5.1 The nature of the problem
5.2 Methods of estimation
5.3 Estimates of the number of Australian heroin users
5.4 Estimated population prevalence of opiate dependence
5.5 Validity of the estimates
5.6 Estimating unmet demand
5.1 The nature of the problemThe number of places in methadone maintenance treatment in Australia has steadily expanded over the past 10 years  from 2,000 in 1985 to approximately 15,000 in 1994. This increase is largely a consequence of the decision taken by the Special Premiers' Conference in 1985 to increase the availability of methadone treatment, although it has also been influenced by other factors. These include fears of HIV infection driving users into treatment, and subsequent changes in treatment policies in some programs that have made methadone treatment more attractive to the client group. Such changes have included easier access, and liberalisation of clinic rules on take-away doses and the consequences of continued illicit drug use.
One of the questions most often asked about the increase in the number of methadone clients is whether it can be expected to continue indefinitely. Behind this question is the further one: how many regular and dependent heroin users are there in Australia? The answer to this question is often seen as especially relevant to assessing how much unmet demand there is for methadone maintenance (and other forms of drug treatment), the assumption being that the discrepancy between the number of heroin users who are currently in treatment and the total number of heroin addicts in the population represents the unmet demand for treatment. The latter assumption is questionable for reasons to be discussed below.
5.2 Methods of estimationThere are a number of obvious difficulties in attempting to estimate the number of regular heroin users there are in the community. Heroin use is illegal and hence difficult to study. It is a stigmatised activity that is usually practiced in private between consenting adults who prefer that others not know about their drug use. There is no universally accepted definition of "regular" or "dependent" heroin use. And there are no well tested and unbiased methods available to produce a credible estimate of the number of people who make up such "hidden populations". A variety of different methods have nonetheless been used in an attempt to estimate their numbers, all of which have their problems .
Top of page
5.2.1 Sample surveys of drug useThe most obvious approach, conducting population surveys of drug use, is not well suited to the task of estimating the number of heroin users in the population. First, household surveys are likely to under-sample heroin users whose lifestyle makes them less likely to live in conventional living arrangements and less likely to participate in household surveys either because of their unavailability at the time the interviewer calls or their reluctance to agree to be interviewed. Second, even if heroin users are sampled and agree to be interviewed, heroin use is likely to be under-reported because it is illegal. Third, the definition of a regular or dependent user is at best crude. In most surveys we only have answers to questions about the frequency of heroin use over some period of time, (e.g. a lifetime or the past year) which is clearly unsatisfactory in a chronic relapsing condition. Fourth, in most household surveys heroin use is a rarely reported event. In the National Drug Strategy surveys, for example, the proportion who have ever used heroin is rarely greater than 1% and the proportion who have used in the past year is smaller still. Consequently, the numbers of heroin users identified in national surveys with a sample size of around 3,000 is very small (e.g. 30 persons who have ever used heroin, and less than 10 who have used in the past year). The resulting estimates of their numbers in the general population are therefore imprecise.
5.2.2 Multiplier methodsA popular way to estimate the number of heroin users in the population has been to multiply the number of heroin users in some accessible population (e.g. persons in treatment for opiate dependence) by a factor (e.g. 6 or 10) that is presumed to reflect the ratio of heroin users in treatment to the numbers of heroin users in the community who are not in treatment. This approach has the advantages that it is simple and easy to understand, and it begins with a count of the number of persons who one can be reasonably confident are regular heroin users (even if only those who have experienced problems as a consequence of their use). It nonetheless has its problems.
First, multiplier methods presuppose that we already know what we need to know, namely, the number of heroin users in the population. Second, such attempts as have been made to estimate the multiplier are usually crude guesses at best and of uncertain value even in the settings in which they were originally derived; their use in new settings is even more questionable. For example, the recommendation to use multipliers of 100-200 for opiate-related deaths derives from data collected on American heroin users in New York in the early 1960s  while the multipliers of 6 to 10 for the number of persons in treatment for opiate dependence are based on data collected in London in the early 1980s . It would be unwise to assume that either of these factors has remained constant in recent times in the same locations, let alone to assume that they are applicable to other countries and cultures. We know, for example, that the death rate among heroin users in many countries has increased with the advent of HIV and other infectious diseases. Similarly, the ratio of treated to untreated heroin users can be expected to differ widely in different health care systems, even over time within the one system as treatment availability, accessibility and attractiveness change. We persist in using these multipliers in the absence of anything else.
5.2.3 Capture-recapture methodsThe most widely used method of estimating the number of heroin users in Australia has been the method of capture-recapture or indicator dilution [115, 116]. This method derives from work in population biology where it has been used to estimate the numbers of fish and other animals in wild populations. It uses at least two samples taken from the population of interest, with members of the first sample being returned to the wild after being marked. In the case of regular heroin users, the method typically involves the use of two or more sets of records (e.g. of arrests or treatment utilisation) as the analogue of samples, and individuals' names (or other unique identifiers) as the equivalent of "marking". The rationale of the method is that the ratio of the original sample size (m) to the total population (N) is the same as the ratio of the number of recaptured individuals (r) in the second sample to the number in the second sample (s). i.e. m/N = r/s.
After algebraic manipulation, the total population size can be estimated by the simple formula:
N = s*m/r.
The principal attractions of this method are that it has a clear mathematical rationale and it produces an indication of its imprecision in the form of a confidence interval around the estimate. Its major disadvantage is that it only provides valid estimates when its underlying assumptions are correct (namely, that all members of the population have an equal chance of being captured, that there are no entrants to or losses from the population in the time between the samples, and that the chances of being captured in the first sample do not influence the chances of being re-captured).
The work of Sandland [117, 118] suggests that the assumptions of the traditional capture-recapture method are frequently false in the case of heroin users, and moreover, that the consequence of their violation is that estimates of the number of heroin users in the population are seriously biased. Sandland has provided improved capture-recapture methods that reduce the seriousness of these problems but their application has been limited by the shortage of sufficiently large data sets to which they can be applied. Nonetheless, the modified capture-recapture method probably provides the best of the available methods.
Top of page
5.3 Estimates of the number of Australian heroin users
1984 to 1987Because no single method is satisfactory, the preferred approach to estimating the number of heroin users in Australia has been the use of multiple methods (of hopefully independent imperfection) that converge upon a credible range of estimates. For example, a series of estimates of the number of heroin users in the Australian population were produced in 1988 by the National Drug Abuse Data System (NDADS). It used a number of different methods to produce a range of estimates of the number of heroin users in Australia in the middle 1980s. These were derived as follows.
- A capture-recapture estimate of the number of heroin users in NSW in 1984 (10,000) was multiplied by a factor of 3 (the estimated ratio of heroin users in NSW to number in the rest of Australia) to estimate that there were 30,000 regular dependent heroin users in Australia.
- The number of opiate-related deaths in Australia in 1986 (namely, 249) was multiplied by the commonly used factors of 100 and 200 to give estimates of between 25,000 and 50,000 regular dependent Australian heroin users.
- A household survey estimate of the percentage of the population that had injected a drug in the past year (1.8%) was used to estimate that there were 172,000 persons who had injected a drug in the past year. This was known to be an overestimate because it included persons who had injected drugs other than heroin. A better estimate can be derived from the percentage of persons who reported using heroin in the previous year in the 1988 National Campaign Against Drug Abuse (NCADA) household survey (the results of which were not available at the time of the NDADS report). The latter survey provides a much lower estimate of the number of persons who had used heroin in the past year, namely 28,000.
- The number of persons in methadone treatment in Australia in 1987 (namely 5,735) was multiplied by two factors. The first was 1.5 (to estimate the number of all persons in opiate treatment) and the second two factors were 6 and 10 (Hartnoll et al's factors) to give estimates of 50,000 to 80,000 regular dependent heroin users.
- The first estimate of the number of regular users (30,000) was multiplied by 2 and 3 (Kozel et al's 1986 estimate of the ratio of irregular to regular heroin users in the U.S.A. ). These gave estimates of 60,000 to 90,000 irregular and "recreational" heroin users (in addition to the 30,000 to 50,000 regular dependent heroin users).
1988-93The NDADS  estimates have been updated as follows using data gathered between 1988 and 1993 (see Table 1).
- The capture-recapture estimate of the number of heroin users in NSW in 1988-89 derived by Kehoe  (namely, 15,000) was multiplied by a factor of 3 to give an estimate of the number of regular heroin users in Australia of 45,000.
- The number of opiate-related deaths in Australia in 1992, namely 492 (National Drug Strategy, 1994) was multiplied by factors of 100 and 200 to give estimates of 49,000 to 98,000 regular heroin users in Australia.
- The 1993 National Drug Strategy household survey estimate of the percentage of the population that had used heroin in the past year (1.6%) was multiplied by the relevant population estimate to give an estimate of 36,000 persons who had used heroin in the past year in Australia.
- The number of persons in methadone treatment in 1991 (approximately 10,000 from Ward ) was multiplied by 1.5 (to estimate the number of all persons in opiate treatment) and then by 6 and 10 (Hartnoll's factors) to give estimates of 90,000 to 150,000 dependent heroin users.
- Multiplying the first estimate of the number of regular users (45,000) by 2 and 3 (Kozel's estimate of the ratios of irregular to regular users in the U.S.A. ) gives estimates of between 90,000 and 135,000 irregular and "recreational" heroin users (in addition to the regular dependent users).
ResultsThe results of this analysis are presented in the following table.
Top of page
Table 1: Estimates of the number of Australian heroin users in 1984-1987 and 1988-1993 from various sources by different methods
|Multiple of % who used heroin in past year|
|Multiple of NSW capture-recapture estimate|
|Multiple of number in treatment|
|Multiple of number of opioid deaths|
|Median estimate regular heroin users*|
|Number of irregular heroin users*|
|Total number of heroin users*|
* to nearest 1,000 Sources:
1 National Household Survey, 1988
2 NDADS, 1988 
3 National Household Survey, 1993
4 Kehoe et al, 1992 
5 Projection of figures in Ward et al, 1992 
6 National Drug Strategy, 1994
5.4 Estimated population prevalence of opiate dependenceIt is necessary to estimate rates of opiate use to take account of changes in the size and composition of the Australian population between the middle 1980s and the early 1990s. This was done as follows:
The median estimates of the numbers of regular and irregular heroin users from Table 1 were divided in each case by estimates of the population in the age groups in which most opiate users are found (namely 15 to 44 years) for the nearest years (1986 and 1990) from data provided by the Australian Institute of Health .
The results of these calculations are presented in Table 2.
These calculations suggest that the estimated prevalence of regular heroin users (per 1,000 of population) had increased from 4.5 in 1986 to 7.2 in 1990. The estimated prevalence of irregular heroin users had increased from 9.9 to 13.8 per 1,000 population, while the estimated prevalence of all heroin users had gone from 14.4 to 21.0 per 1,000 of population. These increases are statistically significant.
Table 2: Estimate population prevalence of regular and irregular heroin use, 1986 and 1990
|Regular heroin users|
|Prevalence per 1,000|
|Irregular heroin users|
|Prevalence per 1,000|
|Total heroin users|
|Prevalence per 1,000|
|Population size 1|
|(15 to 44 years)|
5.5 Validity of the estimatesA number of caveats have to be entered about these estimates.
- First, within each time period, the range of estimates produced by different methods varies widely. In both periods the largest estimate is approximately 2 to 3 times the size of the smallest.
- Second, the imprecision of those estimates which provide some indication of their uncertainty is considerable. For example, there is a very wide 95% confidence interval around the estimates of the number of persons who have used heroin in the past year derived from household surveys. In 1988 the confidence interval ranges between zero and 56,000 while the figure for 1993 ranges between zero and 105,000. Both kinds of variability caution against taking these estimates too seriously.
- Third, the estimate based on multiplying the number of persons in methadone treatment is artefactually increased by the increasing availability of this form of treatment in the period 1985 to 1993 . It accordingly must be discounted in deciding whether the number of heroin users in Australia has increased over this period.
Second, an apparently substantial increase in heroin users also appears to conflict with the fact that the average age of persons in methadone treatment (and other forms of drug treatment) has increased in Australia over the period 1980 to 1992 [123, 124]. However, the average age of methadone clients has not increased by a year per year (as would be necessary if there had been no new recruits to heroin use). Rather the increase has been more like 6 months per year which is consistent with continuing recruitment to heroin use over the past decade or so.
5.6 Estimating unmet demandEven when the considerable uncertainties in these estimates are set aside, it is still unwise to assume that the potential demand for methadone treatment is the simple discrepancy between the estimated number of regular heroin users in the population and the number who are currently in methadone maintenance treatment.
First, not all regular heroin users are interested in treatment in general, or in methadone maintenance treatment in particular. An unknown but probably substantial minority will cease their use without any professional assistance . Second, demand for treatment (or any service) is dynamic: it will be affected by its availability, cost, and attractiveness to potential users. The increased availability of methadone treatment over the past decade has probably contributed to an increased demand. Recent policy changes may be expected to have conflicting effects on demand. Changes in the method of delivery (such as direct costs to users) may be expected to reduce demand while more liberal policies relating to continuing drug use while in treatment and take-away doses can be expected to increase demand.
Accepting these caveats, the current estimates indicate that there may be a substantial unmet demand for methadone treatment. If we ignore the estimate derived from the treatment multiplier and use the median figure in table 1 for the period 1988-1993 then approximately 30% of the estimated 50,000 regular or dependent heroin users in Australia were enrolled in methadone treatment in 1994. This compared with approximately 17% in 1987. These "guesstimates" suggest that although the proportion of regular heroin users enrolled in methadone maintenance treatment has substantially increased over the past five years or so less than half of regular heroin users have been enrolled. The fact that there has been no sign of a slackening in demand for methadone treatment suggests that demand has not been fully met even if, as is likely to be the case, a majority of heroin users are not interested in enrolling in methadone treatment.
Top of page