Identifying Risk Groups in 73,000 Patients with Diabetes Receiving Total Hip Replacement: A Machine Learning...

Area of Science:

• Orthopedic Surgery
• Data Science
• Public Health

Background:

• Diabetes mellitus (DM) is common and impacts total hip arthroplasty (THA) outcomes.
• Identifying specific comorbidity profiles in diabetic THA patients is crucial for improving care.

Purpose of the Study:

• To apply machine learning clustering to define comorbidity profiles in diabetic THA patients.
• To assess the association between identified clusters and postoperative outcomes like discharge disposition and length of stay.

Main Methods:

• Utilized the 2015-2021 National Inpatient Sample database.
• Included 73,606 diabetic patients undergoing THA, analyzing 49 comorbidities and covariates.
• Employed clustering algorithms, logistic regression for non-routine discharge (NRD), and Kruskal-Wallis H testing for length-of-stay (LOS).

Main Results:

• Six distinct patient clusters were identified based on comorbidity profiles.
• A cluster characterized by urinary tract infection and sepsis showed significantly higher NRD risk (OR 7.83) and longest median LOS (9.0 days).
• Other clusters demonstrated varied recovery patterns, with some achieving shorter LOS (2.0 days).

Conclusions:

• Machine learning effectively stratified diabetic THA patients into six unique groups.
• Infection-predominant clusters represent high-risk populations for adverse outcomes.
• This clustering approach offers a novel method for risk stratification and personalized perioperative management in diabetic THA patients.