# On the measurement of cause of death inequality

## TL;DR

This paper introduces a new way to measure inequality in causes of death that accounts for how similar or different causes are, offering a more nuanced view of mortality patterns.

## Contribution

The paper proposes a novel class of heterogeneity measures for cause of death data that incorporate pairwise dissimilarity.

## Key findings

- CoD inequality and diversity generally increased over time in low-mortality countries.
- In some countries like Finland, CoD inequality and diversity moved in opposite directions.
- The new measures are decomposable, allowing analysis of each cause's contribution to overall heterogeneity.

## Abstract

Attempts at assessing heterogeneity in countries’ mortality profiles often rely on measures of cause of death (CoD) diversity. Unfortunately, such indicators fail to take into consideration the degree of (dis)similarity among pairs of causes (e.g. ‘transport injuries’ and ‘unintentional injuries’ are implicitly assumed to be as dissimilar as ‘transport injuries’ and ‘Alzheimer’s disease’)－an unrealistic and unduly restrictive assumption.

We extend diversity indicators proposing a broader class of heterogeneity measures that are sensitive to the similarity between the causes of death one works with. The so-called ‘CoD inequality’ measures are defined as the average expected ‘dissimilarity between any two causes of death’. A strength of the approach is that such measures are decomposable, so that users can assess the contribution of each cause to overall CoD heterogeneity levels—a useful property for the evaluation of public health policies.

We have applied the method to 15 low-mortality countries between 1990 and 2019, using data from the Global Burden of Disease project. CoD inequality and CoD diversity generally increase over time across countries and sex, but with some exceptions. In several cases (notably, Finland), both indicators run in opposite directions.

CoD inequality and diversity indicators capture complementary information about the heterogeneity of mortality profiles, so they should be analysed alongside other population health metrics, such as life expectancy and lifespan inequality.