Impact of the economic recession and subsequent austerity on suicide and self-harm in Ireland: An interrupted time series analysis

New article by Paul Corcoran, Eve Griffin, Ella Arensman, Anthony P Fitzgerald and Ivan J Perry – here.


20 replies on “Impact of the economic recession and subsequent austerity on suicide and self-harm in Ireland: An interrupted time series analysis”

Key Messages

• Five years of economic recession and austerity in Ireland have had a significant negative impact on national rates of suicide and hospital-treated self-harm.
• There were 476 and 85 more male and female suicide deaths and 5029 and 3833 more male and female self-harm presentations to hospital, respectively, in the 5-year period 2008–12, approximately equivalent to an excess of one complete year of suicide and self-harm.
• Men aged 25–44 years were the most affected group in terms of both suicide and self-harm.
• Reliable and well-standardized data are needed on suicide, self-harm and determinants of suicidal behaviour in the population in order to guide policy on how best to mitigate the effects of economic crises on mental health and wellbeing.

The trend extrapolation on which these conclusions are based is surely suspect. How far out should we extrapolate the decline in suicide deaths after 2008? What about the upward trend in the rate before 2000? Should the break in this trend be attributed to the unsustainable boom?

The extrapolation of the 2000-2007 trend is surely a dodgy basis for estimating the number of suicides due to the recession.
If you push this trend far enough forward, Al,ost all suicides would be attributed to the recession.
What about the steep upward trend in the suicide rate before 2000? Should the reversal of this trend after 2000 be attributed to the boom?

Here are the suicide rates per 100,000 population since 2006.

2006 12.8
2007 13.2
2008 13.1
2009 13.8
2010 12.7
2011 13.3
2012 12.8
2013 11.8
2014 11.3

Any effect attributable to the recession seems to have been short-lived.

Fair point to suggest that the boom brought about reversal of the increasing male suicide trend in the late 1990s early 2000s.
Not sure those are correct suicide rates Peter. Finalised data for 2013 and 2014 not available yet and provisional figures are always underestimates.
Female suicide rate showed a shortlived impact – an increase in 2008 then rate fell back to pre-recession levels.
Male rate was decreasing in pre-recession years but recession reversed this and it has not returned to pre-recession levels.

We also thought presuming continuation of the decreasing trend might be naive so we looked at the scenario that the male suicide rate would have stopped decreasing and levelled off after 2007. Compared to this scenario there were approx 300 more suicides.

@Paul Corcoran

While it isn’t your fault, the figures have been misinterpreted by the Irish media (surprise, surprise). There have been several media reports claiming that your figures show a 57 per cent increase in the male suicide rate since Celtic Tiger 1 ended. What your report says, however, is very different, Namely, that the male suicide rate at end 2012 was 57 per cent higher than it would have been if the sharp downward trend during Celtic Tiger 1 had continued beyond 2007.

During Celtic Tiger 1 we were told repeatedly that the focus on growth was causing the suicide rate to rise (e.g., Growth: The Celtic Cancer – published by FEASTA). Now we are being told (correctly) that it actually fell sharply during Celtic Tiger 1. But, that the fall came to an abrupt end after 2007, so much so that the rate (for males) at end 2012 was 57 per cent higher than it would have been had the decline continued.

Of course, this is still a legitimate point for your report to make – if the figures support it.

These are the figures I calculate for Ireland’s suicide rate (number per 100,000) since 1998. They are very similar to Peter Stapleton’s. The figures are for the crude suicide rate. I haven’t time at the moment to calculate age-adjusted suicide rates as you have. But, unlike deaths from natural causes, which obviously are highly correlated with age, I’d be surprised if age-adjustment for suicide rates made that much difference (but, I stand corrected if you show otherwise).

based on occurrence (final figures):

1998 15.31
1999 14.06
2000 14.65
2001 15.52
2002 14.45
2003 14.67
2004 14.19
2005 14.88
2006 12.85
2007 13.19
2008 13.13
2009 13.81
2010 12.69
2011 12.98

based on registration (preliminary figures):

2012 12.85
2013 11.76
2014 11.30

My brief analysis of these figures is as follows:

(a) There was indeed a downward trend from 1998 to 2007 (as you say). That is not in dispute. Although you’ll find very few media reports from the period 1998-2007 acknowledging that the suicide rate was indeed falling.

(b) This indeed flattened out from 2007 to 2011 (as you say). The average suicide rate for the 4-year period 2008-2011 was 13.30 compared with 13.19 in 2007. Even so, this was lower than the suicide rate of 13.78 for the previous 4-year period 2004-2007.

(c) The downward trend MAY (I repeat: ‘may’) have resumed in 2012 and beyond. I agree that it is too early to be certain. I take your point that the final figures (by occurrence) are usually higher than the preliminary figures (by registration). However, I give both above for 2012 and the difference was only 0.13. If the differences between the preliminary and final figures are of a similar magnitude for 2013 and 2014, then indeed the downward trend will have resumed from 2012 on. However, as I say, we can not be certain of this until the final figures for 2013 and 2014 are published.

I summary, the suicide rate fell from 1998 to 2007 (Celtic Tiger 1), flattened out from 2007 to 2011 (recession), and shows tentative (I repeat ‘tentative’) signs of falling again from 2012 on (Celtic Tiger 2). In so far as the suicide rate is affected by economic conditions, all the figures make sense. It is to be hoped that Celtic Tiger 2 (which we are now a couple of years into) has the same effect on the suicide rate as Celtic Tiger 1 from 1998 on. The early figures are encouraging in this regard.

Part of the ’57 per cent higher’ figure you publish seems to be the result of basing the downward trend up to 2007 on quarterly figures rather than annual figures. The annual figures from 1998 to 2007 do show a downward trend (as you say), but only from 15.31 to 13.19 (or 0.24 per year). Projecting this forward to 2012 would put the 2012 suicide rate at 12.02. The actual suicide rate for 2012 was 12.98, or 8 per cent higher, which is a much smaller difference than you find from making a projection based on quarterly figures. Quarterly figures (in anything) are very volatile. I’d find it a bit suspect to base a projection for 2008-2012 on quarterly figures up to 2007. If you made projections for GDP growth based on quarterly figures, you would come up with some fairly strange results.

Also, I’d be interested to know if you used ‘figures by occurrence’ or ‘figures by registration’ for the quarterly suicide rates on which your post-2007 projection is based. If its the latter, I’d find the projection doubly suspect. As far as I know, the CSO only publish quarterly figures ‘by registration’ (but, I stand corrected if you show otherwise). They publish annual figures for both. However, it is possible you obtained quarterly figures ‘by occurrence’ privately from the CSO?

PS I made a critical error when typing the suicide rates – they should be:

based on occurrence (final figures):

2011 13.57
2012 12.98

+ 1
Very fair summary of this complicated topic.
As far as the suicide data are concerned, the difference between the annual summary data in the Quarterly Report on Vital Statitics and the data in the Annual Report on Vital Statitsics (published some two years later) seems to be mainly a reallocation of the deaths due to ‘undetermined intent’ to ‘suicide’. In 2012 the earlier data showed 507 suicides and 82 ‘undetermined intent’, total 589, but the final figures were 541 suicides and 54 ‘undetermined intent’, total 595, a discrepancy of just 1%.
It would thus seem appropriate to place credence on the available figures for 2013 and 2014, which show a significant reduction in the suicide rate.

Informed summary for sure but some issues.
First, you’re right Peter that re-allocating some deaths due to undetermined intent to suicide is the main consequence in going from provisional to finalised figures. The proportion of undetermined intent deaths reallocated to suicide varies from year to year but may be around half. Therefore the indications are strong that 2013 and 2014 are showing a decrease. With 475 suicides and 65 undetermined intent provisionally for 2013 and 459+62 in 2014 we could be looking at just over 500 suicides compared to the finalised figures of 541 in 2012 and 554 in 2011.
Just to caution that credence in the provisional figures means considering the suicide and undetermined intent deaths. Often only the provisional suicide figures are considered thereby leading to flawed conclusions.

Thanks John. You’re exactly right in the way the 57% higher should be interpreted. In response to your points:
Yes, I was provided with access to the finalised data in the CSO offices so I could extract them by quarter, sex, age group etc. At the time the finalised figures for 2012 were not available so we had to use the lower provisional data for that year.
Using quarterly data provided 20 post-recession data points versus 32 pre-recession allowing for a more fine grained look at the trend/change. It did bring more volatility but we adjusted for seasonal variation (more suicide in quarters 2 and 3) which addressed a lot of this.
I felt age-standardising was important in conjunction with using the CSO’s annual population estimates. Mainly this was to address the huge numbers of young adults, especially young men, who left the country. Young men have the highest suicide rate and one quarter of them were no longer in our population thanks to the recession and austerity. Not accounting for that would have obscured the impact on suicide. I wonder what our unemployment rate and our suicide rate would have been without this mass emigration.
I still can’t see how the annual suicide rates you show are correct. They are 10-20% off compared to my own data and multiple extracts of CSO data. This link to a CSO bulletin has an annex towards the end that you could use to compare…

@Paul Corcoran

Thanks for the reply and CSO link.

I’m fairly sure my figures are correct.

I think the difference between my figures and the figures in the CSO link is that my figures include BOTH suicide deaths (X60-X84 since 2007) AND deaths of undetermined intent (Y10-Y34 since 2007), whereas the figures in the CSO link are for suicide deaths (X60-X84 since 2007) only.

If I may use 2011 as an illustration. I give a figure for the suicide rate in 2011 of 13.57 (although I typed it wrong in my first post – corrected in my second post), whereas the figure in the CSO link that you provided is 12.1.

I checked the CSO’s Annual Vital Statistics Report for 2011 and it gives the following figures:

(a) 554 suicide deaths (X60-X84) in 2011

(b) 67 deaths of undetermined intent (Y10-Y34) in 2011

The CSO also gives Ireland’s population in 2011 as 4,574,900.

So, taking just the figure (a) and dividing by the population (in 1,000s) gives:

554 / 4,574.9 = 12.1 (i.e. the figure in the CSO link)

But, adding both figures (a) and (b) and dividing by the population (in 1,000s) gives:

(554 + 67) / 4,574.9 = 621 / 4,574.0 = 13.57 (i.e. the figure I gave)

Its the same for all the other years.

It is a matter of opinion whether any study of suicide statistics should include only suicide deaths (X60-X84 since 2007) or BOTH suicide deaths (X60-X84 since 2007) AND deaths of undetermined intent (Y10-Y34 since 2007). I tend towards the view that both suicide deaths and deaths of undetermined intent should be used, but that’s only an opinion. I think in your report you state that you did the study first using only suicide deaths, but then replicated it using BOTH suicide deaths AND deaths of undetermined intent. I think that is a good approach. What difference did you find in the results when using BOTH suicide deaths AND deaths of undetermined intent as compared with using suicide deaths only?

So, I’m standing by my analysis, in so far as it is based on BOTH suicide deaths AND deaths of undetermined intent. The one element of doubt I have about whether my analysis is correct is the effect of age-standardising the figures. I can see the merit in it and would generally be in favour of it. As I said in my first post, I simply don’t have time at present to do the age-standardisation myself. However, I’m a bit sceptical as to whether it would make much difference to the analysis. This is why:

I calculated these figures for 2011 from the CSO link you provided for suicide deaths (X60-X84 since 2007) only.

age-group 15-24:

number of suicide deaths (X60-X84): 94
population: 579,600
suicide rate (X60-X84): 16.2

age-group 25-44:

number of suicide deaths (X60-X84): 246
population: 1,448,700
suicide rate (X60-X84): 17.0

age-group 45-64:

number of suicide deaths (X60-X84): 180
population: 1,038,300
suicide rate (X60-X84): 17.3

age-group 65-84:

number of suicide deaths (X60-X84): 33
population: 473,400
suicide rate (X60-X84): 7.0

The fact that the suicide rates are very similar for age-groups 15-24, 25-44 and 45-64 make me sceptical about whether age-standardisation would make much difference. This is in contrast to deaths from natural causes where obviously the death rate in age-group 45-64 would be much much higher than in age-group 15-24. So, in the case of natural deaths, age-standardisation is essential for meaningful analysis. But, based on the similarity of the suicide rates in the different age-groups, I’m sceptical as to whether age-standardisation would have much effect on the analysis when it comes to suicide. While the proportion of the population in age-group 15-24 fell, the proportions in age-groups 25-44 and 45-64 rose. So, the one cancels out the other. However, I’m open to correction if the full set of age-standardised figures show otherwise. Do you publish the age-standardised figures anywhere?

Also interested to see in the CSO link you provided that Ireland’s suicide rate in 2010 was the 10th lowest in the EU28 and the 4th lowest of the 22 non-Mediterraenean EU countries. The 6 Mediterraenean EU countries (Greece, Cyprus, Italy, Spain, Malta, Portugal) had by far the lowest suicide rates. Of the 22 non-Mediterraenean EU countries, only the UK, Netherlands and Luxembourg had lower suicide rates than Ireland. A recent Irish Times claim that Ireland had an ‘exceptionally high suicide rate’ is simply not true. I’m not quite sure why these countries have such low suicide rates. It casts doubt on whether economic conditions are the main factor in suicide. Ironically, Greece had the lowest suicide rate in the EU28 in 2010, much lower than Germany. I suspect it is partly due to climate and partly due to these countries being less secular and holding on more to traditional religious values (4 of them are Catholic and 2 of them are Orthodox).

JTO is right to point out that the media have studiously conveyed a misleading impression that the Irish suicide rate is among the highest in Europe.

It is also clear from these data that assertions in the Irish media to an ‘epidemic’ of suicides following the recession are misplaced.

In view of the Greek record on falsifying their fiscal data, it is hard to know what credence should be placed in their suicide data.

I would also doubt the accuracy of the Greek suicide data which is more or less a quote made by a Greek person who worked in suicide research there.
The media do tend to chose the exaggerated ‘epidemic’ and other such terms and forget that our suicide rate is average or below average in Europe although considering young adults then we have one of the highest suicide rates in EU28.
Thanks for clarifying that you were adding in undetermined intent deaths John. I know that is the default approach in the UK whereas I prefer to put primary focus of ‘official’ suicide data supplemented by analysis of suicide + undetermined intent deaths. When we did this is assessing the impact of the recession the number of excess suicides was nearly halved which I think is consistent with your summary earlier.
In part this was to do with the pre-recession trend when official male suicide rates were falling while the male rate of undetermined death was increasing. Taken together there was still evidence of a decrease but it was less pronounced.

On the age-standardising I think if you focus on all suicides in the age range 15-64 years then it makes little once the suicide rate is more or less contact across the age range and this remains so over time (see below for more on that).

But what about when you include over 65s, a growing pop with lower rates. Before the recession we had 660,000 15-24s and 471,000 over 65s, by 2012 we had almost exactly 550,000 of each group.

Adding in the under 15s another growing pop with very low rates is another complication though I recommend excluding them from most analysis of suicide.

With more time I’ll update my figures and do a proper comparison of crude versus standardised but I’d rather do it separately for men and women.

I always look at male and female suicide rates separately. There’s a 4-5 fold difference in their rates, male suicide is highest in young men, female suicide is highest in middle-aged women, men are more affected by economic change both in terms of emigration and suicide.

Below are the male suicide rates per 100,000 by age group:

Suicide rate 2007 2008 2009 2010 2011
15 – 24 years 25 21 22 25 27
25 – 44 years 22 24 29 24 28
45 – 64 years 19 23 27 26 29
65years+over 16 12 12 7 13

The pop in each age group changed over these years, declining for the two younger groups and increasing for the older two groups increase. The overall pop aged 15 plus hardly changed at all.

Pop (thousands) 2007 2008 2009 2010 2011
15 – 24 years 336 324 308 292 280
25 – 44 years 727 726 719 718 705
45 – 64 years 492 503 511 519 525
65 years + over 216 225 233 242 252
15 years+over 1771 1778 1772 1771 1761

Age-standardising deals with age-specific suicide rates showing differing changes over time and with differing trends in the pop size of the age groups whereas crude rates do not.

Sorry, 2nd line should have read…it makes little difference once the suicide rate is more or less constant across the age range.

Paul Corcoran’s It letter claims “thousands more self harm presentations to hospitals”since the recession. In fact data from the 2013 Report of the National Registry of Deliberate Self Harm in Ireland documment a 6% increase in such presentations to emergency departments (ED) with declines year on year from 2011 t0 2013.The meaningfullness of short term variations in such data where the factors determining whether a self harmer arrives at an ED are often fortuitous and “noso-comial” and known to represent only a minority of self harmiing events must surely be questionable.

Incidentally hospital admissions for depression, a most serious risk factor for suicide and serious self harm, fell by almost a quarter during the recession years. Finally let us not forget in all this debate that, particularly when myriad confounders remain unaccounted, nferring causation from association is especially hazardous.

The “thousands more self harm presentations” in 2008-2012 comes from an indepth analysis published in the Intl Jl of Epidemiology

The 2013 Registry reports stated that following two successive decreases from 2011 to 2013, the rate was still 6% higher than in 2007, i.e. the rate had not yet returned to its pre-recession level.

The rate of self harm to EDs was decreasing for men and women pre-recession. Both rates were affected by the recession. If the pre-recession trends continued through 2008-2012 there would have been approx 9000 fewer self harm presentations.

Even compared to a level trend in 2008-2012, there were approx 6500 excess self harm presentations.

The variations come from analysis of about 100,000 self harm presentations over 9 yrs, i.e. almost 1000 per month for 108 months, a scale and time period allowing for meaningful conclusions.

I agree that most self harm events do not come to hospital EDs though we don’t have a good estimate of the proportion.

I believe the proportion of self harm coming to EDs in 2008-2012 most likely fell. Consider for example, affordability of the costs, the longer waiting times in EDs, the closure of some hospital EDs and reduced opening hours for others. Despite all this we saw an excess of thousands of self harm presentations. The true increase in self harm may therefore have been far greater.

Why would hospital admission for depression have fallen? I think it is worth thinking of how a person comes to be admitted for depression. You know this far better than I but unlike hospital EDs they cannot just present themselves but will generally present to their GP (possibly a number of times) and then to the mental health services.

We must consider the burden of costs for those without medical cards presenting to the GPs and also the effects of cutbacks on the capacity of the mental health services meaning people waited longer for appointments to be assessed for possible admission into a service with reduced capacity to care for them. It is possible that the reduction in admissions reflected the reduced capacity of the mental health services.

Ideally we would have data on GP visits for mental health problems, referrals to and assessments by the mental health services, and subsequent referrals for hospital admission. Then we could tease out the situation.

Rightly confounders cause us to avoid inferring causation from association but in this case what are the confounders. Rates of suicide (male) and male and female self harm presentations to hospital EDs were decreasing and then in 2008 there was an increase and a change in these trends for the succeeding few years. Are there any plausible confounding factors that could explain this? I don’t think so.

The fondamental issue in in all this is that had trends 2000-2007 continued post 2007 suicides would have been fewer than they actually were. To attribute this trend reversal to the recession per se seems to me to be questioable given trend changes over longer time sequences pre 2000.

Isn’t it strange that by 2013 and 2014 the suicide rate had fallen below its 2007 level, despite that fact that the recession had not gone away – the unemployment rate was still over 12% and the full rigor of fiscal austerity was still in force?

Comments are closed.