Austerity and Mortality in Spain: The Perils of Overcorrecting an Analytic Mistake

An article by Cabrera de León et al.,¹ which AJPH has retracted (Morabia, p. e17), claimed a marked change in the mortality trend in Spain in 2011, culminating by 2015 in half a million deaths beyond that projected on the 2000 through 2010 trend. This shocking apparent excess mortality was attributed, by comparing with US data, to austerity policies adopted in Spain after 2010.

Following review and acceptance, but before publication, it became clear that the claim involved a serious error. The age-adjusted mortality rates (AAMRs) up to and after 2010 were from different standard populations.² The authors declined an opportunity to recalculate (Morabia, p. e17), instead asserting the overall validity of their data and pointing to a similarity with the crude mortality figures.¹

On publication, although challenged as indicated,² the article was also welcomed as a contribution “to the science that aims to understand how [economic] policies influence health to provide guidance to policymakers,” and so as a contribution to “a public health of consequence.”^3(p985)

A response from Regidor et al. (p. 1043) now attends to the data error (using a consistent standard population). Spanish population AAMRs, in fact, continued to fall after 2010—although there were increases over the previous year in 2012 and 2015. However, the data show a notable slowing in the rate of decrease after 2010. In their model 1, adjusting only for age, Regidor et al. identify the time intervals 2001 through 2007 (before the economic crisis), 2008 through 2010 (during the economic crisis but before austerity), 2011 through 2013 (main austerity period) and 2014 through 2016 (after main austerity) and calculate annual percentage changes (APCs) in AAMRs for these intervals of −2.1, −3.4, −1.8, and −0.3, respectively.

These findings, it should be noted, do not directly refute what is perhaps the most fundamental claim of the retracted article: that mortality in Spain after 2010 was higher than it would have been on the basis of the previous trend. Regidor et al. have not recalculated the figure on the basis of their data, although that would have been useful.

THE ROLE OF AUSTERITY POLICIES

Slowing mortality improvement after 2010 is, of course, not unique to Spain, nor is its attribution to policies of austerity. Other explanations have been proffered, discounting or minimizing the effects of austerity. There is a “highly ideological component”⁴ in the debate that doubtless is unavoidable but that may lead some, consciously or otherwise, to pursue a “public health of consequence” in ways AJPH would not always necessarily intend. In this light, assumptions need to be examined. Fitting linear trends to time intervals (perhaps particularly shorter intervals) and the assumptions involved in projecting a trend beyond its time interval are not unproblematic. Mortality rates cannot decrease in a linear fashion forever, and abrupt changes from one trend to another are generally more likely to reflect the modeling than anything else. All these considerations likely have some bearing on the apparent suspension of disbelief of various parties, allowing the publication of the now-retracted article. How likely was it that austerity could really have contributed to quite so many deaths in such a short time in a population of only 46 million?

Regidor et al. certainly reengage disbelief, regarding not just the mortality trend but also how to account for it. Austerity is now decentered. In their model 2, Regidor et al. further adjust the data for the “probable confounding effect” of the incidence of influenza and heat waves on the AAMR trend for the Spanish population after 2010. The resulting APCs for their time intervals (as for model 1) are −2.0, −2.9, −2.2, and −0.6, respectively—a reduction in the improvement reported for model 1 between 2001 and 2010 and an increase in the improvement reported after 2010.

Taking model 2 at face value, the focus of slowing mortality improvement becomes 2014 through 2016—rather than the previous focus also on 2011 through 2013 (on the basis of model 1). Regidor et al. concluded that influenza activity and the 2015 heatwave “prevent identifying a possible delayed effect of austerity policies in the slowing down of mortality decline in the period 2014 to 2016.” However, here the logic seems awry: model 2, showing a marked slowing of improvement for 2014 through 2016, claims to already adjust for increased influenza and the heat wave. Is it then supposition about some other factors that would prevent attribution of the slowdown in improvement seen in model 2 to austerity?

INTERNATIONAL CONSISTENCY

Apparently so; Regidor et al. belatedly present data for eight other countries, including those “with important austerity policies, such as Greece and Croatia” and those “where austerity policies were less drastic, like Germany.” They report a slowdown in the decline in mortality for 2014 through 2016 across all of them. The authors say, “This may explain why adjusting for the presence of increased influenza activity . . . and heat waves . . . did not completely attenuate the increased mortality . . . in the 2014 through 2016 trend in Spain.” But readers may ask what purpose the further adjustment for influenza and heat waves is now actually serving, considering, it seems, that there is a comparable slowing of mortality decline across countries, regardless of austerity or, indeed, of influenza and heat waves.

Ultimately, however, little of this is convincing—neither the further adjustment for influenza and heat waves (model 2) nor the belatedly introduced international comparison. Model 2 involves a simple assumption of “probable confounding.” But how probable? Why would deaths caused by flu and the heat wave, or at least a fair proportion of them, not be seen to be linked through credible causal pathways to the impacts of austerity? No case is made, and there is no reflection in the section on strengths and weaknesses. Even accepting the assumption, assigning the presence of influenza or deaths from excess heat only to certain years in model 2 seems at best problematic. More generally, it seems implausible that the (unspecified) numbers of deaths from these causes would be sufficient to account for the slowing of mortality improvement in question.

The international comparison focuses on Germany, which has experienced slowing mortality improvement notwithstanding its “less drastic” turn to austerity. But Germany has had a quite distinctive political economy in recent decades, involving substantial wage repression, which might well render it vulnerable to slowing mortality improvement in the context of economic crisis and more limited austerity than seen elsewhere. Rajmil and Fernandez de Sanmamed’s recent AJPH article⁵ took 15 European countries categorized as low, intermediate, and high austerity (2009–2013). Overall, the authors⁵ found greater austerity to have a negative effect on mortality. Most countries saw a decline in mortality between 2011 and 2014. However in 2015, rates increased in both the medium- and high-austerity countries and continued to decline or remain stable in the low-austerity countries—with the exception of Germany, which saw an increase. Germany does indeed seem to be distinctive among low-austerity countries.

DECENTERING AUSTERITY?

The general advocacy of “a public health of consequence,” while of undoubted value, perhaps also carries an element of risk. Regidor et al. helpfully correct erroneous data. They highlight that negative economic growth need not have adverse mortality effects—vital, considering the issue of ecological sustainability. But the consequence the authors seek to see follow from their data, that of decentering austerity as an explanation for slowing mortality decline, seems not well supported by the further adjustments (for influenza and heat waves) and international comparisons introduced.

We should be careful not to meet one fatal suspension of disbelief (in the face of erroneous data) with another one (perhaps stemming from overmodeling data). Policymakers should be helped to understand, on the basis of what we know to be true, that austerity is not only unnecessary, frequently cruel, and not justified by sound economic reason (it is ultimately a form of class war, waged from above) but also has real and significant adverse public health consequences.