Does Multisite Sampling Improve Patient Heterogeneity in Drug Misuse Research?
John Dunn, Cleusa Pinheiro Ferri, Ronaldo Laranjeira
Probability sampling is the gold standard for obtaining representative samples of individuals from a population. However, when drug misusers are the target group various problems arise which mean that probability sampling can become impracticable (Dunn & Ferri, 1999). First, there is the problem of low prevalence. For example, an estimate from the UK suggests that around 2% of the general population are dependent on illicit drugs (Farrell et al, 1998). However, it is likely that less than a quarter of these are dependent on individual substances, such as cocaine. Consequently, to find just one cocaine addict, one might need to interview at least 400 people. To find enough cocaine addicts (300 for example) to allow us to make statistically powerful statements about their characteristics, we might need to interview over 120,000 people. The larger the sample, the longer it will take to complete the interviews, the more expensive the study becomes and since the ratio of cases to non-case is very low, the whole process becomes extremely inefficient.
A second problem with probability sampling is that illegal activities, such as drug misuse, may be hidden or denied, so it can be difficult to find or identify users. Surveys tend to be based on the occupants of private households, so cocaine addicts may missed because they are out buying drugs or involved in illegal activities to finance their use or because they are homeless, in prison or in residential treatment (Dunn & Ferri, 1999). Consequently many research studies that aim to investigate the characteristics of drug users and their patterns of use tend to employ convenience samples using patients from outpatient and inpatient units (Griffin et al, 1989; Kleinman et al, 1990; Dunn et al, 1996). However, selection bias means that drug users recruited from these settings have very different characteristics to those who are not in treatment. Studies suggest that drug users not in treatment are more likely to be polydrug users, be involved in illegal activities but have fewer negative consequences of drug use and less depression (Rounsaville & Kleber, 1985; Carroll & Rounsaville, 1992).