This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Cardio, is properly cited. The complete bibliographic information, a link to the original publication on http://cardio.jmir.org, as well as this copyright and license information must be included.
Smartphone ownership is rising at a stunning rate. Moreover, smartphones prove to be suitable for use in health care due to their availability, portability, user-friendliness, relatively low price, wireless connectivity, far-reaching computing capabilities, and comprehensive memory. To measure vital signs, smartphones are often connected to a mobile sensor or a medical device. However, by using the white light-emitting diode as light source and the phone camera as photodetector, a smartphone could be used to perform photoplethysmography (PPG), enabling the assessment of vital signs.
The objective of this meta-analysis was to evaluate the available evidence on the use of smartphone apps to measure heart rate by performing PPG in comparison with a validated method.
PubMed and ISI Web of Knowledge were searched for relevant studies published between January 1, 2009 and December 7, 2016. The reference lists of included studies were hand-searched to find additional eligible studies. Critical Appraisal Skills Programme (CASP) Diagnostic Test Study checklist and some extra items were used for quality assessment. A fixed effects model of the mean difference and a random effects model of Pearson correlation coefficient were applied to pool the outcomes of the studies.
In total, 14 studies were included. The pooled result showed no significant difference between heart rate measurements with a smartphone and a validated method (mean difference −0.32; 99% CI −1.24 to 0.60;
Smartphone apps measuring heart rate by performing PPG appear to agree with a validated method in an adult population during resting sinus rhythm. In a pediatric population, the use of these apps is currently not validated.
Smartphone ownership rises year by year. Advanced economies still have the highest smartphone ownership rates. Smartphone ownership in countries with an emerging and developing economy, however, is rising at a stunning rate [
Due to their availability, portability, user-friendliness, relatively low price, wireless connectivity, far-reaching computing capabilities, and comprehensive memory, smartphones prove to be suitable for use in health care [
Most of the studies focus on measuring vital signs using a smartphone. To this end, smartphones are mostly connected to a mobile sensor or medical device [
The PPG waveform is influenced by many factors enabling the assessment of vital signs, for example, oxygen saturation, blood pressure, respiratory rate, and heart rate. Promising results show the ability to screen for pathologies related to peripheral vascular disease [
We conducted a systematic literature search of PubMed and ISI Web of Knowledge from January 1, 2009 to December 7, 2016, with the following search key: (smartphone* OR phone* OR ((Applic* OR App*) AND (mobile OR electronic OR software)) OR PPG OR Photoplethysmograph* OR Rheograph*) AND (Electrocardiogr* OR ECG OR EKG or Oximet*) AND ((rate* AND (heart OR pulse)) OR tachycardia* OR beat* OR complex* OR arrhythmia* OR fibrillation*). Only papers in English, German, French, or Dutch were included. The reference lists of included studies were hand-searched to find additional eligible studies.
Studies were included if the measurement of heart rate was conducted with the photo camera of a smartphone by PPG; the measurements were made at a finger, toe, or earlobe; the measurements of the smartphone were compared with an electrocardiogram (ECG), a pulse oximeter, or another validated method to determine heart rate. Studies were excluded if the measurement was conducted with a mobile sensor or medical device connected to a smartphone; the paper did not have heart rate as one of the outcomes; no abstract or full text was available.
Data were extracted by the first author and reviewed by all authors.
Following are study and intervention characteristics extracted from the included studies: first author, study country, study year, sample size, baseline characteristics of participants, age of the participants (mean or range), type of smartphone used, control instrument, duration and conditions of the measurement, and primary outcome measures. The primary outcome measures were the mean difference between heart rate measured by a smartphone and a validated method, the correlation coefficient of the relation between heart rate measurements made by both methods, and the 95% limits of agreement derived from a Bland-Altman plot.
Overall, 1 author was contacted to receive missing data about the heart rate measurements; 2 authors were contacted because of a lack of clarity about the data; and 7 authors were contacted to get access to the full text of the paper; but 2 authors failed to respond to that last request.
Study quality was appraised using the Critical Appraisal Skills Programme (CASP) Diagnostic Test Study checklist [
The quality assessment was performed by the first author and reviewed by the other authors.
In total, 3 different statistics were described, and 2 of them were used for estimation of the pooled result. The first was the mean difference between heart rate measured by a smartphone and a validated method. In case of absence of a mean value and standard deviation in the original paper, it was calculated manually where possible on the basis of the original data.
The second was the Pearson correlation coefficient calculated from the relation between heart rate measured by a smartphone and a validated method. The
The third were the 95% limits of agreement. They were derived from a Bland-Altman plot. Lower and upper limits were calculated starting from the mean difference by respectively subtracting and adding up the standard deviation of the mean difference between both methods, multiplied by a factor of 1.96. In 2 studies, they were calculated manually starting from the mean difference and the described limit of agreement.
The pooled result was estimated using a fixed- or random-effects model. Statistical heterogeneity was tested using the chi-squared test where a significant result indicated statistical heterogeneity. To quantify inconsistency, the I² of Higgins was used. In case of statistical heterogeneity, a random-effects model was used for pooling the results. Due to the small number of included studies, it was not possible to explore heterogeneity by subgroup analysis or meta-regression [
Pearson correlation was used to analyze the relation between different variables (publication year, mean heart rate, and sample size) and the mean difference. The scatter plots of these correlations were drawn.
Statistical analyses were performed using Review Manager Version 5.3 (The Cochrane collaboration, Copenhagen: Denmark: The Nordic Cochrane Centre, 2014), MedCalc 17.4 (MedCalc Software, Ostend: Belgium, 2017), and Microsoft Office Excel 2007 (Microsoft, 2007). Statistical significance level was set at 5%, except for mean difference where statistical significance level was set at 1%.
The mean difference between heart rate measured by a smartphone and a validated method was analyzed in a fixed-effects model (
Search and selection strategy.
Characteristics of included studies.
Author, year, and country | Sample size and age (range or mean [SD]) | Smartphone | Control | Duration and conditions measurement | Outcome measure |
Bolkhovsky et al, 2012, United States [ |
22 subjects, age not specified | Motorola Droid, iPhone 4S | ECGa | 2 × 2 min: supine and sitting up in tilt position (iPhone 4S, n=9); 2 × 5 min: supine and sitting up in tilt position (Motorola Droid, n=13) | Heart rate, heart rate variability |
Drijkoningen et al, 2014, Belgium [ |
28 adults with sinus rhythm during electrophysiological examination, age not specified | Samsung Galaxy S4 | ECG | 60 s | Heart rate, premature atrial ectopic beats identification |
Gregoski et al, 2012, United States [ |
14 adults, 18-59 years | Motorola Droid | ECG, pulse oximeter | 3 × 5 min: sitting, at rest, reading, and playing a video game | Heart rate |
Ho et al, 2014, Taiwan [ |
40 children undergoing ECG monitoring, 3 days to 15 years | iPhone 4S | ECG | 3 × 20 s at finger (or toe) and earlobe | Heart rate |
Koenig et al, 2016, Germany [ |
68 adults (45 patients from a cardiologic outpatient ambulance and 23 healthy controls), 51.7 (18.83) years | iPhone 4S | ECG | 5 min: at rest 2 min: after 3 min of physical exercise (only controls) | Heart rate, heart rate variability |
Kurylyak et al, 2012, Italy [ |
10 adults, 26-60 years | HTC HD2, iPhone 4, Nokia 5800, Samsung Galaxy S i9000 | Pulse oximeter | 2 × 60 s (per smartphone): at rest and after 60 s squatting | Heart rate |
Lagido et al, 2014, Portugal [ |
43 heart failure patients, age not specified | Sony Xperia S | ECG | At rest | Heart rate, heart rate variability |
Losa-Iglesias et al, 2016, Spain [ |
46 healthy adults, 39.3 (7.35) years | Samsung Galaxy Note | Radial pulse, pulse oximeter | 3 × 10-30 s: at rest (resting 10 min before measurements) | Heart rate |
Matsumara et al, 2013, Japan [ |
12 students, 21-24 years | iPhone 4S | ECG | 3 × 3 min: at rest (resting 7 min before measurement), during mental arithmetic, and during mirror tracing | Heart rate, normalize pulse volume |
Nam et al, 2016, United States [ |
11 healthy nonsmoking adults, 20-40 years | HTC One M8 | ECG | 3 × 2 min: breathing at frequencies from 0.1 to 0.5 Hz at increments of 0.1 Hz, breathing at 1 Hz and spontaneous breathing | Heart rate and breathing rate |
Pelegris et al, 2010, UK [ |
50 adults, 21-55 years | HTC Tattoo | Pulse oximeter | 2 × 9 s: well-lit room and average lit room | Heart rate |
Po et al, 2015, China [ |
10 subjects, age not specified | Samsung Galaxy Nexus, LG Optimus P920, Samsung Galaxy S2, Samsung Galaxy Tablet 7.0, Motorala Atrix | Pulse oximeter | 1 × 20 s | Heart rate and root mean square distortion of heart rate |
Scully et al, 2012, United States [ |
1 subject, age not specified | Motorola Droid | ECG | 1 × ?: spontaneous breathing |
Heart rate, respiration rate, oxygen saturation |
Wackel et al, 2014, United States [ |
26 children undergoing an electrophysiology study under general anesthesia, 5-17 years | iPhone 5 | ECG | 2 × ?: during baseline heart rate (34 measurements in 17 children) |
Heart rate |
aECG: electrocardiogram.
Study quality according to Critical Appraisal Skills Programme Diagnostic Test study checklist and extra considerations. Y indicates yes; N indicates no; and C indicates can’t tell.
Study | Validity of results | Utility of results | Extra considerations | |||||||||||||
Q1 | Q2 | Q3 | Q4 | Q5 | Q6 | Q9 | Q10 | Q11 | E1 | E2 | E3 | E4 | E5 | |||
Bolkhovsky et al | Y | Y | Y | Y | N | Y | C | Y | Y | Y | Y | N | Y | N | ||
Drijkoningen et al | Y | Y | Y | Y | Y | Y | C | Y | Y | Y | Y | N | N | N | ||
Gregoski et al | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | N | Y | Y | ||
Ho et al | Y | Y | Y | Y | N | Y | Y | Y | Y | Y | Y | N | N | Y | ||
Koenig et al | Y | Y | Y | Y | N | Y | Y | Y | Y | Y | Y | N | Y | N | ||
Kurylyak et al | Y | Y | Y | Y | N | N | C | Y | Y | Y | Y | N | Y | N | ||
Lagido et al | Y | Y | Y | Y | N | N | C | Y | Y | Y | Y | N | C | N | ||
Losa-Iglesias et al | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | N | N | Y | ||
Matsumara et al | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y | N | ||
Nam et al | Y | Y | Y | Y | N | Y | Y | Y | Y | Y | Y | N | N | N | ||
Pelegris et al | Y | Y | Y | Y | N | N | C | Y | Y | Y | Y | N | N | N | ||
Po et al | Y | Y | Y | Y | N | N | C | Y | Y | Y | Y | N | N | N | ||
Scully et al | Y | Y | Y | Y | N | N | C | Y | Y | Y | Y | N | N | N | ||
Wackel et al | Y | Y | Y | Y | Y | N | Y | Y | Y | Y | Y | N | Y | N |
Critical Appraisal Skills Programme Diagnostic study checklist
Validity of results
Was there a clear question for the study to address?
Was there a comparison with an appropriate reference standard?
Did all patients get the diagnostic test and reference standard?
Is there no possibility that the results of the test have been influenced by the results of the reference standard?
Is the disease status of the tested population clearly described?
Were the methods for performing the test described in sufficient detail?
Utility of results
Can the results be applied to your patients/the population of interest?
Can the test be applied to your patient or population of interest?
Were all outcomes important to the individual or population considered?
Extra considerations
Do both methods measure the same outcome?
Do both methods measure the outcome simultaneous?
Did the investigators motivate their choice for the sample size?
Did the investigators test both methods in different conditions to simulate the possible physiological range of values?
Did the investigators set up cutoff values for the clinical acceptable difference between both methods?
Forest plot for the meta-analysis of mean difference.
Results for heart rate: Pearson correlation coefficient.
Study | Conditions (sample size) | |||
Bolkhovsky et al | iPhone supine (9) | >.99 | Yes | <.001c |
iPhone tilt (9) | >.99 | Yes | <.001c | |
Droid supine (13) | .98 | Yes | <.001c | |
Droid tilt (13) | >.99 | Yes | <.001c | |
Drijkoningen et al | Not specified (28) | .98 | Yes | <.001 |
Gregoski et al | At rest (14) | .99 | Yes | <.001c |
Reading (14) | .99 | Yes | <.001c | |
Video game (14) | .99 | Yes | <.001 | |
Ho et al | App A finger (40) | .81 | No | <.001 |
App A earlobe (40) | .91 | Yes | <.001 | |
App B finger (40) | .75 | No | <.001 | |
App B earlobe (40) | .76 | No | <.001 | |
App C finger (40) | .27 | No | .10 | |
App C earlobe (40) | .46 | No | .003 | |
App D finger (40) | .90 | Yes | <.001 | |
App D earlobe (40) | .98 | Yes | <.001 | |
Koenig et al | 80 randomly chosen intervals at rest or after exercise (68) | >.99 | Yes | <.001c |
Lagido et al | At rest (43) | .94 | Yes | <.001c |
Losa-Iglesias et al | Sitting up (46) | .95 | Yes | <.001 |
Matsumura et al | All conditions (12) | .99 | Yes | <.001c |
Wackel et al | App 1 sinus rhythm (17) | .99 | Yes | <.001c |
App 1 tachycardia (10 succeeded attempts) | .56 | No | .01c | |
App 2 sinus rhythm (17) | .99 | Yes | <.001c | |
App 2 tachycardia (5 succeeded attempts) | −.43 | No | .09c |
a
b
cData based on own calculations.
The correlation between heart rate measurements made by a smartphone and a control instrument was analyzed in a random-effects model (
The correlation between the mean heart rate measured by a validated method, the sample size of the included studies, and the year of publication of the included studies and the mean difference was analyzed in
Forest plot for the meta-analysis of Pearson correlation coefficient.
Results for heart rate: 95% limits of agreement.
Study | Conditions (sample size) | 95% LOAa (BPMb), control—smartphone |
Bolkhovsky et al | iPhone supine (9) | −0.4 to 0.2c |
iPhone tilt (9) | −0.3 to 0.3c | |
Droid supine (13) | −3.4 to 3.0c | |
Droid tilt (13) | −1.7 to 1.1c | |
Gregoski et al | Video game (14) | −3.9 to 3.7c |
Loso-Iglesias et al | Sitting up (46) | −8.5 to 2.0 |
Matsumura et al | All conditions (12) | −1.0 to 1.4 |
Nam et al | At rest, sitting up (11) | −5.6 to 5.5 |
Pot et al | Average all smartphones (10) | −4.1 to 1.2 |
aLOA: limits of agreement.
bBPM: beats per minute.
cData based on own calculations.
Scatter plot comparing correlation between mean heart rate measured by control and mean difference.
Scatter plot comparing correlation between sample size and mean difference.
Scatter plot comparing correlation between year of publication and mean difference.
The meta-analysis of the mean difference showed no statistical difference between the measurement of heart rate by a smartphone and a validated method (mean difference −0.32; 99% CI −1.24 to 0.60;
First, the results of the studies in a pediatric population showed that it is not advisable yet to use these apps in children. A possible cause is that because of the smaller size of children’s fingertips, the pulsatile flow may be less consistently detected. The use of the earlobe as a measuring point may present a possible solution. Children may also have difficulties in containing the appropriate pressure on the camera lens and keeping their finger motionless to make a good measurement [
A second issue is heart rate measurement during periods of arrhythmia [
Third, previous research stated that heart rate measurement can be susceptible to environmental or human factors such as ambient light, motion [
Fourth, it was remarkable that in the included studies the mean difference became more and more negative over time. A plausible explanation is that every paper focuses on (a) certain type(s) of smartphone model(s) or app(s). Consequently, the results cannot be automatically projected to other smartphones and apps [
First of all, to the best of our knowledge, this was the first systematic review and meta-analysis evaluating smartphone apps using PPG to measure heart rate. A comprehensive search strategy was used, including every paper investigating smartphone apps deriving heart rate measurement from a PPG signal. At last, there was a focus on different statistics for assessing agreement between methods.
Nevertheless, there were some limitations of the included studies. First, the methodological quality was often low, reflected by the fact that only 3 studies scored 12 or more out of 14 on the quality assessment questions [
Second, most of the mean heart rates that were reported lay between 70 and 80 beats per minute. As a result, it was not possible to investigate whether smartphones could be used to measure the higher physiological ranges of heart rate.
Third, only 8 of the included studies [
A fourth and last limitation is a high statistical heterogeneity between studies on the level of correlation coefficients. This is likely attributable to clinical heterogeneity caused by differences in patient characteristics (eg, adults vs children), the conditions in which the heart rates were measured (eg, at sinus rhythm vs during a period of tachycardia), and which smartphone or app was used [
All these factors may influence the generalizability of the results.
In addition, there were some limitations specific to the review. The data were extracted by the first author only; however, they were thoroughly reviewed by the other authors, of which one is specialized in cardiology. In addition, 2 studies were excluded because the full text could not be retrieved [
This meta-analysis suggests that heart rate measured by smartphone apps performing PPG agrees with a validated method in an adult population in resting sinus rhythm, provided that during measurement the measuring point was kept still and that appropriate pressure was maintained. In a pediatric population, the use of these apps can currently not be supported, especially not during periods of tachycardia. Future research with a larger and more diverse study population should be conducted. The technology should also be tested in more varied clinical situations evoking variations in normal heart rate and during arrhythmias.
Critical Appraisal Skills Programme
electrocardiogram
photoplethysmography
BDR made most of the contributions to conception and design, to the acquisition of data, and to analysis and interpretation of data. The other authors played an important role in acquiring missing data and revising it critically for important intellectual content. All authors reviewed and approved the data and the final text.
None declared.