Outcomes After Lower Limb Amputation Among Medicare Beneficiaries: Post-Acute Care Settings, Readmission, and Mortality

Abstract

Major lower limb amputation is a life-altering procedure associated with high rates of hospital readmission and mortality, particularly among older adults with complex health conditions. Despite the critical role of post-acute care in supporting recovery after major lower limb amputation, limited evidence exists on how post-acute care access and outcomes vary within this vulnerable population. This dissertation examined post-acute discharge destinations following acute hospitalization and modeled adverse post-discharge outcomes, using national Medicare claims data from 2018 to 2021. Together, the three studies described in this dissertation provide a comprehensive evaluation of post-acute care settings, as well as readmission and mortality outcomes following major lower limb amputation among Medicare fee-for-service beneficiaries. The first study identified disparities in post-acute care discharge destinations following hospitalization for major lower limb amputation based on Medicare beneficiary sociodemographic and clinical characteristics. Despite clinical guidelines suggesting patients receive post-acute care from an inpatient rehabilitation facility after major lower limb amputation, findings demonstrate fewer than one in four beneficiaries were discharged to this post-acute care setting, with discharge patterns varying significantly by age, race/ethnicity, sex, rurality, and clinical complexity. Beneficiaries who are female, Black, Native American, dually eligible for Medicare and Medicaid, or residing in rural areas were less likely to receive post-acute care from an inpatient rehabilitation facility. These findings demonstrate inequities in post-acute care access following lower limb amputation among Medicare beneficiaries. Building on the first study, the second study focused on predicting 180-day hospital readmission after lower limb amputation. A traditional competing risk model was compared to a machine learning model to assess readmission risk. Results showed that machine learning improved prediction accuracy and identified non-linear relationships between beneficiary characteristics and readmission. High-risk subgroups were identified based on clinical complexity, including end-stage renal disease, increased comorbidity burden, prolonged hospitalizations, and an intensive care unit stay. Identifying high-risk beneficiary profiles may guide targeted post-acute care planning, including closer monitoring and increased support following hospital discharge after lower limb amputation. The third study modeled post-discharge mortality following lower limb amputation. A traditional survival model was compared to a machine learning model to predict mortality risk. Both methods demonstrated similar predictive accuracy, but a key assumption of the traditional survival model was violated, therefore the machine learning model was the better choice for prediction. Patients discharged to hospice, skilled nursing facilities, or home without post-acute care services had the highest mortality, while those discharged to inpatient rehabilitation facilities or home health care experienced comparatively lower mortality risk. High-risk subgroups were identified based on key sociodemographic and clinical predictors, including post-acute care discharge setting, increased comorbidity burden, older age, and a diagnosis of end-stage renal disease. Together, these studies highlight variation in post-acute care access and clinical outcomes following major lower limb amputation among Medicare beneficiaries. Machine learning methods improved readmission prediction and identified non-linear risk predictors that could not be captured by traditional statistical models. Mortality prediction was comparable across models; however, a violation of a key assumption in the traditional survival model favored the use of machine learning to predict mortality. These findings demonstrate the utility of machine learning as a flexible modeling approach to support risk-informed clinical decision-making following major lower limb amputation. Results from these analyses also support efforts to develop equitable post-acute care planning strategies and guide Medicare policy toward more standardized and supportive care pathways following major lower limb amputation.