Identification of Patients with Statin Intolerance in a Managed Care Plan: A Comparison of 2 Claims-Based Algorithms
Brandon K Bellows,Amy M Sainski-Nguyen,Cody J Olsen,Susan H Boklage,Scott Charland,Matthew P Mitchell,Diana I Brixner
DOI: https://doi.org/10.18553/jmcp.2017.23.9.926
Abstract:Background: While statins are safe and efficacious, some patients may experience statin intolerance or treatment-limiting adverse events. Identifying patients with statin intolerance may allow optimal management of cardiovascular event risk through other strategies. Recently, an administrative claims data (ACD) algorithm was developed to identify patients with statin intolerance and validated against electronic medical records. However, how this algorithm compared with perceptions of statin intolerance by integrated delivery networks remains largely unknown. Objective: To determine the concurrent validity of an algorithm developed by a regional integrated delivery network multidisciplinary panel (MP) and a published ACD algorithm in identifying patients with statin intolerance. Methods: The MP consisted of 3 physicians and 2 pharmacists with expertise in cardiology, internal medicine, and formulary management. The MP algorithm used pharmacy and medical claims to identify patients with statin intolerance, classifying them as having statin intolerance if they met any of the following criteria: (a) medical claim for rhabdomyolysis, (b) medical claim for muscle weakness, (c) an outpatient medical claim for creatinine kinase assay, (d) fills for ≥ 2 different statins excluding dose increases, (e) decrease in statin dose, or (f) discontinuation of a statin with a subsequent fill for a nonstatin lipid-lowering therapy. The validated ACD algorithm identified statin intolerance as absolute intolerance with rhabdomyolysis; absolute intolerance without rhabdomyolysis (i.e., other adverse events); or as dose titration intolerance. Adult patients (aged ≥ 18 years) from the integrated delivery network with at least 1 prescription fill for a statin between January 1, 2011, and December 31, 2012 (first fill defined the index date) were identified. Patients with ≥ 1 year pre- and ≥ 2 years post-index continuous enrollment and no statin prescription fills in the pre-index period were included. The MP and ACD algorithms were applied to the population, and concordance was examined using individual (i.e., sensitivity, specificity, positive predictive value [PPV], and negative predictive value [NPV]) and overall performance measures (i.e., accuracy, Cohen's kappa coefficient, balanced accuracy, F-1 score, and phi coefficient). Results: After applying the inclusion criteria, 7,490 patients were evaluated for statin intolerance. The mean (SD) age of the population was 51.1 (8.5) years, and 55.7% were male. The MP and ACD algorithms classified 11.3% and 5.4% of patients as having statin intolerance, respectively. The concordance of the MP algorithm was mixed, with negative classification of statin intolerance measures having high concordance (specificity 0.91, NPV 0.97) and positive classification of statin intolerance measures having poor concordance (sensitivity 0.45, PPV 0.21). Overall performance measures showed mixed agreement between the algorithms. Conclusions: Both algorithms used a mix of pharmacy and medical claims and may be useful for organizations interested in identifying patients with statin intolerance. By identifying patients with statin intolerance, organizations may consider a variety of options, including using nonstatin lipid-lowering therapies, to manage cardiovascular event risk in these patients. Disclosures: This study was funded by Regeneron Pharmaceuticals and Sanofi US. Boklage is employed by, and owns stock in, Regeneron, and Charland is employed by Sanofi. Bellows has received fees from Avenir for advisory board membership and grants from Myriad Genetics, Biogen, Janssen, and National Institutes of Health. Brixner reports advisory board and consultancy fees and grants from Sanofi. Mitchell reports consultancy fees from Sanofi. Study concept and design were contributed by Bellows, Boklage, Charland, and Brixner. Bellows, Sainski-Nguyen, and Olsen took the lead in data collection, along with Mitchell. Data interpretation was performed by Mitchell, along with the other authors. The manuscript was written by Bellows, Sainski-Nguyen, and Olsen and revised by all the authors.