Abstract
We identified three distinct self-monitoring of blood pressure (BP) groups among users of consumer-purchased wireless devices. While the majority of users exhibited sporadic engagement, nearly 10% sustained a daily use pattern that was significantly associated with lower BP levels over a 12-month period.
Introduction
Hypertension affects nearly half of all U.S. adults, yet approximately 75% of these individuals do not have their condition under control. While clinical guidelines recommend self-monitoring of blood pressure (SMBP) as a key management strategy, most research on its effectiveness is conducted within the constraints of highly controlled clinical trials.
In the real world, engagement with health technology often declines over time. This study sought to bridge the gap between clinical theory and naturalistic behavior by identifying long-term "usage phenotypes" among individuals using their own wireless monitors. By understanding these patterns, we can better determine if specific self-tracking behaviors correlate with improved physiological outcomes.
Methods
This research involved a secondary analysis of the Health eHeart Study, a prospective "eCohort." The study focused on a subsample of 2,099 individuals who owned a WiFi- or Bluetooth-enabled BP monitor (such as Withings, iHealth, or Qardio) and had at least one full year of observation data. Unlike traditional trials, participants received no specific instructions on device usage, allowing for the observation of naturalistic behavior.
Statistical Analysis
To identify distinct usage patterns, the team applied K-means clustering (via Python 3.7.8). The optimal number of groups was determined using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). Researchers then used multivariate multinomial logistic regression to identify predictors of group membership and linear mixed models to analyze blood pressure trajectories over 52 weeks, adjusting for demographic factors.
Results
Identification of Usage Phenotypes
The analysis identified three primary clusters of users based on their engagement over the year:
- Persistent Daily Use (9.1%): A small minority who maintained a consistent daily habit.
- Persistent Weekly Use (21.2%): Users who tracked their BP approximately once per week.
- Sporadic Use (69.7%): The vast majority of users, whose engagement was inconsistent and declined over time.
Predictors of Persistence
One of the strongest predictors of becoming a "persistent daily user" was the frequency of device use during Week 1. This suggests that the initial period of ownership is a critical window for establishing long-term habits. Additionally, daily users tended to be older, unemployed, and more likely to have existing comorbidities like diabetes or a history of myocardial infarction, suggesting that a higher perceived health threat may drive persistence.
Blood Pressure Outcomes
Engagement frequency was significantly associated with blood pressure levels. Over the 12-month period, the persistent daily use group maintained a mean systolic BP of 126.3 mmHg, compared to 128.8 mmHg for weekly users and 128.7 mmHg for sporadic users ($p < .001$). Furthermore, daily users had a significantly lower proportion of high-BP records (systolic > 130 mmHg) at 33.4%, compared to 43.0% for weekly users.
Discussion
These findings highlight a massive "engagement gap" in digital health; even when individuals purchase their own hardware, sustaining a monitoring habit is the exception rather than the rule. The data suggests that persistent daily monitoring is associated with better cardiovascular health management, as evidenced by lower average BP and fewer high-risk readings.
Limitations and Future Work
The study is limited by selection bias, as the sample was predominantly White, male, and well-educated. Furthermore, the results show correlation rather than causation; it is unclear if daily monitoring causes lower BP or if individuals with better-controlled BP are simply more motivated to track it. Future research should explore intervention strategies to help younger or lower-risk individuals stay engaged and investigate how to integrate this consumer-generated data into clinical workflows.
Conclusion
Identifying distinct usage phenotypes reveals that early behavior is highly predictive of long-term engagement. Encouraging daily self-monitoring habits from the outset may be a powerful tool in the long-term management of hypertension.
Summary updated