Association for Public Policy Analysis & Management

General population surveys (GPS) are the mainstay of epidemiologic research and policy statements on the use of addicting substances. For alcohol, cannabis and tobacco, substances in use by a large share of the population, GPS serve well enough to describe the scale of use, changes over time and distribution across major population groups. The surveys work less well for another purpose for which they also play a major role, at least in the United States, namely the prevalence of, and trends in, frequent use of expensive drugs.

The rise in the number of opioid caused deaths has generated a great interest in the number of those using heroin frequently. Estimates based on the National Survey on Drug Use and Health, the dominant GPS in the United States, are routinely published in prominent research journals and cited in policy. Yet those estimates are implausibly low when compared to other indicators, such as treatment admissions and arrestees reporting frequent heroin use. A substantial literature shows that (1) the sample framework is importantly incomplete since omitting the incarcerated population drops a large number of frequent users of heroin; (2) selective non-response is substantial and (3) under-reporting is comparably problematic. Alternative estimates that incorporate a less statistically strong sample of arrestees (ADAM: Arrestee Drug Abuse Monitoring) are about 5 times as large as those from NSDUH and have been published in leading journals (e.g. Caulkins et al., 2015). We explore the sources of the differences in the estimates and the factors that might explain the unwillingness of the public health research and policy community to accept the implausibility of the NSDUH heroin estimates.