Javier M. Rodriguez

Health, economic crisis, and austerity: A comparison of Greece, Finland and Iceland.

Reports have attributed a public health tragedy in Greece to the Great Recession and the subsequent application of austerity programs. It is also claimed that the comparison of Greece with Iceland and Finland-where austerity policies were not applied-reveals the harmful effect of austerity on health and that by protecting spending in health and social budgets, governments can offset the harmful effects of economic crises on health. We use data on life expectancy, mortality rates, incidence of infectious diseases, rates of vaccination, self-reported health and other measures to examine the evolution of population health and health services performance in Greece, Finland and Iceland since 1990-2011 or 2012-the most recent years for which data are available. We find that in the three countries most indicators of population health continued improving after the Great Recession started. In terms of population health and performance of the health care system, in the period after 2007 for which data are available, Greece did as good as Iceland and Finland. The evidence does not support the claim that there is a health crisis in Greece. On the basis of the extant evidence, claims of a public health tragedy in Greece seem overly exaggerated.

US infant mortality and the President’s party

BACKGROUND Infant mortality rates in the US exceed those in all other developed countries and in many less developed countries, suggesting political factors may contribute. METHODS Annual time series on overall, White and Black infant mortality rates in the US were analysed over the 1965-2010 time period to ascertain whether infant mortality rates varied across presidential administrations. Data were de-trended using cubic splines and analysed using both graphical and time series regression methods. RESULTS Across all nine presidential administrations, infant mortality rates were below trend when the President was a Democrat and above trend when the President was a Republican. This was true for overall, neonatal and postneonatal mortality. Regression estimates show that, relative to trend, Republican administrations were characterized by infant mortality rates that were, on average, 3% higher than Democratic administrations. In proportional terms, effect size is similar for US Whites and Blacks. US Black rates are more than twice as high as White, implying substantially larger absolute effects for Blacks. CONCLUSIONS We found a robust, quantitatively important association between net of trend US infant mortality rates and the party affiliation of the president. There may be overlooked ways by which macro-dynamics of policy impact microdynamics of physiology, suggesting the political system is a component of the underlying mechanism generating health inequality in the USA.

Measuring Recent Apparent Declines In Longevity: The Role Of Increasing Educational Attainment

Independent researchers have reported an alarming decline in life expectancy after 1990 among US non-Hispanic whites with less than a high school education. However, US educational attainment rose dramatically during the twentieth century; thus, focusing on changes in mortality rates of those not completing high school means looking at a different, shrinking, and increasingly vulnerable segment of the population in each year. We analyzed US data to examine the robustness of earlier findings categorizing education in terms of relative rank in the overall distribution of each birth cohort, instead of by credentials such as high school graduation. Estimating trends in mortality for the bottom quartile, we found little evidence that survival probabilities declined dramatically. We conclude that widely publicized estimates of worsening mortality rates among non-Hispanic whites with low socioeconomic position are highly sensitive to how educational attainment is classified. However, non-Hispanic whites with low socioeconomic position, especially women, are not sharing in improving life expectancy, and disparities between US blacks and whites are entrenched. Findings underscore the urgency of an agenda to equitably disseminate new medical technologies and to deepen knowledge of social determinants of health and how that knowledge can be applied, to promote the objective of achieving population health equity.

Black lives matter: Differential mortality and the racial composition of the U.S. electorate, 1970–2004

Excess mortality in marginalized populations could be both a cause and an effect of political processes. We estimate the impact of mortality differentials between blacks and whites from 1970 to 2004 on the racial composition of the electorate in the US general election of 2004 and in close statewide elections during the study period. We analyze 73 million US deaths from the Multiple Cause of Death files to calculate: (1) Total excess deaths among blacks between 1970 and 2004, (2) total hypothetical survivors to 2004, (3) the probability that survivors would have turned out to vote in 2004, (4) total black votes lost in 2004, and (5) total black votes lost by each presidential candidate. We estimate 2.7 million excess black deaths between 1970 and 2004. Of those, 1.9 million would have survived until 2004, of which over 1.7 million would have been of voting-age. We estimate that 1 million black votes were lost in 2004; of these, 900,000 votes were lost by the defeated Democratic presidential nominee. We find that many close state-level elections over the study period would likely have had different outcomes if voting age blacks had the mortality profiles of whites. US black voting rights are also eroded through felony disenfranchisement laws and other measures that dampen the voice of the US black electorate. Systematic disenfranchisement by population group yields an electorate that is unrepresentative of the full interests of the citizenry and affects the chance that elected officials have mandates to eliminate health inequality.

The Implications of Differential Trends in Mortality for Social Security Policy

While increased life expectancy in the U.S. has been used as justification for raising the Social Security retirement ages, independent researchers have reported that life expectancy declined in recent decades for white women with less than a high school education. However, there has been a dramatic rise in educational attainment in the U.S. over the 20th century suggesting a more adversely selected population with low levels of education. Using data from the National Vital Statistics System and the U.S. Census from 1990-2010, we examine the robustness of earlier findings to several modifications in the assumptions and methodology employed. We categorize education in terms of relative rank in the overall distribution, rather than by credentials or years of education, and estimate trends in mortality for the bottom quartile. We also consider race and gender specific changes in the distribution of life expectancy. We found no evidence that survival probabilities declined for the bottom quartile of educational attainment. Nor did distributional analyses find any subgroup experienced absolute declines in survival probabilities. We conclude that recent dramatic and highly publicized estimates of worsening mortality rates among non-Hispanic whites who did not graduate from high school are highly sensitive to alternative approaches to asking the fundamental questions implied. However, it does appear that low SES groups are not sharing equally in improving mortality conditions, which raises concerns about the differential impacts of policies that would raise retirement ages uniformly in response to average increases in life expectancy.

Rejoinder: Time series analysis and US infant mortality: de-trending the empirical from the polemical in political epidemiology

In US Infant Mortality and the President’s Party,1 we analysed US infant mortality data from 1965 through 2010, and found that once one ‘filters’ out a smooth trend from the series, a persistent pattern emerges, with infant mortality rates being above trend during Republican Administrations and below trend during Democratic ones. As we noted in the article and detailed in its online appendix, this pattern is robust to whether we ‘filtered’ the trend using cubic splines, linear polynomials or fractional polynomials. The deviations from trend are visually evident and are easily detectable in the second stage regressions we estimated. Any claims about causality would be substantially stronger were we to understand mechanisms; yet, we believe the pattern we have uncovered merits reporting because it is striking and robust.

Evidence is missing to diagnose a public health tragedy.

Anastasiou et al. [1] cite selected 2013 statistics on vailability of medications, “certain diseases,” and alcool consumption among the young as evidence suggesting hat our analyses incurred in an underestimation of the ecession effects on health. But first, we said in our paper 2] that our data covered since 1990 until 2011, or 2012 n some cases, and second, we presented a comprehenive set of 30 major indicators of heath to avoid this ype of non-systematic selection of statistics. We did not ide a rise of suicides during the recession, but we put t in context. We also emphasized the need to prioriize prevention of suicides. Indeed we suggested “targeted olicies and programs to ameliorate specific worsening rends. That is the case with HIV and suicide prevention” 2]. A sentence in our paper [2] said that published nvestigations reveal consistent evidence that during usiness-cycle recessions “population health tend to deteiorate.” This was indeed a lapsus calami—since we wanted o write “tend to improve” not “tend to deteriorate.” Durng recessions mortality due to major causes of death ends to decrease faster, but mental health tends to eteriorate, and suicides often rise. This was discovered lmost a century ago [3], but it is a fact which many ealth workers and researchers still find surprising. Real-

US life expectancy: The authors reply

US Life Expectancy: The Authors Reply We appreciate the interest of Jennifer Montez and coauthors in our article on US life expectancy trends for the period 1990–2010 (Dec 2015). Overall, we found gains in life expectancy. However, among whites with low socioeconomic status, we reported stagnating life expectancy amongmen and a longevity decline of 1.2 years among women. Stagnation or decline is disturbing. Montez and coauthors object that we describe our estimates as “modest” relative to other published estimates. Yet they are. Earlier researchers reported longevity declines of 3.4 and 5.3 years for white males and females with low socioeconomic status, respectively. Such declines are virtually unprecedented in the developed world. These earlier researchers categorized socioeconomic status by educational attainment (less than a high school diploma, high school graduate, and so on), although the proportion of adults who graduated from high school rose dramatically across the US cohorts they studied. Thus, estimated life expectancy declines in their least educated group might reflect only compositional changes. In contrast, we took a more conservative approach and categorized individuals by their within-cohort rank in the educational distribution. Our finding that longevity declined for white women with low socioeconomic status although educational attainment was rising is, arguably,more disturbing than earlier findings because our results cannot plausibly be driven by compositional effects. The life expectancy decline that we reported contrasts with the longevity gains of more socioeconomically advantaged US groups and contrasts even more sharply with contemporaneous longevity gains in European populations. Clearly, recent public health and medical advances have not been shared equally across US socioeconomic groups. Developing an explicit health equity policy agenda is imperative.

Authors’ response: Politicization with misrepresentation: on de-trending in time series analysis

Professor Catalano referred to our good faith and peer-reviewed empirical article1 as equivalent to ‘dog food, the grocer would have to pull … from the shelf’.2 He attributed political motives to us as the basis for our findings. Call us crazy, but we stand by the view that such criticisms are ad hominen, and far from ad rem. Despite that, and Professor Catalano’s current claim to the contrary, in our initial rejoinder3 we explicitly acknowledged that his commentary also included a number of serious substantive arguments. Indeed, the vast majority of our response involved a detailed discussion of the substantive points he made, most of which he does not acknowledge in his current letter. Instead, Professor Catalano continues to politicize our motives and misrepresents our response in an additional way. Professor Catalano says that we claimed that when we used an ARIMA model to ‘filter’ out a trend from the time series data on infant mortality, we continued to find a statistically significant and quantitatively important association between the residuals from this filtering process and the president’s party. We actually claimed the opposite.3 If one removes the trend from the infant mortality series using standard procedures in the literature such as cubic splines or simple or fractional polynomials, visual inspection of the residuals shows a pattern that strikingly mirrors the pattern of changes in the party of the president (see Figure 2 in our article1). If one removes the trend using an ARIMA model, there is no evident pattern in the residuals. Our disagreement with Professor Catalano is whether, for our research purpose, we should be de-trending using a variety of cubic splines or polynomials, or using an ARIMA model. The figure in our response to Professor Catalano’s commentary was meant to illustrate the difference between our preferred de-trending method and a best-fitting ARIMA model.3 [For fit statistics of nine different ways of de-trending the series see Table 1 and its discussion in the appendix to our paper (available as Supplementary data at IJE online)]. As the figure in our response3 illustrates, fitting an ARIMA process to the infant mortality series would filter out the variation that exists in low and medium frequencies of the data. This is critical. To exemplify the nature of the variation contained in the infant mortality series, we reported the results of our Fourier decomposition series analysis in the appendix to our article1 (available as Supplementary data at IJE online). It formally shows the infant mortality series includes both low frequency variation (the downward trend) and medium frequency variation. This medium-run variation has a period of approximately 16 years, and it mirrors the president’s party, rising during the years in which the president was a Republican and falling during years in which the president was a Democrat. This need not have been true. Professor Catalano misunderstands our cubic spline procedure to de-trend the data and the rationale behind what constitutes a trend and what does not. In theory, a depiction of a long-run trend should be insensitive to short-term, year-specific variation, which is precisely what the residualized series attempts to represent. We did not imply that ‘autocorrelation in a time series divides into good and bad’, as Catalano argues; our concern is that, in this case, the ARIMA model ‘filters’ out the medium-run frequency. In his initial commentary,2 Professor Catalano also suggested that carrying out a comparative vs single-country approach would constitute a best test of our hypothesis. And so we did. As we noted in our response,3 after de-trending the series using a cubic spline (instead of using Box–Jenkins methods), we once again found that the estimated coefficient of the presidential party affiliation was precisely estimated and very similar to the coefficient we reported in our paper. Thus, whatever method one uses to filter the data, and whether one uses a comparative or a single-country approach to test our hypothesis, over the past half-century in the USA, at medium frequencies, there appears to be an association between the president’s party affiliation and national and race-specific infant mortality. This finding showed to be substantial and remarkably robust to a variety of model specifications.1 Whether this robust association is causal, spurious or reflects something else systematically associated with the president’s party affiliation, we do not know, but given the magnitude of its possible implications, and the importance of understanding persistent excess US infant mortality rates, we believe it is worth further investigation.