Estimation of Missed Statin Prescription Use in an Administrative Claims Dataset

• Published in: Journal of managed care & specialty pharmacy Vol. 23; no. 9; pp. 936 - 942
• Main Authors: Wade, Rolin L, Patel, Jeetvan G, Hill, Jerrold W, De, Ajita P, Harrison, David J
• Format: Journal Article
• Language: English
• Published: United States Academy of Managed Care Pharmacy 01-09-2017
• Subjects: Cardiovascular Diseases - chemically induced; Databases, Factual; Drugs, Generic - adverse effects; Drugs, Generic - therapeutic use; Female; Humans; Hydroxymethylglutaryl-CoA Reductase Inhibitors - adverse effects; Hydroxymethylglutaryl-CoA Reductase Inhibitors - therapeutic use; Insurance Claim Review - statistics & numerical data; Male; Managed Care Programs - statistics & numerical data; Middle Aged; Prescription Drugs - adverse effects; Prescription Drugs - therapeutic use; Retrospective Studies
• ISSN: 2376-0540; 2376-1032
• DOI: 10.18553/jmcp.2017.23.9.936

Nonadherence to statin medications is associated with increased risk of cardiovascular disease and poses a challenge to lipid management in patients who are at risk for atherosclerotic cardiovascular disease. Numerous studies have examined statin adherence based on administrative claims data; however, these data may underestimate statin use in patients who participate in generic drug discount programs or who have alternative coverage.
To estimate the proportion of patients with missing statin claims in a claims database and determine how missing claims affect commonly used utilization metrics.
This retrospective cohort study used pharmacy data from the PharMetrics Plus (P+) claims dataset linked to the IMS longitudinal pharmacy point-of-sale prescription database (LRx) from January 1, 2012, through December 31, 2014. Eligible patients were represented in the P+ and LRx datasets, had ≥1 claim for a statin (index claim) in either database, and had ≥ 24 months of continuous enrollment in P+. Patients were linked between P+ and LRx using a deterministic method. Duplicate claims between LRx and P+ were removed to produce a new dataset comprised of P+ claims augmented with LRx claims. Statin use was then compared between P+ and the augmented P+ dataset. Utilization metrics that were evaluated included percentage of patients with ≥ 1 missing statin claim over 12 months in P+; the number of patients misclassified as new users in P+; the number of patients misclassified as nonstatin users in P+; the change in 12-month medication possession ratio (MPR) and proportion of days covered (PDC) in P+; the comparison between P+ and LRx of classifications of statin treatment patterns (statin intensity and patients with treatment modifications); and the payment status for missing statin claims.
Data from 965,785 patients with statin claims in P+ were analyzed (mean age 56.6 years; 57% male). In P+, 20.1% had ≥ 1 missing statin claim post-index; 13.7% were misclassified as nonstatin users; and 14.9% were misclassified as new statin users. MPR was higher in the augmented P+ dataset versus the P+ dataset alone for all patients (79.4% vs. 76.7%, P < 0.001) and new users (61.4% vs. 58.7%, P < 0.001). Similarly, mean PDC was higher in the P+ dataset augmented with LRx versus the P+ dataset alone for all patients (76.0% vs. 74.0%, P < 0.001) and new users (58.5% vs. 56.5%, P < 0.001). Most patients received moderate-intensity statins; few changes in dose, intensity, or discontinuation of statins were observed when the P+ dataset was augmented. The most common reasons for missing data were payment by an alternate third-party program (66.3%) and use of cash, coupon, or discount cards (18.7%).
Augmenting commercial claims data with point-of-sale data provides a more accurate assessment of statin use than claims data alone.
This study was funded by Amgen, which contributed to data interpretation and manuscript preparation. Wade, Hill, and De are employees of QuintilesIMS, which received funding from Amgen for work on this study. Patel and Harrison are employees of Amgen and own Amgen stock/stock options. Study concept and design were contributed by Wade, Hill, Patel, and Harrison. De took the lead in data collection, along with the other authors, and all authors contributed to data analysis. The manuscript was written and revised by all the authors.