A different kind of contextual effect: geographical clustering of cocaine incidence in the USA
- 1Substance Abuse and Mental Health Services Administration, Office of Applied Studies, Rockville, Maryland, USA
- 2The Johns Hopkins University, Bloomberg School of Public Health, Department of Mental Hygiene, Baltimore, Maryland, USA
- Correspondence to: Kenneth R Petronis, Substance Abuse and Mental Health Services Administration, Office of Applied Studies, 5600 Fishers Lane, 16-105, Rockville, MD 20857, USA;
- Accepted 30 March 2003
Study objective: Outline the use of the pairwise odds ratio (PWOR) to quantify the extent to which a binary outcome clusters geographically. Quantify the extent to which first experience with cocaine is spatially correlated within US neighbourhoods and cities. Quantify geographical clustering of first experience with cocaine by neighbourhood context.
Design: Estimate the PWOR of incident cocaine experience at two levels (neighbourhood, city) and compare across years. Within years, estimate the PWOR by neighbourhood disadvantage and test for trend.
Setting: US National Household Survey on Drug Abuse.
Participants: Civilian, non-institutionalised household residents of the United States age 12 years and older interviewed in person during 1979, 1988, 1990, 1991, 1992, 1993.
Main results: First experience with cocaine clusters within US neighbourhoods and cities. There is some evidence that the spatial correlation of first experience with cocaine increases with percentage of neighbourhood households living in poverty.
Conclusions: The gradient in spatial correlation of incident cocaine experience by neighbourhood poverty level is consistent with current theories of concentrated disadvantage. The possibility that the direction of the poverty gradient might change over the course of a drug epidemic is discussed.
- PWOR, pairwise odds ratio
- NHSDA, National Household Survey on Drug Abuse
- PSU, primary sampling unit
- SES, socioeconomic status
- CWE, chief wage earner
- ALR, alternating logistic regressions