We used data collected by the Bureau of Alcohol and Drug Use Prevention, Care and Treatment (BADUPCT) of the NYC Department of Health and Mental Hygiene (DOHMH). Individual-level naloxone recipient data was collected from 170 OOPPs using a standardized naloxone recipient form. The 170 OOPPs comprised syringe service programs (n = 14), correctional health services (n = 8), shelters (n = 7), drug treatment programs (n = 50), healthcare facilities (n = 48), substance use-related community-based organizations (n = 7), drug treatment programs (n = 50), DOHMH (n = 5), multi-component priority programs (n = 14) and others (n = 17). Data were collected between April 1, 2018, and March 31, 2019. The quarterly counts of naloxone kits received by NYC residents from OOPPs were stratified by the naloxone recipients’ neighborhood of residence designated by United Hospital Fund (UHF) neighborhoods, racial/ethnic group (16). UHF neighborhood borders are contiguous with ZIP Codes, allowing us to assign ZIP Code-level residence data to each neighborhood without overlaps. The racial/ethnic groups included in the OOPP dataset were mutually exclusive and defined as: (1) Latino/Hispanic of any race, (2) non-Latino Black, (3) non-Latino White, and (4) non-Latino Other. The non-Latino Other category included: Asian, American Indian/Alaska Native, Pacific Islander/Native Hawaiian, two or more races, other, and don’t know.
We obtained annual, neighborhood-level counts of all-type overdose deaths and opioid-related overdose deaths from the DOHMH’s Bureau of Vital Statistics and the Office of the Chief Medical Examiner. However, the number of all-type and opioid-related overdose deaths were not stratified by racial/ethnic group or quarter due to data suppression guidelines.
We obtained other neighborhood-level characteristics from the United States Census American Community Survey (ACS), including the number of residents who identify as either non-Latino Black, non-Latino White, Latino or non-Latino Other, and the percentages of residents in poverty and residents with a Bachelor’s degree or higher in 2018 (US Census Bureau and American Community Survey, 2020). Neighborhood-level characteristics were created by aggregating characteristics from ZIP codes to UHF neighborhoods. Incarceration rates for different neighborhoods, defined as the rate of current imprisonment among people who identified a specific neighborhood as their resident neighborhood at intake, were calculated using methodology from the Prison Policy Initiative (18).
We conducted a multilevel negative binomial regression model nested by UHF neighborhoods to assess the difference in naloxone receipt rates across racial/ethnic categories. The outcome variable, quarterly naloxone receipt rate in a neighborhood, was defined as the number of naloxone kits received by individuals in each racial/ethnic group according to neighborhood and quarter, with racial/ethnic stratified population sizes defined as the offset. The independent variable of interest was categorical, representing the four mutually exclusive racial/ethnic groups noted above. Other covariates included in the multivariate models were neighborhood-level annual opioid-related overdose death rate, incarceration rate, percentage of residents in poverty, and the percentage of residents over 25 years old with a Bachelor’s degree or higher. We also conducted a sensitivity analysis where we replaced the opioid-related overdose death rate with the all-type overdose death rate in the model. Given the similarity in results, the latter results are not shown.
Lastly, we used the Getis-Ord Gi* statistic for neighborhood-level geospatial analysis. Using the Gi* statistic and z-scores, we identified geospatial clustering of racial/ethnic-specific naloxone receipt rates. Hot spots were clusters of ≥ 2 adjacent UHF neighborhoods with significantly higher naloxone receipt rates than the expected rate. Cold spots were clusters of ≥ 2 adjacent UHF neighborhoods with statistically significantly lower naloxone receipt rates than the expected. We performed separate geospatial analyses for each racial/ethnic group to evaluate within-group variation in the distribution of naloxone receipt rate across neighborhoods in NYC. This analysis used R version 1.0.143, SAS version 9.4, and ArcGIS version 10. The Brown University School of Public Health and NYC DOHMH Institutional Review Boards approved and considered this study exempt.