The life expectancy gap in the United States
My colleague Dr. Corinne Riddell is a post-doctoral fellow at McGill, and she has been researching the gap in life expectancy at birth between blacks and whites in the US over the last several decades.1 Life expectancy at birth is a metric commonly used to compare health disparities between different populations. She took mortality data from the National Vital Statistics System from 1969 to 2013, categorized by race, sex, age group, year, state, and cause-of-death. The six cause of death categories were cancer, cardiovascular disease, communicable disease, non-communicable disease, injuries, and all other causes. She then decomposed the gap in life expectancy between blacks and whites, to explore what causes of death contributed most to the disparity. Her findings will be published as a journal article, and she has also developed an interactive shiny app where you can explore these data and results.
For this project, I coded and ran a Bayesian model that was used to smooth the rates and impute missing values. An observed mortality rate may appear to be zero, in the sense that zero deaths may occur within a (age, sex, state, race, year, cause-of-death) strata. This becomes more common when population counts are small. However, the theoretical mortality rate for any population is never zero, it just may be very small. Therefore, we conceptualized the mortality rate as a latent measure that is not observed directly, but rather the resulting number of deaths are observed. Also, any strata that had between one and nine were censored (or rather, coarsened, as we knew the true number was between 1 and 9). We therefore smoothed the mortality rates over time, using ‘neighbouring’ years to inform the underlying true mortality rate and using a truncated Poisson for the coarsened strata ‘cells.’ When the paper and shiny app are public, I will publish the JAGS code used for this model.
1 This work builds on her supervisor’s previous work: Harper S, MacLehose RF, Kaufman JS. 2014. Trends in the black-white life expectancy gap among US states, 1990–2009. Health Affairs, 33(8): 1375-1382.