Refining current methods: Including comorbidity adjustment in burden of disease estimates

Abstract

Abstract When quantifying years lost due to disability (YLD), a correction for multimorbidity (or comorbidity, COMO for short, in the Global Burden of Disease study) is necessary to avoid overestimating the morbidity-related burden of disease. Such an overestimation occurs if the calculation of disease frequencies does not consider that persons are counted more than once in a cross-sectional year. In such cases, the time spent with illness at an individual level can add up to a value of more than 1 year. To avoid this, a microsimulation is used that corrects the YLD to a maximum value of 1. A standard approach is to generate a synthetic data set by age, gender and region. Using prevalence estimates and based on (independent) Bernoulli experiments, a vector of diseases is randomly assigned to these (pseudo) individuals. The assignment is usually independent, which means that the presence of one disease has no influence on the presence of another disease. At this level, the individual YLD are then proportionally corrected (the maximum value is 1) and extrapolated to the population. However, there is evidence that disease clusters exist at the individual level. As part of the further development of such multimorbidity correction methods, a new approach will be tested and applied to generate a more realistic population of interest. This considers correlation patterns (1st step) between diseases when generating the synthetic data (2nd step) on the basis of Bernoulli experiments. For this purpose, the existing approaches need to be enriched by (pairwise) correlations between diseases, which requires the inclusion of additional epidemiological data sources. It is assumed that this will lead to an additional correction of the YLD, as they are still overestimated under an independent assignment.