Exploring User Behavior, Profiles, and Generation of Missed Reading Alerts in Long-Term Users of a Technology-Enabled Intervention for Self-Monitoring of Blood Pressure in Public Primary Care Setting in Singapore: Longitudinal Observational Study

Shilpa Tyagi, Keith Chiaw Meng Sng, David Wei Liang Ng, Valerie Teo, Chun Yen Beh, Evon Oh, Jeremy Cong En He, Siyu Heng, Gerald Choon‐Huat KohJournal of Medical Internet Research 2025 · DOI

TL;DR

This study investigates user behavior and identifies distinct profiles of long-term users in a nationwide telehealth program for hypertension management. It reveals that younger age and longer program duration are associated with generating more "missed reading" alerts, and proposes tailored interventions based on identified user profiles to improve engagement and blood pressure control.

What Problem Does This Paper Address?

Telehealth interventions for chronic disease management, like self-monitoring of blood pressure (BP), show promise but often struggle with sustained user engagement and high attrition rates. Understanding why users disengage or generate alerts for missed readings is crucial for designing more effective and sustainable digital health solutions. This paper specifically addresses the challenge of understanding long-term user behavior within a large-scale, real-world telehealth program for hypertension, focusing on the generation of "Missed Reading" (MR) alerts and identifying different user profiles to inform better intervention design.

Methods

The researchers employed a longitudinal observational study design, analyzing backend data from the Primary Technology Enhanced Care for Hypertension Program (PTEC-HT) in Singapore. Key methods included:

  • Logistic Regression: Used to identify factors associated with generating MR alerts during a specific "index month."
  • Latent Class Analysis (LCA): A statistical technique to identify unobserved subgroups (latent classes) of participants based on their observed behaviors (MR alert generation, BP control status, and submission frequency). This allowed for profiling distinct user types.
  • Generalized Estimating Equations (GEE): Applied to analyze temporal trajectories of MR reminders and their conversion rates over a 12-month period, accounting for repeated measures within participants.

Datasets

The study utilized backend data from the PTEC-HT system, a nationwide telehealth initiative implemented across public primary care clinics in Singapore. The dataset comprised information from 491 participants who had been enrolled in the program for at least one year. The data covered a 12-month observation period preceding the index month of August 2023, with specific segments used for different analyses (e.g., a 6-month segment for LCA). No new benchmarks were introduced; the study leveraged existing program data.

Key Insights and Ideas

  • Age and Duration Matter: Younger participants and those with longer durations on the PTEC-HT program were more likely to generate MR alerts. This suggests that while the program aims for long-term engagement, adherence patterns might shift over time, with younger users potentially needing more proactive support.
  • Distinct User Profiles Emerge: The LCA identified three key user profiles:
  • Compliant Triers: Low MR alerts but poor BP control. These users are adhering to the system but not achieving desired health outcomes.
  • Compliant Achievers: Low MR alerts and good BP control. This is the largest group, representing successful engagement and outcomes.
  • Non-Compliant Achievers: High MR alerts but good BP control. These users are achieving targets despite generating many alerts, suggesting a potential disconnect between alert generation and actual adherence or a need for different engagement strategies.
  • Alert Generators vs. Non-Generators: Users who generated MR alerts consistently had higher counts of MR reminder messages but lower conversion rates (i.e., they were less likely to submit a reading after receiving a reminder) compared to those who did not generate alerts. This points to potential "alert fatigue" or reduced responsiveness to repeated notifications.

Key Results

  • Age: Younger participants (OR 0.97, 95% CI 0.95-0.99) were significantly more likely to generate MR alerts.
  • Program Duration: Each additional month on the PTEC-HT program increased the odds of generating an MR alert by 11% (OR 1.11, 95% CI 1.01-1.22).
  • User Profiles:
    • "Compliant Achievers" were the most prevalent group (74.9%).
  • "Non-Compliant Achievers" had the highest proportion of MR alert generators (70.1%).
  • "Compliant Triers" showed a consistent trend of low MR alert probability and low BP control.
  • Temporal Trajectories: MR alert generators consistently received more MRRA and MRRB messages but had lower conversion rates over the 12 months preceding the index month.

Limitations and Future Work

The authors acknowledge several limitations:

  • Psychosocial Factors: The study did not capture psychosocial factors like patient motivation, technology literacy, or perceptions of telehealth, which are known to influence adherence. This is a limitation of relying solely on system usage data.
  • Sample Size Calculation: A formal sample size calculation was not performed, as data from all eligible participants were extracted.
  • External Factors: The analysis did not adjust for external factors such as policy changes or participant life events that could influence engagement.
  • Generalizability: Findings may not be fully generalizable to healthcare systems with different infrastructures, patient demographics, or care delivery models, as the study was conducted in Singapore.

Future work suggested includes:

  • Qualitative Research: Exploring the reasons behind declining engagement and alert fatigue through qualitative studies.
  • Temporal Analysis of Compliance: Investigating whether compliance with BP submission predicts sustained BP control over longer periods.
  • Tailored Interventions: Developing and testing interventions specifically designed for each identified user profile (e.g., intensified coaching for "Compliant Triers," sustained engagement for "Compliant Achievers," and adherence-focused strategies for "Non-Compliant Achievers").
  • Optimizing Messaging: Refining message content, channels, and frequency to improve conversion rates and combat alert fatigue.

Summary updated

Abstract

Our study reported that younger participants and longer program durations were linked to higher MR alert generation. The identification of distinct user profiles suggests that tailored intervention features could enhance engagement and BP control. The study underscores the importance of monitoring compliance patterns and optimizing message content to improve conversion rates. These insights contribute to the understanding of telehealth engagement dynamics and support targeted interventions for hypertension management.
BibTeX
@article{tyagi2025exploring,
  author = {Shilpa Tyagi and Keith Chiaw Meng Sng and David Wei Liang Ng and Valerie Teo and Chun Yen Beh and Evon Oh and Jeremy Cong En He and Siyu Heng and Gerald Choon‐Huat Koh},
  title = {{Exploring User Behavior, Profiles, and Generation of Missed Reading Alerts in Long-Term Users of a Technology-Enabled Intervention for Self-Monitoring of Blood Pressure in Public Primary Care Setting in Singapore: Longitudinal Observational Study}},
  journal = {Journal of Medical Internet Research},
  year = {2025},
  doi = {10.2196/74051},
}
Figure