Continuous Monitoring of Heart Rate Variability in Free-Living Conditions Using Wearable Sensors: Exploratory Observational Study

Abstract

Background: Wearable physiological monitoring devices are promising tools for remote monitoring and early detection of potential health changes of interest. The widespread adoption of such an approach across communities and over long periods of time will require an automated data platform for collecting, processing, and analyzing relevant health information.

Objective: In this study, we explore prospective monitoring of individual health through an automated data collection, metrics extraction, and health anomaly analysis pipeline in free-living conditions over a continuous monitoring period of several months with a focus on viral respiratory infections, such as influenza or COVID-19.

Methods: A total of 59 participants provided smartwatch data and health symptom and illness reports daily over an 8-month window. Physiological and activity data from photoplethysmography sensors, including high-resolution interbeat interval (IBI) and step counts, were uploaded directly from Garmin Fenix 6 smartwatches and processed automatically in the cloud using a stand-alone, open-source analytical engine. Health risk scores were computed based on a deviation in heart rate and heart rate variability metrics from each individual's activity-matched baseline values, and scores exceeding a predefined threshold were checked for corresponding symptoms or illness reports. Conversely, reports of viral respiratory illnesses in health survey responses were also checked for corresponding changes in health risk scores to qualitatively assess the risk score as an indicator of acute respiratory health anomalies.

Results: The median average percentage of sensor data provided per day indicating smartwatch wear compliance was 70%, and survey responses indicating health reporting compliance was 46%. A total of 29 elevated health risk scores were detected, of which 12 (41%) had concurrent survey data and indicated a health symptom or illness. A total of 21 influenza or COVID-19 illnesses were reported by study participants; 9 (43%) of these reports had concurrent smartwatch data, of which 6 (67%) had an increase in health risk score.

Conclusions: We demonstrate a protocol for data collection, extraction of heart rate and heart rate variability metrics, and prospective analysis that is compatible with near real-time health assessment using wearable sensors for continuous monitoring. The modular platform for data collection and analysis allows for a choice of different wearable sensors and algorithms. Here, we demonstrate its implementation in the collection of high-fidelity IBI data from Garmin Fenix 6 smartwatches worn by individuals in free-living conditions, and the prospective, near real-time analysis of the data, culminating in the calculation of health risk scores. To our knowledge, this study demonstrates for the first time the feasibility of measuring high-resolution heart IBI and step count using smartwatches in near real time for respiratory illness detection over a long-term monitoring period in free-living conditions.

Keywords: PPG; community; data collection; data platform; devices; health risk; heart rate; heart rate variability; monitoring; observation study; photoplethysmography; physiological; physiological monitoring; remote monitoring; sensor; sensors; smartwatch; wearable; wearable devices; wearable sensors; wearables.

Figure 1

Data collection and algorithm workflow. The smartwatch is paired to a smartphone app, and data are uploaded to the data server. Data cleaning corrects sensor-specific artifacts and removes periods where data quality is poor or there is not enough data to make a representative calculation. Metrics are standardized by computing a z score, which corrects for physical activity using an individual’s healthy baseline data and eliminates between-person variation. The output of the algorithm is a risk score.

Figure 2

RDF per participant for a convenience sample of 59 individuals who provided smartwatch data between January 2023 and the end of August 2023. Labels on the y-axis indicate the average RDF over the 8-month monitoring period per participant. Each row illustrates daily RDF values per participant. Symbols × indicate participants who left the study and + indicate participants who joined in the last 15 weeks of the study. RDF: raw data fraction.

Figure 3

VDF per participant for a convenience sample of 59 individuals who provided smartwatch data between January 2023 and the end of August 2023. Labels on the y-axis indicate the average VDF over the 8-month monitoring period per participant. Each row illustrates daily VDF values per participant. The participant order along the y-axis corresponds to the y-axis participant ordering shown in Figure 2. Symbols × indicate participants who left the study and + indicate participants who joined in the last 15 weeks of the study. VDF: valid data fraction.

Figure 4

IBI ADF per participant for a convenience sample of 59 individuals who provided smartwatch data between January 2023 and the end of August 2023. Labels on the y-axis indicate the average ADF calculated from the data measured over the 8-month monitoring period per participant. Each row illustrates daily ADF values per participant. The participant order along the y-axis corresponds to the y-axis participant ordering shown in Figure 2. Symbols × indicate participants who left the study and + indicate participants who joined in the last 15 weeks of the study. ADF: artifact data fraction; IBI: interbeat interval.

Figure 5

Self-reported health symptoms and illness SDF per participant for a convenience sample of 59 individuals who provided data between January 2023 and the end of August 2023. Labels on the y-axis indicate the percentage of daily survey reports submitted relative to the number of days in the 8-month monitoring period per participant. The color map indicates symptom survey reports submitted per day per participant. The participant order along the y-axis corresponds to the y-axis participant ordering shown in Figure 2. Symbols × indicate participants who left the study and + indicate participants who joined in the last 15 weeks of the study. SDF: survey data fraction.

Figure 6

Example metrics measured from a participant over a period of 2 weeks in April 2023. All values were calculated in 5-minute epochs except symptom scores, which were calculated from self-report of individual symptoms. RDF and VDF are averaged by hour. The risk score remained below the threshold (dotted gray line) and symptom scores were zero for general and respiratory symptom categories throughout the 2 weeks. The risk scores were detected from analysis of heart rate variability-derived and activity metrics. The symptom scores were calculated from symptom ratings on a scale of 0 (no symptoms) to 3 (severe symptoms) for the following categories: allergies, runny nose, sore throat, cough, shortness of breath, fever, fatigue, headache, body ache, loss of taste and smell, and gastrointestinal symptoms. ADF: artifact data fraction (calculated for IBI metric); HF: log of high-frequency power; IBI: interbeat interval; LF: log of low-frequency power; RDF: raw data fraction; RMSSD: root mean square of consecutive heartbeat intervals; SDNN: standard deviation of IBI; Step: step count; VDF: valid data fraction; z: standardized metric.

Figure 7

The number of influenza and COVID-19 illness and elevated risk score events observed by month and normalized US county metrics showing the percent of wastewater samples with detectable COVID-19 virus (noise added). Elevated risk scores were detected from analysis of heart rate variability-derived and activity metrics of a convenience sample of 59 participants who provided smartwatch and health survey data between January 2023 and the end of August 2023. Influenza and COVID-19 illnesses were self-reported by participants through daily surveys and may or may not have been ascertained by diagnostic tests.

Figure 8

Metrics from a participant who initially reported influenza and ultimately tested positive for COVID-19 in August 2023. Elevated risk scores relative to a threshold of 15 (dotted gray line) are observed prior to and coinciding with reported symptoms. The risk scores were detected from analysis of heart rate variability-derived and activity metrics. Plots of step count, IBI, RDF, VDF, IBI ADF, standardized IBI metrics, risk score, and health survey symptom score are shown for a period of 1 month. Markers indicating self-reported influenza and COVID-19 illnesses are included alongside numeric scores of general (any type) and respiratory-related symptoms. The symptom scores were calculated from symptom ratings on a scale of 0 (no symptoms) to 3 (severe symptoms) for the following categories: allergies, runny nose, sore throat, cough, shortness of breath, fever, fatigue, headache, body ache, loss of taste and smell, and gastrointestinal symptoms. ADF: artifact data fraction; HF: log of high-frequency power; IBI: interbeat interval; IBI: interbeat interval; LF: log of low-frequency power; RDF: raw data fraction; VDF: valid data fraction; z: standardized metric.