Mapping Multimorbidity Progression

Mapping Multimorbidity Progression Among 190 Diseases

Background

Globally, with the aging population and the significant increase in the negative impacts of chronic diseases, multimorbidity, the coexistence of multiple long-term conditions, has become an increasingly severe health challenge. Understanding the accumulation process of multimorbidity in individuals can help researchers better comprehend its pathogenesis, assist healthcare providers in intervening or screening for other conditions when the first symptom appears, and enable decision-makers to formulate integrated care management approaches to better meet the health needs of patients.

Research Motivation

Current multimorbidity clustering methods are mainly based on the frequency of common disease combinations, making it difficult to understand how diseases develop over time. Most association-based clustering methods tend to exaggerate the most popular diagnoses, resulting in too many unrelated diseases being grouped into a large cluster. More importantly, these association-based clustering methods can be misleading when determining the trajectories of multimorbidity progression. This method does not allow for bidirectional trajectories between diseases, where one disease can lead to another or be influenced by another. Therefore, researchers proposed a more informative method to study patterns of multimorbidity progression.

Source of the Paper

This study was written by Shaosha Chen et al. and published in Volume 4 of “Communications Medicine” in 2024.

Research Methods

Study Design and Population

The study was conducted among 502,413 adult participants from the UK Biobank, aged between 37 and 73 years, with an average age of 57.1 years. The study took place from March 2006 to December 2010, with a follow-up period of 12.7 years. The study received ethical approval from the UK’s National Information Governance Committee and the North West Multi-center Research Ethics Committee. All participants provided informed consent via electronic signature at the baseline assessment.

Disease Status and Baseline Covariates

The study focused on diseases with a prevalence greater than 1% in both males and females, identifying a total of 190 of the most common diseases (154 in females and 160 in males). Baseline covariates included sociodemographic factors, health behaviors, health status, all disease statuses, factors affecting health status, and interactions with health services.

Data Analysis

The study used Targeted Maximum Likelihood Estimation (TMLE) to estimate the impact of one disease on the development of another, analyzing the causal relationships of diseases in females and males. This covered 23,562 directional pairwise causal effects (females) and 25,440 directional pairwise causal effects (males). Diseases were clustered using hierarchical clustering algorithms, and the progression of multimorbidity was clustered into multimorbidity constellations using a self-tuning k-means clustering method.

Workflow

The study first estimated the causal relationships between each pair of diseases, then identified the most influential and most affected diseases by plotting marginal spectra. Next, clustering algorithms were used to group diseases with similar patterns. Finally, a visual clustering diagram composed of disease nodes was plotted, and cluster stability was evaluated.

Research Results

Disease Causal Effect Diagram

The study showed that the most influential diseases, whether in females or males, were often chronic diseases such as hypertension and diabetes, while acute diseases had gender-specific characteristics. Bidirectional multimorbidity progression exhibited significant clustering trends, crossing chapters of the International Classification of Diseases, suggesting complex mechanisms between diseases.

Influential and Affected Diseases

The study identified the top ten influential diseases and the top ten affected diseases for each gender, finding that most diseases showed consistency between genders, but some acute diseases had gender-specific characteristics. Additionally, some diseases had a significant impact on others in a short period.

Clustering Results

Through clustering analysis, the study identified 26 disease clusters in females and 28 in males. These disease clusters reflected shared disease development mechanisms. Analysis of bidirectional disease progression indicated that diseases were more likely to exhibit bidirectional progression within the same International Classification of Diseases chapter.

Multimorbidity Constellations

The study identified ten multimorbidity progression constellations in females and nine in males, showing gender differences. For example, the hypertension-asthma constellation showed many pathways from hypertension to asthma, while the chronic obstructive pulmonary disease-circulation constellation demonstrated multimorbidity progression between respiratory and cardiovascular diseases.

Research Conclusions

The study proposed a framework for multimorbidity progression based on disease causal relationships, providing a foundation for future targeted interventions. The clustering of diseases in the study offers a systematic perspective for updating disease classification systems and discovering potential shared therapeutic strategies. By identifying the roles of different diseases in multimorbidity, the study lays a foundation for future research aimed at preventing and managing multimorbidity.

Highlighted Research Points

1. Causal Relationship Analysis of Multimorbidity: For the first time, an in-depth analysis of the causal progression of 190 diseases.
2. Disease Development Mechanisms: Clustering of bidirectional multimorbidity progression identified shared pathogenesis.
3. Multimorbidity Constellations: Identification of multimorbidity constellations for each gender, showing gender-specific disease progression pathways.

Scientific Value and Application Value

This study provides important information for the prevention and management of multimorbidity, aiding in the development of community-based multimorbidity screening strategies and reorganizing services to better meet patient needs.

These findings are not only significant for further understanding the development mechanisms of multimorbidity but also have profound implications for formulating better medical intervention strategies. In this way, healthcare services can be better optimized, making them more personalized and precise to effectively address the challenges