Variation in multimorbidity between and within age groups

Abstract The chronic conditions associated with age are an increasing challenge to individual and public health in many countries. In general, older age groups have a greater average multiplicity of chronic conditions (greater “multimorbidity”). But the variation (not only the average) of multimorbidity within groups of increasing age seems not to have been studied empirically. The variation in multimorbidity in groups of increasing age could, in principle, increase or decrease. We use the enhanced Charlson Comorbidity Index (eCCI) as a quantitative index of a person’s concurrent chronic medical conditions (multimorbidities) in 39 categories, weighted by severity. We demonstrate systematic relationships between the mean eCCI and the variance of eCCI of people by age group in de-identified electronic health records of 238,156 adults served by four United States health systems (HSs), two in Chicago, two in New York City. The relation of variance to mean eCCI across age groups is described by a generalization of Taylor’s power law. Within almost every age group, the frequency distribution of eCCI per individual is well approximated by a negative binomial distribution. The means and the variances of the fitted negative binomial distributions within age groups account approximately for the relations of mean and variance of eCCI across age groups. The clear findings that variance of eCCI within an age group rises with the mean eCCI, and that the variance exceeds the mean, imply that increasing average multimorbidity brings more variable or heterogeneous multimorbidity. Both clinical care and health policy need to take account of the variation or heterogeneity of multimorbidity among older adults.