The story so far

I recently posted an article describing the way in which the Office for National Statistics (ONS) has somehow managed to overlook the majority of the health risk posed by cold weather. In it, I drew attention not only to a shift in the editorial stance taken in ONS bulletins, but also a fundamental change in the way the ONS chooses to frame risk when cold weather is involved. Both changes led me to suspect that the ONS has either deliberately or unwittingly moved to a position that promotes the incorrect idea that the temperature-related health risk is already on the increase.

However, there were legitimate reasons for the ONS to have revised its methodology and, for that reason, I have thus far been reluctant to brand what the ONS has done as either a cover up or flagrant attempt to deceive. Specifically, the new ONS methodology (in the guise of the Standards for Official Statistics on Climate-Health Interactions, or SOSCHI) at least has the benefit of removing deaths that are not strictly due to cold weather. For example, even though cold weather will always be implicated, it is nevertheless simplistic to say that all excess deaths due to the flu virus are caused by cold (take, for example, the confounding impact of an ineffective vaccine). I may not approve of the crude and flawed solution found for this problem (basically, ignore the death unless icicles are involved), and I certainly disapprove that the ONS fails to acknowledge in its bulletins the obvious switch from tracking the full climate-related public health risk towards tracking extreme weather health risk only; nevertheless, I have remained sympathetic to the technical difficulties the ONS faces.

The plot thickens

However, such sympathies have been sorely tested after discovering the amendments and enhancements that have been made to the SOSCHI framework as a consequence of the UN Statistics Division, 57^th Session’s adoption. The whole point of SOSCHI was for it to be applicable globally, so there can be no surprise that the session addressed temperature-related deaths that are particular to tropical environments. However, the specific adjustments made are very telling in their detail.

Firstly, in respect of Indicator 44 (Incidence of cases of climate-related diseases), it is telling that the UN and the ONS were in agreement that allowance should be made for:

• Incidence of diarrheal disease cases attributable to (a) heat and (b) rainfall
• Incidence of malaria cases attributable to (a) heat and (b) rainfall
• Incidence of other vector-borne diseases of national importance attributable to (a) heat and (b) rainfall

This eagerness to allow for tropical, vector-borne and waterborne diseases contrasts markedly with the ONS’s efforts to exclude cold-related virus deaths, such as those caused by an influenza outbreak. When dealing with the cold, the ONS was happy to strip away the confounding virus-mediated mortality in order to obtain what it thinks is a more reliable climate signal, even though this will downplay the health burden due to cold. But when it comes to heat-related death, the disease-mediated mortality is not treated as a potentially confounding component but instead as a core indicator of climate risk, to be added without reservation.

One can argue that one mediation confounds more than the other, but the fact remains that this uneven handling is bound to skew the assessment of the health burden. Not that the ONS is all that interested in accurately capturing that burden. SOSCHI is designed to be a simple-to-use system that captures the extreme weather impact caused by climate change, and the deliberate and calculating nature of the asymmetric treatment of disease-mediated mortalities suggests that the resulting bias towards the heat-related health burden is as intended. It is treated as an acceptable side-effect of seeking a supposedly climate-only health statistic.

If this were all that the ONS and UN had done, I might still give them the benefit of the doubt, but the asymmetry doesn’t end there.

In respect of Indicator 45 (Incidence of heat- and cold-related illnesses or excess mortality), it is telling that ‘Mortality from suicide attributable to excess heat (proxy for mental health)’ should be included whilst no mention is made of suicide attributable to excess or (more to the point) chronic cold. Are they seriously suggesting that climate-influenced mental health only becomes a problem during excess heat? Can they not see that the privations endured during either extreme or prolonged periods of cold can be equally challenging to mental health? It’s not just about getting hot and bothered!

Furthermore, in respect of Indicator 45, they say that ‘incidence of occupationally-related health outcomes of excess heat’ should be included. No mention is made of the outcomes of excess cold. Are they saying that the occupationally-related health outcomes of cold can be ignored, or do they actually believe that no such outcomes could possibly exist?

The bottom line is that every adjustment or enhancement agreed with the ONS during the UN Statistics Division’s 57^th Session just so happens to have the effect of exclusively boosting heat-related death counts, no matter how indirect the deaths may be. Meanwhile, all previous ONS efforts had been to exclude cold-related deaths on the grounds that they may be spurious. There comes a point when the odour of rodent cannot be ignored.

Finally, in respect of Indicator 46 (Climate-induced air pollution) the UN and the ONS were in agreement that allowance should be made for:

• Mortality attributable to short-term effects of outdoor air quality (PM2.5)
• Mortality attributable to effects of wildfire smoke (PM2.5)
• Mortality attributable to long-term effects of outdoor air quality

Of the above, the ONS claims to have the ability to accurately model the first two, but I have my doubts. The uncertainties involved in modelling the attribution of such mortality to climate change are not inconsiderable. I may be wrong, since epidemiologists can be very clever, but I see the challenge of accurate attribution as being very high, and my baseline level of trust is very low. Take, for example, wildfire smoke, for which the existing studies attempting to quantify the climate signal in mortality statistics require so many layered levels of modelling that the results have been recognised as statistically insignificant. As for the prospects of ever being able to attribute the long-term effects of outdoor air quality, I certainly don’t share the ONS’s confidence of being able to do this any time in the near future, if ever.

So where are we now?

To summarise, in terms of the SOSCHI framework’s biases, we have so far:

a) A failure to recognise that the exposure-risk curves for the cold-related and heat-related threats are completely different, resulting in a failure to capture mortality caused by relatively mild but prolonged cold spells.

b) The treatment of disease-mediated death during cold spells as a confounder requiring dismissal, whilst simultaneously treating disease-mediated death during hot spells as a core indicator demanding inclusion.

c) The inclusion of mental health risk associated with hot spells but no corresponding inclusion of the same where cold spells are concerned.

d) The inclusion of occupational health risk associated with hot spells but no corresponding inclusion of the same where cold spells are concerned.

e) The inclusion of mortalities related to air quality that rely upon highly uncertain and unreliable mathematical modelling.

The ONS and UN would, no doubt, be able to offer plenty of technical justifications for the above, but the fact remains that the net effect is to introduce an operational bias in the health burden statistics that is music to the ears of the likes of the Guardian, particularly since the ONS now obliges with press bulletins that are so friendly to the heat death narrative, emphasising as they do a closing gap rather than a still-falling net risk.

Time to put on the tinfoil hat

Scientific understanding, data and statistical methods are bound to improve over time. But when it comes to climate change, it seems odd that every historical improvement in data collection and analysis, whether it be in relation to temperature records, rising sea-level, mortality, or whatever, has resulted in a heightened perception of risk. Without exception, an adjustment has meant things turning out to be worse than we thought. This latest raft of ‘improvements’ by the ONS follow in that fine tradition and I, for one, would not blame anyone who has looked into it in any way, if they were to come away suspecting that there is a thumb on the scales. Every time it is a case of adding cold-related mortality, the response seems to be ‘Well, we should, but it’s just too difficult’. Meanwhile, when it comes to heat-related mortality, the response seems to be ‘Well, we shouldn’t really, but it’s so easy that it would be rude not to’.

Call me a conspiracy theorist, but I am beginning to think there is just too much clumsy work going on here for it to be perfectly innocent.