Date of Award

January 2025

Document Type

Thesis

Degree Name

Master of Public Health (MPH)

Department

School of Public Health

First Advisor

Shuangge Ma

Abstract

Recent research on human disease networks (HDNs) has revealed the importance of jointly analyzing multiple diseasesand their interconnections, yet many existing approaches focus on molecular-based data and do not fully capture non-molecular relationships. Building on the concept of the phenotypical HDN, we leverage Medicare inpatient claims data to construct a comprehensive disease co-occurrence network. In our framework, two diseases are deemed interconnected if their observed probability of co-occurrence significantly deviates from that expected under independence. We adopt the weighted gene co-expression network analysis (WGCNA) framework to build our network, using a ϕ-correlation measure and a hard threshold to ensure a scale-free topology that highlights hub diseases and the overall network structure. Unlike previous studies that emphasize static snapshots, our primary objective is to investigate how the connectivity structure of this network varies over time. To this end, we apply an Age–Period–Cohort (APC) effect framework to unravel the distinct influences of patient age, temporal period, and generational cohort on co-occurring disease patterns. By decomposing the overall temporal variation into these longitudinal factors, we capture the dynamic shifts in network connectivity that mirror changes in epidemiological patterns, diagnostic practices, and treatment guidelines. Furthermore, our analysis incorporates a data-driven clustering method that groups diseases based on the evolution of their functional connectivity curves, thereby revealing latent patterns that would be obscured in a purely cross-sectional analysis. Our work not only advances the construction of phenotypical HDNs beyond molecular ones and static ones but also provides critical insights for public health policymakers and clinicians. By integrating longitudinal analyses with robust network and clustering methodologies, we offer a comprehensive view of how disease interrelationships evolve across different demographic segments and time periods. These insights are pivotal for anticipating shifts in disease burdens, informing targeted intervention strategies, and ultimately improving clinical outcomes in the face of complex, evolving comorbidity patterns.

Comments

This thesis is restricted to Yale network users only. It will be made publicly available on 06/16/2026

Download

Share

COinS