Project Lead: Rachael Rosen, UW Medicine

Data Science Lead: Valentina Staneva

Lower limb amputation (LLA) is a significant health event that is often an indicator for worsening chronic disease and increased risk of further health decline—especially among older adults.  After amputation, recovery depends not only on medical care but also on access to supportive post-acute care services, such as inpatient rehabilitation, skilled nursing, or home health. These services help individuals manage chronic conditions, regain mobility, and reduce the risk of hospital readmission. However, access to high-quality care is shaped by a complex mix of personal, structural, and community-level factors—including race, rurality, socioeconomic status, geography, and health history.

This project uses machine learning to study hospital readmissions after LLA among Medicare beneficiaries. Unlike traditional models that often assume simple, linear relationships, machine learning can identify how multiple factors interact in complex and non-linear ways. By comparing different predictive algorithms to conventional approaches, this work aims to identify which combinations of patient, facility, and community characteristics are most associated with readmission risk. The goal is to inform more equitable approaches to recovery and care planning after limb loss.

Image Caption: This heatmap is a mock and abridged illustration of predicted hospital readmission risk following LLA. It does not represent actual data, but is intended to demonstrate how demographic, geographic, clinical, and care-setting factors can interact to shape recovery outcomes.

The figure visualizes variation in readmission risk across combinations of group identity, post-acute care (PAC) setting, and amputation level.

PAC settings:

• IRF = Inpatient Rehabilitation Facility
• SNF = Skilled Nursing Facility
• HH = Home Health
• Home = Discharged home without formal services

Amputation levels:

TF = Transfemoral (above-the-knee)

TT = Transtibial (below-the-knee)