JMIR Publications

ABSTRACT

Background:

Digital health applications generate rich behavioral data, yet how users transition through behavioral states remains poorly understood. Existing approaches show limited capacity to capture and represent the evolving, nuanced nature of real-world behavioral dynamics, which is essential for informing personalized behavior change interventions.

Objective:

This study aims to understand how users’ behaviors evolve over time within digital health applications by developing a data-driven approach that captures transition dynamics across behavioral features. We also aim to validate the modeled transitions against observed user data, analyze feature transition pathways, and derive interpretable insights.

Methods:

We analyzed 32 weeks of data from 36,574 users from a population health programme run by the Health Promotion Board (HPB) in Singapore. We developed GraphBeTraM, a graph-based framework that models behavioral transitions as shortest paths through user similarity graphs. By aggregating paths over time, it captures users’ progression toward target behavioral states and quantifies direction, magnitude, and timing of change. We validated modeled transitions against real-world data by comparing (a) transition matrices and (b) feature-level pathways, using Spearman correlation, Fisher z-transformed means, and cosine distance. We focused on high-variance features and used heatmaps to visualize patterns of change. Finally, we characterized transitions between behavioral states, and distinguished generalizable and context-specific features along transition pathways.

Results:

We observed strong alignment between observed and modeled transition matrices with very strong correlation (Spearman ρ = 0.82, P < .001) and low cosine distance (0.05), which suggests that real-world behavioral transitions can be effectively represented through shortest paths with GraphBeTram. Feature-level pathway validation showed consistent patterns of change across both personalized and real-world pathways. Physical activity features such as weekly Moderate to Vigorous Physical Activity (MVPA) emerged as stable generalizable signals of change on a population level, with strong mean correlation (ρ_personalized=0.97; ρ_real-world=0.79) and low mean cosine distance (cosine_personalized=0.24; cosine_real-world=0.41) across both personalized and real-world transitions. In contrast, features related to personal preferences, including temporal activity patterns (e.g., proportion of weekday to weekend MVPA), purchase preferences for healthy foods/drinks, and engagement indicators (e.g, last contact with the programme/app) exhibited context-specific relevance and reflected change at a more individual level. An in-depth transition analysis between distinct behavioral states, like from an active state with prolonged sedentary periods to an active state with healthy eating habits, revealed interpretable patterns in onset, magnitude, and rate of change. Specifically, certain features like preferences for healthy food/drink purchases emerged later in the trajectory but then showed sharper and faster transitions once these behaviors began to shift. These insights highlight how GraphBeTraM can guide nudging strategies in alignment with natural behavior change dynamics, including both stable and later-onset patterns.

Conclusions:

GraphBeTraM provides a structured, interpretable framework for modeling behavioral transitions in digital health applications and captures key dynamics that support personalized intervention design.