CHARACTERIZING THE COMPLEXITY OF MULTIMORBIDITY IN OLDER ADULTS USING A NETWORK APPROACH

There is no evidence that any single multimorbidity measure captures the multifactorial nature of health and function in older adults. Current measures are generally based on the concept of multimorbidity as two or more co-occurring chronic diseases, while symptoms and geriatric syndromes are frequently excluded. To better characterize the complexity of multimorbidity, we used a network approach to evaluate the co-occurrence of diseases, symptoms, and geriatric syndromes in older adults. We estimated sex-stratified networks of 26 self-reported diseases, symptoms, and geriatric syndromes in 9,267 participants aged 65+ from the Health and Retirement Study. A network depicts relationships among interacting elements. Diseases, symptoms, and geriatric syndromes were represented as points (“nodes”) and pairwise correlations were represented as lines. Network structure was evaluated using centrality measures, crisp clustering, and fuzzy clustering. Cluster analysis grouped strongly co-occurring nodes. Crisp clustering placed each node into one group, while fuzzy clustering provided probabilities of belonging to multiple groups. Symptoms and geriatric syndromes played a more important role, measured by degree centrality, in network structure than diseases. Diseases, symptoms, and geriatric syndromes clustered differently in males and females. Fuzzy cluster analysis revealed symptoms and geriatric syndromes were more likely than diseases to occur in multiple clusters. The approach to measuring multimorbidity based solely on diseases likely underestimates complexity by disregarding symptoms and geriatric syndromes. The network approach revealed how diseases, symptoms, and geriatric syndromes commonly co-occur in older adults. This work illustrates the potential of network methods to improve the measurement of multimorbidity in older adults.