User:Adam Harangozó (NIHR WiR)/sandbox/disease clusters

Clusters of long-term conditions

Introduction

When particular long-term conditions co-occur within an individual or groups of individuals they are called clusters. Understanding the clusters or patterns of long-term conditions that commonly occur together can help in identifying common risk factors and suggest the best treatment and prevention approaches. This use of the term ‘cluster’ is distinct from the geographical clustering of disease such as disease clusters of infectious disease.

Some conditions are more likely to be closely associated with others due to their pathophysiology. For example, common cardiovascular conditions such as heart disease and stroke commonly cluster with conditions such as hypertension and obesity because people with hypertension and obesity are at greater risk of developing cardiovascular disease. However, there are many different patterns of clusters and some clusters have been identified where the causal pathway is not as obvious, such as the clustering of depression with several physical health conditions. The nature of clusters also depends on the methods used to identify them in any given population. The biosocial concept of understanding the way in which diseases co-occur together, and the social and environmental factors that promote and enhance the negative effects of disease interaction, is referred to as syndemics.

Definition

Clusters of long-term conditions can be understood as patterns of frequently co-existing long term conditions. Several statistical methods have been used to identify clusters of long-term conditions (not to be confused with disease clusters which refers to diseases appearing within a particular geographical location or period) with two or more medical conditions in each clusters which tend to occur (cluster) together in the population.

Compared to looking at each individual condition, these clusters of long-term conditions may differ in how they affect health, health-care visits, health expenditure, and social determinants of health, as well as quality of life.

Methods

Standard methods for doing this include latent class analysis, hierarchical clustering with multiple correspondence analysis, partitioning around medoids using Jaccard distance, and Bayesian networks. These methods are applied on a cross-section of data while disease clusters have also been found using their trajectories.