Problems in determining occurrence rates of multimorbidity

Abstract

This article describes methodological decisions that have to be made when studying multiple pathology and presents appropriate analytical techniques. The main question of this article is: how can comorbidity and multimorbidity be operationalized with respect to the number and type of diseases studied, and which analytic approaches are available for the evaluation of multiple pathology? Choices regarding the number and type of diseases studied have great impact on the observed incidence and prevalence rates of comorbidity and multimorbidity. These rates are largely dependent on age, sex, and other determinants. In addition to crude descriptive measures, odds ratios and relative risks can be used to study comorbidity, whereas multimorbidity can be studied using observed/expected ratios. While basic analyses of comorbidity can be performed using standard statistical packages, two additional programs were developed for the analysis of the distribution of multimorbidity and statistically unexpected comorbidity, respectively. As some analyses are addressing multicomparisons, external validity testing is recommended.

Introduction

Since the introduction of the term comorbidity [1] a few decades ago, the amount of research in this area has strongly increased. Most of these studies were focused on either the consequences of multiple pathology or the causes of specific combinations or chains of diseases. Specific causes of multimorbidity have, to our knowledge, been little studied so far [2]. Multiple pathology is a highly prevalent health problem and is found to be as frequent as 10% in 0–19-year-olds to 78% in people aged 80 or older [3]. Because of the ageing of the population, and the related increase of the prevalence of chronic diseases, these proportions are expected to rise further. Although there has been increasing interest in research on multiple pathology during recent years, few articles have been published on the methodology of comorbidity and multimorbidity research. As a result, researchers in this field often have to find out by trial and error the most feasible ways to study and analyze multiple pathology.

This article describes some methodological choices and different analytical strategies that have to be considered. The consequences of these strategies are illustrated with examples using data from the Registration Network Family Practices (RNH), a comprehensive general practice-based data system 4, 5 (see Appendix), and from recent studies [3] on the determinants of multimorbidity. The focus is on research using comorbidity or multimorbidity as the dependent variable, but much of the reasoning may also be applicable to studies in which comorbidity or multimorbidity is used as an intermediate or independent variable. This overview may be useful for researchers in critical reading of articles reporting on multiple pathology and in starting studies in this area.