New means, new measures: assessing prescription drug‐seeking indicators over 10 years of the opioid epidemic

Abstract

Background and aims Prescription drug‐seeking (PDS) from multiple prescribers is a primary means of obtaining prescription opioids; however, PDS behavior has probably evolved in response to policy shifts, and there is little agreement about how to operationalize it. We systematically compared the performance of traditional and novel PDS indicators. Design Longitudinal study using a de‐identified commercial claims database. Setting United States, 2009–18. Participants A total of 318 million provider visits from 21.5 million opioid‐prescribed patients. Measurements We applied binary classification and generalized linear models to compare predictive accuracy and average marginal effect size predicting future opioid use disorder (OUD), overdose and high morphine milligram equivalents (MME). We compared traditional indicators of PDS to a network centrality measure, PageRank, that reflects the prominence of patients in a co‐prescribing network. Analyses used the same data and adjusted for patient demographics, region, SES, diagnoses and health services. Findings The predictive accuracy of a widely used traditional measure (N + unique doctors and N + unique pharmacies in 90 days) on OUD, overdose and MME decreased between 2009 and 2018, and performed no better than chance (50% accuracy) after 2015. Binarized PageRank measures however exhibited higher predictive accuracy than the traditional binary measures throughout 2009‐2018. Continuous indicators of PDS performed better than binary thresholds, with days of Rx performing best overall with 77–93% predictive accuracy. For example, days of Rx had the highest average marginal effects on overdose and OUD: a 1 standard deviation increase in days of Rx was associated with a 6–8% [confidence intervals (CIs) = 0.058–0.061 and 0.078–0.082] increase in the probability of overdose and a 4–5% (CIs = 0.038–0.043 and 0.047–0.053) increase in the probability of OUD. PageRank performed nearly as well or better than traditional indicators of PDS, with predictive performance increasing after 2016. Conclusions In the United States, network‐based measures appear to have increasing promise for identifying prescription opioid drug‐seeking behavior, while indicators based on quantity of providers or pharmacies appear to have decreasing utility.

Comparing measures of centrality in bipartite patient-prescriber networks: A study of drug seeking for opioid analgesics

“…For instance, eigenvector centrality is known to localize to a small set of nodes in large networks [ 49 ]. PageRank, on the other hand, has been proven robust and informative in the current context according to the literature [ 17 , 21 ]. We normalize all centrality measures across network components to keep centrality estimates at a consistent scale, and we take the natural log transformation of each centrality measure to account for their right-skewed distribution of centrality estimates.…”

“…One common measurement tracks whether a patient had a multiple provider episode (MPE), defined by obtaining opioid prescriptions from multiple doctors simultaneously (commonly four) over a short period of time (commonly 90 days) [ 14 , 15 ]. However, classifying prescription drug-seeking using MPE results in estimates with high specificity but low sensitivity: very few patients obtain opioid prescriptions from multiple providers over a short period of time [ 17 ]. Yet, lower thresholds for measuring opioid-seeking through MPEs are equally problematic, as many patients who obtain opioid prescriptions from two or three doctors during a short time period often do so for legitimate (and often urgent) medical reasons.…”

“…This would result in patients seeking opioids for nonmedical use holding prominent structural positions within patient-provider opioid prescription networks [ 23 ]. Based on this observation, two studies found that network centrality in a patient-provider opioid prescription network is strongly associated with other common measures of nonmedical prescription opioid use and that centrality is significantly more predictive of future opioid overdoses than are most traditional measures [ 17 , 21 ].…”

“…To confirm and compare the effectiveness of bipartiteness-aware node prominence index in identifying opioid drug seekers, the present paper studies four variants of bipartite centrality index, i.e., HITS [ 40 ], CoHITS [ 37 ], BGRM [ 39 ], and BiRank [ 38 ], in a real-world social network of patient-provider relationships. We also project the bipartite patient-provider network into a unipartite patient-patient network and apply PageRank to produce a baseline [ 17 , 21 ]. Lacking the ground truth of drug seeking behavior, we use opioid overdose diagnosis as the outcome and evaluate how strongly each index is associated with the outcome for patients receiving opioid prescriptions.…”

Commentary on Perry et al.: New means, new measures—without discarding all the previous ones!

“…Starting from the observation that there is little agreement in the literature about how to best operationalize prescription drug‐seeking (PDS), Perry et al . compared the performances of multiple PDS indicators during 10 years of the opioid epidemic in the United States [1]. The authors illustrated the value of a network centrality measure (PageRank) that they had introduced recently [2].…”