Table 2. Comparison of prediction performance using any of the Centers for Medicaid & Medicaid Services (CMS) high-risk opioid use measures vs. elastic net in the validation sample (n = 114,253) over a 12-month period^a.

	Any CMS measureb		High risk in elastic net using different thresholdsc
Risk subgroups (n, % of the cohort)	Low risk (n = 110,171, 96.4%)	High risk (n = 4,082, 3.57%)	Top 1 percentile (n = 2,207, 1.93%)	Top 5th percentile (n = 11,633, 10.18%)	Top 10th percentile (n = 23,541, 20.6%)
Number of actual OUD (% of each subgroup)	412 (0.4)	155 (3.8)	186 (8.4)	391 (3.4)	475 (2.0)
Number of actual non-OUD (% of each subgroup)	109,759 (99.6)	3,927 (96.2)	2,021 (91.6)	11,242 (96.6)	23,066 (98.0)
NNE	270	26	11	29	49
% of all OUD over 12 months (n = 567) captured	72.7	27.3	32.8	69.0	83.8

Abbreviations: NNE: number needed to evaluate; OUD: opioid use disorder

^a: The CMS measures were based on a 12-month period rather than three months. To compare CMS measures, beneficiaries were thus required to have at least a 12-month period of follow up and the resulting sample size was smaller than the sample size in the main analysis. If classifying beneficiaries with any of the CMS high-risk opioid use measures as OUD, the remaining will be consider as non-OUD.

^b: The 2019 CMS’ opioid safety measures are meant to identify high-risk individuals or utilization behavior.¹ These measures include 3 metrics (1) high-dose use, defined as >120 MME for ≥90 continuous days, (2) ≥4 opioid prescribers and ≥4 pharmacies, or (3) concurrent opioid and benzodiazepine use ≥30 days.

^c: For elastic net, we presented high-risk groups using different cutoff thresholds of prediction probability: individuals with (1) predicted probability in the top 1 percentile (0.95); (2) predicted probability in the top 5^th percentile (0.77) or above; and (3) predicted probability in the top 10^th percentile (0.61) or above. If classifying beneficiaries in the high-risk group of OUD, the remaining will be consider as non-OUD.