Temporal multimorbidity patterns and cluster identification: a longitudinal analysis of administrative data.

Area of Science:

• Epidemiology
• Network Science
• Health Services Research

Background:

• Multimorbidity presents significant analytical and clinical complexity due to intricate disease interactions.
• Understanding population-level disease co-occurrence is crucial for enhancing disease prevention, management, and healthcare delivery.

Purpose of the Study:

• To analyze multimorbidity patterns using a large-scale longitudinal cohort.
• To identify disease co-occurrence, prevalence, and association networks.
• To detect clusters of diseases representing potential multimorbidity profiles.

Main Methods:

• Utilized linked administrative data from 1,347,820 individuals in British Columbia, Canada, over 20 years.
• Employed a directed network-based approach to assess disease prevalence (frequency) and non-random associations (lift).
• Applied a community detection algorithm to identify multimorbidity disease clusters.

Main Results:

• Mood and anxiety disorders and hypertension were identified as prevalent disease predecessors, varying by age group.
• Lift networks highlighted significant non-random disease associations, suggesting potential etiological links, shared risk factors, or overlapping disease constructs.
• Identified disease clusters often centered on a single disease, indicating potential multimorbidity profiles for patient subgrouping.

Conclusions:

• The network-based analysis provides valuable insights complementing traditional surveillance methods.
• Flagging specific disease patterns can guide further research into their impact on patient function, mortality, and healthcare utilization.