Detecting Public Health Crises: APC-Detrended Methodology and Residuals in a 25-Country, 35-Year Mortality Matrix

Anja Leist, University of Luxembourg
Louis Chauvel, University of Luxembourg
Herbert Smith, University of Pennsylvania

Background. Mortality-by-cause analyses as done in the Global Burden of Disease 2016 update are helpful to monitor progress of public health improvements within and across countries. However, separate analyses by cause and country miss larger patterns of public health crises that are restricted to certain cohorts and periods. Those public health crises may go unnoticed even if they affect several countries and thus come with some threat potential. We propose a new method to detect cohorts with increased mortality at certain ages and periods.

Method. We develop an analytical and visualizing technique based on established Age-Period-Cohort-Detrended (APCD) methodology (Chauvel and Schröder 2014). After detecting all-cause mortality increases, plotting the resulting age-period coefficients and APCD residuals in equilateral Lexis diagrams, mortality patterns can be distinguished as age, period, or cohort trends and fluctuations. Age-period interactions are plotted as ‘big red spots’. We employ the new technique in data from the Human Mortality Database, spanning 25-60 years of age, calendar years 1975-2010, and 25 countries.

Results. We detect age-period interactions of young-adult cohorts in the early 1990s in Spain, other southern European countries and the U.S. Additional analyses with WHO mortality data show that mortality increases are mostly due to increased HIV/AIDS mortality.

Discussion. Country-specific explanations, such as political frustrations in Spain, have been proposed to explain the 1990s increases in HIV/AIDS mortality. However, the new technique suggests that increases in HIV/AIDS mortality were more likely to be due to specific behaviors of cohorts of certain ages in a certain period. We discuss limitations of the method, such as detecting social class mortality differences of affected cohorts. Altogether, the new technique offers intuitive and efficient handling of large amounts of age-country-year mortality information. The method can further be applied in the fields of education, longevity, and demography at large.


Presented in Session 1196: Mortality and Longevity