NDARC Technical Report No. 197


Problems related to heroin use, such as dependence, blood borne virus transmission, premature death from overdose and crime, negatively affect the community in ways that are disproportionate to the relatively small proportion of Australian adults who are dependent on heroin. An important question to ask (and answer) is therefore: how many people use heroin? Such an exercise was undertaken in 2000, when Hall and colleagues estimated the population prevalence of opioid dependence using 1997/1998 data.

In recent years, however, we have seen significant changes in the heroin market in Australia. Early in 2001, anecdotal reports were received of a sharp reduction in the availability of heroin. These reports were initially confirmed by research in convenience samples (Day et al., 2003; Weatherburn, Jones, Freeman, & Makkai, 2001), and later confirmed in the monitoring systems that had documented the increasing heroin availability during the previous five years (Darke, Topp, Kaye, & Hall, 2002; Topp, Kaye, Bruno, Longo, Williams, O'Reilly, Fry, Rose, Williams et al., 2002). A common issue for consideration has been the extent to which changes in the availability of heroin may have impacted upon the number of persons using the drug. This report seeks to examine which methods for estimating the number of heroin users are most appropriate to use when the numbers of heroin users may be rapidly changing in response to marked changes in availability of their drug of choice.

Achieving this aim is not simple. Making estimates of a hidden population, such as heroin users, is difficult for a range of reasons in the best of circumstances. Added to these difficulties are complications introduced by marked changes in drug supply.

Given the illegal and stigmatised nature of heroin use, it is not a simple task to estimate the number of dependent heroin users in Australia. There are no widely accepted "gold standard" methods for estimating the size of this "hidden population". The favoured strategy is to apply a variety of different estimation methods of varying validity to different data sources, looking for convergence in the estimates. We carried out an evaluation of data sources used to produce estimates of the number of heroin users, and of methods that might be able to measure the size of this population during relatively short periods (i.e. annually).

The aims of the current study were to examine methods with which to estimate the number of regular heroin users in NSW (for discussion of Australian estimates, please see Degenhardt et al., 2004). In particular, we wished to consider:

  1. Are some methods less sensitive than others to changes in the supply and use of heroin (and potentially the number of regular heroin users)?;
  2. Are there problems with the use of some estimation methods in the context of rapid suspected changes in the size of a hidden population?
  3. Do different sources of data provide information about different subsets of the group of regular heroin users?

The estimates of the number of the number of regular heroin users are presented in detail in Degenhardt et al (2004). In considering these questions, we have examined the estimates of NSW regular heroin users over the period 1997-2002 derived from different sources and methods.

Estimates were generated from secondary analyses of existing databases. We made estimates of the number of current regular heroin users. The number of opioid dependent persons in total is likely to include the number of persons maintained upon opioid pharmacotherapy as well as our estimates of the number of regular heroin users. The data sources that were used in the current study included:

  1. national data on the number of opioid induced deaths per year compiled by the Australian Bureau of Statistics;
  2. data on ambulance attendances at suspected drug overdoses provided by the NSW ambulance service;
  3. the NSW Health Department’s heroin pharmacotherapy client database provided by the Pharmaceutical Services Branch (PSB);
  4. data on arrests for drug offences provided by the NSW Police Service; and
  5. data provided by the Alcohol and Drug Information Service (ADIS) on calls received related to heroin use.

Three indirect estimation methods were used to make annual estimates of the number of active, regular heroin users with the following estimation methods:

  1. Multiplier methods using data on the number of opioid overdose fatalities, entrants to opioid pharmacotherapy, arrests and ambulance callouts to suspected opioid overdoses;
  2. Capture-recapture estimates based on methadone and police arrest data; and
  3. Back-projection methods using data on fatal opioid overdose and first time entrants to methadone maintenance to estimate the incidence and prevalence of heroin dependence.

Clear increases in the scale of harms related to heroin use were documented in the latter half of the 1990s, with sharp reductions from 2001. These changes were relatively consistent across the data sources examined, with the exception of the total number of persons in opioid pharmacotherapy, which has steadily increased over time as new heroin dependent persons have entered treatment, and others have remained stabilised on it.

Capture recapture methods tended to “average” changes across time in the indicators of current heroin use, and the resulting estimates did not provide estimates sensitive to the changes observed in indicators of use. Back projection methods, although more sensitive to changes in the indicators used, were limited in their ability to accurately model changes that occurred due to their recency . Back projection methods will probably be able to better model recent trends after a longer period of time has elapsed. In contrast, multiplier methods were able to capture these short term changes. As a result, the range of multipliers developed for use in this study formed the basis of the estimates derived.

There has been a significant drop in the estimated number of current, regular heroin users in NSW (Table A). This drop was sustained in 2002. Detailed estimates of the number in NSW and in Australia, as well as stratification by age and gender, are provided in the companion report (Degenhardt et al., 2004).
These estimates do not necessarily imply a similar reduction in the number of opioid dependent people. There has been an increasing number of persons in pharmacotherapy in NSW; in 2002, almost 15 000 persons were in pharmacotherapy for opioid dependence (Table A).

In periods where there are dramatic and sudden changes to the extent of drug supply in the community, making estimates of potential changes in the population of users is a difficult process. Methods that use data from a long window period (such as capture recapture), or which require data on longer term trends in drug use and its progression (such as back projection), may be less able to provide estimates that correspond to known changes in trends in existing indicator data. Multiplier methods, because of their relatively straightforward use of these indicator data, may be better suited to indirect estimates of the population in these periods. It will be of use to repeat this estimation exercise in two to three years.

Citation: Degenhardt, L., Rendle, V., Hall, W., Gilmour, S., and Law, M. (2004) Estimating the size of a heroin using population after a marked reduction in heroin supply, Sydney: National Drug and Alcohol Research Centre.



L. Degenhardt, V. Rendle, W. Hall, S. Gilmour, M. Law
Date Commenced
10 Sep 2004
Resource Type
Technical Reports