Original Research

Digital Health
Volume 8: 1–9
Commercial smartwatch with pulse oximeter © The Author(s) 2022
Article reuse guidelines:
detects short-time hypoxemia as well as sagepub.com/journals-permissions
DOI: 10.1177/20552076221132127
standard medical-grade device: Validation study journals.sagepub.com/home/dhj

Jakub Rafl1 , Thomas E Bachman1, Veronika Rafl-Huttova1 ,

Simon Walzel1 and Martin Rozanek1

Abstract

Objective: We investigated how a commercially available smartwatch that measures peripheral blood oxygen saturation
(SpO2) can detect hypoxemia compared to a medical-grade pulse oximeter.
Methods: We recruited 24 healthy participants. Each participant wore a smartwatch (Apple Watch Series 6) on the left wrist
and a pulse oximeter sensor (Masimo Radical-7) on the left middle finger. The participants breathed via a breathing circuit
with a three-way non-rebreathing valve in three phases. First, in the 2-minute initial stabilization phase, the participants
inhaled the ambient air. Then in the 5-minute desaturation phase, the participants breathed the oxygen-reduced gas mixture
(12% O2), which temporarily reduced their blood oxygen saturation. In the final stabilization phase, the participants inhaled
the ambient air again until SpO2 returned to normal values. Measurements of SpO2 were taken from the smartwatch and the
pulse oximeter simultaneously in 30-s intervals.
Results: There were 642 individual pairs of SpO2 measurements. The bias in SpO2 between the smartwatch and the oximeter was 0.0%
for all the data points. The bias for SpO2 less than 90% was 1.2%. The differences in individual measurements between the smartwatch
and oximeter within 6% SpO2 can be expected for SpO2 readings 90%–100% and up to 8% for SpO2 readings less than 90%.
Conclusions: Apple Watch Series 6 can reliably detect states of reduced blood oxygen saturation with SpO2 below 90% when
compared to a medical-grade pulse oximeter. The technology used in this smartwatch is sufficiently advanced for the indi-
cative measurement of SpO2 outside the clinic.
Trial Registration: ClinicalTrials.gov NCT04780724

Keywords

Wearables, oxygen saturation, pulse oximetry, reflectance mode, hypoxemia, hypoxic gas mixture, Apple Watch
Submission date: 17 March 2022; Acceptance date: 22 September 2022

Introduction
Recently, consumer wearables have created the vision of new Earlier feasibility studies focused on activity monitoring and
possibilities for personal care.1–5 Routine monitoring of bio-
logical signals such as heart rate or sleep pattern using wearable
1
devices is an emerging trend in health monitoring outside the Department of Biomedical Technology, Faculty of Biomedical Engineering,
Czech Technical University in Prague, Kladno, Czech Republic
clinic and in-home care with a multi-billion dollar potential.6,7
The COVID-19 pandemic and its aftermath will only emphasize Corresponding author:
Jakub Rafl, Department of Biomedical Technology, Faculty of Biomedical
this trend.8,9 Nevertheless, the clinical applicability of wearables Engineering, Czech Technical University in Prague, nam. Sitna 3105, CZ-272
must be separated from consumer curiosity.10–14 Currently, the 01 Kladno, Czech Republic.
role of smartwatches in health care is investigated and discussed. Email: rafl@fbmi.cvut.cz

Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial
4.0 License (https://creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work
without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 DIGITAL HEALTH

chronic disease self-management.15,16 Recent prospective error.36 Similarly, a study with a wrist-worn reflectance pulse
studies have looked at the use of smartwatch technology in oximeter under development found its performance was worse
a range of medical applications such as the detection of than finger-based oximeters and it was not able to detect hypox-
atrial fibrillation,17,18 sleep monitoring,19 post-admission emia.37 Also, Hermand et al. reported a commercial smartwatch
recovery in pediatric patients with respiratory diseases,20 failed to provide trustworthy SpO2 values, especially during
monitoring women during pregnancy21 or pre-habilitation induced oxygen desaturation.38 On the other hand, a study by
prior to abdominal cancer surgery.22 Several studies were Lauterbach et al. tested a commercial smartwatch in a normoba-
also interested in using smartwatch data in the detection of ric hypoxia chamber and found only minimal differences in
viral infections such as COVID-19.23,24 However, a previous SpO2 measured by the smartwatch compared to a standard
study warned that a smartwatch did not have sufficient accur- pulse oximeter,26 with the largest difference for the lowest
acy in measuring blood pressure or pulse oximetry compared inspiratory oxygen fraction. Other recent studies have also
to clinical standards.13 reported positive results on the accuracy of wrist SpO2 measure-
Pulse oximetry as a method of indirect measurement of per- ments by commercial devices, but most of them did not focus on
ipheral blood oxygen saturation (SpO2) is a relatively new metric hypoxia.39–42
in smartwatches, but it is becoming routinely available in new Thus, there are currently a few studies available that evaluate
models,25 allowing convenient SpO2 monitoring at home or, wrist SpO2 measurement with mixed results. Concerns about
with some restrictions due to movement, outdoors without the measurement accuracy remain and, as new smartwatch models
need for a dedicated pulse oximeter. In addition, the smart- are launched, further studies are desirable.27 A question persists
watch’s SpO2 sensor does not need to be attached to a finger whether wrist-worn devices, and smartwatches in particular, can
to complicate daily activities. This might be useful not only to monitor SpO2 even in low blood oxygen levels well enough to
athletes in training or mountaineers in high altitudes but more provide early warning of desaturation episodes.
importantly to patients suffering from cardiovascular diseases, This study aims to compare the measurement of periph-
lung diseases such as chronic obstructive pulmonary disease eral blood oxygen saturation using a very popular smart-
(COPD), or dealing with the consequences or concerns of watch to a medical-grade pulse oximeter at normal and
COVID-19.26,27 In particular, the ability of smartwatches to potentially hypoxic levels.
measure SpO2 without conscious user intervention might help
to detect intermittent hypoxemia associated with sleep apnea, a
chronic health disorder that results in neurocognitive dysfunction Methods
and cardiovascular problems.28–30 The prospective single-arm interventional study was
Pulse oximetry is an optical method that evaluates approved by the Ethical Review Board of the Faculty of
changes in light absorption at multiple frequencies due to Biomedical Engineering, Czech Technical University in
the oxygen content in arterial blood. Levels of SpO2 95% Prague (No. B1/2021). The study was registered with
or higher are considered normal, whereas SpO2 below ClinicalTrials.gov (identifier NCT04780724).
90%, even if transient, is considered clinically relevant.31
Standard medical pulse oximeters, including portable oxi-
meters, use transmission pulse oximetry, in which the Recruitment
light sources and the photodetector are positioned on the Twenty-four healthy student volunteers (mean ± SD: age 24
opposite sides of the measurement site (usually a thin ± 2 years, height 181 ± 8 cm, mass 77 ± 11 kg) were
place such as a fingertip or an earlobe) and the light recruited for the study. They were only included if they
passing through the site is evaluated. Smartwatches, and did not suffer from any disease of the cardiovascular
other wrist-worn devices, for practical reasons, utilize system and had no injury to the upper limbs or hands that
reflectance pulse oximetry, in which the light sources and could affect the peripheral perfusion. In addition, partici-
the photodetector are positioned on the same side of the pants were excluded for pregnancy, diabetes, hypotension,
measurement site and the light reflected into the photo- hypertension, acute asthma or any other acute respiratory
detector from the tissue is evaluated. Reflectance pulse oxi- disease. None of the participants used nail polish or had
meters face less light absorption and thus have less power false nails at the time of the measurement. Participants
consumption, can be placed at diverse measurement loca- were required to stay at least 30 min at rest before entering
tions, and the absence of moving parts increases their resist- the laboratory. All participants provided written informed
ance to motion artifacts.32,33 However, in practice, the consent before their enrollment into the study.
reflectance mode can exhibit a low signal-to-noise ratio
and be sensitive to ambient light sources.34 At the wrist,
the performance of the reflectance pulse oximeter depends Experiment setup and protocol
on the exact placement of the sensor.34,35 In a study with Upon arrival at the laboratory, Apple Watch Series 6 (Apple
an experimental reflectance pulse oximeter system, SpO2 Inc., Cupertino, CA, USA)—further referred to as the
measurement at the wrist showed an unacceptably large smartwatch—was placed on a participant’s left wrist and
Rafl et al. 3

the sensor of a medical-grade pulse oximeter Radical-7 simultaneous smartwatch and oximeter measurements.
(Masimo Corp., Irvine, CA, USA)—further referred to as Uncertainties in the estimates of the bias and 95% limits
the oximeter—was attached to the left middle finger of of agreement are expressed as 95% confidence intervals.
the participant. During the experimental procedure, SpO2 The standard deviation was calculated using the modified
readings were taken by hand from the smartwatch and Bland–Altman method for multiple observations per indi-
oximeter simultaneously. Participants were sitting at rest vidual when the measured quantity changes over the
throughout the experiment, and they were asked to keep period of observation.45 In addition, we evaluated the root
their hands still on the table with their wrist and palm mean square difference between smartwatch and oximeter
down and flat and avoid any movement according to the paired measurements as
instructions of the smartwatch manufacturer. 






































A simple breathing circuit with a three-way non- (SpO2,smartwatch − SpO2,oximeter )2
rebreathing valve was assembled for the experiment. It Arms =
n
allowed the participant to inhale the hypoxic gas mixture
(12% O2) from a polyethylene Douglas bag or the where n is the number of evaluated pairs of SpO2
ambient air and to exhale into the ambient air outside the measurements.44
Douglas bag. The gas composition was monitored continu- Further, the differences in the smartwatch and oximeter
ously by a Datex Ohmeda S/5 patient monitor measurements were evaluated with respect to study time,
(Datex-Ohmeda Inc., Madison, WI, USA) with a sensor that is, to evaluate the relative response rate of the two
placed between the three-way valve and the participant. A devices. To do this we averaged the measurements of all
disposable antibacterial filter separated the participant participants at each study time for the smartwatch and for
from the breathing circuit. the oximeter. The mean SpO2 values across all participants
There were three phases of the experimental procedure. and iterations of the experimental procedure were used to
During the first 2 min, in the initial stabilization phase, par- graphically compare the average time courses of the
ticipants inhaled the ambient air via the breathing circuit. pooled smartwatch data and the pooled oximeter data. A
Two SpO2 readings were taken (times 0:45 min and 1:15 two-tailed paired t test was used to evaluate the statistical
min of the experiment). Then, in the 5-minute desaturation difference between the smartwatch data and the oximeter
phase, participants inhaled the hypoxic gas mixture from data at each measurement time. P value less than 0.05
the Douglas bag. Readings of SpO2 were taken every 30 was considered statistically significant. Only the observa-
s (from time 2:45 min to time 6:45 min of the experiment). tions, where simultaneous readings from both devices
The final stabilization phase followed when the participants were available, were included in the analysis. All data
inhaled the ambient air and SpO2 was recorded every 30 s were analyzed in Matlab 2021a (MathWorks, Natick,
(from time 7:30 min) until SpO2 returned to normal values. MA, USA) after transcription from the log.
Typically, three or four readings were taken in the final sta-
bilization phase. Each participant underwent the experi-
mental procedure twice. There was a delay of a minimum Results
of 1 h between the two iterations of the experimental pro-
cedure to address possible slow washout of test gas. Agreement between devices
The study was conducted in the Laboratory of special
equipment for ICU of the Czech Technical University in
Data processing and analysis Prague, Department of Biomedical Engineering, Kladno,
We concluded that the number of participants enrolled in Czech Republic, during February and March 2021 at an
the study and the number of paired SpO2 observations altitude of 405 m (1330 ft). All 24 volunteers (five
would meet the basic recommendations of the Food and women and nineteen men, all Caucasian, aged 20–28
Drug Administration and the International Organization years) completed the experiment with two iterations of
for Standardization (ISO 80601-2-61) for study design for the experimental procedure and two measuring devices,
in vivo accuracy testing of pulse oximeters (10 or more so there were 48 series of paired measurements available.
healthy subjects, 200 or more paired measurements).43,44 As in some cases, one of the devices did not provide a
We used the Bland–Altman analysis to compare the valid reading, there were 1284 valid paired readings in
agreement between simultaneous smartwatch and oximeter total out of a possible number of 1364. The SpO2 readings
SpO2 measurements. The Bland–Altman analysis looks at ranged between 76% and 100%. Most (75%) were between
two parameters, the bias and 95% limits of agreement. 90% and 100%, 24% between 80% and 89% and 1% below
The bias is quantified as the mean difference in the paired 80%.
measurements. The 95% limits of agreement, calculated The presented Bland–Altman plot is based on 642 indi-
as the mean difference ± 1.96 standard deviations, deter- vidual data points calculated from all complete pairs of
mine the range of expected difference in future pooled SpO2 readings (Figure 1). The bias (mean
4 DIGITAL HEALTH

difference) in SpO2 between the smartwatch and oximeter the higher average SpO2 measured by the smartwatch
was 0.0% for all the data points. The 95% confidence during induced desaturation. Generally, there are differ-
limits of the bias were −0.2% and 0.3%, indicating that ences between the reaction times of pulse oximeters to
there was no statistically significant bias between the meas- sudden hypoxia.46 During the experiments, we also
uring devices. The 95% limits of agreement were estimated observed a faster return of smartwatch values than oximeter
to be −5.8% and 5.9%. The most extreme individual differ- in the final stabilization phase after the desaturation phase,
ences between the smartwatch and oximeter SpO2 measure- but not being the primary concern of our study, there were
ments were −9% and 17%. The Arms evaluated across the not enough data to evaluate for this.
pooled SpO2 readings was 3.0%. The same approach was
used to analyze the data after splitting into SpO2 90%–
100% and SpO2 less than 90%. The results are summarized Comparison with prior work
in Table 1. As shown, the absolute bias was greater for Several studies have evaluated smartwatches in hypoxemia.
SpO2 measurements under 90%. In their analysis, Lauterbach et al. compared a different
smartwatch Garmin fē nix® 5X Plus (Garmin, Olathe, KS,
USA) with a medical-grade pulse oximeter Model 7500
Average response of devices (Nonin Medical BV, Amsterdam, the Netherlands) in a cus-
The time series of average smartwatch and oximeter mea- tomized chamber that allowed to change and maintain the
surements show the absolute differences between the inspiratory oxygen fraction. Twenty-three volunteers
means of SpO2 measurements were small (Figure 2). The breathed a gas mixture under normobaric conditions with
difference between the means of the smartwatch and oxim- inspiratory oxygen fractions between 14% and 21%. The
eter ranged from −0.64% (study time 195 s) to 0.74% study reported SpO2 bias (smartwatch−oximeter) only
(study time 480 s) with the minimum absolute difference 0.7%–0.8% for higher values of the inspiratory oxygen
of the means 0.22% (study time 450 s). None of the differ- fraction, but 3% for the smallest inspiratory oxygen frac-
ences between paired smartwatch and oximeter measure- tion. Two explanations were offered for the bias increase
ments at any study time reached a statistically significant by the authors of the study; first, elevated PaCO2 levels
difference. resulting in increased other hemoglobin derivatives in the
bloodstream, and second, hypoxia-mediated vasoconstric-
tion that altered blood flow in fingers compared to the
Discussion wrist.26 Hermand et al. compared a smartwatch from the
same manufacturer (Garmin Forerunner 245) with a
Principal results medical-grade oximeter on 10 healthy participants during
The main finding of our study is that SpO2 measurement by normoxia and normobaric hypoxia when the inspiratory
Apple Watch Series 6, a consumer product, did not differ on oxygen fraction was gradually reduced to 10.5%. The
average from SpO2 measurement by Masimo Radical-7 total observed bias of the smartwatch was 5.4%, and the
pulse oximeter, a medical device. The average absolute dif- bias for the lowest oxygen fraction was even 13.2%. The
ference or bias between smartwatch and oximeter SpO2 authors concluded the smartwatch was not a reliable alter-
measurements, evaluated for all pooled data, in two native to medical-grade oximeters.38 A study with another
ranges and at the individual study times, was less than smartwatch (Withings ScanWatch) by Kirszenblat and
1% SpO2. This is the resolution in which the SpO2 values Edouard reached opposite findings. Measurements of
are displayed on both devices. SpO2 in 14 healthy participants were compared with arterial
At low-oxygen levels, the smartwatch tended to measure blood oxygen saturation (SaO2) determined with a
higher SpO2 values than the oximeter, and this difference co-oximeter at various stable levels of oxygen saturation.
averaged approximately 1% SpO2 for readings less than The total bias found was 0.98% (right wrist) and 1.56%
90%. The time chart (Figure 2) illustrates a very similar (left wrist), and overall accuracy was adequate to medical-
response of both devices for the “average patient,” with grade oximeters.40 Our results, i.e., the negligible bias at
the average difference between SpO2 reported by the smart- higher saturation and the small bias with decreased satur-
watch and oximeter at the end of the desaturation phase ation, generally correspond to those of Lauterbach et al.
being only 0.26%, and −0.23% upon recovery. The time and Kirszenblat and Edouard although we detected a
series in Figure 2 also suggests that the response of the smaller bias for lower inspiratory oxygen fraction (12% in
smartwatch to sudden desaturation may be slower than our study vs. 14%) and somewhat lower measured SpO2
the response of the oximeter. The smartwatch required a values than Lauterbach et al. We also suggest that the differ-
15-s period for a single SpO2 measurement compared to ences reflect different devices used in the studies.
the 2–4-s averaging time of the oximeter, so the smartwatch Two recent studies examined the SpO2 measurement
reading lagged behind the oximeter readings during the using Apple Watch Series 6 compared to medical-grade
continuous SpO2 decrease. This may have contributed to pulse oximeters.41,42 The studies on subjects at rest
Rafl et al. 5

Figure 1. Differences between simultaneous SpO2 readings of the smartwatch (Apple Watch 6) and oximeter (Masimo Radical-7) across
different ranges of oxyhemoglobin saturation. Pooled SpO2 measurements were analyzed for all participants grouped. The solid line is the
mean difference of the measurements (bias). Dashed lines are the 95% limits of agreement. The area of markers is proportional to the
number of measurements.

Table 1. Comparison of measurement bias and agreement.

SpOa2, % Biasb (95% CI), % Lower LOA (95% CI), % Upper LOA (95% CI), % Arms, %

Entire range 0.0 (−0.2 to 0.3) −5.8 (−6.2 to −5.4) 5.9 (5.5–6.3) 3.0

<90 1.2 (0.7 to 1.7) −5.3 (−6.1 to −4.4) 7.6 (6.7–8.4) 3.4

90–100 −0.3 (−0.6 to 0.1) −5.8 (−6.2 to −5.4) 5.1 (4.7–5.5) 2.8
a
[(smartwatch + oximeter)/2].
b
[smartwatch − oximeter].
LOA: 95% limits of agreement.

included both healthy participants and diseased participants differences in individual SpO2 measurements against
with lung or cardiovascular diseases. Both studies reported direct measurements of SaO2 by co-oximetry under pro-
a bias (smartwatch−oximeter) of less than 1% and no sig- gressive normobaric hypoxia. The 95% limits of agreement
nificant differences between subject groups (healthy or dis- reported by Kolb et al. were (−6.5%, 5.6%) and (−7.6%,
eased). However, neither of the two studies induced 9.8%) for SpO2 finger measurements when SaO2 was
hypoxemia in the subjects, and they contained very few above 85% and under 85%, respectively.47 Others also
SpO2 measurements below 90%. reported individual readings may differ as much as 6%.48
The differences between Apple Watch Series 6 and In a more recent study, narrower 95% limits of agreement
Masimo Radical-7 within 6% SpO2 can be expected for (−1.8%, 1.8%) were reported by Louie et al. for a nonmo-
individual measurements for SpO2 readings 90%–100% tion SpO2 measurement when SaO2 was above 90%.49
and up to 8% for SpO2 readings less than 90%. This The root mean square difference is the standard metric
again is consistent with Lauterbach et al. and Kirszenblat for assessing accuracy in pulse oximetry that combines
and Edouard who reported 95% limits of the agreement bias and precision of the SpO2 measurement when com-
up to 8.6% and 6.6%, respectively. The differences in indi- pared to co-oximetry. Accuracy better or equal to 4.0%
vidual SpO2 measurements between the smartwatch and SpO2 is required in general.44 Typically, Arms ≤ 3.0 and
oximeter are also similar to what was reported as Arms ≤ 3.5 are expected for transmittance and reflectance
6 DIGITAL HEALTH

Figure 2. The time courses of the mean of all smartwatch SpO2 measurements (Apple Watch 6) and the mean of all oximeter SpO2
measurements (Masimo Radical-7) across all 24 participants. Data are mean ± SEM.

sensors, respectively.43 Medical-grade oximeters have an mixture. The results could be different in the case of
accuracy of 2%–3% according to manufacturers or 3%–4% chronic elderly patients with very long or extreme desatura-
according to what was reported in clinical studies.32,50 tions. However, the contribution of wearables for such
Numerous studies however reported that the accuracy of patients in real-world situations will not be a detailed ana-
pulse oximeters deteriorates as blood oxygen saturation lysis of the severity of the condition, but rather a warning
decreases.47,49,51,52 The Arms metric has also been utilized of an aggravated trend in the chronic problem or a sudden
when comparing SpO2 measurements. Verkruysse et al. com- major change. Our results suggest that SpO2 monitoring
pared contactless photoplethysmography with a median of using wearables could be, due to its ability to detect the
measurements taken by standard pulse oximeters in healthy magnitude and speed of desaturation, a useful tool in self-
adults under normoxic conditions and also hypoxic conditions care outside the clinic.
where the inspiratory oxygen fraction was about 15%.53 They We did not evaluate SaO2 in our study as this would
estimated Arms ≤ 2.5% for short-time segments and even Arms require arterial blood sampling and greatly complicate the
≤ 1.7% when discarding short-time errors. Hahnen and her col- experiment. It was demonstrated that SpO2 overestimates
leagues investigated the accuracy of a handheld portable device saturation compared to SaO2.52,54,55 Due to the inaccessibil-
for vital sign measurements on 85 participants and reported ity of actual SaO2 values, we chose 12% O2 and the
Arms was 3.1% for SpO2 when compared to a medical-grade 5-minute duration of the desaturation phase as the limit to
vital signs monitor.13 In this context our results, Arms < 3.0% avoid a frequent decrease of SpO2 below 80% and
for saturation of 90% and greater and Arms < 3.5% for satur- prevent transient cognitive effects that may be associated
ation under 90%, seem within the expected range with some with deep hypoxia.56 The reduced oxygen fraction we
of the 6%–8% span likely attributable to the Masimo device. used under normobaric conditions corresponds approxi-
Even with the large uncertainty between paired measurements, mately to the partial pressure of oxygen at an altitude of
the smartwatch seems reliable in detecting relevant drops in 4400 m and the results of our study may therefore not be
SpO2 below 90% even of short duration. applicable to areas of higher altitude or to SpO2 below
80% in general.
The steady decline of the SpO2 levels at the end of the
Limitations desaturation phase (Figure 2) suggests that the desaturation
Our study has numerous limitations. The study included phase needed to be extended to reach the plateau. This
only healthy young volunteers and short-time desaturation may have better explained whether there was some time
induced by the low-oxygen level of the inhaled gas delay in the smartwatch readings compared to the oximeter
Rafl et al. 7

readings. Nevertheless, our focus was primarily on whether Conclusions

the smartwatch can provide an alert of the same quality as
Apple Watch Series 6, as a representative of wearables, pro-
repeated SpO2 measurements with a medical-grade pulse
vides reliable SpO2 values as compared to a medical-grade
oximeter and thus be a useful screening method for detect-
pulse oximeter, at both normal oxygen levels and induced
ing hypoxia.
desaturation with SpO2 below 90%. The SpO2 monitoring
Finally, in our study, we used one type of smartwatch
technology used in this smartwatch is sufficiently advanced
from a single manufacturer. This must be considered
for the indicative measurement of SpO2 outside the clinic
when generalizing our observations to other smartwatches
and can detect states of reduced blood oxygen saturation.
in the rapidly evolving market. Smartwatches from other
manufacturers may show differences in performance,
Acknowledgements: The authors thank Lenka Horakova, MD,
even if they use the same principle of reflectance pulse
for the medical supervision of the experiments.
oximetry, as several hardware and software factors can
affect the PPG signal, including the geometry of the light
emitter and light detector or denoising.57 Smartwatch per- Contributorship: JR, TEB, VRH and MR conceptualized the
formance may also vary between users at rest and while study. JR and VRH administered the study. VRH and SW
executed the study and acquired the data. JR and TEB
active. We measured participants at rest, as required by
performed data analysis, interpretation, and visualization. JR and
the manufacturer. The results may not correspond to mea- TEB drafted the manuscript. All authors revised and edited the
surements during or just after sporting activities due to manuscript. All authors approved the final version of the
motion artifacts, which could be the subject of further manuscript.
study.
Declaration of Conflicting Interests: The authors declared no
Future perspectives potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.
The availability and convenience of measuring biological
signals using wearable devices such as smartwatches offer
the potential to expand patient care options in chronic ORCID iDs: Jakub Rafl https://orcid.org/0000-0001-5102-9354
Veronika Rafl-Huttova https://orcid.org/0000-0001-7370-5667
disease management. The clinical standard so far has
been isolated measurements under the supervision of
health professionals, which are taken with a relatively Data availability statement: The data underlying this article will
large time lag and then compared with the prevalence of be shared on reasonable request to the corresponding author.
the clinically relevant events in the population. Wearables
allow long-term and continuous monitoring of trends or, Ethical approval: The Ethical Review Board of the Faculty of
on the contrary, detection of abnormal fluctuations in indi- Biomedical Engineering, Czech Technical University in Prague
viduals9,11,58 and thus more quickly assess the change in approved the study (No. B1/2021).
their health status over time. Wearables are not intended
to replace medical devices, but they need sufficient accur- Funding: The authors disclosed receipt of the following financial
acy to provide an approximate assessment of an indivi- support for the research, authorship, and/or publication of this
dual’s condition.59 The risk is both overreacting to article: This work was supported by the Czech Technical
clinically irrelevant fluctuations in monitored signals and University in Prague [grant numbers SGS20/202/OHK4/3T/17,
neglecting serious changes related to real health complica- SGS22/202/OHK4/3T/17].
tions.14 In particular, while portable pulse oximeters with
transmission technology have been shown to be comparable Guarantor: JR
to patient monitors,60 data contradict SpO2 measurement
with commercial smartwatches as this feature is relatively Informed consent: All participants provided written informed
new. The results of our study are intended to help fill this consent before their enrollment in the study.
gap. They suggest that smartwatch technology for measur-
ing SpO2 has matured enough to be considered part of
References
patient care. This can help detect hidden, but potentially
1. Ehrler F and Lovis C. Supporting elderly homecare with
serious problems such as sleep apnea, which is a growing
smartwatches: advantages and drawbacks. Stud Health
problem with possible cognitive impacts,56 or in the early
Technol Inform 2014; 205: 667–671.
detection of acute exacerbations of chronic conditions 2. Reeder B and David A. Health at hand: a systematic review of
such as COPD.14 We further suggest that the exact require- smart watch uses for health and wellness. J Biomed Inform
ment of each of these potential health care applications need 2016; 63: 269–276.
to be articulated and wearable devices evaluated against 3. Tana J, Forss M and Hellstén T. The use of wearables in
those requirements. healthcare—challenges and opportunities. Arcada Working
8 DIGITAL HEALTH

Papers. https://www.theseus.fi/handle/10024/140584. (2017, pandemic in Finland: longitudinal monitoring through smart-

accessed 23 July 2021). watch technology. PLoS One 2021; 16: e0246494.
4. Kumari P, Mathew L and Syal P. Increasing trend of wear- 22. Waller E, Sutton P, Rahman S, et al. Prehabilitation with wear-
ables and multimodal interface for human activity monitoring: ables versus standard of care before major abdominal cancer
a review. Biosens Bioelectron 2017; 90: 298–307. surgery: a randomised controlled pilot study (trial registration:
5. Isakadze N and Martin SS. How useful is the smartwatch NCT04047524). Surg Endosc 2022; 36: 1008–1017.
ECG? Trends Cardiovasc Med 2020; 30: 442–448. 23. Mishra T, Wang M, Metwally AA, et al. Pre-symptomatic
6. Phaneuf A. Latest trends in medical monitoring devices and detection of COVID-19 from smartwatch data. Nat Biomed
wearable health technology, https://www.businessinsider. Eng 2020; 4: 1208–1220.
com/wearable-technology-healthcare-medical-devices (2021, 24. Quer G, Radin JM, Gadaleta M, et al. Wearable sensor data
accessed 3 May 2021). and self-reported symptoms for COVID-19 detection. Nat
7. GlobeNewswire. Global wearable medical devices markets Med 2021; 27: 73–77.
report 2021: Market is expected to reach $24.38 billion in 25. Sawh M. SpO2 and pulse ox wearables: why blood oxygen is
2025 at a CAGR of 24% - Long-term forecast to 2030. the big new health metric, https://www.wareable.com/
https://www.globenewswire.com/en/news-release/2021/06/ wearable-tech/pulse-oximeter-explained-fitbit-garmin-
14/2246369/28124/en/Global-Wearable-Medical-Devices- wearables-340 (2020, accessed 3 May 2021).
Markets-Report-2021-Market-is-Expected-to-Reach-24-38- 26. Lauterbach CJ, Romano PA, Greisler LA, et al. Accuracy and
Billion-in-2025-at-a-CAGR-of-24-Long-term-Forecast-to- reliability of commercial wrist-worn pulse oximeter during
2030.html (2021, accessed 23 July 2021). normobaric hypoxia exposure under resting conditions. Res
8. Lee SM and Lee D. Healthcare wearable devices: an analysis of key Q Exerc Sport 2021; 92: 549–558.
factors for continuous use intention. Serv Bus 2020; 14: 503–531. 27. de Barros GM, de Barros GM, dos Anjos MS, et al.
9. Ates HC, Yetisen AK, Güder F, et al. Wearable devices for the Smartwatch, oxygen saturation, and COVID-19: trustworthy?
detection of COVID-19. Nat Electron 2021; 4: 13–14. ABCS Health Sci 2021; 46: e021101.
10. Piwek L, Ellis DA, Andrews S, et al. The rise of consumer 28. White DP. Sleep apnea. Proc Am Thorac Soc 2006; 3: 124–
health wearables: promises and barriers. PLoS Med 2016; 128.
13: e1001953. 29. Uddin MB, Chow CM and Su SW. Classification methods to
11. Aliverti A. Wearable technology: role in respiratory health detect sleep apnea in adults based on respiratory and oximetry
and disease. Breathe (Sheff) 2017; 13: e27–e36. signals: a systematic review. Physiol Meas 2018; 39: 03TR01.
12. Raja JM, Elsakr C, Roman S, et al. Apple watch, wearables, 30. Chen Y, Wang W, Guo Y, et al. A single-center validation of
and heart rhythm: where do we stand? Ann Transl Med the accuracy of a photoplethysmography-based smartwatch
2019; 7: 417. for screening obstructive sleep apnea. Nat Sci Sleep 2021;
13. Hahnen C, Freeman CG, Haldar N, et al. Accuracy of vital 13: 1533–1544.
signs measurements by a smartwatch and a portable health 31. World Health Organization. Pulse oximetry training manual,
device: validation study. JMIR Mhealth Uhealth 2020; 8: https://www.who.int/patientsafety/safesurgery/pulse_
e16811. oximetry/who_ps_pulse_oxymetry_training_manual_en.pdf?
14. Lu L, Zhang J, Xie Y, et al. Wearable health devices in health ua=1 (2011, accessed 16 March 2022).
care: narrative systematic review. JMIR Mhealth Uhealth 32. Nitzan M, Romem A and Koppel R. Pulse oximetry: funda-
2020; 8: e18907. mentals and technology update. Med Devices (Auckl) 2014;
15. Ahanathapillai V, Amor JD, Goodwin Z, et al. Preliminary 7: 231–239.
study on activity monitoring using an android smart-watch. 33. Chung S-C and Sun C-C. Signal-enhancement reflective pulse
Healthc Technol Lett 2015; 2: 34–39. oximeter with Fresnel lens. Opt Commun 2016; 375: 9–14.
16. King CE and Sarrafzadeh M. A survey of smartwatches in 34. Lee H, Ko H and Lee J. Reflectance pulse oximetry: practical
remote health monitoring. J Healthc Inform Res 2018; 2: 1– issues and limitations. ICT Express 2016; 2: 195–198.
24. 35. Kiruthiga A, Annamol A, Balamugesh T, et al. Reflectance
17. Perez MV, Mahaffey KW, Hedlin H, et al. Large-scale assess- pulse oximetry for blood oxygen saturation measurement
ment of a smartwatch to identify atrial fibrillation. N Engl J from diverse locations—a preliminary analysis. In: 2018
Med 2019; 381: 1909–1917. IEEE international symposium on medical measurements
18. Avram R, Ramsis M, Cristal AD, et al. Validation of an algo- and applications (MeMeA) proceedings, Rome, Italy, 11–13
rithm for continuous monitoring of atrial fibrillation using a June 2018, pp. 666–671: IEEE.
consumer smartwatch. Heart Rhythm 2021; 18: 1482–1490. 36. Longmore SK, Lui GY, Naik G, et al. A comparison of
19. Asgari Mehrabadi M, Azimi I, Sarhaddi F, et al. Sleep track- reflective photoplethysmography for detection of heart rate,
ing of a commercially available smart ring and smartwatch blood oxygen saturation, and respiration rate at various ana-
against medical-grade actigraphy in everyday settings: instru- tomical locations. Sensors (Basel) 2019; 19: 1874.
ment validation study. JMIR Mhealth Uhealth 2020; 8: 37. Santos M, Vollam S, Pimentel MAF, et al. The use of wear-
e20465. able pulse oximeters in the prompt detection of hypoxemia
20. Kruizinga MD, Moll A, Zhuparris A, et al. Postdischarge and during movement: diagnostic accuracy study. J Med
recovery after acute pediatric lung disease can be quantified Internet Res 2022; 24: e28890.
with digital biomarkers. Respiration 2021; 100: 979–988. 38. Hermand E, Coll C, Richalet J-P, et al. Accuracy and reliabil-
21. Niela-Vilén H, Auxier J, Ekholm E, et al. Pregnant women’s ity of pulse O2 saturation measured by a wrist-worn oximeter.
daily patterns of well-being before and during the COVID-19 Int J Sports Med 2021; 42: 1268–1273.
Rafl et al. 9

39. Guber A, Epstein Shochet G, Kohn S, et al. Wrist-sensor pulse 50. Watson JN, Mannheimer PD and Kelley S. Nellcor™ pulse
oximeter enables prolonged patient monitoring in chronic oximetry motion testing. White paper, Medtronic, USA,
lung diseases. J Med Syst 2019; 43: 230. https://www.medtronic.com/content/dam/covidien/library/us/
40. Kirszenblat R and Edouard P. Validation of the withings scanwatch en/product/pulse-oximetry/nellcor-oximetry-motion-testing-
as a wrist-worn reflective pulse oximeter: prospective interventional white-paper.pdf (2016, accessed 16 March 2022).
clinical study. J Med Internet Res 2021; 23: e27503. 51. Gehring H, Duembgen L, Peterlein M, et al. Hemoximetry as
41. Pipek LZ, Nascimento RFV, Acencio MMP, et al. the “gold standard”? Error assessment based on differences
Comparison of SpO2 and heart rate values on Apple Watch among identical blood gas analyzer devices of five manufac-
and conventional commercial oximeters devices in patients turers. Anesth Analg 2007; 105: S24–S30.
with lung disease. Sci Rep 2021; 11: 18901. 52. Ross PA, Newth CJL and Khemani RG. Accuracy of pulse
42. Spaccarotella C, Polimeni A, Mancuso C, et al. Assessment of oximetry in children. Pediatrics 2014; 133: 22–29.
non-invasive measurements of oxygen saturation and heart 53. Verkruysse W, Bartula M, Bresch E, et al. Calibration of con-
rate with an Apple Smartwatch: comparison with a standard tactless pulse oximetry. Anesth Analg 2017; 124: 136–145.
pulse oximeter. J Clin Med 2022; 11: 1467. 54. Kelly AM, McAlpine R and Kyle E. How accurate are pulse
43. U.S. Department of Health and Human Services, Food and oximeters in patients with acute exacerbations of chronic
Drug Administration. Pulse oximeters - premarket notification sub- obstructive airways disease? Respir Med 2001; 95: 336–340.
missions [510(k)s]: Guidance for industry and Food and Drug 55. Kohyama T, Moriyama K, Kanai R, et al. Accuracy of pulse
Administration staff, https://www.fda.gov/regulatory-information/ oximeters in detecting hypoxemia in patients with chronic
search-fda-guidance-documents/pulse-oximeters-premarket- thromboembolic pulmonary hypertension. PLoS One 2015;
notification-submissions-510ks-guidance-industry-and- 10: e0126979.
food-and-drug (2013, accessed 16 March 2022). 56. Bickler PE, Feiner JR, Lipnick MS, et al. Effects of acute, pro-
44. ISO 80601-2-61:2017. Medical electrical equipment—Part found hypoxia on healthy humans: implications for safety of
2-61: Particular requirements for basic safety and essential tests evaluating pulse oximetry or tissue oximetry perform-
performance of pulse oximeter equipment. ance. Anesth Analg 2017; 124: 146–153.
45. Bland JM and Altman DG. Agreement between methods of 57. Charlton PH, Pilt K and Kyriacou PA. Establishing best prac-
measurement with multiple observations per individual. J tices in photoplethysmography signal acquisition and process-
Biopharm Stat 2007; 17: 571–582. ing. Physiol Meas 2022; 43: 050301.
46. Rafl J, Kulhanek F, Kudrna P, et al. Response time of indir- 58. Buekers J, Theunis J, De Boever P, et al. Wearable finger
ectly accessed gas exchange depends on measurement pulse oximetry for continuous oxygen saturation measure-
method. Biomed Tech (Berl) 2018; 63: 647–655. ments during daily home routines of patients with chronic
47. Kolb JC, Farran P, Norris SR, et al. Validation of pulse oxim- obstructive pulmonary disease (COPD) over one week: obser-
etry during progressive normobaric hypoxia utilizing a port- vational study. JMIR Mhealth Uhealth 2019; 7: e12866.
able chamber. Can J Appl Physiol 2004; 29: 3–15. 59. Nemcova A, Jordanova I, Varecka M, et al. Monitoring of
48. Batchelder PB and Raley DM. Maximizing the laboratory heart rate, blood oxygen saturation, and blood pressure
setting for testing devices and understanding statistical using a smartphone. Biomed Signal Process Control 2020;
output in pulse oximetry. Anesth Analg 2007; 105: S85–S94. 59: 101928.
49. Louie A, Feiner JR, Bickler PE, et al. Four types of pulse oxi- 60. Li X, Dunn J, Salins D, et al. Digital health: tracking phy-
meters accurately detect hypoxia during low perfusion and siomes and activity using wearable biosensors reveals useful
motion. Anesthesiology 2018; 128: 520–530. health-related information. PLoS Biol 2017; 15: e2001402.