Changes in mortality inequalities over two decades: register based study of European countriesBMJ 2016; 353 doi: https://doi.org/10.1136/bmj.i1732 (Published 11 April 2016) Cite this as: BMJ 2016;353:i1732
All rapid responses
Response to Scanlan
We broadly agree with Scanlan’s exposition of the mathematical complexities of changes in absolute and relative inequalities in mortality. These are indeed one important reason why, as stated in our paper, “[t]here is no agreement among researchers or policy makers on what measures to use for monitoring progress towards the reduction or elimination of health inequalities”. One unresolved issue, correctly highlighted by Scanlan, is whether, and if so how, measures of inequalities in health should be made independent of the prevalence of the health problem. Scanlan has usefully proposed such a measure (described in references 3-6 of his response to our paper), although to the best of our knowledge this has not yet been formally evaluated.
In the mean-time, it is important to realize that there is a reality with real lives, real deaths, and real values attached to lives lived and lives lost, behind these mathematics. As illustrated in our paper, when average death rates go down, relative inequalities in mortality tend to go up and absolute inequalities in mortality tend to go down. We can try to remove this paradox mathematically by devising a measure that is independent of the average death rate, but does one really want to ignore the fact that deaths have become more infrequent? Some people may argue that as a result of declining death rates we have become more averse to premature death (and therefore lay emphasis on the fact that relative inequalities have increased), whereas others may argue that simply less lives are lost now than in the past (and therefore lay emphasis on the fact that absolute inequalities have declined). This illustrates that there is more to measuring health inequalities than mathematics, and that normative judgements do have a role in assessing whether health inequalities have increased or not.
Competing interests: No competing interests
Response to McCartney
McCartney’s response to our paper highlights a few uncertainties in the interpretation of our findings that we already mostly addressed in the Discussion section of our paper. We agree that we cannot be entirely certain that the trends observed in Barcelona and Turin accurately represent the trends in Spain and Italy as a whole, and we also agree that problems in the educational classification for England & Wales render the results for this country less certain that those for most other countries. However, narrowing absolute inequalities in mortality among men are also seen in, e.g., Sweden, France and Switzerland, so our over-all conclusion that “absolute inequalities have decreased in several countries” is likely to be valid.
In our study, as well as in several other related studies cited in our paper, we have found evidence suggesting that changes over time in the magnitude of inequalities in mortality are partly driven by changes in the distribution of “downstream” determinants such as smoking, excessive alcohol consumption, and medical care. We agree with McCartney that inequalities in mortality, smoking, excessive alcohol consumption and medical care ultimately depend on “upstream” or “fundamental” causes, but these do not manifest themselves clearly in our data, as illustrated by the fact that substantial reductions in mortality among the low educated have occurred in all Western European countries despite almost universally widening income inequalities.
Competing interests: No competing interests
Interpreting the apparent absolute reduction in educational inequalities in mortality in much of Europe: a response to Mackenbach et al.
The recent paper by Mackenbach et al. describing trends in educational inequalities in mortality across Europe is an important and welcome contribution to the ongoing task of reducing the unjust and avoidable mortality gradients within populations.
They conclude that absolute inequalities across nations ranked by educational attainment have declined in almost all nations for middle-aged men (except Finland) and several nations and cities for women (Scotland, England & Wales (E&W), Barcelona and Turin), but relative inequalities have almost ubiquitously increased (with the exception of French men and women in Barcelona). These findings were broadly confirmed using occupational groups as the measure of social class. We discuss some methodological and interpretative issues that arise and then comment on their implications.
First, there is at times a conflation of cities in Italy and Spain (Turin and Barcelona) with the overall countries. There is evidence that suggests that there are similarities between the city and national estimates for a single point in time, but there is also evidence from the same papers that regional and city estimates of health inequalities can be markedly different from the overall national estimate (for example, see Figure 1 in Federico et al.3). Interpreting trends in health inequalities for entire southern European countries is therefore challenging, based on the data presented.
Second, the data available in the UK for allocating people to educational groups is generally imperfect and difficult to compare meaningully across Europe. This is further compromised by the substantial changes in educational attainment evident within many of the countries included in the analyses that make confounding by age within the broad age strata used in this paper a risk given that educational attainment for different birth cohorts will confer markedly different social exposures.
Specifically, the 1991 Census data for E&W and Scotland (which are used to allocate people to different educational groups) asked only about qualifications gained after the age of 18 years, and therefore the available data only distinguish between International Standard Classification of Education (ISCED) levels 5 and 6 (although a reasonable assumption can be made that non-respondents have lower levels of education). A very small percentage of these populations reported attainment at ISCED 5 or 6 (c15% in E&W and Scotland ). This means that calculations of rate ratios (RRs) and indices of inequality are very reliant on the difference in outcomes for this very small group compared to the rest of the population. The lack of social stratification for the vast majority of the population makes estimates of inequalities across the population highly problematic and uncertain.7 Although the use of occupational group data could be helpful to cross check the findings, it is similarly limited by stratification into only two groups (and, unlike for the educational data, there is no consideration of the proportion of the populations in these groups through the calculation of indices of inequality for occupational groups in the webappendix). We are less familiar with the data sources for other included populations and therefore cannot comment on whether these issues are more widespread than the UK.
The specific problems with 1991 educational attainment data in the UK mean that it is difficult to draw substantive conclusions regarding trends from the 1990s, given that the estimates will be very sensitive to the experiences of a very small proportion of the population.7 However, it is the decline in absolute mortality inequalities in E&W and Scotland from 1991 (alongside the trends in city data which may not be representative of trends in national data as noted above) which provide much of the evidence of a widespread decline in absolute inequalities (see Figure 3 in Mackenbach et al.1), which leaves the overall conclusion that there are many nations with a decline in absolute mortality inequalities more doubtful.
Third, in relation to the interpretation of these trends (which we are much less certain about), Mackenbach et al. suggest that they can be largely explained by the changing prevalence of health behaviours (such as smoking) or clinical treatments.1 However, the importance of trends in the fundamental causes of health inequalities – income, wealth and power inequalities – and the political contexts in which these change, is completely absent from the paper, and any explanation of trends which fails to consider such forces will only ever by a partial, and potentially misleading, view.
1. Mackenbach JP, Kulhanova I, Artnik B, et al. Changes in mortality inequalities over two decades: register based study of European countries. BMJ 2016; 353: i1732.
2. Regidor E, Reques L, Belza MJ, et al. Education and mortality in Spain: a national study supports local findings. Int J Public Health 2016 61:139–145.
3. Federico B, Mackenbach JP, Eikemo TA, et al. J Epidemiol Community Health 2013; 67: 603–609.
4. Schnieder SL. The application of the ISCED‐97 to the UK’s educational qualifications. In: Schneider SL (Ed.). The International Standard Classification of Education: an Evaluation of Content and Criterion Validity for 15 European Countries. Mannheim, MZES, 2008.
5. Educational attainment statistics. Luxembourg, Eurostat, 2016 [downloaded from http://ec.europa.eu/eurostat/statistics-explained/index.php/Educational_... on 27th June 2016].
6. Moreno-Betancur M, Latourche A, Menvielle G, Kunst AE, Rey G. Relative Index of Inequality and Slope Index of Inequality: a Structured Regression Framework for Estimation. Epidemiology 2015;26: 518–527
7. Flanagan L, McCartney G. How robust is the calculation of health inequality trends by educational attainment in England and Wales using the Longitudinal Study? Public Health 2015; 129: 621-8.
8. Popham F, Boyle P. Assessing socio-economic inequalities in mortality and other health outcomes at the Scottish national level. Edinburgh, Scottish Collaboration for Public Health Research and Policy, 2010.
9. Phelan JC, Link BG, Tehranifar P. Social Conditions as Fundamental Causes of Health Inequalities Theory, Evidence, and Policy Implications. Journal of Health and Social Behavior 2010; 51(1): S28-S40.
10. Krieger N. Theories for social epidemiology in the 21st century: an ecosocial perspective. Int. J. Epidemiol. 2001; 30(4): 668-677.
11. McCartney G, Collins C, MacKenzie M. What (or Who) causes health inequalities: theories, evidence and implications? Health Policy 2013; 113: 221– 227.
Competing interests: No competing interests
Though purporting to systematically measure progress in reducing health inequalities, the article by Mackenbach et al. will only impede understanding of health inequalities issues.
So far in human history, little is said about whether health inequalities are increasing or decreasing, or the role of particular policies in such increases or decreases, has been statistically sound. For, with minor exceptions,  those studying such issues have not attempted to distinguish between (a) the extent to which observed patterns are functions of the effects of changes in the prevalence of an outcome on the measures employed and (b) the extent to which observed patterns reflect something meaningful about underlying processes.
For reasons relating to the shapes of distributions of factors associated with risks of experiencing an outcome, the rarer the outcome the greater tend to be relative differences between rates at which advantaged and disadvantaged groups experience it and the smaller tend to be relative differences between rates at which such groups avoid it. Thus, for example, as mortality declines, relative differences in mortality tend to increase while relative differences in survival tend to decrease; as rates of appropriate healthcare increase, relative differences in receipt of care tend to decrease while relative difference in non-receipt of care tend to increase.
Similarly, relative differences in adverse health and healthcare outcomes tend to be larger, while relative differences in the corresponding favorable outcomes tend to be smaller, where the adverse outcomes are comparatively rare than where they are comparatively common.[3-6]
Absolute differences tend also to be affected by the prevalence of an outcome, though in a more complicated way than the two relative differences. Roughly, as an outcome goes from being rare to being common, absolute differences tend to increase; as an outcome goes from being common to being very common, absolute differences tend to decrease. As the frequency of an outcome changes, the absolute difference tends to change in the same direction as the smaller relative difference. [3-6] Since persons relying on relative differences to appraise health inequalities typically examine the larger relative difference (often without even recognizing the possibility that the other relative difference would yield an opposite conclusion about direction of a change in inequality much less the tendency for this to occur systematically) such persons usually reach opposite conclusions from persons relying on absolute differences.
The absolute difference and both relative differences may all change in the same direction, in which case one may infer that there occurred a meaningful change in the strength of the forces causing the outcome rates to differ. But in the common situations where all measures change in accordance with the aforementioned patterns, one can only determine whether those forces have increased or decreased by employing a measure unaffected by the prevalence of an outcome (such as that described in references 3-6).
For the most part, health inequalities research has suffered from the failure of those conducting it to understand that these patterns exist at all. But the lead author of this article is one of the few people who have recognized the patterns (if not the distribution-related basis for them), as I have discussed here previously. Yet nothing in the article suggests recognition of these patterns, even as it discusses the potential role of particular policies in observed patterns. One cannot make sound appraisals of such roles without consideration of the patterns by which measures tend to be affected by the prevalence of an outcome.
One other aspect of the article warrants particular mention. Like another recent article by Professor Mackenbach, the article apparently accepts the reasoning of Harper et al. that choice between the relative difference and the absolute difference reflects a normative judgment.
Those addressing the putatively normative choice between the absolute difference and the relative difference they happen to be examining invariably have failed to mention the relative difference in the opposite outcome. Yet, anytime a relative difference and the absolute difference have changed in opposite direction, the unmentioned relative difference will necessarily have changed in the opposite direction of the mentioned relative difference and the same direction as the absolute difference. Thus, those discussing choosing between a relative difference and the absolute difference have already chosen, without evident justification, to rely on the relative difference that shows an opposite result from the absolute difference rather than the relative difference that shows the same result as the absolute difference.
But the forces causing adverse outcome rates of advantaged and disadvantaged groups to differ are exactly the same forces causing the corresponding favorable outcome rates to differ. And there is no rational basis for choosing the relative measure that says those forces are growing stronger over the relative measure that says they are growing weaker. The crucial factor, however, is that because the measures tend to changes solely because the prevalence of the outcome changes, a factor also applicable to the absolute difference, they cannot indicate whether the forces causing rates to differ are growing stronger or weaker without consideration of way the measure tends to be systematically affected by changes in the prevalence of the outcome.
Thus, for example, there is little purpose in discussing a change in an absolute difference without consideration of whether it changed less than, more than, or the same as what one should expect to occur solely because of a change in the prevalence of an outcome.
Because many forces play have a role in causing adverse (and the corresponding favorable) outcome rates of advantaged and disadvantaged groups to differ, it is possible that some may increase over a particular period while others decrease over the same period. But there can be only one answer to the question of whether the net effect of those forces has increased or decreased. Scientific judgments may have a role in answering that question. Normative judgments do not. See discussion regarding Table 1 of reference 3, Table 5 of reference 4, and Table 1 of reference 5.
1. Mackenbach JP, Kulhánová I, Artnik B, et al. Changes in mortality inequalities over two decades: register based study of European countries. BMJ 2016;353:i1732
http://dx.doi.org/10.1136/bmj.i1732 (Accessed 9 May 2016)
2. Carr-Hill R, Chalmers-Dixon P. The Public Health Observatory Handbook of Health Inequalities Measurement. Oxford: SEPHO; 2005: http://www.sepho.org.uk/extras/rch_handbook.aspx (Accessed 9 May 2016)
3. Scanlan JP. The mismeasure of health disparities. J Public Health Manag Pract. 2016;22 (July/Aug.) (in press)
4. Scanlan JP. Race and mortality revisited. Society. 2014;51:327-346. http://jpscanlan.com/images/Race_and_Mortality_Revisited.pdf (Accessed 9 May 2016)
5. Scanlan JP. Measuring health and healthcare disparities. Proceedings of the Federal Committee on Statistical Methodology 2013 Research Conference. 2014 (May).
http://fcsm.sites.usa.gov/files/2014/05/J4_Scanlan_2013FCSM.pdf (Accessed 9 May 2016)
6. Scanlan JP. The mismeasure of health disparities in Massachusetts and less affluent places. Quantitative Methods Seminar, Department of Quantitative Health Sciences, University of Massachusetts Medical School (Nov. 18, 2015).
Abstract: http://jpscanlan.com/images/UMMS_Abstract.pdf (Accessed 9 May 2016)
PowerPoint: http://jpscanlan.com/images/Univ_Mass_Medical_School_Seminar_Nov._18,_20... (Accessed 9 May 2016)
7. Scanlan JP. The monitoring of health inequalities has never been sound II. BMJ Aug. 18, 2014 (responding to Marmot M, Goldblatt P. Importance of monitoring health inequalities. BMJ 2013;347:f6576). http://www.bmj.com/content/347/bmj.f6576/rr/76285 (Accessed 9 May 2016)
8. Mackenbach JP. Should we aim to reduce relative or absolute inequalities in mortality? Eur J Pub Health 2015;25(2):185.
9. Harper S, King NB, Meersman SC, Reichman ME, Breen N, Lynch J. Implicit value judgments in the measurement of health inequalities. Milbank Q 2010;88:4-29. http://onlinelibrary.wiley.com/enhanced/doi/10.1111/j.1468-0009.2010.005... (Accessed 9 May 2016)
Competing interests: No competing interests