OBJECTIVES: Prevalence of diabetes in peripheral artery disease patients is high and these patients are at increased risk for major cardiovascular events. This study’s aim was to use a retrospective cohort to assess healthcare resource use (HRU) and costs among critical limb ischemia (CLI) patients with diabetes.

METHODS: Using a major US database comprised of integrated administrative claims and electronic health records (2007-15), we estimated annual all-cause HRU and total healthcare costs for a sample of 3,189 CLI adults ≥50 years. CLI was characterized by rest pain, ulceration or gangrene. HRU and costs were calculated from medical and pharmacy claims for 1 year following first diagnosis of CLI (index date). Patients who died in the post-index period were included. We stratified patients into 2 cohorts: with and without pre-index diabetes diagnosis. Reverse Engineering and Forward Simulation (REFS^TM), a hypothesis free machine learning platform that uses Bayesian network inference, was used to build an ensemble of prediction models for hospitalization and annual total healthcare costs of CLI patients.

RESULTS: Nearly 50% of CLI patients had comorbid diabetes. Diabetics had more hospitalizations (mean [SD]: 1.5 [2.0] vs 1.1 [1.6], p<0.001) and higher costs ($70,808 [$102,568] vs $43,319 [$76,751], p<0.001) compared to non-diabetics. REFS^TM selected factors with the highest frequency in both models. Those predictive of hospitalization were (OR, SD): cellulitis and abscess (2.1, 0.04), beta blockers non-selective (2.1, 1.3), and other aftercare (1.6, 0.1). Predictors of increased costs were (% change in cost, SD): south region (1.2, 0.03), chronic skin ulcers (2.0, 0.1), and chronic kidney disease (1.9, 0.2).

CONCLUSIONS: HRU and costs were higher for CLI patients with comorbid diabetes. Factors driving these increases in the overall CLI population may be related to the increased complexity of comorbid diabetes and may provide an opportunity for cost savings via timely care decisions in diabetes and CLI.