Optimizing identification of major adverse limb and cardiovascular events among adults with peripheral artery disease in administrative data.

OBJECTIVE: Atherosclerotic cardiovascular diseases (ASCVD) are the commonest cause of death. Peripheral artery disease (PAD) is an ASCVD that significantly increases risk of death and reduces quality of life; however, characterization of its consequences is limited because it requires manual adjudication of major adverse cardiovascular events (MACEs) and major adverse limb events (MALEs). Hence, we developed and validated a system to quantify MACE and MALE using administrative data.

METHODS: In a multihospital single health care system (2016-2023), we identified adult index vascular surgery clinic visits for PAD. We randomly sampled the patients for clinical adjudication of their electronic health record (EHR) data to identify long-term MACE (myocardial infarct, nontraumatic stroke, cardiac death) and MALE (major revascularization, amputation). We achieved consensus through a modified Delphi process with three rounds of EHR review by five experts, generating the gold standard diagnosis. We compared the accuracy of hospitalization diagnosis (International Classification of Diseases) and/or procedure (Current Procedural Terminology) codes to the EHR review gold standard diagnoses. Testing parameters identified the optimal administrative coding strategies when compared to the gold standard diagnoses. The predictive increment was measured by net reclassification indices (NRI).

RESULTS: We included 620 patients (mean age, 70 ± 12 years; 40% female; 89% White race; 16% frail). Throughout long-term follow-up (median, 2.7 years; interquartile range, 1.3-4.7 years), clinical adjudication identified 13% MACE and 22% MALE. Administrative strategies yielded similar rates for MACE and MALE. For MACE, diagnosis-only codes optimized identification with an F1 score of 60.3% and 0.55 Mathews Correlation Coefficient. For MALE, procedure-only codes optimized identification with an F1 score of 95.2% F1 score and Matthews correlation coefficient of 0.94. The NRI was 33% for MACE and 2% for MALE.

CONCLUSIONS: Administrative codes can accurately identify ASCVD outcomes among patients with PAD. MACE identification is inferior to MALE identification, but optimized with International Classification of Diseases codes only, driven by limitations in positive predictive value and NRI. MALE identification, which is limited to procedure-based events, is optimized with Current Procedural Terminology codes alone, with minimal false negatives.