Scale-Up Methods as Applied to Estimates of Heroin use

Abstract

The feasibility of using the network scale-up method to estimate heroin use is described. A random sample was asked “How many people do you personally know” who use heroin, and how many in other subpopulations – robbery, assault, burglary, auto-theft victims, binge drinkers, and marijuana users – whose size is more accurately known. A model estimated the overall number of persons each respondent knew and the size of each subpopulation. Estimates of the subpopulation are compared with known subpopulation sizes to assess the plausibility of the model. Data came from the 1999 survey evaluating the “Fighting Back” substance prevention program. Fourteen sites with clear political boundaries were used (n = 5892). Heroin use varied from city to city. Rates estimated for heroin use correlated .832 with the level of respondents' sense of “crime in their neighborhood.” The average ratio between the known populations and the estimates is .943. Members of each subpopulation, especially drug users, tended to know more people within their own subpopulation.