Digital Cousins in Arthroplasty: An Interpretable Clustering Framework.

Area of Science:

• Orthopedic Surgery
• Medical Informatics
• Data Science

Background:

• Digital twins are proposed for personalized medicine.
• Identifying clinically similar patients ('digital cousins') is a more feasible near-term goal using routine clinical data.
• Arthroplasty outcome analysis can benefit from interpretable patient clustering.

Purpose of the Study:

• To explore interpretable clustering of baseline arthroplasty data.
• To establish a methodological foundation for outcome analysis using digital cousins.
• To assess the utility of patient clustering for identifying clinically similar individuals.

Main Methods:

• Extracted hip arthroplasty records from the MIMIC-IV database.
• Separated data into primary and revision surgery cohorts.
• Clustered patients using standardized variables (age, BMI, Charlson Comorbidity Index).
• Reserved diagnosis categories for post hoc interpretation.

Main Results:

• Generated clinically readable patient profiles through clustering.
• Highlighted distinct patterns between primary and revision arthroplasty surgery.
• Demonstrated the feasibility of the digital cousin approach.

Conclusions:

• Interpretable clustering provides a useful foundation for digital cousin identification.
• The digital cousin framework is a more realistic application than full digital twins in this context.
• Findings support using patient similarity for outcome interpretation in complex arthroplasty cases.