Recent public health studies made headlines reporting that life expectancy had declined over time for some subgroups in the U.S., particularly for White women with less than a high school education. These findings were understandably alarming, leading some to make analogies to the losses in life expectancy in Russia after the collapse of the Soviet Union.
Now, a recent editorial in the International Journal of Epidemiology urges caution in interpreting these mortality patterns. Co-authors Dr. Jennifer Dowd, associate professor of epidemiology at the CUNY School of Public Health, and Dr. Amar Hamoudi, assistant professor of public policy at the Sanford Institute of Public Policy at Duke University propose an alternative explanation, whereby lower life expectancies and higher mortality rates emerge as an artifact of a phenomenon they call “lagged selection bias.” In this case, lagged selection bias arises from the fact that access to education expanded dramatically for Whites in the United States over the first half of the twentieth century. As a result of that expansion, most Whites who died twenty years ago came of age when high school completion was the norm only for the most socially advantaged, whereas most of those who are dying today grew up in environments where a high school diploma was much more accessible. Thus the meaning of having less than a high school education with regards to baseline social circumstances has changed dramatically over the two time periods.
Comparing life expectancy for those with less than high school across this time period, the authors argue, is akin to measuring the average temperature of the USA two decades ago to Alaska today, and calling the resulting difference a “decline” in temperature over time. In the same way the difference between these two average temperatures may not evidence that anyone is actually experiencing more cold than they used to, so the difference in life expectancies may not indicate that anyone is actually dying any younger.
Using historical data on education and mortality in the U.S., the paper shows how under reasonable assumptions about stable mortality differences by early life social class, life expectancy for those with less than a high school education can appear to “fall” even though in fact mortality has improved for every group. Failure to acknowledge lagged selection bias can have important policy implications, since it can misattribute explanations for these trends to contemporary changes in policy or the economy when the trends are actually due to changing social dynamics around entry into education decades prior. Lagged selection bias can also apply in other contexts where there is lag between the timing of the exposure and outcome examined, and the dynamics of selection into the exposure have changed over time. Thus lagged selection bias may be important for interpreting life expectancy changes across U.S. counties that have seen significant out-migration over time, as well as numerous other applications. The authors note ways in which typical methods to deal with this non-comparability fail to address the underlying problem, and encourage serious attention to this empirical issue in the future, particularly in the area of health disparities by educational attainment.
Article: “Is life expectancy really falling for groups of low socio-economic status? Lagged selection bias and artefactual trends in mortality”, by Dr. Jennifer B. Dowd and Dr. Amar Hamoudi, is published in International Journal of Epidemiology (2014) 43(4): 983-988 DOI: 10.1093/ije/dyu12: published by Oxford journals.