Using the Health Care Cost Institute claims database, we identified all individuals from 2008 to 2017 who were under 65 years of age, covered by employer-sponsored health insurance, and who did not have a cancer diagnosis or hospice claim within a year. We excluded patients who moved to different states within a month or were enrolled in multiple health plans within a month. Using the Center for Disease Control’s most current cumulative list of opioids at the National Drug Code (NDC) level, we then identified individuals with at least 1 claim. We excluded buprenorphine (used mainly to treat opioid addiction) and any opioids that did not appear in any claims from 2008 to 2017. Our denominator is average member months (total member months of all qualifying individuals within a year, divided by 12) to account for changes in year-to-year total enrollment. In the text, we refer to our utilization amount as “per person” which is the numerator of interest (aggregated sum of a specified utilization metric) divided by (average member months multiplied by 12).

Our analysis of utilization specifically measured “pills” which measures the typical number of tablets, patches, or otherwise specified delivery method supplied to patients. The other utilization metric is filled days which is the estimated amount of days a drug should last as written on a prescription label. The potency of opioids differs significantly across the commonly prescribed classes, so we also examined morphine milligram equivalents (MMEs) which standardizes strength across all opioids. MMEs are calculated by multiplying the (units of drug) by the (gram specific strength) by the (MME conversion factor).

The ‘other’ category, which accounts for a small share of total opioid use, includes butorphanol, dihydrocodeine, hydromorphone, levorphanol, meperidine, methadone, opium, oxymorphone, pentazocine, propoxyphene, and tapentadol. Drug categories display the combined results for the short acting and long acting prescriptions for all opioids except oxycodone. The major change in composition of the other category from 2008 to 2011 was the elimination of propoxyphene in November 2010 due to unintended cardiac issues related to the drug.

Data was suppressed at the state level for Hawaii and Vermont due to low data availability. Certain drug categories were condensed into the “other” category for states where opioid utilizer counts were low. Opioid utilization at the national level contains data from all 50 states.