Monday, November 06, 2006

Australian illicit drug consumption data

Much of the illicit drug consumption data that you read about in the national press is constructed by the University of New South Wales’ group NDARC (National Drug and Alcohol Research Centre). This is the sort of data you see in newspaper headlines about methamphetamines like ‘ice’ or the number of illicit drug users in Australia.

For the most part this data is derived from NDARC’s IDRS (Illicit Drug Reporting System) or EDRS (Ecstasy and Related Drugs System). This is a very expensive data collection operation costing millions of dollars and funded by the Australian Government Department of Health and Ageing.

It is poor quality data. In my view it is not only subject to error – it is positively misleading and is very likely to be a poor guide to drug policy.

It is very difficult to determine what this data actually represents but it probably does not represent illicit drug consumption in Australia despite the name of the session I attended on Sunday, at the current APSAD Conference, titled ‘Australian Drug Trends 2006’ which is based on the most recent compilation of annual reporting by NDARC.

This session was opened by a mild-mannered critic of the NDARC data, Professor Jake Najman, from the University of Queensland. Using reference household survey data Professor Najman found that illicit drug users in Australia have a lower unemployment rate than the general community and earn more. By way of contrast NDARC in surveying Australian drug users select a sample of 914 intravenous drug users from around Australia of whom 77% are unemployed, 51% had a prior prison history and 44% were in treatment for their addiction. This sample seems to have almost nothing in common with the patterns of illicit drug consumption as they actually are in Australia.

The sample NDARC looks at is a ‘sentinel group of users’ who distinguish themselves by requiring treatment of some sort, a group of key experts (such as nightclub workers close to the drug scene) and ‘indicator data’ from other sources that represent measures of harm and use. These disparate sources of data input are collated in some unspecified way (they are ‘triangulated’ to use the jargon of NDARC).

Professor Najman recommended that NDARC get more information about the characteristics of the users it interviews and to compare them with national drug survey estimates in order to make the estimates they do obtain more representative. Otherwise the NDARC data will substantially overstate the dimensions of illicit drug problems simply because it is based at the problem end of drug use markets – those unemployed, with gaol records or significant involvement with the criminal justice system. My own view is that, because the NDARC data is so unrepresentative, this is unlikely to yield significant improvements to try to relate it to broader population data. The issue of statistical error is paramount – this will mask any advantage gained from a mapping onto such unrepresentative users

A better suggestion would be for the Department of Health and Ageing to prod NDARC to hire some statisticians who know how to derive a sample that represents what it should be representing - the population of illicit drug users in Australia.

NDARC’s response to these sorts of criticisms is to say that they know the data is unrepresentative but representativeness is not the intention. What is sought is a quick measure of newly emerging trends. But you do want representativeness – how do you know that the biased sample selected by NDARC gives any useful information at all about emerging trends.

An interesting dimension of the NDARC approach is that national estimates are seemingly a simple arithmetic average of usage data in each state without any weighting to reflect the dominance of drug users in high population states such as NSW and Victoria. It is trivial to weight the various state measures by state population but even this is not done.

If workers in a non-health field produced this type of work they would be exposed to strong criticism. My advise to drug researchers concerned with illicit drug consumption data is to use the triennial National Health Survey data which is a mile ahead of NDARC’s work in terms of quality.


Anonymous said...

Earlier this year there was a newspaper report on illicit drug use in Aus that was based on NDARC reports. Unfortunately, I can't find a link to the report but at the time I was astounded (incredulous) at the high figure cited: from memory it was about one in ten of ALL Australians. I wondered if the journalist had made a mistake or NDARC had major sampling problems. I went to NDARC's website to find out how they could have calculated such a statistic but all that I found out was NDARC provides very little (easily accessible) information on how they collect data and construct stats.

hc said...

Its collected as I say from about 900 injecting drug users at the criminal-unemployed end of the drug user spectrum. Totaloly unrepresentative and hard to see what it could be conceivably used for.

Anonymous said...

Obtaining more government funding? What’s the bet that it’s not just NDARC but a couple of government agencies that obtain their funding based on the high levels publicized by NDARC?

But seriously, how can they get away with this? It’s worth a letter to the NDARC's director with recommendations that (a) thou shall be explicit about one's methods including explaining said hocus pocus method of extrapolation of local to national and (b) thou should seriously consider altering one's methods to those that meet at least the bare minimum of scientific standards.