AMPREDICT MoRe: Predicting Mortality and Re-amputation Risk after Dysvascular Amputation

Area of Science:

• Vascular Surgery
• Health Informatics
• Predictive Analytics

Background:

• Dysvascular amputation, often due to diabetes and peripheral arterial disease, carries significant risks of mortality and re-amputation.
• Existing prediction models face implementation barriers due to data requirements.
• Electronic Health Records (EHR) offer a rich, accessible data source for clinical prediction.

Purpose of the Study:

• To develop and validate a novel prediction model, AMPREDICT MoRe, for death and re-amputation post-dysvascular amputation.
• To utilize only readily available EHR predictors to overcome implementation barriers.
• To support clinical decision-making at the time of amputation level selection.

Main Methods:

• Retrospective cohort study of 9,221 patients undergoing incident unilateral dysvascular amputation (transmetatarsal, transtibial, or transfemoral).
• Utilized Veterans Affairs Corporate Data Warehouse data from October 2015 to September 2021.
• Employed multinomial logistic regression and LASSO (least absolute shrinkage and selection operator) for variable selection and model fitting.
• External validation performed on 20% of the cohort.

Main Results:

• The final AMPREDICT MoRe model included 23 EHR-based predictors.
• Observed outcome distributions: No Death/No Re-amputation (57.7%), No Death/Re-amputation (22.9%), Death/No Re-amputation (14.3%), Death/Re-amputation (5.1%).
• Model demonstrated moderate overall discrimination (M index 0.70), with stronger prediction for death (0.79) than re-amputation (0.67), and best discrimination between no adverse outcomes and both adverse outcomes (0.82).

Conclusions:

• The AMPREDICT MoRe model is successfully developed and validated for predicting death and re-amputation after dysvascular amputation.
• Its reliance on EHR data facilitates practical application in clinical settings.
• Future decision support tools can be developed without requiring patient interviews, streamlining clinical workflows.