Monday 27 August 2012

Moral panic ‘du jour’: drugged driving [updated]

Today’s “moral panic” is drugged driving, on the back of stories like this one in the Tabloid Herald titled “Tests reveal most crash drivers had taken drugs,” to which there is a whole lot less than meets the eye.

First, the sample size is just 453--being the number of crashed drivers for whom the Ministry of Transport had blood samples. One presumes, but is not told, that a blood sample is generally only taken if the police on the scene consider it possible the drivers involved were under the influence of something more than just euphoria. Meaning many crashed drivers (i.e., some number not determined by the study, or the journalists who (mis)reported it) had not taken anything more stimulating before crashing their vehicle than their evening or morning meal.

So the headline might more accurately read:

“Tests reveal most crash drivers from whom police collected and saved blood samples had taken drugs.”

But that’s not the only problem here. The Tabloid headline screams DRUGS!!! (cue: moral panic) but buried in the story itself is news that:

Drugs were detected in the systems of 258 drivers… Of that group, 156 were found to be on drugs not administered by a medical professional.

Meaning nearly one-quarter of those who were supposed to be terrifyingly out of control were driving not having taken DRUGS!!! (cue: moral panic), but having taken the sort of drugs prescribed by your local family doctor. Maybe they’re just bad drivers on aspirin?  Who knows, the details are completely unreported.

So the connotations so carefully suggested by the Tabloid’s headline don’t fly. It might have been more accurate, for “most” readers who rely on reading headlines for their news, to report,

“Tests reveal some crash drivers from whom police collected and saved blood samples had taken drugs.”

But that’s not exactly news.

But there are further problems for those who do read further on. We are also told by the Tabloid that “drivers with more than the legal limit of alcohol in their system made up just over half of the 453 samples analysed.” With “more than half” meaning “most” to most people, you might have thought (as Stats Chat noted) that the headline might have involved alcohol rather than drugs.

So the headline might even more accurately read:

“Tests reveal most crash drivers from whom police collected and saved blood samples were drunk.”

But since there’s already a continuing and grossly out-of-control moral panic about that particular problem, the story would neither have made page one nor been re-(mis)reported with all the associated moral panic on talkback, blogs and Twitter. Which would not have given the alleged journalists the bonus they were obviously seeking in writing their story.

Yet even with this headline we have a problem. Because we’re also told “fifty-three per cent of the alcohol group had drugs in their system” -- i.e., 53% of “more than half” of the 453 whose blood featured in the study. So that’s perhaps a full third of that 453, with (we’re also told) “90 crashes caused by people with both alcohol and cannabis.”

So which had the most causal influence? The alcohol, the cannabis, or just being shite drivers?

Because the presence of a drug or some alcohol in a person's blood does not mean that they were impaired by that drug or alcohol. Correlation is not causation. To draw the conclusion desired by the authors of this latest moral panic, the study would seek to need to show the causal link between the crashes and impairment caused by the recreational pharmaceutical or the alcohol. Which hasn’t even been attempted here.

So perhaps our headline should just say:

“Crashes caused by lots of things.”

But that’s never going to sell a newspaper, is it.

Still, the more you look at it, this study and the story about it looks increasingly less worthy of being re-reported, and more use to statistics and journalism students as an example of what not to do.

To be fair, the politicians, talkback hosts and Tabloid journalists are not the only ones playing fast and loose with the story’s too few relevant figures.  David Farrar, that well-known self-described Stats Guru, weighs in as well, opining that “the data above indicates a much much higher presence of drugs in drivers who have crashed than in the normal population.” But of course it does nothing of the sort—even if the sample size itself was truly representative of drivers causing crashes, which is presumed by the Tabloid staff but not shown, there is no comparison whatsoever performed on the “normal population” with which to compare this selective statistic.

And not could there be.  Because few if any survey respondents in the “normal population” are going to accurately report to a person carrying a clipboard the presence of drugs in their system.

It’s all just so much total bollocks.

UPDATE: 8-page report on the “research” here.  The study targeted drug use by drivers from three groups.

1. Drivers hospitalised following a crash (a section of the driving population not previously studied).
2. Drivers who were found to be impaired by the Police Compulsory Impairment Test and had used methadone.
3. Drivers whose blood samples were sent to ESR under the drug-driving legislation but no drugs were detected in their blood.

The sample of 453 drivers reported in The Tabloid was group 1 drivers only: i.e., drivers hospitalised following a crash who were “deemed” (not necessarily by the courts, but by the MoT) to be at fault for the crash. Blood samples were not received from all hospitalised drivers, or even from all drivers, but only those deemed responsible from whom a blood sample had been taken to test for alcohol…

The number of “drug drivers” reported by The Tabloid has been further enhanced by the alleged journalists responsible for writing the story failing to notice that a very large large proportion of the drugs detected (around 210)  were “very likely” to have been administered in the course of hospitalisation.

'Which rather makes a mockery of the whole moral panic, doesn’t it.

And as the report itself notes,

The presence of a drug in a person’s blood does not mean that they were impaired by that drug.

2 comments:

Gavin said...

"Because the presence of a drug or some alcohol in a person's blood does not mean that they were impaired by that drug or alcohol. Correlation is not causation. To draw the conclusion desired by the authors of this latest moral panic, the study would seek to need to show the causal link between the crashes and impairment caused by the recreational pharmaceutical or the alcohol. Which hasn’t even been attempted."

Yeah it has. There are a ton of studies showing the causal link between drug use and impairment. You're just too busy working up an ideological boner to bother looking them up.

James Gray said...

No, Gavin, there is a causal link between *some* drug use, and impairment.

Caffiene, nicotine, and paracetamol spring to mind as commonly used drugs not known to cause impairment.

Are you suggesting we start arresting those who drink coffee on the way to work?