 |
Gambling with lies |
Loyal readers may recall that Public Health England misused a study of Swedish hospital patients to make the claim that 409 suicides a year are linked to problem gambling. The science was so shoddy that the claim was shelved by its successor, the Office for Health Improvement and Disparities, who then used the same study and made the same mistakes to claim that "up to 496" suicides are linked to gambling. Although government agencies have been careful to use terms like "linked to" and "associated with", campaigners and journalists have been less careful with their language.

But, as I mentioned last year, one of the authors of the Swedish study used the same dataset for her PhD thesis and concluded that gambling disorder was not an independent risk factor for suicide among the hospital patients. It turns out that people in hospital with a range of psychiatric problems suffer from a lot of issues associated with premature mortality and that you can't use an inherently high-risk group to extrapolate across the entire population of a different country. Who knew?
Along with the co-author of the original study, she has now published
a new study which comes to the same conclusion: people with gambling disorder
are more likely to commit suicide but this is because of various co-morbidities, not gambling disorder.
Individuals with gambling disorder had an increase in levels of
mortality and suicide mortality compared to age, gender and
municipality-matched controls. However, gambling disorder itself was not
at the 0.05 alpha-level statistically associated with neither suicide
nor general mortality when controlling for somatic and psychiatric
comorbidities, gender, age and socioeconomic status. Thus individuals
with gambling disorder suffer from increased mortality and suicide
mortality and reasons for these appear to be multifactorial motivating
careful suicide risk assessment and screening for somatic comorbidities
in individuals with gambling disorder.
Or, to put it another way...
In the regression model gambling disorder was not significantly
associated with mortality, this was predicted by socioeconomic status,
increasing age, low education level, somatic comorbidity, substance use
disorder and previous intentional self-harm in men and for women by
increasing age and somatic comorbidity.
Of course, this is only one study and one cohort of people. But it is the same cohort of people that PHE and OHID used to come up with their spurious statistics (and those spurious statistics were then used to come up with equally useless claims about the cost of gambling to the health service).
If you extrapolate the findings from this study, you get the result that no suicides are causally linked to problem gambling in the UK. Stick that on a t-shirt.