The Government’s drug testing programme in prisons is distorted by perverse incentives, and fails to make the best use of the data collected.

I have made no secret of my criticisms of random mandatory drugs testing in prisons, first introduced in 1995 by Michael Howard. It underestimates the use of heroin in prisons, provides an incentive for prisoners to switch from cannabis to heroin to avoid detection, and uses a crude measure (the percentage positive for any illegal drug) as a performance indicator for prisons.
 
In December 2008, the Home Office advisory committee which I chair (the Surveys, Design and Statistics Subcommittee, SDSCC) published a report which made recommendations for the better exploitation of existing sets of data.
 
Data made available by David Hanson of the Ministry of Justice in August 2008 covered the results of random mandatory drug testing (rMDT) for prisons in England and Wales up to 2006-07, and categorised by the day of collection. This allowed positive rates for specific drugs to be compared, for the first time, during/after the weekend against other weekdays. Prisons perform 14 per cent of rMDTs during the weekend.
 
Analysis of these data for the SDSCC raised new questions, including the low rate of prescribed methadone for prisoners. But subsequent discussions with officials from the Ministry of Justice produced neither answers nor data, and demonstrated an apparent reluctance to use the random drugs testing results in novel ways: either to monitor the roll-out in prisons of the Integrated Drug Treatment System, or to assess how effectively prisons use the other form of mandatory testing available to them – targeted mandatory drug testing (tMDT). This form of testing is not random, but is intelligence-driven.
 
A letter to Jack Straw, the Secretary of State for Justice, in August this year, was equally unrewarding. But a series of questions in parliament by Andrew Pelling MP has proved far more productive.

My analysis of these answers will be published on this website in a series of articles this week. My conclusions are:

• Opiate use in prisons is underestimated, and may be rising
• Smaller prisons do fewer targeted tests, relative to their size, but are more adept at selecting whom to test.
• Methadone substitution for the treatment of heroin addicts is rolling out slower than it ought to be. 
• Uncorrected for the extent of prescribed methadone, randomised testing gives the false impression that opiate use in prisons is falling, when it may actually be rising
• Unless rMDTs are properly analysed to monitor the availability of prescribed methadone in prisons, it ought to be abandoned on cost as well as human rights grounds.
• The performance indicators set for prisons are inept because they do not distinguish between cannabis and opiate users; and Ministers seldom make this important distinction when answering parliamentary questions on their prison drug strategy. 

Parliamentary answers to Andrew Pelling MP and others provide a means of addressing two questions. First, how well has the drug treatment system in prisons managed the roll-out of methadone substitution for heroin-addicted prisoners?
 
Second, how effective is the intelligence used by prison officers in targeting mandatory tests at prisoners they have reason to believe may be using drugs? This question can be answered by examining how much higher is the percentage of positive tests in tMDT than in rMDT. If the right prisoners are being targeted, a greater proportion of tests should be positive than when prisoners are selected for testing at random.
 
How might we expect these questions to be answered? Scientists need to set results in the context of plausible prior expectations.
 
First, the methadone question. In Scotland, random testing has been abandoned, but by the time it was, about one in six Scottish prisoners was receiving methadone. That is consistent with the fact that in England and Wales in the late 1990s, 29 per cent of adult inmates were estimated to have a history of injection drug misuse, and, when asked, about half of class A drug users say they want help. And so, we can reckon that prison inmates’ need for methadone substitution will be at least 15 per cent.
 
Second, the quality of prisons’ intelligence systems. Random testing is much more likely to detect cannabis than heroin, because cannabis stays in the urine for up to two weeks, while heroin does so for only two to three days. The result is that random tests underestimate heroin use by a factor of 3.5.
And so, targeted testing should produce a much higher number of positive tests for opiates than for cannabis, if the right prisoners are being targeted and tested in a timely way. The ratio, based simply on the timing question, would be around 3.5. Any higher ratio than that implies that intelligence-led targeting is actually working: lower ratios imply it is not. This question can therefore tell us something useful about how well prison staff know what is going on in their prisons.
 
Tomorrow I shall analyse what the data tells us about the roll-out of the drug treatment programme, and the availability of methadone. On Wednesday I shall look at what they tell us about intelligence within prisons.
 
(Sheila Bird chairs the Surveys, Design and Statistics Subcommittee of the Home Office’s independent Scientific Advisory Committee.)