The use of structured data elements to identify ASCVD patients with statin-associated side effects: Insights from the Department of Veterans Affairs

Abstract

Accurate identification of patients with statin-associated side effects (SASEs) is critical for health care systems to institute strategies to improve guideline-concordant statin use. The objective of this study was to determine whether adverse drug reaction (ADR) entry by clinicians in the electronic medical record can accurately identify SASEs. We identified 1,248,214 atherosclerotic cardiovascular disease (ASCVD) patients seeking care in the Department of Veterans Affairs. Using an ADR data repository, we identified SASEs in 15 major symptom categories. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were assessed using a chart review of 256 ASCVD patients with identified SASEs, who were not on high-intensity statin therapy. We identified 171,189 patients (13.71%) with documented SASEs over a 15-year period (9.9%, 2.7%, and 1.1% to 1, 2, or >2 statins, respectively). Statin use, high-intensity statin use, low-density lipoprotein cholesterol, and non-high-density lipoprotein cholesterol levels were 72%, 28.1%, 99mg/dL, and 129mg/dL among those with vs 81%, 31.1%, 84mg/dL, and 111mg/dL among those without SASEs. Progressively lower statin and high-intensity statin use, and higher low-density lipoprotein cholesterol and non-high-density lipoprotein cholesterol levels were noted among those with SASEs to 1, 2, or >2 statins. Two-thirds of SASEs were related to muscle symptoms. Sensitivity, specificity, PPV, NPV compared with manual chart review were 63.4%, 100%, 100%, and 85.3%, respectively. A strategy of using ADR entry in the electronic medical record is feasible to identify SASEs with modest sensitivity and NPV but high specificity and PPV. Health care systems can use this strategy to identify ASCVD patients with SASEs and operationalize efforts to improve guideline-concordant lipid-lowering therapy use in such patients. The sensitivity of this approach can be further enhanced by the use of unstructured text data.