ORIGINAL ARTICLE

COMPARING THE VALIDITY OF A GPS MONITOR AND A SMARTPHONE APPLICATION TO MEASURE PHYSICAL ACTIVITY

Manolis Adamakis, PhD1 1Department of Physical Education and Sport Science, National and Kapodistrian University of Athens, Greece

Background: A recent approach to increasing physical activity levels, managing weight and improving health has been via technological advances, such as the Web 2.0 , wearable activity trackers and smartphones. These approaches might be effective due to reduced cost, user- friendly environment and real-time feedback provided. Many of these monitors and smartphone applications are marketed to provide personal information on the level of physical activity, however little or no information is available regarding their validity.

Aims: The purpose of this study was to compare the criterion validity for distance travelled and total energy expenditure (TEE) between a commercially available GPS monitor, Garmin Forerunner 310XT, and a freeware GPS application for Android smartphones, Runkeeper, under semi- structured activity settings.

Methods: A single, healthy and physically active participant took part in all trials. The same protocol was repeated on 40 occasions (20 walking and 20 running sessions). The participant wore the Garmin GPS on the left wrist and a smartphone with the Runkeeper application activated on the left arm. Distance was compared against an objectively measured distance with three different methods and energy expenditure estimates for each monitor was evaluated relative to criterion values concurrently obtained from the portable metabolic system Cosmed K4b2. Differences from criterion measures were expressed as a mean absolute percent error and were evaluated using repeated measures ANOVA and Bland-Altman plots.

Results: For overall group comparisons, the mean absolute percent error values for distance were 0.30% and 0.74% for Garmin (walking and running), while higher values were calculated for Runkeeper (3.28% during running and 4.43% during walking), which significantly overestimated distance in both conditions. For energy expenditure estimation, significant differences were observed for both monitors (pB0.001). Garmin significantly underestimated energy expenditure compared to the criterion method in both conditions by 17%, while Runkeeper significantly overestimated it by 6.29% during running and 35.52% during walking.

Conclusions: The present study offers initial evidence for the validity of GPS of wearable activity monitors and smartphone applications for measuring distance travelled. However, estimates of energy expenditure were poor, except for Runkeeper during running which provided acceptable error.

Journal MTM 6:2:28Á38, 2017 doi:10.7309/jmtm.6.2.4 www.journalmtm.com

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 28 ORIGINAL ARTICLE

Introduction Even though GPS monitors are accurate for dis- Every year approximately 3.2 million deaths and tance estimation, commercially available applica- 32.1 million illnesses are associated with physical tions that use the GPS sensor of the smartphones inactivity.1 In most western societies people between are sparsely tested for their accuracy. Bauer12 for 18 and 35 years belong in the high-risk group of example compared ten GPS applications over a 1 becoming either overweight or obese, due to lack of km distance, with the Adidas miCoach been the physical activity (PA).2 In order to reverse these most accurate and Runkeeper having a divergence statistics, individuals are recommended to accumu- of 20 m. late at least 30 min/day, 5 days/week of moderate intensity PA or 15 min/day, 5 days/week of vigorous Nowadays these monitors and applications have the intensity PA (75 min/week), or a combination of ability to estimate, using anthropometric and GPS both.3 data, the energy expenditure (EE) of individuals during exercise. However, no published research, 13 A recent approach to increasing PA levels, manag- only a Master’s dissertation and a conference 14 13 ing weight and improving health has been via paper have validated these outcomes. Mallula technological advances, such as the Web 2.0 tech- compared a Garmin Forerunner 405CX with Nike nologies, activity trackers and smartphones.4 application for iPod and found that the application Griffiths and colleagues5 proposed that these ap- was more accurate in EE, while the device had a more proaches might be effective due to reduced cost, valid estimation of distance and speed. Furthermore, 14 user-friendly environment, real-time information Adamakis and Zounhia in a preliminary study and feedback provided. Sufficient evidence exists compared a Garmin Forerunner 310XT with the to support the positive effect of these programs with Runkeeper application for Android smartphones, the use of new technologies, especially when these with the Garmin monitor having the best results in are combined with other research approaches, such distance, speed and EE, while Runkeeper overesti- as face to face interventions. For example, King and mated all exercise parameters. In conclusion, Bort- 15 colleagues6 used three mobile applications to Roig and colleagues suggested that well designed change the physical (in)activity and sedentary studies are needed that comprehensively assess physi- behavior of 95 underactive adults during an 8 cal activity measurement accuracy, however till these week intervention program, and the results provid- days few researches have validated GPS monitors and ed initial support for promoting PA and reducing smartphone GPS-enabled applications. sedentary behavior. Also Laurson, Welk and Eisen- mann7 asked 111 children to wear pedometers over The purpose of this study was to compare the criterion a seven day period and record their steps. Children validity for distance travelled and total energy expen- who wore the pedometers longer appeared more diture (TEE) between a commercially available GPS active, with a significant increase of steps per monitor, Garmin Forerunner 310XT, and a freeware minute. GPS application for Android smartphones, Run- keeper, under semi-structured activity settings. Global Positioning System (GPS) technology, even though it is quite new in PA and exercise, is a low Method cost, objective and discreet way to track individuals’ movement.8 Most previous studies have concluded Research design that it is a valid way to measure distance travelled. An experimental research design was used. The For example, Specht and Szot9 tested six GPS independent variable was the type of GPS monitor: receivers and found that logging receivers, such as Garmin Forerunner 310XT and freeware Runkeeper Garmin Forerunner 310XT, demonstrated the high- Android application. The primary outcome was est accuracy in determining positions. In another distance and the secondary outcome was total energy similar research, four wearable GPS monitors were expenditure (TEE), during walking and running. compared while four adults walked a distance of 1.24 km and the most accurate device was Garmin Forerunner 205.10 Using a similar methodological Participant approach, Lee et al.11 chose four low cost GPS One single, healthy and physically active individual, minotors and found that all devices were valid for who could run at least 5 km continuously, took part distance estimation, with Garmin 60 to be the most in all trials. The same protocol was repeated on 40 accurate. occasions (20 walking and 20 running sessions)

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 29 ORIGINAL ARTICLE

during 40 consecutive days. Relevant characteristics The program has full and efficient functionality such as height (190 cm), weight (80 kg), gender after downloading, with no need for additional (male) and age (33 years) were entered into each software download being necessary. Furthermore monitor separately. The anthropometric measures it can record and upload exercise data to a were obtained at the beginning of the data collec- computer database after been registered to an tion session. Standing height was measured to the online system. It has the ability to track running nearest 0.1 cm with the use of a wall mounted and walking sessions separately and calculate Harpenden Stadiometer (Harpenden, London, energy expenditure based on these activities. Addi- UK), using standard procedures. Body mass was tionally, we made sure from its description that the measured by participants in light clothes and bare application relied solely on GPS for tracking feet on an electronic platform scale (Tanita Corp., the user. Finally, Runkeeper has been identified as Tokyo, Japan) to the nearest 0.1 kg. Following the application with the most applied behavior anthropometric measurements, the participant was changing techniques (eight in number) in the asked to lay down in bed for 10 minutes and then market.18 fitted with the portable indirect calorimeter (Cosmed K4b2) to measure resting energy expendi- Garmin Forerunner 310XT: Garmin, software ver- ture (REE). REE was measured for 15 minutes with sion 4.50 (Garmin Ltd., USA; https://buy.garmin. the subject quiet, but awake. The first 5 min as well com/en-US/US/into-sports/running/forerunner-310 as the last minute of measurement were eliminated XT/prod27335.html), is a GPS-enabled training and and the REE was obtained from the average of 9 heart rate monitor for multisport athletes. Its physi- min. REE measurement was performed after a 10- cal dimensions are 545619 mm, with display hour fast, following previously published guide- size 3320 mm and lightweight, 72 grams. It is lines.16 The REE was expressed as kilocalories water resistant up to 50 meters, with a 20-hour (kcal) per minute by dividing the TEE by 9 and battery life and a memory history of 1000 laps. the estimation was 1.05 kcal  min1. Forerunner tracks time, distance, average and lap speed and pace, heart rate with a premium heart rate monitor, on land and estimated calorie burn. Instruments The main method for EE estimation on Garmin Smartphone - Galaxy S4 mini (Samsung fitness monitors uses the Firstbeat algorithm.19 The Electronics Co., Ltd., Suwon, South Korea): The calculation takes into account the user’s inputted mini uses ’s Android variables including gender, height, weight and Jelly Beam 4.2.2 mobile operating system equipped fitness class. It then combines the data with heart with a built-in GPS receiver, accelerometer and rate information from the heart rate strap. More gyroscope. This phone is small, 125 by 61 by 9 specifically, it evaluates the time between heartbeats mm, and lightweight, 107 grams. (beat to beat) to determine estimated Metabolic Runkeeper Android application: Runkeeper, soft- Equivalent (MET), which in turn is used to deter- ware version 4.4.3 (FitnessKeeper, Inc.; http://Run- mine actual work expenditure. This method is inexpensive and with relatively high accuracy, with keeper.com/) is one of the most commercial 19 applications for Android and iOS smartphones.17 a marginal error of 7-10%. It is considered a Runkeeper is a free of charge application, which reliable method for estimating the rate of oxygen calculates average pace and speed, laps’ speed, route consumption (VO2) and EE, however it may underestimate these parameters by 6% to 13% distance, elevation and estimated calorie burn for a 20 variety of fitness activities, in high accuracy and respectively. Furthermore, this prediction method real time, using the in-built GPS sensor of the may be considered sufficiently accurate to deter- mine the average VO2 in field use, but it does not Android phone. EE is calculated taking into account 21 distance computed, speed, time, elevation, acti- allow precise estimation of VO2. vity type, body weight measurements and REE. Body measurements are entered manually in the application settings. Objective-criterion measurements Distance: The criterion distance was measured by: It has been downloaded 10 Á 50 million times and a. Leica DISTOTM D810 laser range finder, which has a total of 4.5 out of 5 rate out of 421,124 users provides typical measuring accuracy 91 mm and in the Google Play store (as accessed on 21/9/2016). has a range up to 200 m,22 b. Calibrated measuring

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 30 ORIGINAL ARTICLE

wheel, and c. Google Maps Distance Calculator Experimental procedures (http://www.daftlogic.com/projects-google-maps- The study was conducted at an urban outdoor distance-calculator.htm). The three-way estimated environment, on a wooded paved trail with no grade distance course was 3.58 km. (0%) around an Athens central park (Figure 1). This is an area of 49 acres, which formerly was hosting the Energy Expenditure: Indirect calorimetry was uti- military camp ‘Goudi’ and currently is used as a lized in this study as the criterion method for 23 recreational and sports park. Before the start of TEE. The accuracy of estimated TEE by the two every session, both GPS monitors were activated and monitors was compared to that measured breath by the GPS signal was ensured to be acceptable for breath with Cosmed K4b2 (Cosmed S.r.I., Rome, adequate measurement. The participant’s personal Italy) portable metabolic analyzer. Cosmed allows data were entered manually in the smartphone measurement of oxygen consumption under free- application and the GPS monitor prior to the start living conditions providing valid and reliable results 24 of the test. The chest strap for measuring heart rate in the general population. Previous research, in for the GPS unit as well as the calorimeter was which Cosmed was compared with the Douglas placed around the chest of the participant, while 25,26 airbag method and the traditional metabolic Garmin was place on the left wrist and the smart- 27 analyzer Medgraphics D-Series, showed that phone on the left arm. The calorimeter was placed Cosmed provided comparable energy expenditure on a harness over the subject’s shoulders and estimates in stable, submaximal exercise intensities strapped around the subject’s chest and back. The and the results of oxygen consumed, carbon dioxide face mask was placed covering the subject’s mouth eliminated and respiratory quotient ranged within and nose so as to not allow air to escape. acceptable limits of agreement. It comprises an analyzer unit and a facemask. Volume and gas The 3.58 km test began at the quarter mile mark calibrations (4% CO2, 16% O2) were performed (point A, depicted on Figure 1) and was run or before each trial by following manufacturer recom- walked on an out and back course ending at mendations. the measured ending (point B, depicted on

Figure 1: The 3.58 km park route.

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 31 ORIGINAL ARTICLE

Figure 1). The calorimeter needed a brief calibra- Data were analysed using descriptive (mean, stan- tion before being started. This was conducted on the dard deviation) and inferential statistics. A two-way trail prior to the start of the 3.58 km course. When repeated measures ANOVA (measurement meth- the calibration was complete, the 3 devices (Gar- odtype of activity) was performed to detect min, Runkeeper and calorimeter) were started differences in the dependent variables between meth- simultaneously (9 2 sec) and the participant began ods at two intensity levels. Post-hoc analyses using to walk or run. One researcher started the calorim- paired t-tests with the Bonferroni correction were eter and the GPS unit, and the participant started conducted to examine specific differences between the Runkeeper. The participant walked, ran at a the two GPS monitors and the criterion methods. comfortable pace, and self-selected speed. At the Mean absolute percent errors (MAPE) were also end of the 3.58 km course all measuring devices calculated to provide an indicator of overall mea- were stopped simultaneously (9 2 sec) by the same surement error (MAPE[(monitor measurement - researcher who started them. The data from all actual measurement) / actual measurement]100) devices were collected and recorded on a data sheet and was used as an outcome measure. A smaller and, afterwards, data were transferred to the online MAPE represents better accuracy, and less than 3% websites of each monitor. is considered acceptable for distance30 and 10% for TEE.28 This method is a more conservative estimate In total, the participant conducted 20 running and of error that reflects the true error in estimation and 20 walking sessions. The average time needed in provides the most appropriate indicator of overall order to complete the 3.58 km course running was error. Further analyses included Bland-Altman plots, 21 min 5 sec955 sec and walking 38 min 22 sec952 which were calculated to examine the level of sec. The average heart rate, as measured by agreement between each monitor and the criterion Garmin’s heart rate strap, was 155.6593.95 bpm methods across all dependent variables.31 Limits of for running and 99.6097.91 bpm for walking. agreement were calculated as 92SDfromthe overall mean bias between the dependent variables Each monitor uses different outcomes measures to and each GPS monitor. summarize EE data. The Runkeeper provides esti- mates of TEE, while the Garmin reports estimates of activity EE. In order to provide comparable Results estimates in the EE results, it was necessary to add Distance REE to the activity EE values for the Garmin, Table 1 provides descriptive statistics (means, SD) adding the measured REE obtained prior to the for all of the different monitors compared with the activity protocol to the estimated EE. This ensured measured values for distance in the two conditions, that we had comparable outcome measures of TEE walking and running. Separate 3 (2 GPS mon- for both monitors. This procedure was implemented itorscriterion)2 (walking or running) repeated in previous studies28 and ensured that we had measures ANOVA were conducted to examine comparable outcome measures of TEE for both differences in distance obtained from Garmin and monitors. Runkeeper. There was no significant interaction effect between the 3 measurement methods and the 2 types of activities [F(1,21)1.58, p0.23, Statistical analysis h20.08] and no significant main effect between The statistical analysis was conducted with the use walking and running [F(1,19)2.94, p0.10, of the statistical package SPSS version 21.0 (IBM h20.13]. However there was a significant main SPSS Corp., Armonk, NY, USA) and the signifi- effect between the 3 measurement methods cance level was set at pB0.05. Data analysis was based on Bruton, Conway and Holgate29 recom- mendations. Before the main procedures, variables Mean SD Mean SD were screened for accuracy of data entry, missing walking walking running running values, potential outliers and distribution (skewness Measured 3.58 .00 3.58 .00 and kurtosis) for running and walking separately. Runkeeper 3.74 .08 3.70 .11 No missing values were observed and the box plots, Garmin 3.59 .04 3.58 .02 skewness and kurtosis analysis indicated that no extreme values existed and data were approximately normally distributed. Table 1: Descriptive statistics for Distance (km)

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 32 ORIGINAL ARTICLE

Figure 2: Mean absolute percent error (9SD) for distance estimation.

[F(1,21)64.44, pB0.001, h20.77]. Post-hoc while higher values for Runkeeper during running within-subjects contrasts with Bonferroni correc- (mean bias0.12 km; 95% CI0.17 to 0.07 tion revealed that Runkeeper significantly over- km) and walking (mean bias0.16 km; 95% estimated distance [F(1,19)73.75, pB0.001, CI0.20 to -0.12 km) were observed. The slopes h20.80], while Garmin provided almost the for the fitted lines were significant for all measure- same results as the measured distance ments; Runkeeper walking (slope1.93, pB0.001), [F(1,19)0.57, p0.46, h20.03]. Runkeeper running (slope1.96, pB0.001), Gar- min walking (slope1.78, pB0.001) and Garmin The MAPE (computed as the average absolute running (slope1.05, pB0.001), which suggests value of the errors relative to the measured significant patterns of proportional systematic bias for distance) observed for Garmin during the running these monitors. condition was 0.30% and for the walking condition was 0.74%. A larger MAPE was observed for TEE Runkeeper during running (3.28%) and walking (4.43%) (Figure 2). Table 2 provides descriptive statistics (means, SD) for all of the different monitors compared with the Bland-Altman plot analysis showed the distribution of error and assisted with testing for proportional Mean SD Mean SD systematic bias in the estimates. The plots show the walking walking running running residuals of the various estimates on the y-axis (measured distance Á monitor) relative to the mean Cosmed 198.55 6.46 314.05 4.51 of two methods (x-axis). The plots (see Figure 3) Runkeeper 268.80 23.74 333.70 25.01 revealed the narrowest 95% limits of agreement for Garmin 184.60 38.26 259.30 21.95 Garmin during running (mean bias0.00; 95% CI0.01 to 0.01 km) and Garmin during walking Table 2: Descriptive statistics for estimated TEE (kcal), (mean bias-0.01 km; 95% CI0.02 to 0.01 km), with added measured REE for Garmin

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 33 ORIGINAL ARTICLE

measured values from the Cosmed in the two Bland-Altman plot analysis showed the distribution conditions, walking and running. Separate 3 (two of error and assisted with testing for proportional GPS monitorscriterion)2 (walking or running) systematic bias in the estimates. The plots (see repeated measures ANOVA were conducted to Figure 5) revealed the narrowest 95% limits of examine differences in TEE obtained from Garmin agreement for Garmin during walking (mean and Runkeeper application. There was a significant bias13.95 kcal; 95% CI4.45 to 32.35 kcal) interaction effect between the 3 measurement meth- and for Runkeeper during running (mean bias ods and the 2 types of activities [F(2,38)16.10, 19.65 kcal; 95% CI31.10 to 8.20 kcal), while pB0.001, h20.46], a significant main effect higher values for Garmin during running (mean between walking and running [F(1,19)267.47, bias54.75 kcal; 95% CI43.78 to 65.72 kcal) and pB0.001, h20.93] and a significant main effect Runkeeper during walking (mean bias70.25 between the 3 measurement methods kcal; 95% CI81.77 to 58.73 kcal) were [F(2,38)128.74, pB0.001, h20.87]. Post-hoc observed. The slopes for all fitted lines were within-subjects contrasts with Bonferroni correc- significant; Runkeeper walking (slope1.73, tion revealed that Runkeeper significantly over- pB0.001), Runkeeper running (slope1.75, estimated TEE for both conditions pB0.001), Garmin walking (slope1.94, 2 p 0.001) and Garmin running (slope2.03, [F(1,19)44.72, pB0.001, h 0.70], while Garmin B p 0.001), which suggests significant patterns of significantly underestimated TEE for both condi- B proportional systematic bias with these monitors. tions [F(1,19)15.14, p0.001, h20.44].

The MAPE (computed as the average absolute Discussion value of the errors relative to the measured distance) The present study aimed to examine the criterion observed for Garmin during the running condition validity of Garmin Forerunner 310XT GPS monitor was 17.39% and for the walking condition was and freeware Android application Runkeeper for 17.32%. A smaller MAPE was found for Runkeeper distance and TEE estimation, during walking and during running (6.29%) and larger for walking running. To our knowledge, this is the first study (35.52%) (Figure 4). that has tried to compare the accuracy of a

Figure 3: Bland-Altman plots for distance estimation.

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 34 ORIGINAL ARTICLE

Figure 4: Mean absolute percent error (9SD) for TEE estimation. commercially available GPS and a freeware smart- validity and reflected the error that individuals phone application under semi-structured settings. could expect if they were tracking their personal Even though Garmin Forerunner 310XT and Run- activity estimates. Overall distance error estimates keeper have been validated for distance estimation, were similar to results from previous evaluations of the algorithms that they use for EE prediction have wearable monitors and smartphone applica- never been validated previously. Furthermore, stud- tions,12,14 with overall MAPE values typically ies, which compare wearable activity monitors and ranging from 0.30% to 4.43% for both Garmin smartphone PA applications are scarce in the and Runkeeper. Garmin had the smallest MAPE in international literature and only recently some both conditions and was the most precise estimate similar attempts have been conducted.32,33 based on the 95% limits of agreement. Runkeeper had higher errors, overestimating distance; however, A unique feature of this study was the naturalistic these errors of 3% to 4% are acceptable. Taking into design that sought to replicate free-living over- account the small difference between the lower and ground movement. In contrast with traditional upper limit of agreement of 0.10 km, this applica- validation studies,34 the participants were given tion may also be considered valid for distance the option to select the intensity of walking and counting. running they preferred during exercise. The results of the present study support the accuracy of these On the other hand, TEE error estimates were larger, methods for distance recording, however TEE ranging from 6.29% (Runkeeper during running) to estimates had large errors. 35.52% (Runkeeper during walking). Garmin un- derestimated TEE, whereas Runkeeper overesti- A further advancement, as indicated by Bai et al.,35 mated it. The slopes for the fitted lines in the was that both individual-level and group-level Bland-Altman plots were all significant due to at accuracy in distance and TEE estimation were least two outliers in each graph, which might have evaluated. MAPE values and 95% limits of agree- significantly biased the outcomes of these analyses. ment provided a useful indicator of individual-level Since no previous study has validated the EE of

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 35 ORIGINAL ARTICLE

Figure 5: Bland-Altman plots for TEE estimation (adding REE for Garmin).

GPS devices and applications, no direct comparable not improved with the addition of heart rate results could be made. Most previous studies, which measures to traditional GPS device, in accordance examined the validity of accelerometer-based wear- with previous research findings in accelerometer- able activity trackers (FitbBit, Jawbone, Nike, based monitors.37 A possible explanation for the etc.) came to similar conclusions.28,35,36 Runkeeper limited accuracy of EE prediction could be that the during running gave a MAPE less than 10%, algorithm used could not accurately compute EE making it accurate in TEE under this condition, based on distance and heart rate, even though these while MAPE during walking was significantly high- initial raw data were very accurate. Technical er. Garmin was more stable during the two condi- assistance was sought from the company to ascer- tions, with an overall medium to large MAPE of tain specific information regarding the algorithm 17%. used to determine EE, however this information was not disclosed. Mean bias and repeated measure ANOVA provided alternative indicators of group-level accuracy. The The present study had some limitations. Only one mean bias and repeated measure ANOVA results healthy participant performed all activities and thus we did not account for different potential con- favored only Garmin for distance estimation, which founding factors such as BMI, monitor placement, provided group-level validity. Both the monitor and gender and age, which could potentially influence the application had low group-level accuracy for accuracy. In addition, we examined the accuracy TEE estimation, differing significantly from the during walking and running in outdoor terrain, so criterion measurement. the results cannot be generalized in other settings. Lastly, these estimates were obtained for a specific A novel finding is that Garmin’s TEE estimation distance and may not reflect accuracy for longer was less accurate than expected. Garmin, in order distances, in example for a marathon race. to estimate TEE, uses the Firstbeat algorithm,19 which combines the data obtained from the heart rate strap. Previous research showed that this might Conclusion underestimate EE by 6% to 13%.20 In this study, the In conclusion, the present study offers initial error was 17%. The accuracy of EE predictions was evidence for the validity of GPS technology of PA

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 36 ORIGINAL ARTICLE

monitors and smartphone applications for measur- 5. Griffiths F, Lindenmeyer A, Powell J, Lowe P, ing distance travelled during walking and running. Thorogood M. Why are health care interventions However, estimates of EE were low, except for delivered over the internet? A systematic review of the Runkeeper during running which provided accept- published literature. J Med Internet Res 2006;8(2):e10. doi: 10.2196/jmir.8.2.e10 able error. This limits their use for monitoring energy balance, and therefore as a weight manage- 6. King AC, Hekler EB, Grieco LA, Winter SJ, Sheats ment tool. These results can assist consumers, JL Buman MP, et al. Effects of three motivationally researchers and health care providers to make targeted mobile device applications on initial physical evidence-based choice for a GPS PA monitor to activity and sedentary behavior change in midlife and measure distance during exercise. People who own older adults: A randomized trial. PLoS ONE Android smartphones have a valid alternative for 2016;11(6):e0156370. doi: 10.1371/journal.pone.0156370 distance estimation during walking and running 7. Laurson KR, Welk GJ, Eisenmann JC. Estimating with this freeware GPS application. Caution should physical activity in children: Impact of pedometer wear be made when mixing the results from different time and metric. JPhysActHealth2015;12:124Á31. activities into one estimated measure, due to http://dx.doi.org/10.1123/jpah.2013-0111. probable cancelling of overestimation and underes- timation from these activities, which may lead to an 8. Duncan MJ, Badland HM, Mummery WK. Applying illusion of improved accuracy. GPS to enhance understanding of transport-related physical activity. J Sci Med Sport 2009;12:549Á56. doi: 10.1016/j.jsams.2008.10.010 No competing interests 9. Specht M, Szot T. Accuracy analysis of GPS sport All authors have completed the Unified Competing receivers in dynamic measurements. An Nav Interest form at www.icmje.org/coi_disclosure.pdf 2012;19:165Á76. doi: 10.2478/v10367-012-0013-9 (available on request from the corresponding author) and declare: no support from any organisa- 10. Wieters KM, Kim JH, Lee C. Assessment of tion for the submitted work; no financial relation- wearable global positioning system units for physical ships with any organisations that might have an activity research. J Phys Act Health 2012;9:913Á23. interest in the submitted work in the previous 3 11. Lee K, Kim JY, Putti K, Bennett DH, Cassady D, years; no other relationships or activities that could Hertz-Picciotto I. Use of portable Global Positioning appear to have influenced the submitted work. System (GPS) devices in exposure analysis for time- location measurement. Kor J Envir Health Sc References 2009;35(6):461Á67. doi: 10.5668/JEHS.2009.35.6.461 1. World Health Organization. Prevalence of insufficient physical activity: Situation and trends. 2014. Availiable 12. Bauer C. On the (in-)accuracy of GPS measures of at http://www.who.int/gho/ncd/risk_factors/physical_ smartphones: A study of running tracking applica- activity_text/en/ (accessed 25 June 2016). tions. In: Proceedings of MoMM ’13 International Conference on Advances in Mobile Computing & 2. Reither EN, Hauser RM, Yang Y (2009). Do birth Multimedia. New York, NY, USA: ACM 2013:335- cohorts matter? Age-period-cohort analyses of the 40. doi:10.1145/2536853.2536893. obesity epidemic in the . Soc Sci Med 2009;69(10):1439Á48. doi: 10.1016/j.socscimed.2009. 13. Mallula CL. Comparing Garmin Forerunner 405CX 08.040 GPS and Nike iPod to accurately measure energy expenditure, distance and speed of overground run- 3. Piepoli MF, Hoes AW, Agewall S, Albus C, Brotons ning. Master’s Thesis, Department of Health, Physical C, Catapano AL, et al. European Guidelines on Education, Recreation and Dance, Cleveland State cardiovascular disease prevention in clinical prac- University. 2010. Availiable at http://rave.ohiolink. tice: The Sixth Joint Task Force of the European edu/etdc/view?acc_numcsu1273708884 (accessed 3 Society of Cardiology and Other Societies on September 2016). Cardiovascular Disease Prevention in Clinical Prac- tice. Eur Heart J 2016;37:2315Á81. doi: 10.1093/ 14. Adamakis M, Zounhia K. Comparing the validity eurheartj/ehw106 and output of the Garmin Forerunner 310XT and Runkeeper android application in an urban environ- 4. Bardus M, Smith JR, Samaha L, Abraham, C. Mobile ment. In: Dasheva D, Antala B, Djobova S, Kuleva phone and web 2.0 technologies for weight manage- M, (Eds.), Physical Education and sport Á Compe- ment: A systematic scoping review. J Med Internet Res tences for life: Book of abstracts of the 9th FIEP 2015;17(11):e259. doi: 10.2196/jmir.5129 European congress. Sofia, Bulgaria: FIEP 2014:59Á60.

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 37 ORIGINAL ARTICLE

15. Bort-Roig J, Gilson ND, Puig-Ribera A, Contreras measurement system. Med Sci Sports Exerc RS, Trost SG. Measuring and influencing physical 2001;333(Suppl. 5):S300. activity with smartphone technology: A systematic review. Sports Med 2014;44(5):671Á86. doi: 10.1007/ 26. Eisenmann JC, Brisko N, Shadrick D, Welsh S. s40279-014-0142-5 Comparative analysis of the Cosmed Quark b2 and K4b2 gas analysis systems during submaximal ex- 16. Compher C, Frankenfield D, Keim N, Roth-Yousey ercise. J Sports Med Phys Fitness 2003;43(2):150Á55. L. Best practice methods to apply to measurement of resting metabolic rate in adults: a systematic review. 27. Schrack JA, Simonsick EM. Ferrucci L. Comparison of J Am Diet Assoc 2006;106(6):881Á903. doi: 10.1016/ the Cosmed K4b2 portable metabolic system in measur- j.jada.2006.02.009 ing steady-state walking energy expenditure. PLoS ONE 2010;5(2):e9292. doi: 10.1371/journal.pone.0009292 17. Duarte L, Ribeiro P, Guerreiro T, Carrico L. Defining a design space for persuasive cooperative 28. Lee JM, Kim Y, Welk GJ. Validity of consumer- interactions in mobile exertion applications. Collab based physical activity monitors. Med Sci Sports Tech 2014;8658:105Á12. Exerc 2014;46(9):1840Á48. doi: 10.1249/MSS. 0000000000000287 18. Middelweerd A, Mollee JS, van der Wal CN, Brug J, te Velde SJ. Apps to promote physical activity among 29. Bruton A, Conway JH, Holgate ST. Reliability: adults: A review and content analysis. Int J Behav What is it, and how is it measured? Physiotherapy Nutr Phys Act 2014;11(97):1Á9. doi: 10.1186/s12966- 2000;86(2):94Á9. 014-0097-9 30. Crouter SE, Schneider PL, Karabulut M, Bassett 19. Firstbeat Technologies. An energy expenditure esti- DRJr. Validity of 10 electronic pedometers for mation method based on heart rate measurement. measuring steps, distance, and energy cost. Med Sci Firstbeat Technologies Ltd, 2012. Availiable at Sports Exerc 2003;35(8):1455Á60. http://www.firstbeat.com/userData/firstbeat/Energy_ Expenditure_Estimation.pdf (accessed 16 August 31. Bland JM, Altman DG. Statistical methods for 2016). assessing agreement between two methods of clinical measurement. Lancet 1986;8:307Á10. 20. Montgomery PG, Green DJ, Etxebarria N, Pyne DB, Saunders PU, Minahan CL. Validation of heart rate 32. Adamakis M. Preliminary study of consumer-level monitor-based predictions of oxygen uptake and activity monitors and mobile applications for step energy expenditure. J Strength Cond 2009;23(5):1489Á counting under free living conditions. J Mob Technol 95. doi: 10.1519/JSC.0b013e3181a392777 Med 2016;5(3):1Á8. doi: 10.7309/jmtm.5.3.x

21. Smolander J, Ajoviita M, Juuti T, Nummela A, Rusko 33. Case MA, Burwick HA, Volpp KG, Patel MS. H. Estimating oxygen consumption from heart rate Accuracy of smartphone applications and wearable and heart rate variability without individual calibra- devices for tracking physical activity data. JAMA tion. Clin Physiol Funct Imaging 2011;31(4):266Á71. 2015;313(6):625Á26. doi: 10.1001/jama.2014.17841 doi: 10.1111/j.1475-097X.2011.01011.x 34. Boyce G, Padmasekara G, Blum M. Accuracy of 22. Leica Geosystems (2015). Leica DISTOTM D810 touch: pedometer technology. J Mob Technol The smartest solution to measuring and documenting. Med 2012;1(2): 16Á22. doi: 10.7309/jmtm.13 Available at http://www.leica-geosystems.com/en/ Leica-DISTO-D810-touch_104560.htm (accessed 35. Bai Y,Welk GJ,Nam YH, Lee JA, Lee JM, Kim Y,et al. 15 August 2016). Comparison of consumer and research monitors under semistructured settings. Med Sci Sports Exerc 2015; 23. Hills AP, Mokhtar N, Byrne NM. Assessment of 48(1):151Á158. doi: 10.1249/MSS.0000000000000727 physical activity and energy expenditure: an overview of objective measures. Front Nutr 2014;1(5):1Á16. 36. Ferguson T, Rowlands AV, Olds T, Maher C. The doi: 10.3389/fnut.2014.00005 validity of consumer-level, activity monitors in healthy adults worn in free-living conditions: A 24. McLaughlin JE, King GA, Howley ET, Bassett cross-sectional study. Int J Behav Nutr Phys Act DRJr, Ainsworth BE. Validation of the COSMED 2015;12:42. doi: 10.1186/s12966-015-0201-9 K4b2 Portable Metabolic System. Int J Sports Med 2001;22:280Á84. 37. Wallen MP, Gomersall SR, Keating SE, Wisloff U, Coombes JS. Accuracy of heart rate watches: Im- 25. Parr BB, Strath SJ, Bassett DR, Howley ET. plications for weight management. PLoS ONE 2016; Validation of the Cosmed K4b2 portable metabolic 11(5):e0154420. doi: 10.1371/journal.pone.0154420

#JOURNAL OF MOBILE TECHNOLOGY IN MEDICINE VOL. 6 | ISSUE 2 | AUGUST 2017 38