Agreement between Fitbit Versa 4 and a reference standard on time-domain nocturnal heart rate variability: A pilot study

Abstract

Background

Reduced heart rate variability (HRV) has been associated with impaired autonomic nervous system (ANS) adaptation, suggesting that HRV could serve as a biomarker for identifying ANS imbalances that may adversely affect an individual's overall health. Many wearable devices used outside medical settings provide measures of HRV, but clinimetric data on their reliability and agreement remain limited. The primary aim of this study was to assess the ability of the Fitbit Versa 4, a consumer-based wearable, to measure nocturnal HRV compared to a reference standard HRV measure obtained from an ambulatory ECG device, the Firstbeat Bodyguard 3.

Methods

Eight participants simultaneously wore the Fitbit Versa 4 and Firstbeat Bodyguard 3 beginning at least one hour before sleep, throughout their sleep period, and for at least one hour after waking, over a minimum of three nights. HRV was quantified using the root mean square of successive differences (RMSSD). A t-test assessed differences in median HRV between devices, while a Linear Mixed Model and Bland-Altman analysis evaluated relative difference and agreement, respectively.

Results

The median RMSSD from Fitbit (21.660 ms [15.927, 22.821]) was not significantly different (p = 0.3858) from the median RMSSD from Firstbeat (20.558 ms [13.648, 21.828]. Linear Mixed Model results indicated that, on average, Firstbeat measurements were approximately 92.3% of those obtained from Fitbit.

Conclusions

Strong correlation and acceptable agreement between Fitbit Versa 4 HRV measurements and reference standard HRV measurements derived from ECG provide preliminary support for using Fitbit Versa 4 to measure nocturnal HRV outside laboratory settings.

Keywords

Heart rate variability chronic pain wearable technology autonomic nervous system sleep

Introduction

Heart rate variability (HRV) measures variation in the time interval between adjacent heartbeats and has become a widely used biomarker of autonomic nervous system (ANS) balance.¹ High variability in heart rate is associated with efficient adaptation of the ANS to physical and psychological challenges (sometimes called stress), whereas low variability in heart rate may indicate insufficient ANS adaptation to challenge and illness.² Resting HRV, as quantified by the root mean square of successive differences (RMSSD), ranges from less than 20 ms to over 70 ms in healthy adults.³ However, this variability is highly individual and can be influenced by several factors, including age, sex, physical fitness, and general health status. HRV parameters in both the time and frequency domains have been associated with cardiovascular morbidity and mortality,^4,5 diabetes,⁶ depression,^7,8 stress levels,^8,9 pain,^10,11 sleep quality,¹² and athletic performance.^13–15 Therefore, HRV measures have the potential to indicate imbalances in the ANS that may also affect an individual's overall health status and function.

Heart rate variability is measured using electrocardiogram (ECG) or photoplethysmography (PPG) applied to the chest and/or limbs.¹ While multi-lead ECG is considered the gold standard for assessment, this method does not provide for the continuous and routine collection of HRV data.¹⁶ Wearable devices utilizing ECG to measure HRV can be generally placed into three categories: wired chest monitors, clothing garments, and chest/shoulder/arm straps. Multiple systematic reviews suggest that wired chest monitors (e.g., Bittium FarosTM, Firstbeat Bodyguard 2, and Actiheart) demonstrate the best reliability and validity compared to multi-lead ECG, making these devices a suitable alternative to the gold standard.^17–19 The Firstbeat Bodyguard 2 in particular demonstrated a very strong correlation (r > 0.95, p < 0.001) for RMSSD when compared to a reference 2-lead ECG²⁰ and correctly detected 99.95% of all heartbeats, which improved to 99.98% when artifact correction was applied.²¹

Multiple studies have evaluated the reliability and validity of using rings or smart watches equipped with PPG to assess HRV compared to the gold standard of ECG. Apple, Samsung, and Garmin smart watches demonstrate acceptable accuracy for specific resting or nocturnal time-domain measures of HRV, but not for measures recorded during moderate to intense activity.^22,23 In another laboratory-based study designed to validate six wearable devices (i.e., Apple Watch S6, Garmin Forerunner 245 Music, Polar Vantage V, Oura Ring Generation 2, WHOOP 3.0 and Somfit) for assessing HRV against the gold-standard electrocardiogram, the authors found mixed results (ICCs ranging from 0.24 to 0.99).²⁴ All six devices underestimated HRV (RMSSD), with absolute bias ranging from 0.7 to 33.1 ms. These findings suggest that PPG recordings of HRV may be useful, but only when assessed in resting or nocturnal states due to the susceptibility of PPG to motion noise.^25,26

Fitbit is a smart watch utilizing PPG that can provide a measure of nocturnal HRV and is one of the most popular brands among consumers and researchers.^27–29 Despite its widespread use, only one study has investigated the clinimetrics of any Fitbit device for measuring HRV. Hermans et al. assessed the validity of the Fitbit Charge 4, using the Polar H10 chest strap as the reference standard, for measuring HRV over seven consecutive days (24 h per day) in patients with chronic obstructive pulmonary disease (COPD) and age- and sex-matched healthy controls.³⁰ The study reported a strong correlation between HRV measures from the Fitbit Charge 4 and the reference device in COPD patients (r = 0.83), and a moderate correlation in healthy controls (r = 0.64).²⁷ A key strength of this study was the inclusion of a matched healthy control group. A potential limitation of this study is that the Polar H10 is a relatively new model that has not been as rigorously evaluated for concurrent validity³¹ and requires the wearer to simultaneously don a chest strap and wristwatch to collect data.

The resting RMSSD measure reported by Fitbit is one of the most widely used time domain measures of HRV, and is best at describing the short-term beat-to-beat variations reflective of parasympathetic activity.³² The ability of Fitbit to track variations in parasympathetic activity could be particularly useful in patients with pain, who experience increases in heart rate, blood pressure, and respiratory rate when pain is elevated.³³ While evidence generally supports the collection of HRV at night and in a supine position using ECG,^34,35 definitive clinimetric studies are lacking on the ability of Fitbit to provide reliable and valid measures of nocturnal HRV, particularly in a sample of patients with pain. Therefore, the purpose of this investigation was to compare nocturnal HRV measures obtained from the Fitbit Versa 4 to a validated reference standard measure obtained from the Firstbeat Bodyguard 3 in a sample of adults with and without shoulder pain.

Methods

Participant population

Institutional Review Board approval (STU-2021-0495) was acquired prior to the inception of this study, and all experiments were conducted in accordance with the Helsinki Declaration.³⁶ All participants were fully informed of the experimental procedures and risks, and gave their written consent before any testing was conducted. Each participant was screened for eligibility via email or phone call. Participants were recruited as part of a larger study with inclusion criteria of: 1) 18 years of age or older and 2) scheduled to undergo a primary shoulder arthroplasty. Exclusion criteria were: 1) presence of central or peripheral neurological conditions impairing sensation, 2) non-English speaking, and 3) pregnant or may become pregnant during the duration of the study. Healthy controls were also recruited as part of the larger study (no history of shoulder pain or dysfunction in the preceding 36 months). Between July 2023 and January 2024, eight participants (seven with shoulder pain and one healthy control) were prospectively recruited and agreed to participate in the study. We performed a separate correlation analysis excluding the one healthy control participant, which yielded a similar result. Therefore, we included the control participant in the final analysis.

Measurements

Data was collected using one of three Fitbit Versa 4 (Google LLC, Mountain View, California, USA) watches and one of three Firstbeat Bodyguard 3 (Firstbeat Analytics, Jyväskylä, Finland) devices checked out to the participant. Participants were also given Firstbeat ECG electrodes and a Fitbit power charger. In addition to a comprehensive handout detailing operation instruction for both wearables, participants were given detailed verbal instructions on proper wearing techniques and wearing schedules. Participants were instructed to wear both devices concurrently for a minimum of one hour prior to bedtime, throughout the night, and for at least one hour upon waking, over the course of at least three nights. Nocturnal HRV was measured simultaneously on both devices for a total of 26 nights across 8 participants.

In this study, we focused on determining the agreement of RMSSD values reported by the Fitbit Versa 4 with those reported by Firstbeat Bodyguard 3. RMSSD is calculated by first determining the time intervals between consecutive heartbeats in milliseconds. Next, these intervals are squared, averaged, and finally, the square root of that average is taken to yield the RMSSD value.³² Fitbit provides to the user a singular measure of nocturnal HRV calculated as the median of 5-min RMSSD recordings collected during sleep. Nocturnal RMSSD is the only time-domain HRV metric produced and displayed by the device. Among the time-domain measures of HRV, RMSSD has the advantage of computation simplicity and reasonable accuracy for short and ultrashort recodings.³⁷ As a result, it is a preferred measure for the short-term recordings provided by wearable sensors.

Data were stored on the individual devices until they were returned via mail to the investigators, at which point Fitbit data was uploaded to the Fitbit cloud and accessed through Fitabase, and Firstbeat data were downloaded to a CSV file. ECG signals were sampled at 256 samples per second, allowing for high temporal resolution and thus accurate HRV computation.^38,39 The sampled ECG signals were transferred to Kubios HRV Software (version 4.1.0, Kubios Oy, Kuopio, Finland) to extract HRV (expressed as RMSSD in ms).⁴⁰ Kubios provides flexible preprocessing options for extracting RMSSD values from ECG data to enhance heartbeat detection. The Kubios automatic beat detection algorithm identifies artifacts based on differences between successive RR intervals using an adaptive (time-varying) threshold.⁴¹ Ectopic and misplaced beats are corrected via interpolation, while abnormally long or short RR intervals are adjusted by removing superfluous R-waves and recalculating intervals. A threshold-based correction also evaluates each RR interval against a median-filtered local average, where values deviating beyond a user-defined threshold are excluded. In this study, we applied the “Medium” setting, corresponding to a 0.25 s threshold. Additionally, Kubios applies detrending using the smoothness priors method to mitigate slow, nonstationary trends in heart rate, with a default cutoff frequency of 0.049 Hz, which was used in our analysis.⁴² The detrending helps avoid errors in HRV analysis due to slow nonstationary trends that may be present in the mean heart rate. This detrending approach has the effect of a time-varying filtering where the cut-off frequency can be adjusted by changing the level of a smoothing parameter.⁴²

The details of the preprocessing or any filtering of the optically detected heartbeats by Fitbit are not publicly available. Available documentation indicates that RMSSD is calculated from the longest sleep period over the past 24 h, and only sleep periods greater than 3 h are considered. As noted previously, Fitbit computes RMSSD values for 5-min intervals during sleep, but only when its proprietary algorithm deems the heart rate data acceptable. This is evidenced by occasional missing RMSSD values during sleep periods. Rather than relying on Fitbit's internal selection criteria, our analysis includes only the 5-min epochs for which RMSSD values are reported, with Firstbeat data aligned to these same intervals.

Statistical analysis

Normality was examined for continuous data by the Shapiro-Wilk test, and normal data were expressed as the mean plus/minus standard deviation. Non-normal data were expressed as median with first and third quartiles (Q1 and Q3). Wilcoxon signed-rank test was performed to compare the median RMSSD of the reference standard Firstbeat Bodyguard 3 with RMSSD measured by Fitbit Versa 4. Spearman correlation analysis was conducted to test the correlation between Fitbit and Firstbeat for the median RMSSD of each night. A linear mixed-effects model (LMM) with log transformation was used to investigate the difference between Firstbeat and Fitbit RMSSD 5-min epochs, accounting for the correlation among repeated measurements from the same subject. In an effort to address movement artifacts or variability in device fit across participants, we removed outliers from the final analysis that exceeded a threshold of ±3 standard deviations from the mean. We selected this threshold since values exceeding the mean by more than three standard deviations were supraphysiological and unlikely to represent genuine physiological fluctuations in heart rate. Thirty-two, or 1.51%, data points were ultimately removed from our final analysis.

Finally, to assess the agreement between the Fitbit and the Firstbeat median RMSSD, we performed a Bland-Altman analysis accounting for multiple observations per participant.^43,44 We calculated the mean of the difference between the Fitbit and Firstbeat-derived RMSSD values (bias), the standard deviation (SD) of the differences, and the 95% confidence interval of the agreement.⁴⁵ We also calculated the mean absolute difference between the Fitbit and Firstbeat-derived RMSSD values (absolute bias). For all tests, a p-value less than 0.05 was considered statistically significant. All the data analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

Detailed participant demographics and clinical characteristics can be found in Table 1. The median RMSSD from Fitbit was 21.660 ms [15.101, 30.671] and the median RMSSD from Firstbeat was 20.558 ms [12.741, 30.710], resulting in an absolute bias of 1.102 ms. There was no significant difference between Fitbit and Firstbeat median RMSSD values, p = 0.3858. Results of the Spearman rank correlation demonstrated a strong correlation between Fitbit median RMSSD and Firstbeat median RMSSD values for each night (r = 0.944; p < 0.0001). Following log transformation to satisfy assumptions of the linear mixed model, the results revealed, on average, a magnitude of difference of −0.07684 on the log scale, which corresponds to Firstbeat measurements being approximately 92.3% of those from Fitbit on average. This finding suggests that if Fitbit measured a hypothetical RMSSD 5-min epoch at 100 ms, Firstbeat would measure that same epoch at 92.3 ms. Although the difference is statistically significant (p < 0.0001), the relative difference in 5-min epochs is only 7.7%, indicating an interquartile range of 1.16 and 2.36 ms, which is unlikely to be clinically meaningful given the typical variability in heart rate measurements. Figure 1 shows a scatter plot of the 5-min RMSSD values recorded during deep sleep (n = 2178), and Figure 2 represents a scatter plot of the median RMSSD values of the entire night (n = 26). A Bland-Altman plot analysis shows acceptable agreement between Fitbit and Firstbeat, with only one median RMSSD value falling ±1.96 standard deviations outside the limit of agreement (Figure 3).

Figure 1.

Scatter plot of the 5-min median RMSSD values recorded during sleep. Each data point represents a 5-min epoch of median RMSSD measured throughout each night from all participants.

Figure 2.

Scatter plot of the Fitbit median RMSSD and Firstbeat median RMSSD for each night. Each data point represents the median RMSSD of an entire night of sleep for a participant.

Figure 3.

Bland-Altman plot showing the agreement between Fitbit and Firstbeat for measuring nocturnal HRV.

Table 1.

Demographic and clinical characteristics.

	Participants (n = 8)
Age (years)	64.13 ± 6.81
Female (n)	5
BMI (kg/m²)	28.18 ± 2.92
Height (inches)	66.14 ± 3.13
Weight (pounds)	176.43 ± 29.84
Ethnicity
- Hispanic	2
- Non-Hispanic	6
Race
- African American	1
- Caucasian	5
- Hispanic	2
Nocturnal Fitbit RMSSD (ms) (median [Q1, Q3])	21.660 [15.101, 30.671]
Nocturnal Firstbeat RMSSD (ms) (median [Q1, Q3]	20.558 [12.741, 30.710]
Difference in Median Fitbit and Firstbeat (ms)	1.102 (n = 2178)

Values are mean ± one standard deviation unless otherwise noted.

Abbreviations: BMI – Body Mass Index; RMSSD – Root mean square of successive differences; Q1 – First quartile; Q3 – Third quartile.

Discussion

The primary objective of this study was to compare nocturnal HRV measures obtained from the Fitbit Versa 4 and the Firstbeat Bodyguard 3. The results from this study demonstrated no significant difference and a strong correlation between devices for median nocturnal RMSSD values, as well as acceptable agreement based on the Bland Altman Plot. This finding is critical since the Fitbit user only sees the median nocturnal RMSSD value. As seen in Figure 1, after removal of the outliers, the 5-min epochs from the Fitbit Vera 4 appeared to be more susceptible to extreme values. One possible explanation for the extreme values in the Fitbit data is the placement on the participant's wrist, as opposed to being securely fastened to the chest like the Firstbeat device. This difference in positioning makes the Fitbit more susceptible to movement artifacts and may lead to decreased contact with the skin. Another explanation could be variability in the tightness of the Fitbit watch strap, potentially resulting in inconsistent contact with the skin.

The results of the present study are consistent with previous research conducted on older adults with chronic obstructive pulmonary disease (COPD), which found a strong correlation (r = 0.83) between HRV measures from a Fitbit Charge 4 and the reference device, Polar H10.³⁰ The authors concluded that the Fitbit Charge 4 significantly underestimated the average median HRV by 3 ms in patients with COPD compared to 7 ms in healthy patients. They hypothesized that the enhanced accuracy and stronger correlation observed in patients with COPD, compared to healthy controls, may be attributed to the accelerated aging and skin thinning associated with COPD, which could facilitate a more efficient reflection of light during photoplethysmographic measurements. Results from our analysis suggested that the Fitbit Versa 4 overestimated median HRV on average by 0.76 ms in patients with shoulder pain, which is smaller in magnitude but opposite in direction to the previous study. Although skin thinning did not likely impact our results since no patients in the present study had COPD, it is possible that subjects in both studies shared a mechanism of autonomic nervous system dysfunction. This suggestion is supported by subjects in both studies reporting the presence of chronic conditions and RMSSD values in the low 20 s.

Although medication effects, inflammation, and sleep disturbance are all likely to influence HRV, our results suggest that Fitbit is still able to provide accurate median nocturnal RMSSD values compared to a reference standard in those with suspected autonomic dysfunction. RMSSD is an established measure of parasympathetic nerve activity that is useful for tracking autonomic nervous system responses.⁴⁶ Jarczok et al., in a proof of concept study involving 9550 working adults from 19 study sites, have shown that reduced RMSSD can reflect elevated risk across a range of established cardiovascular risk factors, namely, hyperglycemia, hyperlipidemia, and inflammation.⁴⁷ Also, Schuman et al.⁴⁸ have reported that baseline RMSSD predicted symptom improvement in veterans suffering from post-traumatic stress disorder undergoing HRV biofeedback. Most recently, a study has shown the efficacy of RMSSD in tracking physiological stress.⁴⁹ These studies provide support for the use of clinical devices to track RMSSD in patients with autonomic dysfunction in a variety of conditions, and in a meaningful way, to use this data to impact patient prognosis or management decisions.

To the best of our knowledge, research evaluating the agreement of Fitbit to a reference device for measuring HRV is limited to the present study and a recent study by Hermans et al.³⁰ For other metrics, systematic reviews have reported that specific Fitbit models showed high inter-device reliability for steps, distance, energy expenditure, and sleep, and acceptable validity for measures of steps and time in bed or asleep.^50,51 Fitbit devices tend to overestimate daily step counts in cohorts with better functional exercise tolerance (e.g., healthy adults^50–52 or cancer survivors⁵³), but not in those with lower exercise tolerance (e.g., patients living with Parkinson's Disease⁵⁴ or COPD⁵⁵). In two studies validating Fitbit against reference measures for heart rate (HR), one found that the Fitbit significantly underestimated HR on average in a cohort of healthy adolesents,⁵⁶ while another found no significant difference in healthy adults.⁵⁷ Further research is needed to investigate the accuracy of step counts, HR, and HRV in patients with comorbidities affecting the autonomic nervous system, as accuracy could differ from that seen in healthy controls. Additionally, it is important to highlight that sex differences exist in the autonomic control of the heart, as reflected by heart rate variability measures.⁵⁸ Future studies should account for age, sex, and the presence of comorbidities affecting the autonomic nervous system when investigating the agreement and reliability of wearable devices for measuring HRV.

The present study is not without limitations. Our sample size was small (n = 8); however, our analysis included data over 26 nights, and post-hoc power analysis confirmed sufficient power, β = 0.84. The comparisons between the two devices are limited to HRV and cannot be generalized to other metrics in the absence of comparative data. Additionally, our study population consisted primarily of older adults with symptomatic shoulder pathology, limiting the generalizability of our findings. Further, because we did not sync the Fitbit watch to the cloud until the devices were returned to us, there were some technical issues related to data extraction from the Fitbit devices. For example, if the device at any point was fully drained of battery charge, data for an entire night of sleep was lost. To address this potential for missing data, we instructed participants to fully charge the device prior to returning the device via first-class mail. Finally, we used a validated medical-grade device (Firstbeat) as a reference standard for the Fitbit; however, even these devices are susceptible to measurement error outside of laboratory settings.

Conclusion

To our knowledge, this is the first study to examine both the correlation and agreement between the consumer-grade Fitbit Versa 4 wearable and a validated reference standard device for measuring nocturnal HRV. Our findings suggest that the Fitbit Versa 4 demonstrates acceptable agreement with Firstbeat Bodyguard 3 in a population of adults with shoulder pain. Future studies should explore the feasibility and acceptability of using physiological data collected from the Fitbit Versa 4 to inform management decisions in real-world settings.

Footnotes

Acknowledgment

We would like to thank Bansari Thakkar, PT, ScD, for her assistance with participant recruitment, scheduling, and data collection.

ORCID iDs

Andrew Nasr

Khosrow Behbehani

Christopher Pierson

Robin B. Jarrett

Ethical considerations

The approval of the Institutional Review Board of the University of Texas Southwestern Medical Center, Dallas, was acquired before the inception of this study, and all experiments were conducted in accordance with the Helsinki Declaration.

Author contributions

A. Nasr participated in patient enrollment and data collection, downloaded and processed the data, performed data analysis, and wrote the initial versions of the manuscript. K. Behbehani provided technical assistance in downloading and processing the data and collaborated in the writing of the manuscript versions. J. Wang provided advice on the statistical analyses and reviewed the results of these analyses as well as the manuscript drafts. C. Pierson participated in the development of the study methods, data collection and storage, and patient enrollment. T. L. Greer and R.B. Jarrett contributed to the discussions of the data analysis results and edited the manuscript drafts. J. Zafereo took part in the development of the study methods, patient enrollment and data collection, uploading and organizing the collected data, drafting the manuscript versions, as well as the review and discussion of the results of the data analyses.

Funding

This work was supported in part by the Southwestern School of Health Professions Collaborative Research Grant Program.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Guarantor

A. K. Behbehani

References

Cardoso

Moctezuma-Ramirez

, et al.

Heart rate variability measurement through a smart wearable device: another breakthrough for personal health monitoring?

Int J Environ Res Public Health 2023; 20: 7146.

Giovanni

. Root mean square of the successive differences as marker of the parasympathetic system and difference in the outcome after ANS stimulation. In: Theodoros

(eds) Autonomic nervous system monitoring. Rijeka: IntechOpen, 2020, Ch. 2.

Nunan

Sandercock

Brodie

. A quantitative systematic review of normal values for short-term heart rate variability in healthy adults. Pacing Clin Electrophysiol 2010; 33: 1407–1417.

Song

Zhang

, et al. Usefulness of the heart-rate variability complex for predicting cardiac mortality after acute myocardial infarction. BMC Cardiovasc Disord 2014; 14: 59.

Guzzetti

La Rovere

Pinna

, et al. Different spectral components of 24h heart rate variability are related to different modes of death in chronic heart failure. Eur Heart J 2005; 26: 357–362.

Benichou

Pereira

Mermillod

, et al. Heart rate variability in type 2 diabetes mellitus: a systematic review and meta-analysis. PLoS One 2018; 13: e0195166.

Glassman

Bigger

Gaffney

, et al. Heart rate variability in acute coronary syndrome patients with major depression: influence of sertraline and mood improvement. Arch Gen Psychiatry 2007; 64: 1025–1031.

Kemp

Quintana

Felmingham

, et al. Depression, comorbid anxiety disorders, and heart rate variability in physically healthy, unmedicated patients: implications for cardiovascular risk. PLoS One 2012; 7: e30777.

Hjortskov

Rissen

Blangsted

, et al. The effect of mental stress on heart rate variability and blood pressure during computer work. Eur J Appl Physiol 2004; 92: 84–89.

10.

Forte

Troisi

Pazzaglia

, et al. Heart rate variability and pain: a systematic review. Brain Sci 2022; 12: 153.

11.

Tracy

Ioannou

Baker

, et al. Meta-analytic evidence for decreased heart rate variability in chronic pain implicating parasympathetic nervous system dysregulation. Pain 2016; 157: 7–29.

12.

Burton

Rahman

Kadota

, et al. Reduced heart rate variability predicts poor sleep quality in a case-control study of chronic fatigue syndrome. Exp Brain Res 2010; 204: 71–78.

13.

Plews

Laursen

Kilding

, et al. Heart-rate variability and training-intensity distribution in elite rowers. Int J Sports Physiol Perform 2014; 9: 1026–1032.

14.

Chen

Yeh

Lee

, et al. Parasympathetic nervous activity mirrors recovery status in weightlifting performance after training. J Strength Cond Res 2011; 25: 1546–1552.

15.

Edmonds

Sinclair

Leicht

. Effect of a training week on heart rate variability in elite youth rugby league players. Int J Sports Med 2013; 34: 1087–1092.

16.

Billman

. Heart rate variability - a historical perspective. Front Physiol 2011; 2: 86.

17.

Palmer

Distefano

Leneman

, et al. Reliability of the BodyGuard2 (FirstBeat) in the detection of heart rate variability. Appl Psychophysiol Biofeedback 2021; 46: 251–258.

18.

Dobbs

Fedewa

MacDonald

, et al. The accuracy of acquiring heart rate variability from portable devices: a systematic review and meta-analysis. Sports Med 2019; 49: 417–435.

19.

Board

Ispoglou

Ingle

. Validity of telemetric-derived measures of heart rate variability : a systematic review. Journal of Exercise Physiology - Online 2016; 19: 64–84.

20.

Bogdány

Boros

Szemerszky

, et al. Validation of the firstbeat TeamBelt and BodyGuard2 systems. Magyar Sporttudományi Szemle 2016; 17: 5–12.

21.

Parak

Tarniceriu

Renevey

, et al. Evaluation of the beat-to-beat detection accuracy of PulseOn wearable optical heart rate monitor. Annu Int Conf IEEE Eng Med Biol Soc 2015 2015: 8099–8102. DOI: 10.1109/EMBC.2015.7320273

22.

Theurl

Schreinlechner

Sappler

, et al. Smartwatch-derived heart rate variability: a head-to-head comparison with the gold standard in cardiovascular disease. Eur Heart J Digit Health 2023; 4: 155–164.

23.

Sarhaddi

Kazemi

Azimi

, et al. A comprehensive accuracy assessment of samsung smartwatch heart rate and heart rate variability. PLoS One 2022; 17: e0268361.

24.

Miller

Sargent

Roach

. A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults. Sensors (Basel) 2022; 22: 6317.

25.

Georgiou

Larentzakis

Khamis

, et al. Can wearable devices accurately measure heart rate variability? A systematic review. Folia Med (Plovdiv) 2018; 60: 7–20.

26.

Barrios

Oldrati

Santini

, et al. Evaluating the accuracy of heart rate sensors based on photoplethysmography for in-the-wild analysis. In: Proceedings of the 13th EAI international conference on pervasive computing technologies for healthcare. Trento, Italy: Association for Computing Machinery, 2019, pp.251–261.

27.

https://www.statista.com/forecasts/997195/most-used-ehealth-tracker-smart-watches-by-brand-in-the-us (2024, accessed 08-14-2024).

28.

Feehan

Geldman

Sayre

, et al. Accuracy of Fitbit devices: systematic review and narrative syntheses of quantitative data. JMIR Mhealth Uhealth 2018; 6: e10527.

29.

Evenson

Goto

Furberg

. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act 2015; 12: 159.

30.

Hermans

Arents

Blondeel

, et al. Validity of a consumer-based wearable to measure clinical parameters in patients with chronic obstructive pulmonary disease and healthy controls: observational study. JMIR Mhealth Uhealth 2024; 12: e56027.

31.

Hinde

White

Armstrong

. Wearable devices suitable for monitoring twenty four hour heart rate variability in military populations. Sensors (Basel) 2021; 21: 1061.

32.

Shaffer

Ginsberg

. An overview of heart rate variability metrics and norms. Front Public Health 2017; 5: 58.

33.

Yeater

Cruz

Cruz-Almeida

, et al. Autonomic nervous system dysregulation and osteoarthritis pain: mechanisms, measurement, and future outlook. Curr Rheumatol Rep 2022; 24: 175–183.

34.

Kerkering

Greenlund

Bigalke

, et al. Reliability of heart rate variability during stable and disrupted polysomnographic sleep. Am J Physiol Heart Circ Physiol 2022; 323: H16–H23.

35.

Chuangchai

Pothisiri

. Postural changes on heart rate variability among older population: a preliminary study. Curr Gerontol Geriatr Res 2021; 2021: 6611479.

36.

Schuklenk

. Helsinki Declaration revisions. Issues Med Ethics 2001; 9: 29.

37.

Munoz

van Roon

Riese

, et al. Validity of (ultra-)short recordings for heart rate variability measurements. PLOS ONE 2015; 10: e0138921.

38.

Zhou

Lindsey

Snyder

, et al. Sampling rate requirement for accurate calculation of heart rate and its variability based on the electrocardiogram. Physiol Meas 2024; 45: 025007.

39.

Abboud

Barnea

. Errors due to sampling frequency of the electrocardiogram in spectral analysis of heart rate signals with low variability. In: Computers in Cardiology 1995; 1995: 461–463.

40.

Tarvainen

Niskanen

Lipponen

, et al. Kubios HRV–heart rate variability analysis software. Comput Methods Programs Biomed 2014; 113: 210–220.

41.

Lipponen

Tarvainen

. A robust algorithm for heart rate variability time series artefact correction using novel beat classification. J Med Eng Technol 2019; 43: 173–181.

42.

Tarvainen

Ranta-aho

Karjalainen

. An advanced detrending method with application to HRV analysis. IEEE Trans Biomed Eng 2002; 49: 172–175.

43.

Carstensen

Simpson

Gurrin

. Statistical models for assessing agreement in method comparison studies with replicate measurements. Int J Biostat 2008; 4: Article 16.

44.

Bland

Altman

. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 2007; 17: 571–582.

45.

Bland

Altman

. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–310.

46.

Minarini

. Root mean square of the successive differences as marker of the parasympathetic system and difference in the outcome after ANS stimulation. In: Aslanidis

(ed.) Autonomic Nervous System Monitoring - Heart Rate Variability. Rijeka: IntechOpen, 2020, pp.1–14.

47.

Jarczok

Koenig

Wittling

, et al. First evaluation of an Index of low vagally-mediated heart rate variability as a marker of health risks in human adults: proof of concept. J Clin Med 2019; 8: 1940.

48.

Schuman

Ognyanova

Ginsberg

, et al. Brief report: baseline HRV time domain parameters predict trauma and depression symptom change in veterans with PTSD undergoing biofeedback. Appl Psychophysiol Biofeedback 2024; 49: 629–635.

49.

Paula Levi

Chance Nicholson

Heaton

, et al. Intensive care unit nursing stress: measuring heart rate variability and workplace trauma. Nursing Research and Reviews 2025; 15: 91–109.

50.

https://www.fitabase.com/research-library/ (accessed 14 August 2024).

51.

Tedesco

Sica

Ancillao

, et al. Validity evaluation of the fitbit Charge2 and the garmin vivosmart HR+ in free-living environments in an older adult cohort. JMIR Mhealth Uhealth 2019; 7: e13084.

52.

Farina

Lowry

. The validity of consumer-level activity monitors in healthy older adults in free-living conditions. J Aging Phys Act 2018; 26: 128–135.

53.

De Groef

Asnong

Blondeel

, et al. Accuracy of consumer-based activity trackers as measuring tool and coaching device in breast and colorectal cancer survivors. Support Care Cancer 2023; 31: 96.

54.

Ginis

Goris

De Groef

, et al. Validation of commercial activity trackers in everyday life of people with Parkinson's disease. Sensors (Basel) 2023; 23: 4156.

55.

Blondeel

Demeyer

Janssens

, et al. Accuracy of consumer-based activity trackers as measuring tool and coaching device in patients with COPD and healthy controls. PLoS One 2020; 15: e0236676.

56.

de Zambotti

Baker

Willoughby

, et al. Measures of sleep and cardiac functioning during sleep using a multi-sensory commercially-available wristband in adolescents. Physiol Behav 2016; 158: 143–149.

57.

Haghayegh

Khoshnevis

Smolensky

, et al. Accuracy of PurePulse photoplethysmography technology of fitbit charge 2 for assessment of heart rate during sleep. Chronobiol Int 2019; 36: 927–933.

58.

Koenig

Thayer

. Sex differences in healthy human heart rate variability: a meta-analysis. Neurosci Biobehav Rev 2016; 64: 288–310.