Page 9
crosswalk provided by the Department of Housing and Urban Development. ARCOS data are reported as rates of morphine equivalent opioid transactions (called “sales” in this report) per 100,000 residents. There are several important limitations to statistical use of ARCOS data. First, not all opioids are reported into ARCOS, particularly opioids that are not Schedule III. Second, being an administrative data collection, entries into ARCOS may not be consistent and may be correlated with the prevalence of opioid misuse. For example, an investigation by the DEA Inspector General16 in 2002 found that ARCOS reports “are limited in their value…because of problems of completeness, accuracy, and timeliness.” It is unclear to what extent data collection has improved since 2002. One study of psychostimulants found that ARCOS had a high reliability when compared to a state-run prescription drug monitoring program, suggesting that whatever the issues are with absolute measurement, the distribution of ARCOS data may be appropriate.17 Finally, while research has found a correlation between legitimate use of opioids for therapy and opioid misuse18 , not all individuals that misuse opioids obtain their drugs through prescriptions. This measure will be biased to the extent that non-medical flows of opioids do not mirror prescription flows. Opioid-Related Hospitalizations Data on hospital stays and emergency department visits related to opioids were drawn from the State Inpatient Databases (SID) and State Emergency Department Databases (SEDD), collected as part of the Healthcare Cost and Utilization Project from the Agency for Healthcare Quality and Research. States voluntarily submit patient data to the SID and SEDD, following ICD codes. Hospital inpatient stays and emergency department visits are non-duplicative: patients admitted to the hospital after visiting the ER 14 are considered a hospital stay and removed from the ED visit. Data were available for 32 states, for the years 2011 through 2014. Patient records were aggregated to the county of patient residence, and calculated as rate per 100,000 residents. Separate data are reported for several categories of substances. Table A1 reports the ICD-9 codes used in this report. Table A1. ICD-9 Codes for Hospital and Emergency Department Data ICD-9-CM diagnosis codes Description 304.00–304.03 Opioid type dependence 304.70–304.73 Combinations of opioids with any other 305.50–305.53 Nondependent opioid abuse 760.72 Narcotics affecting neonate 965.00 Poisoning by opium 965.01 Poisoning by heroin 965.02 Poisoning by methadone 965.09 Poisoning by other opiates and related narcotics E850.0 Heroin poisoning, accidental E850.1 Methadone poisoning, accidental E850.2 Other opiates and related narcotics poisoning, accidental E935.0 Heroin, adverse effects Medicare Part D Opioid Prescriptions Data on Medicare Part D opioid prescriptions come from the CMS Patient Drug Event (PDE) file, and are for all claims processed between January 1st 2006 through December 31st, 2016. The PDE file is an administrative data source, based on claims from beneficiaries enrolled in Medicare Part D. Data were tabulated by the address of the Medicare beneficiary. Prescriptions are included for all opioid-containing prescriptions. A complete list of medications is available upon request. Data were calculated as Medicare Part D opioid prescriptions per 100,000 population. Death Due to Accidental Drug Poisoning Data on deaths due to accidental drug poisoning come from small area estimates produced by the CDC.19 The data contain deaths due to any substances, excluding alcohol and tobacco. Details can be found here: https://www.cdc.gov/nchs/data-visualization/drug-poisoning-mortality/. Data were calculated as age adjusted death rates per 100,000 population. Economic Measures Poverty rates were drawn from U.S. Census Bureau’s Small Area Income and Poverty Estimates (SAIPE), which use small area estimation techniques to augment data collected from the American 15 Community Survey. Unemployment rates were drawn from the Bureau of Labor Statistics. Data for the employment-to-population ratio were drawn from the Census Bureau’s County