Introduction:
The clinical and economic impact of statin intolerance (SI) in high CV risk patients is unknown due, in part, to a lack of consensus in its definition. We sought to define and validate an SI algorithm for use in an administrative database (AD) among high-CV risk patients
Methods:
Adults with ≥1 qualifying change (See Table 1) in statin therapy and ≥1 prior diagnosis of hyperlipidemia, hypercholesterolemia, or mixed dyslipidemia were identified from the AD of the Health Alliance Plan at Henry Ford Health System (HFHS). A sample of 1000 patients was drawn from the pool of eligible adults and stratified by high CV risk based on presence of comorbid conditions including diabetes, coronary heart disease, and peripheral artery disease. Statin utilization and adverse events data were abstracted both from the AD and the HFHS electronic medical record (EMR). SI was defined using both a primary definition inclusive of all possible statin related adverse events and a secondary definition that included only musculoskeletal events. SI was categorized as absolute (AI) or titration (TI) intolerance. The performance of the AD algorithm was assessed using measures of concordance (Cohen’s kappa [κ]) and accuracy (sensitivity, specificity, positive predictive value [PPV]) with the EMR as reference.
Results:
A total of 353 patients (48% female, 44% Caucasian, mean (SD) age 63 (12) years) were identified as high CV risk with 33% having a history of CHD, 77% diabetes and 2% PAD. Forty-two percent of patients were on simvastatin, 35% atorvastatin, 11% lovastatin, 7% rosuvastatin and 6% pravastatin/fluvastatin. Table 1 characterizes the validation sample. SI was identified in 19.3% and 20.7%, AI in 3.1% and 2.8%, and TI in 16.7% and 18.7% of patients in the EMR and AD, respectively. The algorithm identifying any SI had robust concordance (κ=0.73), good sensitivity (80.9%) and PPV (75.3%). The TI algorithm performed better (κ=0.78, sensitivity=86.4%, PPV=77.3%) than the AI algorithm (κ=0.56, sensitivity=54.5%, PPV=60.0%). Specificity was high (>94%) across all 3 algorithms.
Conclusion:
This study successfully defined SI among high-CV risk patients using an evidence-based validated algorithm. To our knowledge, this is the first such algorithm for use in AD to be made available to decision-makers.
On Minimal Copulas under the Concordance Order
