A Machine Learning Model for Prediction of Amputation in Diabetics.

Area of Science:

• Medical Informatics
• Clinical Prediction Models
• Machine Learning in Healthcare

Background:

• Diabetic foot ulcer (DFU) and subsequent lower extremity amputation are linked to poor survival outcomes.
• Predicting major amputation risk in hospitalized DFU patients is crucial for improving patient prognosis.

Purpose of the Study:

• To develop and validate a predictive model for major lower extremity amputation in hospitalized patients with DFU.
• To identify key clinical factors contributing to major amputation risk in DFU patients.

Main Methods:

• Utilized the National Inpatient Sample (NIS) database (2008-2014) for patient selection.
• Employed machine learning algorithms, including decision trees (CTREE) and random forests, for model development.
• Analyzed International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) and AHRQ comorbidity codes.

Main Results:

• Identified 326,853 inpatients with DFU; 5.9% underwent major amputation.
• Top predictors for amputation included gangrene (OR=11.8), peripheral vascular disease (OR=2.9), weight loss (OR=2.6), systemic infection (OR=2.5), and osteomyelitis (OR=1.7).
• The predictive model achieved 77.8% accuracy on testing data, with an Area Under the Curve (AUC) of 0.84.

Conclusions:

• A validated clinical algorithm using machine learning accurately predicts major lower extremity amputation risk in hospitalized DFU patients.
• The model demonstrates potential for early identification of high-risk individuals, enabling timely intervention.
• This tool can aid clinicians in managing DFU patients and mitigating amputation risk.