Mortality Rate: A Curve-Ball

From The Office For Nobbled Statistics

When trying to understand the incidence of death over time we often resort to calculation of a crude rate such as deaths per 100k population. If we are lucky we can enhance this to age-adjusted mortality provided we have robust estimates of the age profile over the years. Though mightily ubiquitous these rate figures are based upon an assumption of linearity; that is to say if we get 10 deaths in 100,000 population (MR = 0.01%) we shall expect to see 20 deaths in 200,000 (MR = 0.01%) if nothing has changed other than the population increasing over time. But does that assumption of linearity necessarily hold true for a nation developing over a period of 50 years?

In the attached slide I’ve done something quite geeky and plotted out the weekly all cause death counts for England & Wales against the population estimate for that week. Many are going to ask how weekly population estimates were obtained and the answer is from ONS mid-year estimates for each year since 1970 that have been subj…