Detecting Suspected Prescribing Cascades by Prescription Sequence Symmetry Analysis of Nationwide Real-World Data

Abstract

A prescription cascade is a subsequent event that occurs when an adverse drug event is misinterpreted as a new medical condition, resulting in the prescription of a potentially unnecessary medication to treat this new condition. This study was designed to test the hypothesis of whether prescribing cascades exist and their magnitude among older individuals by using prescription sequence symmetry analysis (PSSA). A retrospective population-based cohort study based on data from Taiwan's National Health Insurance Research Database (NHIRD), a nationwide claims-based database that covers 99% of the population in Taiwan. Patients receiving newly initiated index and marker drugs in the outpatient setting between 2014 and 2016 were included. Those who received index and marker drugs on the same date were excluded. PSSA measures the propensity for a marker drug (eg, thyroxine) to be prescribed after an index drug (eg, amiodarone) has been initiated, where the index drug is suspected of inducing a side effect (eg, hypothyroidism) and the marker drug is used to treat the side effect. We evaluated 14 PSSA sets as potential prescription cascade-related drug pairs (index and marker drug), including 2 confirmatory analyses (eg, amiodarone and thyroxin), to check the validity of the PSSA results and 12 exploratory analyses of cardiovascular medication-related prescribing cascades (eg, statins and antidepressants). The observation periods for sequences of incident marker drug use were 1year before and 1year after the use of the incident index drug, and the symmetry of prescriptions filled for the 2 periods was examined. The results of the PSSA were performed as the crude sequence ratio (SR), defined as the ratio of patients initiating the marker drug after the index drug (causal group) to those initiating the marker drug before the index drug (noncausal group), and the adjusted SR (aSR) with 95% confidence intervals (CIs). Among 12 potential prescription cascade-related drug pairs in exploratory analysis, 9 of them reached statistical significance, and aSR ranged from 1.02 to 1.46. Dihydropyridine calcium channel blocker-induced edema showed the highest aSR (1.46, 95% CI 1.45-1.48), and statin-induced muscle pain showed the lowest aSR (1.02, 95% CI 1.02-1.03). Our study supported the application of PSSA in detecting prescribing cascades using a nationwide claims database. Nevertheless, most previously reported prescribing cascades among cardiovascular medications were shown to have a low effect size of sequence ratios among older individuals.