,219 patients with self-reported diabetes; overall adherence to ADA standards was 33.72%.
• Identified five patient clusters based on complexity, insurance, and demographics.
• Clusters B (moderate-complexity, publicly insured female), C (low-complexity, privately insured female), and E (moderate-complexity, publicly insured male) showed nonadherence.

Conclusions:

• Pharmacists can target specific patient groups for diabetes interventions.
• Publicly insured elderly (Groups B and E) and privately insured middle-aged females (Group C) are key targets.
• Interventions can include resource navigation and reminders to improve diabetes care adherence.