Methods estimating the prevalence of problem drug use

Abstract

0 d rug treatment program data. His aim was to challenge conentional ‘wisdom’ regarding the interpretation of these data, hich had inferred that heroin use had peaked in the United tates, and had then begun to decline. Hunt’s analysis sugested that incident heroin use was somewhat higher than simple interpretation of the data suggested, and that in maller US cities, initiation of heroin use was continuing o increase. We now largely take for granted the application f sophisticated statistical and mathematical models for estiating trends in drug use, but at the time this analysis was oth novel and challenging to conventional interpretations of outine data collections. There are two important criticisms of the analyses perormed by Hunt. First, the estimates crucially depend on the vailability of a good estimate of the rate at which new heroin sers enter treatment. Hunt used information collected in the rug treatment programs to estimate what he called this ‘lag’ between the onset of dependence symptoms and treatment entry (Kessler et al., 2001). Finally, a significant proportion of drug users do not ever seek treatment for their drug use (Kessler et al., 2001): a factor that Hunt acknowledges but downplays. The implication of this is that the population that Hunt was estimating was the number of new heroin userswho would subsequently enter treatment (who would presumably also be regular or dependent users) within 6–7 years of initiating use. The total number of new heroin users each year would therefore be underestimated as the estimate would not include those heroin users who died, became abstinent without the aid of formal treatment, entered treatment more than 6–7 years after initiating use, or never entered treatment. This is an important point in the interpretation of the results that is not made appropriately clear. A second important criticism again relates to the rate that new heroin users enter drug treatment. Hunts assumed that istribution. To be fair, Hunt did this in such a manner as o avoid the obvious pit-falls, and also included some good iscussion of the problem. However, it does mean that the ssumption in Hunt’s analyses was that all new heroin users ill enter drug treatment within 6 or 7 years. This assumpthis ‘lag’ distribution had not changed over time, and used the data from the treatment programs to justify this assumption. The problem with this is firstly that drug treatment program data would probably not have been a good method for detecting such time trends in the ‘lag’ distribution because of the r a a b r c a u ion was probably not a valid one: cohort studies of heroin sers have suggested that even within that time period, a mall but significant proportion of heroin users will have ied or become abstinent without receiving treatment for their eroin use (Hser, Hoffman, Grella, & Anglin, 2001; Vaillant,