Following the heart: what does variation of resting heart rate tell about us as individuals and as a population

Abstract

Resting heart rate (RHR) and heart rate variability (HRV) reflect the autonomic control of cardiac chronotropic activity, and they associate with cardiovascular fitness, acute and chronic health status, and mental stress. Relatively low RHR and relatively high HRV are generally seen as marks of better health, performance, and recovery levels. Nevertheless, the values are highly individual and comparison between individuals is not straightforward. On the other hand, evolution of wearable devices has made it possible to follow the course of individual RHR and HRV as long-term time series, which in turn enables observation of how behavioral, societal and seasonal factors affect RHR and HRV at individual and population scale. In this article, data measured by the Oura ring is used to study how alcohol and training affect these values, and moreover, how societal and seasonal factors affect us as a population.

Full text

Following the Heart - What Does Variation of Resting Heart Rate
Tell about Us as Individuals and as a Population
Heli Koskimäki∗
Hannu Kinnunen∗
heli.koskimaki@ouraring.com
hannu.kinnunen@ouraring.com
Oura Health
Oulu, Finland
Salla Rönkä
salla.ronka@bitfactor.
Bitfactor
Oulu, Finland
Benjamin Smarr
benjamin.smarr@ouraring.com
University of California, Department
of Psychology & Oura Health
Berkeley, CA, USA
ABSTRACT
Resting heart rate (RHR) and heart rate variability (HRV) reect
the autonomic control of cardiac chronotropic activity, and they as-
sociate with cardiovascular tness, acute and chronic health status,
and mental stress. Relatively low RHR and relatively high HRV are
generally seen as marks of better health, performance, and recovery
levels. Nevertheless, the values are highly individual and compari-
son between individuals is not straightforward. On the other hand,
evolution of wearable devices has made it possible to follow the
course of individual RHR and HRV as long-term time series, which
in turn enables observation of how behavioral, societal and seasonal
factors aect RHR and HRV at individual and population scale. In
this article, data measured by the Oura ring is used to study how
alcohol and training aect these values, and moreover, how societal
and seasonal factors aect us as a population.
CCS CONCEPTS
•Human-centered computing →Mobile devices
;
Empirical
studies in ubiquitous and mobile computing
;
•Applied com-
puting →
Consumer health;Health informatics;Sociology;
•Com-
puter systems organization →Embedded hardware.
KEYWORDS
heart rate, heart rate variability, wearables, Oura ring, seasonal
variation, training, alcohol
ACM Reference Format:
Heli Koskimäki, Hannu Kinnunen, Salla Rönkä, and Benjamin Smarr. 2019.
Following the Heart - What Does Variation of Resting Heart Rate Tell about
Us as Individuals and as a Population. In Adjunct Proceedings of the 2019 ACM
International Joint Conference on Pervasive and Ubiquitous Computing and
the 2019 International Symposium on Wearable Computers (UbiComp/ISWC
’19 Adjunct), September 9–13, 2019, London, United Kingdom. ACM, New
York, NY, USA, 4 pages. https://doi.org/10.1145/3341162.3344836
∗These authors contributed equally to this research.
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specic permission and/or a
fee. Request permissions from permissions@acm.org.
UbiComp/ISWC ’19 Adjunct, September 9–13, 2019, London, United Kingdom
©2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-6869-8/19/09. . . $15.00
https://doi.org/10.1145/3341162.3344836
Figure 1: Age (years) composition of the Oura ring user pop-
ulation, n=57278.
1 INTRODUCTION
Resting heart rate (RHR) and heart rate variability (HRV) are bio-
signals related to sympathetic and parasympathetic nervous system
activity, and known markers of health. For example, heart rate (HR)
has a direct role in blood pressure regulation in an interplay with
stroke volume and peripheral resistance. HR and wearable sensors
have been a natural combination since the invention of heart rate
monitors in the 1970’s. While the rst solutions were targeted to
athletes to assess their acute exercise intensity, over the past decade
trend has shifted towards optimizing your training based on your
resting HRV [6], and even evaluation of sleep and recovery [1].
HRV consists of periodic and aperiodic changes in the duration
of the cardiac cycle and provides a powerful means of observing the
interplay between the sympathetic and parasympathetic nervous
systems. The beat of a healthy heart shows signicant variation
among the time intervals between heartbeats and it is used as a
general marker of mental and physical well-being [
8
]. Nightly sleep
provides an excellent measurement point because external condi-
tions are typically constant and confounding factors are controlled
without extra eorts.
There are several known behavioral and environmental factors
that aect an individual’s RHR. Acute causes include intense exer-
cise, alcohol, ambient temperature and sickness. Long-term inu-
encers include age, aerobic training and cardiovascular tness and
health. For example, a single intense or exceptionally long exercise
session can increase RHR for several hours, yet participation in
regular training will lower the baseline of RHR. In addition to these
1178

UbiComp/ISWC ’19 Adjunct, September 9–13, 2019, London, United Kingdom Koskimäki and Kinnunen, et al.
12345678910
Subject number
40
45
50
55
60
65
70
75
80
Resting heart rate
Figure 2: Boxplot of mean resting heart rates (RHR) over a
1-week period among 10 individual subjects. Each subject re-
ported alcohol intake during 1 day only, and the minimum
amount was 3 doses. In every subject the highest RHR value
occurred after the alcohol intake. Moreover, in 8 out of 10
subjects, the alcohol night stands out as an outlier in the
boxplot (red asterisk).
short and long term responses, woman’s RHR varies according
to menstrual cycle phase. Similar responses are also seen in HRV,
typically in the opposite direction. It has not been explored thor-
oughly how much seasonal variability there is in RHR and HRV,
yet seasonal variation in cardiovascular disease and its’ risk factors
have been reported, risks peaking in winter and lowering during
the summer [
9
]. Among others, environmental temperature, light
exposure, infections, bad diet and lack of physical activity are men-
tioned as risk factors that could link to seasonal variation. Distance
from the equator seems to enhance the seasonality [4].
PPG has been used for several decades to derive oxygen satura-
tion, typically with a ngertip sensor. More recently, PPG sensors
embedded into wrist bracelets and rings have been utilized to assess
HR and HRV for various applications. The accuracy seems to vary
depending on the application. The main challenge is clearly sensitiv-
ity to motions [
10
]. Subsequently, best accuracy has been obtained
in nocturnal measurements during sleep where wrist PPG has been
reported to reach 5-6 ms mean absolute errors (corresponding to
12-15 ms standard deviation), in inter-beat-intervals compared to
the R-R intervals derived via ECG devices [
3
]. Finger PPG seems
to have potential for better beat-to-beat detection accuracy than
wrist PPG [
2
], which is likely due to the close proximity of arteries
on the palmar side of the nger.
The Oura ring is a scientically validated, wearable sleep tracker
which estimates RHR, HRV, and sleep stages based on nocturnal
PPG (250 Hz), 3-D accelerometer (50 Hz) and skin temperature [
5
],
[
1
], [
7
]. From RHR and HRV accuracy point of view, a very high
between-subjects agreement in night-time average RHR and HRV
between the Oura ring and the ECG (Faros cardiac monitor) have
been reported: r
2
= .998 and r
2
= .967 for RHR and HRV, respectively,
where HRV was assessed using the time domain parameter rMSSD
(root mean square of successive dierences) [5].
12345678910
Subject number
20
40
60
80
100
120
140
Heart rate variability
Figure 3: Boxplot of mean heart rate variability (HRV) in the
same way as with RHR on the left. In every subject the low-
est HRV value occurred after the alcohol intake. Moreover,
in 8 out of 10 subjects the alcohol night stands out as an out-
lier in the boxplot (red asterisk below). Subjects are the same
as in Figure 2.
2 STUDY
The experimental part of this study is divided into two parts: 1) RHR
and HRV responses to reported daily choices (including alcohol in-
take and physical exercise in this study) among individual subjects,
and 2) RHR and HRV responses to lifestyle, societal and seasonal
factors across a population. In the rst part a small sub-set of Oura
ring users have been selected to demonstrate if the eect of alcohol
and training to RHR and HRV can be detected from among other
potential confounding factors. One calendar week of 10 individual
users was analyzed in each case, where seven night-time average
values of RHR and HRV were collected from each subject.
•
Alcohol response: a week of RHR and HRV data including
the night after a day when each user reported at least 3 doses
of alcohol
•
Training response: a week of data including one or two
nights following a day with at least 30 minutes of high in-
tensity exercise (jogging was considered high intensity, as
an example) with no alcohol intake during that week.
For both experiments, every subject chose a week according to
their own schedules, and they gave their consent to report the day
of alcohol intake and days of exercise participation. All subjects
followed the guidance. Exact amount of alcohol, exact duration and
type of exercise were not collected: this decision was made respect-
ing subject privacy and in order to include normal-life variation in
the resulting data.
In the second part of this study the focus shifts from an individual
to a population scale. In this study all the Oura ring users were
included during a time window starting from July 2016 (n=1415)
and ending at March 2019 (n=57278). For every day, the average
nocturnal RHR is calculated, so that the contribution of a single user
to the overall variance is small especially in the latter part of the
time window. The gender composition was 36 percent female and
64 percent male, majority being between 30 and 50 years, yet 20-30-
year and 50-60-year old people were also well represented (Figure
1). Moreover, a clear majority of Oura users live in the Northern
1179

Following the Heart UbiComp/ISWC ’19 Adjunct, September 9–13, 2019, London, United Kingdom
12345678910
Subject number
20
40
60
80
100
120
140
Heart rate variability
Figure 4: Boxplots of resting heart rate variability among 10
individuals during a training week.
Table 1: Nightly HRV (rMSSD in ms) during a training week.
Red, bold numbers represent the nights following a subjec-
tively reported exercise day.
SUB MON TUE WED THU FRI SAT SUN
1 84 76 65 68 69 78 71
6 70 56 59 59 56 40
1 92 74 45 54 85 55 69
10 68 55 45 47 43 42 36
7 30 32 29 45 35 37 32
2 139 109 87 106 96 105 101
9 80 74 67 68 74 75 71
8 59 59 58 48 53 44 52
5 59 44 44 37 46 21 43
4 32 32 41 30 25 26 18
hemisphere; approximately 60 percent in US and Canada, 30 percent
in Europe and the rest in all continents including Australia and
New Zealand from the Southern hemisphere of our planet Earth.
3 RESULTS AND DISCUSSION
3.1 Eect of daily choices (i.e. alcohol and
training response) at individual level
In Figures 2 and 3 the individual alcohol responses are shown. Even
though the RHR and HRV are both highly individual values - for
example in this group between-individual mean HRV values vary
from 30 to 110 ms - the eect of alcohol has similar characteristic
in nearly all individuals, manifesting as an increase of RHR and
decrease of HRV in the night with alcohol compared to the rest
nights of the week (the highest RHR and lowest HRV occur for every
subject after the alcohol intake night). To visualize the magnitude
of this eect a box plot was used; in a box plot, an observation that
falls outside normal distribution is marked as an outlier point. In
the case of alcohol intake, these nights stand out as outliers; e.g.
in Figure 2, for 8 out of 10 individuals, meaning that the eect of
alcohol exceeds the variation observed during the rest of the week.
Figure 4 and Table 1 could have capacity to reveal a response
in resting HRV to daily exercise. Nevertheless, it becomes clear
that dierent subjects respond in a dierent way to the reported
exercise days. For only 2/10 individuals the HRV after an exercise
day is considered as an outlier (Subjects 2 and 9). For some other
subjects the lowest HRV is not even the day after the training day,
suggesting that these individuals may have been accustomed to the
training mode, intensity and duration they performed during the
given week. In general, a given absolute intensity of exercise has a
bigger impact to unt or unaccustomed individuals. The usability of
the relation between training intensity and the impact in HRV as a
potential resilience or tness marker remains to be studied later in
higher detail. Moreover, HRV is susceptible to various other stress
and lifestyle factors choices, including habitual physical activity
outside structured exercise that may also positively aect HRV.
3.2 Lifestyle, society and seasonality eect at
population level
As shown and discussed in the previous section, a mixture of daily
lifestyle factors aects our health and recovery, and some of their
eects are highly individual by nature. In order to study the health
related outcomes broadly, such as the impacts of lifestyle, societal
impositions, or natural phenomena including seasonality, long term
data and large number of people will be a pre-requisite. However,
clean individual long-term data can be dicult to achieve without
a burdensome requirement for subjects to respond to queries and
keep diary - solutions which are subjective by nature and do not
lend themselves to scale. By contrast, wearable device data can be
easily scaled into the population level, see for example Figure 5
which gives an example outcome from a data collection that did
not require any additional eort from individual users. This kind of
solutions illustrate a huge potential awaiting exploration of natural
human dynamics, as well as culturally or regionally specic public
health investigations.
3.2.1 Distinct dates (New Year). Looking at the RHR curve pre-
sented in Figure 5, there are three distinct upward peaks where an
abrupt increase in RHR is observed. A closer look at the date axis
reveals that those peaks are exactly one year apart from each other,
and they are all from New Year’s Eves, a public holiday associated
with staying up late and drinking across the Western society. These
two lifestyle factors behind this particular increase in RHR are quite
obvious, and as indicated by the big response in RHR, the negative
health impact at the population level can be huge. It would be in-
teresting to study more closely the relationship between adverse
health outcomes and RHR.
3.2.2 Seasonal highs and lows. Another interesting observation is
the yearly uctuation in RHR with a peak in winter and a trough at
the end of the summer season. The magnitude of seasonal variation
can be observed to be roughly 2 bpm: In 2018, August averages
1.6 bpm lower than March 2018, and December averages 2.1 bpm
higher than August. The reasons behind these uctuations may
be of biological, societal, and lifestyle origin. On one hand, the
role of biological factors, such as seasonal variation in Sun light
and ambient temperature are supported by the observation that in
people living at the Northern hemisphere, RHR start to decline in
April and reach its minimum in August, before autumnal equinox.
Interestingly, looking at subjects from New Zealand and Australia
separately, exactly opposite phase with regards to the length of the
day was observed. Since the response in RHR is delayed in respect
to change of day length at both hemispheres, ambient temperature
1180

UbiComp/ISWC ’19 Adjunct, September 9–13, 2019, London, United Kingdom Koskimäki and Kinnunen, et al.
Jan 2017 Jul 2017 Jan 2018 Jul 2018 Jan 2019
60
62
64
66
68
Resting heart rate
8
12
16
20
24
Duration of Night (h)
Jan 2017 Jul 2017 Jan 2018 Jul 2018 Jan 2019
Date
60
62
64
66
68
Resting heart rate
8
12
16
20
24
Duration of Night (h)
Figure 5: Nocturnal mean resting heart rate among users of Oura ring for people from the Northern (above, n = 1400 - 56
000) and Southern hemisphere (below, n = 120 - 1500). Duration from sunset to sunrise in New York (above) and Melbourne,
Australia (below) plotted with red line to mark the seasonal variation in daylight.
could really be another, potentially complementing input. Seasonal
variations can also be linked to time spent outside, amount of phys-
ical activity, and general health status including u. However, some
clear observations emphasize the role of lifestyle eects. For ex-
ample, the rate of increase in RHR in December exceeds the speed
expected from the change in the day length, leading us to specu-
late whether elevated RHR is also linked to holiday season stress.
There is also a drop in RHR from December to January when a new,
healthier life is started by people at a large scale. If we consider
the societal inputs further, summer holiday season should decrease
work related stress. All in all, it seems obvious that the annual
trends are due to both seasonality and societal factors, which fur-
ther experiments can help to identify. Distance from the equator
could not be assessed in our data in detail. With reference to the
reported seasonality in cardiovascular disease and observed links
to the distance from equator [
4
], it could be a topic of future study.
3.2.3 Weekly rhythm. Figure 5 also illustrate a regular weekly
rhythm where HR peaks at weekends. Quantifying the magnitude
of this weekly rhythm, night between Saturday and Sunday aver-
ages 1.6 bpm higher than the night between Monday and Tuesday
in our data. It is a well known social pattern that sleep timing is
delayed during weekends: the phenomenon is also called social
jetlag as it results from sleep debt accumulated during the week
when social factors make people wake up earlier than their biologi-
cally preferred wake-up time. Delayed sleep during the weekends
increases RHR because sleep is misaligned from the time when
body metabolism is the lowest due to circadian regulation. Alcohol
consumption across the weekdays also peaks during Friday and
Saturday nights, and a portion of population apparently increase
their physical activity levels at weekends as well.
4 CONCLUSIONS
In this paper we have shown that data generated from the Oura
Ring sensor platform can be used to identify health and lifestyle
related eects both for individuals and across large populations. We
provide evidence that individuals interested in regulating their RHR
or HRV can identify acute impacts from alcohol consumption and
from exercise, as well as monitor long term changes, as in changing
resilience with rising tness after sustained training. At the same
time, because of the large population contributing anonymized
data, the Oura Ring is also providing insights into population scale
phenomena. For instance, here we quantify the eect of New Years
on RHR, and also provide the highest resolution of human physi-
ological response to seasons of which we are aware. Many more
avenues suggest themselves for exploration, including other acute
and longitudinal eects for individuals, and social cost explorations
and basic human naturalism at the population level. We can even
imagine experiments in quantifying social policy impacts, as in
changing work schedules on RHR by region.
REFERENCES
[1]
Massimiliano de Zambotti, Leonardo Rosas, Ian M Colrain, and Fiona C Baker.
2019. The sleep of the ring: comparison of the
¯
OURA sleep tracker against
polysomnography. Behavioral sleep medicine 17, 2 (2019), 124–136.
[2]
J. Harju, Tarniceriu A, Parak J, Vehkaoja A, Yli-Hankala A, and Korhonen I. 2018.
Monitoring of heart rate and inter-beat intervals with wrist plethysmography in
patients with atrial brillation. Physiol. Meas. 39 (2018). Issue 065007.
[3]
Parak J, Tarniceriu A, Renevey P, Bertschi M, Delgado-Gonzalo R, and Korhonen
I. 2015. Evaluation of the beat-to-beat detection accuracy of PulseOn wearable
optical heart rate monitor. Conf Proc Annu Int Conf IEEE Eng Med Biol Soc (2015),
8099–102.
[4]
Pell JP and Cobbe SM. 1999. Seasonal variations in coronary heart disease. QJM
Int J Med. 92 (1999), 689–96. Issue 12.
[5]
HO Kinnunen and H Koskimäki. 2018. 0312 The HRV Of The Ring-Comparison
Of Nocturnal HR And HRV Between A Commercially Available Wearable Ring
And ECG. Sleep 41 (2018), A120.
[6]
Antti M Kiviniemi, Arto J Hautala, Hannu Kinnunen, and Mikko P Tulppo.
2007. Endurance training guided individually by daily heart rate variability
measurements. European journal of applied physiology 101, 6 (2007), 743–751.
[7]
Heli Koskimäki, Hannu Kinnunen, Teemu Kurppa, and Juha Röning. 2018. How
Do We Sleep: a Case Study of Sleep Duration and Quality Using Data from Oura
Ring. In Proceedings of the 2018 ACM International Joint Conference and 2018
International Symposium on Pervasive and Ubiquitous Computing and Wearable
Computers. ACM, 714–717.
[8]
U. Rajendra Acharya, K. Paul Joseph, N. Kannathal, C. M. Lim, and J. S. Suri. 2006.
Heart rate variability: a review. Med. Biol. Eng. Comput. 44 (2006), 1031–1051.
[9]
Stewart S, Keates AK, Redfern A, and McMurray JJV. 2017. Seasonal variations
in cardiovascular disease. Nat Rev Cardiol. 14 (2017), 654–64. Issue 11.
[10]
A. Schäfer and J. Vagedes. 2013. How accurate is pulse rate variability as an
estimate of heart rate variability? A review on studies comparing photoplethys-
mographic technology with an electrocardiogram. Int. J. Cardiol. 166 (2013),
15–29.
1181