- Email: [email protected]
Validity of Commercial Activity Trackers in Children With Congenital Heart Disease
Accepted Manuscript Validity of commercial activity trackers in children with congenital heart disease Christine Voss, PhD, Ross F. Gardner, Paige H. Dean, BSc, Kevin C. Harris, MD, MHSc PII:
S0828-282X(16)31147-3
DOI:
10.1016/j.cjca.2016.11.024
Reference:
CJCA 2325
To appear in:
Canadian Journal of Cardiology
Received Date: 12 August 2016 Revised Date:
14 November 2016
Accepted Date: 28 November 2016
Please cite this article as: Voss C, Gardner RF, Dean PH, Harris KC, Validity of commercial activity trackers in children with congenital heart disease, Canadian Journal of Cardiology (2017), doi: 10.1016/ j.cjca.2016.11.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
TITLE PAGE
RI PT
Validity of commercial activity trackers in children with congenital heart disease
Christine Voss, PhD; Ross F Gardner; Paige H Dean, BSc; Kevin C Harris, MD, MHSc
Dr. Kevin Harris Children’s Heart Centre BC Children’s Hospital 1F27 – 4480 Oak Street Vancouver, BC V6H 3V4 Phone: 604-875-3878 Fax: 604-875-3463 Email: [email protected]
TE D
M AN U
Correspondence:
SC
Division of Cardiology, Department of Pediatrics, University of British Columbia, Canada (all authors)
EP
Short title: Activity tracker validation in children with CHD
AC C
Word count excluding brief summary, word count, and short title: 4,734
1
ACCEPTED MANUSCRIPT
BRIEF SUMMARY In this validation study of a commercial activity tracker in children, including those with CHD,
RI PT
we found that the device provides results that were generally reflective of overall physical activity levels and can identify those meeting physical activity guidelines. These findings are important given the increasing emphasis on physical activity promotion and monitoring in
AC C
EP
TE D
M AN U
SC
children with cardiovascular risk factors.
2
ACCEPTED MANUSCRIPT
ABSTRACT Background: Increasing physical activity levels is a high priority to optimize long-term health in
RI PT
children with congenital heart disease (CHD). Commercial activity trackers have been validated in adults and are increasingly used to measure and promote physical activity in pediatric populations, but they have not been validated in children.
Methods: In 30 children with CHD aged 10-18yrs, we assessed the validity of physical activity
SC
form the wrist-based Fitbit Charge HRTM against hip-based ActiGraph accelerometers under free
M AN U
living conditions for 7 days. We assessed the association between devices by intra-class correlation coefficients and Bland-Altman plots. Receiver-operating curves were used to identify Fitbit step cut-points.
Results: There was a strong association between the two devices for daily steps across 138
TE D
analysed person-days (ICC=0.855, p<0.001), but poorer agreement for time spent in physical activity intensities (ICCs<0.7). Daily Fitbit steps of ≥12,500 identified meeting physical activity guidelines defined as ≥60 min MVPA/day. Fitbit devices recorded more steps than
EP
accelerometers (-2,242 steps/d, 95%LoA of -7,738 to 3,253). Between-device differences were greater in boys vs. girls. Fitbit devices were worn for longer than accelerometers (-36 min/d,
AC C
95%LoA -334 to 261), but overall differences in wear time explained little of the variance in step differences (7%, p=0.048).
Conclusions: Commercial activity trackers provide opportunities to remotely monitor physical activity in children with CHD, but absolute values may differ from accelerometers. These findings are important given the increasing emphasis on physical activity promotion and monitoring in children with cardiovascular risk factors.
3
ACCEPTED MANUSCRIPT
KEY WORDS
AC C
EP
TE D
M AN U
SC
RI PT
Congenital Heart Disease, Children, Pediatric, Physical Activity, Fitbit, ActiGraph
4
ACCEPTED MANUSCRIPT
INTRODUCTION Physical activity is increasingly recognized as an important behaviour to optimize long-term cardiovascular health and quality of life in children with congenital heart disease (CHD),1,2 who
RI PT
are at increased cardiovascular risk.3 Commercial activity trackers hold promise as a tool to measure, prescribe and promote physical activity. Several studies are already taking advantage of commercial trackers to measure and/or promote physical activity in both healthy and clinical
SC
pediatric populations,4-6 including children with CHD.7 While the validity and reliability of
M AN U
commercial trackers has been extensively documented in adults,8 no previous research has established the validity of such devices in children. This is an important shortcoming, given that it is unclear if children’s sporadic bursts of activity9 are adequately quantified by commercial trackers that are designed for adults and optimized to capture walking.10 The objective of the current study was to assess the validity of physical activity from a commercial activity tracker
Sample
EP
METHODS
TE D
against a research-grade accelerometer under free living conditions in children with CHD.
AC C
We included a convenience sample of 40 participants aged 10-18yrs recruited from the Children’s Heart Centre at BC Children’s Hospital in Vancouver, BC and provincial partnership clinics. We obtained institutional ethics approval, parental consent and participant assent. Cardiac diagnoses were: 1) CHD; categorized as mild, moderate or severe11, 2) cardiac transplant, 3) electrophysiological disorders, or 4) none (e.g. innocent murmur). Nurses measured height (0.1 cm) and weight (0.1 kg). BMI was expressed according to WHO norms12 and categorized according to International Obesity Task Force criteria.13
5
ACCEPTED MANUSCRIPT
Criterion measure: ActiGraph accelerometer We fitted participants with an ActiGraph accelerometer (GT3X+, GT9X; ActiGraph LLC, Pensacola, FL; sampling at 30Hz) to be worn over the right hip and instructed them to wear it
RI PT
continuously for the next 7 days and to only remove it for water-based activities,
showering/bathing and sleep. The ActiGraph is widely used to objectively measure physical
SC
activity in children under free-living conditions.14
Test measure: Fitbit activity tracker
M AN U
We chose Fitbit Charge HRTM (Fitbit) for our study. Studies in children have exclusively used Fitbit models (without validation),4-7 and we expected that the appeal of using one of the market leaders would benefit compliance in our study. Wristband size and placement were in accordance with manufacturer’s guidelines. We asked participants to wear the Fitbit continuously during the
TE D
same 7 days as the accelerometer. We created anonymous Fitbit user profiles (online
valid in children).
Data processing
EP
‘dashboard’) and disabled displays of ‘calories burned’ and ‘distance travelled’ (unlikely to be
AC C
We used ActiLife v.6.13.2 (ActiGraph LLC, Pensacola, FL) to analyse accelerometry files (15s epoch). Valid days were defined as ≥600 min/d wear time (allowing for ≤60min of zero activity) and moderate-to-vigorous physical activity was calculated (MVPA; ≥2296 CPM; includes vigorous).15 Sedentary time was expressed as a percentage relative to wear time.16 Mean physical activity values were based on ≥1 valid days (inclusion criterion for validation analyses); we also used conservative wear time criteria of ≥3 weekdays and ≥1 weekend day(s) to estimate general physical activity levels.17 We defined meeting guidelines as mean daily MVPA ≥60 min/d.18 For
6
ACCEPTED MANUSCRIPT
Fitbit data, we used the ‘Data Export’ function on the ‘dashboard’ to extract steps and physical activity intensities (sedentary, light, fair, and very). MVPA was defined as ‘fair + very’ and sedentary time was expressed relative to wear time. A sub-sample were managed through
RI PT
Fitabase (Small Steps Labs LLC, San Diego, CA; www.fitabase.com), a fee-for-service online platform that facilitates access to multiple Fitbit devices simultaneously and provides high
SC
resolution minute-by-minute data exports.
We created a person-day dataset of matched ActiGraph and Fitbit data on valid accelerometry
M AN U
days (≥600 min/d). For a subset of person-days with high-resolution data (n=20), we estimated Fitbit wear time using a modified data analyses package in ‘R’ that processes minute-by-minute accelerometry data (‘accelerometry’, v. 2.2.5, May 2015). First, we re-produced the ActiLife wear time results for accelerometry data by this method, and yielded similar results (815 vs. 809
TE D
min/d). Second, we repeated step 1 using accelerometry steps rather than ‘counts’ to establish that steps can serve as a wear time metric, and yielded similar results (821 vs. 815 min/d). Third, we analysed the minute-by-minute Fitbit steps to determine Fitbit wear time.
EP
Statistical Analysis
AC C
Descriptive statistics (frequencies (%), mean±SD, median (IQR)) were calculated for applicable variables. Between-sex differences were assessed by independent t-tests, Mann-Whitney U Ranked Sum test, or Pearson’s chi-squared test. Associations and agreement between devices were assessed by intra-class correlation coefficients (ICC) and Bland-Altman plots. The between-device difference (bias and 95% Limits of Agreement, LoA) was defined as ‘criterion measure (accelerometer)’ minus ‘test measure (Fitbit)’. The mean absolute percent error was calculated as ‘mean between-device difference / mean criterion measure *100’. Receiver
7
ACCEPTED MANUSCRIPT
operating curves (ROC) were drawn using Fitbit steps as the classifier and meeting physical activity guidelines (≥60 min/d of MVPA from accelerometry) as the true-status reference; the coordinate with the greatest sum of sensitivity and specificity was used to identify the Fitbit step
RI PT
cut-point with discriminatory value.19 Because meaningful between-day variations in physical activity levels are expected within individuals, we treated valid days as independent observations and used a person-day dataset using actual days rather than hypothetical mean days for validation
SC
analyses. We conducted sensitivity analyses using a reduced dataset (mean values per person) to rule out undue influences of related observations within individuals. Analyses were carried out
M AN U
using Stata (v.14.1, Stata Corp LP, College Station, TX) or R (x64 3.3.1, The R Foundation for Statistical Computing) and significance was set at p<0.05.
RESULTS
TE D
Sample characteristics
Thirty participants had at least one day of matching accelerometry and Fitbit data. Four participants did not wear both devices at the time and 6 did not have any valid data, which is
EP
unsurprising considering that non-compliance with accelerometer wear is commonly observed in adolescents.20,21 There were no significant differences for sample characteristics between the
AC C
analytical sample and those who were excluded, nor were there significant sex-differences (Table 1). Similarly to what was previously reported for children with CHD,22 23% were overweight or obese, which is comparable to national data in healthy children.21 Mean MVPA (based on ≥1 valid day) was 45 min/d, which was similar to national data for similarly aged children and adolescents (~50 min).21 Physical activity levels were not meaningfully different when computed according to conservative wear time inclusion criteria of ≥3 weekdays and
8
ACCEPTED MANUSCRIPT
≥1weekend day (e.g. 47 min/d MVPA, available for n=24 individuals). We acknowledge a lack of expected sex-differences in MVPA, which was likely explained by inter-personal differences and the wide age range of our relatively small sample. Because the aim of the current study was
RI PT
to assess the agreement between two devices, we do not consider the heterogeneous physical activity levels of our sample a limitation. We defined adherence to physical activity guidelines as
SC
mean daily MVPA ≥60 min/d and 25% met physical activity guidelines.
Validity: association and agreement between ActiGraph accelerometer and Fitbit
M AN U
We included 138 valid person-days (mean per person: 4.6 days, range: 1-7). We conducted detailed assessments between Fitbit and accelerometer results for physical activity metrics, including steps and time spent in physical activity intensities. The strongest between-device agreement was observed for daily steps (ICC=0.855, p<0.001; Figure 1), which did not vary
TE D
notably between sexes or age groups. However, the Fitbit devices recorded more steps than the accelerometers (10,383±5,321 vs. 8,141±4,023 steps/d; p<0.001). The mean bias was -2,242 steps/d, although between-device differences were highly varied (95%LoA of -7,738 to 3,253;
EP
Figure 1). The mean absolute percent error was 28%, which is higher than the ≤10% error that is desired for valid step estimates under free-living conditions.23 The mean bias for daily steps was
AC C
greater in boys (-3,320 steps/d; 95%LoA -7,643 – 1,003; 39% MAP) than girls (-1,389 steps/d; 95%LoA -7,133 – 4,356; 18% MAP; Figure 1). An example day illustrating differences in hourly and cumulative steps between devices is found in Figure 2.
Compared with steps, agreement between devices was notably lower for time spent in physical activity intensities (all ICCs<0.66; online supplement Table S1 and Figure S1). Discrepant results are likely explained by methodological differences. For example, the Fitbit records only
9
ACCEPTED MANUSCRIPT
MVPA that is accrued in bouts of ≥10 min duration,24 which is in line with adult physical activity guidelines25 but not current guidelines for children18 and thus not how we calculated MVPA based on accelerometry. This explains person-days where the accelerometer registered
RI PT
some MVPA but the Fitbit did not. In contrast, the Fitbit registered substantially more MVPA than the accelerometer on other person-days, which was almost exclusively observed in younger children. The proprietary Fitbit algorithm categorises MVPA as ≥3METs, which in turn is
SC
derived from estimated caloric expenditure based on movement, age, sex, height and weight.24 Energy cost of activity is greater in children than in adults,26 and it is unclear if and how the
M AN U
Fitbit algorithm corrects for this in our younger participants. Fitbit also states that the heart rate feature does ‘a better job of recognizing active minutes’,26 but it is not disclosed how this is done. Sedentary time was the only activity intensity metric that yielded agreement between devices that was comparable to steps (ICC=0.845, p<0.001). However, we faced challenges
TE D
regarding data availability and the need for additional data manipulation when the Fitbit was not worn over night (see online supplement for details) which currently limits the usefulness of this metric. All analyses were performed on person-day datasets that were unadjusted for multiple
EP
observations per person; we found similar results using a reduced person-level data-set (not
AC C
shown), confirming that no meaningful within-person effects necessitated adjustments.
Diagnostic values of Fitbit devices to identify adherence to physical activity guidelines A ROC curve was created using Fitbit steps as the classifier and meeting physical activity guidelines (≥60 min MVPA based on accelerometry) as the true-status reference (Figure 3). The ROC area under the curve was 0.82 (95%CI 0.74-0.90). The coordinate with the greatest sum of sensitivity and specificity (cut-point that identifies meeting guidelines) was 12,562 steps/d.
10
ACCEPTED MANUSCRIPT
Device wear time analyses and its effect on validity
RI PT
Out of a possible 20 participants who had their devices managed through Fitabase, high resolution minute-by-minute Fitbit data were available for 15 participants and 55 person-days. Infrequent Fitbit syncing resulted in some loss of high resolution data. Validation results were
SC
similar in this sub-sample compared with the entire sample (ICC=0.908; bias -2,236). On average, Fitbits were worn for 36 minutes longer than ActiGraphs (Fitbit 866 min/d vs.
M AN U
accelerometer 830 min/d), although differences in wear time were highly varied (95%LoA -334 to 261). Girls wore both devices for longer (844 min/d vs. Fitbit 906 min/d) compared with boys (ActiGraph: 813 min/d vs. Fitbit 821 min/d), although neither sex-difference was statistically significant (ActiGraph p=0.288 and Fitbit p=0.060). Differences in wear time between devices
TE D
tended to be greater in girls (-62 min/d, 95%LoA -294 to 170) than in boys, but the variance was also greater in boys(-7.8 min/d, 95%LoA -360 to 344). Within either device, wear time was not related to steps (ActiGraph r2=0.015, p=0.368; Fitbit r2=0.003, p=0.708). Between-device
EP
differences in wear time was a significant correlate of between-device differences in steps, but explained only 7% of the variance (r2=0.072, p=0.048). On days were the two devices had
AC C
similar wear time (n=12; ±30 min) the bias in steps was -1,854 (95%LoA -4,815 to 1,107).
DISCUSSION
In children with CHD, we assessed the validity of the Fitbit Charge HRTM - a wrist-worn commercial tracker - to estimate physical activity. We demonstrated a strong correlation between the Fitbit and the ActiGraph for daily steps, and importantly, we identified a cut-point of ~12,500
11
ACCEPTED MANUSCRIPT
Fitbit steps/d to correspond to meeting physical activity guidelines (≥60 minutes MVPA/d from accelerometry), which will have utility for clinical practice and future research.
RI PT
Physical activity – a cornerstone of health Physical activity during childhood is of paramount importance for optimal cardio-metabolic health27 and mental well-being,28 and active children are more likely stay active as adults.29 We
SC
are facing a physical inactivity crisis where fewer than 1 in 10 Canadian children are meeting physical activity guidelines.21 For children with CHD, who are at high cardiovascular risk as a
M AN U
consequence of their condition,3 there has been an urgent call for action to promote physical activity to optimize long-term cardiovascular health and quality of life in these individuals.1,2 Innovative strategies are needed to engage these children and their families in health promotion.
Utility of commercial activity trackers in children in clinical practice and research
TE D
Exercise training programs are effective at improving physical activity levels in children with CHD,30 but access to such programs is typically limited to those residing near metropolitan areas. Considering the relevance of technology for our young patients, the role of commercial activity
EP
trackers for remote / home-based physical activity promotion is already being explored.4,6,7
AC C
Commercial trackers such as the Fitbit offer several advantages over research-grade activity devices: 1) aesthetically pleasing and user-friendly hardware; 2) appealing user-experience through the online ‘dashboard’; 3) cost-effective and readily available; 4) no specialist software and/or data processing expertise necessary; 5) wrist-based, a preferred device location in children and youth.31 In addition, services such as Fitabase can streamline the remote management of multiple users at once. This is important to effectively facilitate exercise prescription, where
12
ACCEPTED MANUSCRIPT
continuous remote monitoring of participant activity levels are needed in order to provide tailored feedback and recommendations.
RI PT
Validity of Fitbit devices in children The target consumer group for commercial trackers are adults, whose most common activity is walking.32 Fitbit algorithms are “designed to look for motion patterns most indicative of people
SC
walking”.10 No previous research has validated the Fitbit in children, but adult studies
consistently demonstrate good validity for steps under free-living conditions.8,33-35 Children
M AN U
engage in sporadic bouts of high intensity activity,9 and it is unclear to what extent, if any, these differences in activity patterns may impact algorithms to accurately identify physical activity. We found agreement between devices was most consistent for daily steps – a physical activity metric that most closely resembles total activity rather than intensity. The poorer agreement for time spent in physical activity intensities is in part explained by the Fitbit only recording MVPA
TE D
bouts of ≥10 min duration (in line with adult guidelines),24 and the interpretation of Fitbit MVPA being further complicated by undisclosed proprietary algorithms drawing on estimates of caloric
EP
expenditure that may not be valid in children.
AC C
Sex differences in Fitbit validity: wear time or physical activity? Between-device step differences in steps were nearly twice as great in boys as in girls. We investigated whether this was explained by poorer compliance in boys. Girls tended to wear both devices for longer than boys, but it is the girls who were less consistent with wearing both devices equally, and clearly favouring the Fitbit. Regardless, wear time differences overall explained only a small portion of the variance in step differences (7%), and a sub-analyses of person-days with similar device wear time still registered a substantial between-device step
13
ACCEPTED MANUSCRIPT
difference (-1,854). Thus, we do not believe that device wear time is an important factor affecting our validation results.
RI PT
Physical activity volume and patterns could instead play a role in our observed sex-differences. Our physical activity data hints at well-documented sex-differences in physical activity, with boys generally being more active than girls.36 In addition to more sporadic bursts of high
SC
intensity activity in children compared with adults,9 boys commonly play games, whereas girls frequently walk for leisure.37 Walking is an activity that Fitbits are optimized to capture,10 which
M AN U
could explain the more favourable step results in girls. The notion that boys may engage more in higher intensity, sustained play/games is supported by our finding that between-device agreement for MVPA was good for boys (ICC=0.824), but poorer for the sample overall (ICC=0.657; Table S1).
TE D
Strengths, limitations and future directions
This is the first validation study of commercial activity trackers in children. Our analyses identified the daily Fitbit steps as the most suitable metric for use in children, in line what has
EP
been found for adults.8 Our sample was a convenience sample of patients attending pediatric
AC C
cardiology clinics and most had CHD. We previously found that our CHD patients are as active as healthy Canadians of the same age and sex, and thus believe the current findings are of relevance to the wider pediatric community.
We acknowledge that there was considerable variance in our results. For example, we found that while the Fitbit recorded on average 2,242 more steps than accelerometers did, 95%LoA indicated that there was considerable variability. A certain degree of variance is to be expected in
14
ACCEPTED MANUSCRIPT
a context where children are tasked with collecting behavioural data under free living conditions, and are acceptable in light of the strong correlations we observed.
RI PT
It is a limitation that we did not assess the validity of the Fitbit against a direct measure of physical activity (e.g. direct observation) and/or energy expenditure (e.g. indirect calorimetry), but instead against a widely used accelerometer14 and associated MVPA cut-points that are based
SC
on indirect calorimetry in children.15 Adult validation studies reported different device
performance between controlled, laboratory-based treadmill walking and running versus free-
M AN U
living conditions;8,35 thus, validation under free-living conditions is clearly warranted. However, (unknown) differences in device wear time can impact validation studies under such conditions. We used algorithms to identify non-wear from a lack of movement signal, but different devices may have different thresholds as to what acceleration signal constitutes a step, for example. More
TE D
direct measures of wear time are desirable, such as skin sensors in the newer ActiGraph models (optimized for wrist placement). For Fitbits, the continuous heart rate feature has potential to identify wear time, but access to the fee-for-service Fitabase and advanced data processing skill
EP
are required.
We did not assess intra- or inter-device reliability of the Fitbit, but this has been reported
AC C
elsewhere.e.g. 38 There are currently no published studies of the Fitbit Charge HRTM, likely due to the discrepancy between the fast-paced release of new commercial trackers and the lag in research and peer-reviewed scientific publication. Fewer adult studies have assessed the newer wrist-based Fitbit models, and although they performed well, they were typically inferior to the hip-based models.e.g. 35 It is possible that hip-based Fitbit models yield more comparable results
15
ACCEPTED MANUSCRIPT
to the ActiGraph in children. However, children favour wrist-based over hip-based devices,31 and the importance of compliance should be carefully evaluated against the need for accuracy.
RI PT
CONCLUSION In children with CHD, commercial activity trackers can provide general estimates of physical activity using daily step counts, and a cut-point of ~12,500 Fitbit steps/d relates to meeting
ACKNOWLEDGEMENTS
M AN U
prescribe and monitor physical activity in children.
SC
physical activity guidelines (defined as ≥60 min MVPA/d). Fitbits may be useful to remotely
We are grateful to the participants and their families who participated in this research. We thank the nurses, administrative staff and research staff at the Children’s Heart Centre and our
FUNDING SOURCES
TE D
partnership clinics for their support.
EP
We received funding form the Heart and Stroke Foundation of Canada (F1504) and the Child &
AC C
Family Research Institute, University of British Columbia.
REFERENCES
1. Takken T, Giardini A, Reybrouck T, et al. Recommendations for physical activity, recreation sport, and exercise training in paediatric patients with congenital heart disease: a report from the Exercise, Basic & Translational Research Section of the European Association of Cardiovascular Prevention and Rehabilitation, the European Congenital Heart and Lung Exercise Group, and the Association for European Paediatric Cardiology. Eur J Prev Cardiol. 2012; 19:1034-65.
16
ACCEPTED MANUSCRIPT
2. Longmuir PE, Brothers JA, de Ferranti SD, et al. Promotion of physical activity for children and adults with congenital heart disease: a scientific statement from the American Heart Association. Circulation. 2013; 127:2147-59.
RI PT
3. Kavey RE, Allada V, Daniels SR, et al. Cardiovascular risk reduction in high-risk pediatric patients: a scientific statement from the American Heart Association Expert Panel on Population and Prevention Science; the Councils on Cardiovascular Disease in the Young, Epidemiology and Prevention, Nutrition, Physical Activity and Metabolism, High Blood Pressure Research, Cardiovascular Nursing, and the Kidney in Heart Disease; and the Interdisciplinary Working Group on Quality of Care and Outcomes Research: endorsed by the American Academy of Pediatrics. Circulation. 2006; 114:2710-38.
M AN U
SC
4. Gibson C, August K, Greene J, et al. A televideo exercise and nutrition program for children with acute lymphoblastic leukemia in maintenance therapy: design and methods. Open Access Journal of Clinical Trials. 2015; 7:77-84. 5. Hayes LB, Van Camp CM. Increasing physical activity of children during school recess. J Appl Behav Anal. 2015; 48:690-5. 6. Hooke MC, Gilchrist L, Tanner L, Hart N, Withycombe JS. Use of a Fitness Tracker to Promote Physical Activity in Children With Acute Lymphoblastic Leukemia. Pediatr Blood Cancer. 2016; 63:684-9.
TE D
7. Jacobsen RM, Ginde S, Mussatto K, et al. Can a Home-based Cardiac Physical Activity Program Improve the Physical Function Quality of Life in Children with Fontan Circulation? Congenit Heart Dis. 2016; 11:175-82.
EP
8. Evenson KR, Goto MM, Furberg RD. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act. 2015; 12:159.
AC C
9. Welk GJ, Corbin CB, Dale D. Measurement issues in the assessment of physical activity in children. Res Q Exerc Sport. 2000; 71:S59-73. 10. Fitbit Inc. How does my tracker count steps? Available at: https://help.fitbit.com/articles/en_US/Help_article/1143. Accessed on August 9, 2016. 11. Warnes CA, Williams RG, Bashore TM, et al. ACC/AHA 2008 Guidelines for the Management of Adults with Congenital Heart Disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (writing committee to develop guidelines on the management of adults with congenital heart disease). Circulation. 2008; 118:e714-833.
17
ACCEPTED MANUSCRIPT
12. de Onis M, Onyango AW, Borghi E, et al. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ. 2007; 85:660-7.
RI PT
13. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000; 320:1240-3. 14. Sherar LB, Griew P, Esliger DW, et al. International children's accelerometry database (ICAD): design and methods. BMC Public Health. 2011; 11:485.
SC
15. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. Calibration of two objective measures of physical activity for children. J Sports Sci. 2008; 26:1557-65.
M AN U
16. Gabel L, McKay HA, Nettlefold L, Race D, Macdonald HM. Bone architecture and strength in the growing skeleton: the role of sedentary time. Med Sci Sports Exerc. 2015; 47:36372. 17. Trost SG, Pate RR, Freedson PS, Sallis JF, Taylor WC. Using objective physical activity measures with youth: how many days of monitoring are needed? Med Sci Sports Exerc. 2000; 32:426-31.
TE D
18. Canadian Society for Exercise Physiology. 24-Hour Movement Guidelines for Children and Youth. Available at: http://www.csep.ca/CMFiles/Guidelines/24hrGlines/Canadian24HourMovementGuidelines 2016.pdf. Accessed on August 9, 2016. 19. Akobeng AK. Understanding diagnostic tests 3: Receiver operating characteristic curves. Acta Paediatr. 2007; 96:644-7.
EP
20. Sirard JR, Slater ME. Compliance with wearing physical activity accelerometers in high school students. J Phys Act Health. 2009; 6 Suppl 1:S148-55.
AC C
21. Colley RC, Garriguet D, Janssen I, et al. Physical activity of Canadian children and youth: accelerometer results from the 2007 to 2009 Canadian Health Measures Survey. Health Rep. 2011; 22:15-23. 22. Tamayo C, Manlhiot C, Patterson K, Lalani S, McCrindle BW. Longitudinal evaluation of the prevalence of overweight/obesity in children with congenital heart disease. Can J Cardiol. 2015; 31:117-23. 23. Tudor-Locke C, Sisson SB, Lee SM, et al. Evaluation of quality of commercial pedometers. Can J Public Health. 2006; 97 Suppl 1:S10-5, S-6. 24. FitBit. What are active minutes? http://goo.gl/C6RTnZ. 2015; 18
ACCEPTED MANUSCRIPT
25. Canadian Society for Exercise Physiology. Canadian Physical Activity Guidelines for Adults 18-64 years. Available at: http://csep.ca/CMFiles/Guidelines/CSEP_PAGuidelines_adults_en.pdf. Accessed on August 9, 2016.
RI PT
26. Harrell JS, McMurray RG, Baggett CD, et al. Energy costs of physical activities in children and adolescents. Med Sci Sports Exerc. 2005; 37:329-36.
SC
27. Ekelund U, Luan J, Sherar LB, et al. Moderate to vigorous physical activity and sedentary time and cardiometabolic risk factors in children and adolescents. JAMA. 2012; 307:70412. 28. Biddle S, Asare M. Physical activity and mental health in children and adolescents: a review of reviews. British Journal of Sports Medicine. 2011; 45:886-95.
M AN U
29. Malina RM. Tracking of physical activity and physical fitness across the lifespan. Res Q Exerc Sport. 1996; 67:S48-57. 30. Duppen N, Takken T, Hopman MT, et al. Systematic review of the effects of physical exercise training programmes in children and young adults with congenital heart disease. Int J Cardiol. 2013; 168:1779-87.
TE D
31. Fairclough SJ, Noonan R, Rowlands AV, et al. Wear Compliance and Activity in Children Wearing Wrist- and Hip-Mounted Accelerometers. Med Sci Sports Exerc. 2016; 48:24553.
EP
32. Dai S, Carroll DD, Watson KB, et al. Participation in Types of Physical Activities Among US Adults--National Health and Nutrition Examination Survey 1999-2006. J Phys Act Health. 2015; 12 Suppl 1:S128-40.
AC C
33. Tully MA, McBride C, Heron L, Hunter RF. The validation of Fibit Zip physical activity monitor as a measure of free-living physical activity. BMC Res Notes. 2014; 7:952. 34. Ferguson T, Rowlands AV, Olds T, Maher C. The validity of consumer-level, activity monitors in healthy adults worn in free-living conditions: a cross-sectional study. Int J Behav Nutr Phys Act. 2015; 12:42. 35. Kooiman TJ, Dontje ML, Sprenger SR, et al. Reliability and validity of ten consumer activity trackers. BMC Sports Sci Med Rehabil. 2015; 7:24. 36. Nader PR, Bradley RH, Houts RM, McRitchie SL, O'Brien M. Moderate-to-vigorous physical activity from ages 9 to 15 years. JAMA. 2008; 300:295-305.
19
ACCEPTED MANUSCRIPT
37. Fakhouri TH, Hughes JP, Burt VL, et al. Physical activity in U.S. youth aged 12-15 years, 2012. NCHS Data Brief. 2014:1-8.
RI PT
38. Dontje ML, De Groot M, Lengton RR, Van der Schans CP, Krijnen WP. Measuring steps with the Fitbit activity tracker: An inter-device reliability study. J Med Eng Tech. 2015; 39:286-90.
FIGURE LEGENDS
SC
Figure 1. Association between daily steps measured by accelerometer and Fitbit
M AN U
Scatter plot (A) and Bland-Altman plot (B) for the overall sample (n=138 person-days, n=30 individuals). Note the strong correlation between the two devices (A, dotted line), while a significant bias suggests that the Fitbit measured more steps than the accelerometer. There was systematic bias, whereby the differences between devices were greater at higher total steps. Note the outliers likely indicate that the accelerometer was worn for longer than the Fitbit, which we confirmed in wear time analyses.
Figure 2. Example day of a participant’s hourly steps by accelerometer and Fitbit Note how the Fitbit registers more steps at almost every hour of the day, which significantly amplifies the total cumulative steps for the day.
AC C
EP
TE D
Figure 3. Receiver-operating curve identifying a Fitbit step cut-point. Fibit steps ≥12,500 steps/d identify those who meet physical activity guidelines (defined as ≥60 min/d based on accelerometry).
20
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
S0828-282X(16)31147-3
DOI:
10.1016/j.cjca.2016.11.024
Reference:
CJCA 2325
To appear in:
Canadian Journal of Cardiology
Received Date: 12 August 2016 Revised Date:
14 November 2016
Accepted Date: 28 November 2016
Please cite this article as: Voss C, Gardner RF, Dean PH, Harris KC, Validity of commercial activity trackers in children with congenital heart disease, Canadian Journal of Cardiology (2017), doi: 10.1016/ j.cjca.2016.11.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
TITLE PAGE
RI PT
Validity of commercial activity trackers in children with congenital heart disease
Christine Voss, PhD; Ross F Gardner; Paige H Dean, BSc; Kevin C Harris, MD, MHSc
Dr. Kevin Harris Children’s Heart Centre BC Children’s Hospital 1F27 – 4480 Oak Street Vancouver, BC V6H 3V4 Phone: 604-875-3878 Fax: 604-875-3463 Email: [email protected]
TE D
M AN U
Correspondence:
SC
Division of Cardiology, Department of Pediatrics, University of British Columbia, Canada (all authors)
EP
Short title: Activity tracker validation in children with CHD
AC C
Word count excluding brief summary, word count, and short title: 4,734
1
ACCEPTED MANUSCRIPT
BRIEF SUMMARY In this validation study of a commercial activity tracker in children, including those with CHD,
RI PT
we found that the device provides results that were generally reflective of overall physical activity levels and can identify those meeting physical activity guidelines. These findings are important given the increasing emphasis on physical activity promotion and monitoring in
AC C
EP
TE D
M AN U
SC
children with cardiovascular risk factors.
2
ACCEPTED MANUSCRIPT
ABSTRACT Background: Increasing physical activity levels is a high priority to optimize long-term health in
RI PT
children with congenital heart disease (CHD). Commercial activity trackers have been validated in adults and are increasingly used to measure and promote physical activity in pediatric populations, but they have not been validated in children.
Methods: In 30 children with CHD aged 10-18yrs, we assessed the validity of physical activity
SC
form the wrist-based Fitbit Charge HRTM against hip-based ActiGraph accelerometers under free
M AN U
living conditions for 7 days. We assessed the association between devices by intra-class correlation coefficients and Bland-Altman plots. Receiver-operating curves were used to identify Fitbit step cut-points.
Results: There was a strong association between the two devices for daily steps across 138
TE D
analysed person-days (ICC=0.855, p<0.001), but poorer agreement for time spent in physical activity intensities (ICCs<0.7). Daily Fitbit steps of ≥12,500 identified meeting physical activity guidelines defined as ≥60 min MVPA/day. Fitbit devices recorded more steps than
EP
accelerometers (-2,242 steps/d, 95%LoA of -7,738 to 3,253). Between-device differences were greater in boys vs. girls. Fitbit devices were worn for longer than accelerometers (-36 min/d,
AC C
95%LoA -334 to 261), but overall differences in wear time explained little of the variance in step differences (7%, p=0.048).
Conclusions: Commercial activity trackers provide opportunities to remotely monitor physical activity in children with CHD, but absolute values may differ from accelerometers. These findings are important given the increasing emphasis on physical activity promotion and monitoring in children with cardiovascular risk factors.
3
ACCEPTED MANUSCRIPT
KEY WORDS
AC C
EP
TE D
M AN U
SC
RI PT
Congenital Heart Disease, Children, Pediatric, Physical Activity, Fitbit, ActiGraph
4
ACCEPTED MANUSCRIPT
INTRODUCTION Physical activity is increasingly recognized as an important behaviour to optimize long-term cardiovascular health and quality of life in children with congenital heart disease (CHD),1,2 who
RI PT
are at increased cardiovascular risk.3 Commercial activity trackers hold promise as a tool to measure, prescribe and promote physical activity. Several studies are already taking advantage of commercial trackers to measure and/or promote physical activity in both healthy and clinical
SC
pediatric populations,4-6 including children with CHD.7 While the validity and reliability of
M AN U
commercial trackers has been extensively documented in adults,8 no previous research has established the validity of such devices in children. This is an important shortcoming, given that it is unclear if children’s sporadic bursts of activity9 are adequately quantified by commercial trackers that are designed for adults and optimized to capture walking.10 The objective of the current study was to assess the validity of physical activity from a commercial activity tracker
Sample
EP
METHODS
TE D
against a research-grade accelerometer under free living conditions in children with CHD.
AC C
We included a convenience sample of 40 participants aged 10-18yrs recruited from the Children’s Heart Centre at BC Children’s Hospital in Vancouver, BC and provincial partnership clinics. We obtained institutional ethics approval, parental consent and participant assent. Cardiac diagnoses were: 1) CHD; categorized as mild, moderate or severe11, 2) cardiac transplant, 3) electrophysiological disorders, or 4) none (e.g. innocent murmur). Nurses measured height (0.1 cm) and weight (0.1 kg). BMI was expressed according to WHO norms12 and categorized according to International Obesity Task Force criteria.13
5
ACCEPTED MANUSCRIPT
Criterion measure: ActiGraph accelerometer We fitted participants with an ActiGraph accelerometer (GT3X+, GT9X; ActiGraph LLC, Pensacola, FL; sampling at 30Hz) to be worn over the right hip and instructed them to wear it
RI PT
continuously for the next 7 days and to only remove it for water-based activities,
showering/bathing and sleep. The ActiGraph is widely used to objectively measure physical
SC
activity in children under free-living conditions.14
Test measure: Fitbit activity tracker
M AN U
We chose Fitbit Charge HRTM (Fitbit) for our study. Studies in children have exclusively used Fitbit models (without validation),4-7 and we expected that the appeal of using one of the market leaders would benefit compliance in our study. Wristband size and placement were in accordance with manufacturer’s guidelines. We asked participants to wear the Fitbit continuously during the
TE D
same 7 days as the accelerometer. We created anonymous Fitbit user profiles (online
valid in children).
Data processing
EP
‘dashboard’) and disabled displays of ‘calories burned’ and ‘distance travelled’ (unlikely to be
AC C
We used ActiLife v.6.13.2 (ActiGraph LLC, Pensacola, FL) to analyse accelerometry files (15s epoch). Valid days were defined as ≥600 min/d wear time (allowing for ≤60min of zero activity) and moderate-to-vigorous physical activity was calculated (MVPA; ≥2296 CPM; includes vigorous).15 Sedentary time was expressed as a percentage relative to wear time.16 Mean physical activity values were based on ≥1 valid days (inclusion criterion for validation analyses); we also used conservative wear time criteria of ≥3 weekdays and ≥1 weekend day(s) to estimate general physical activity levels.17 We defined meeting guidelines as mean daily MVPA ≥60 min/d.18 For
6
ACCEPTED MANUSCRIPT
Fitbit data, we used the ‘Data Export’ function on the ‘dashboard’ to extract steps and physical activity intensities (sedentary, light, fair, and very). MVPA was defined as ‘fair + very’ and sedentary time was expressed relative to wear time. A sub-sample were managed through
RI PT
Fitabase (Small Steps Labs LLC, San Diego, CA; www.fitabase.com), a fee-for-service online platform that facilitates access to multiple Fitbit devices simultaneously and provides high
SC
resolution minute-by-minute data exports.
We created a person-day dataset of matched ActiGraph and Fitbit data on valid accelerometry
M AN U
days (≥600 min/d). For a subset of person-days with high-resolution data (n=20), we estimated Fitbit wear time using a modified data analyses package in ‘R’ that processes minute-by-minute accelerometry data (‘accelerometry’, v. 2.2.5, May 2015). First, we re-produced the ActiLife wear time results for accelerometry data by this method, and yielded similar results (815 vs. 809
TE D
min/d). Second, we repeated step 1 using accelerometry steps rather than ‘counts’ to establish that steps can serve as a wear time metric, and yielded similar results (821 vs. 815 min/d). Third, we analysed the minute-by-minute Fitbit steps to determine Fitbit wear time.
EP
Statistical Analysis
AC C
Descriptive statistics (frequencies (%), mean±SD, median (IQR)) were calculated for applicable variables. Between-sex differences were assessed by independent t-tests, Mann-Whitney U Ranked Sum test, or Pearson’s chi-squared test. Associations and agreement between devices were assessed by intra-class correlation coefficients (ICC) and Bland-Altman plots. The between-device difference (bias and 95% Limits of Agreement, LoA) was defined as ‘criterion measure (accelerometer)’ minus ‘test measure (Fitbit)’. The mean absolute percent error was calculated as ‘mean between-device difference / mean criterion measure *100’. Receiver
7
ACCEPTED MANUSCRIPT
operating curves (ROC) were drawn using Fitbit steps as the classifier and meeting physical activity guidelines (≥60 min/d of MVPA from accelerometry) as the true-status reference; the coordinate with the greatest sum of sensitivity and specificity was used to identify the Fitbit step
RI PT
cut-point with discriminatory value.19 Because meaningful between-day variations in physical activity levels are expected within individuals, we treated valid days as independent observations and used a person-day dataset using actual days rather than hypothetical mean days for validation
SC
analyses. We conducted sensitivity analyses using a reduced dataset (mean values per person) to rule out undue influences of related observations within individuals. Analyses were carried out
M AN U
using Stata (v.14.1, Stata Corp LP, College Station, TX) or R (x64 3.3.1, The R Foundation for Statistical Computing) and significance was set at p<0.05.
RESULTS
TE D
Sample characteristics
Thirty participants had at least one day of matching accelerometry and Fitbit data. Four participants did not wear both devices at the time and 6 did not have any valid data, which is
EP
unsurprising considering that non-compliance with accelerometer wear is commonly observed in adolescents.20,21 There were no significant differences for sample characteristics between the
AC C
analytical sample and those who were excluded, nor were there significant sex-differences (Table 1). Similarly to what was previously reported for children with CHD,22 23% were overweight or obese, which is comparable to national data in healthy children.21 Mean MVPA (based on ≥1 valid day) was 45 min/d, which was similar to national data for similarly aged children and adolescents (~50 min).21 Physical activity levels were not meaningfully different when computed according to conservative wear time inclusion criteria of ≥3 weekdays and
8
ACCEPTED MANUSCRIPT
≥1weekend day (e.g. 47 min/d MVPA, available for n=24 individuals). We acknowledge a lack of expected sex-differences in MVPA, which was likely explained by inter-personal differences and the wide age range of our relatively small sample. Because the aim of the current study was
RI PT
to assess the agreement between two devices, we do not consider the heterogeneous physical activity levels of our sample a limitation. We defined adherence to physical activity guidelines as
SC
mean daily MVPA ≥60 min/d and 25% met physical activity guidelines.
Validity: association and agreement between ActiGraph accelerometer and Fitbit
M AN U
We included 138 valid person-days (mean per person: 4.6 days, range: 1-7). We conducted detailed assessments between Fitbit and accelerometer results for physical activity metrics, including steps and time spent in physical activity intensities. The strongest between-device agreement was observed for daily steps (ICC=0.855, p<0.001; Figure 1), which did not vary
TE D
notably between sexes or age groups. However, the Fitbit devices recorded more steps than the accelerometers (10,383±5,321 vs. 8,141±4,023 steps/d; p<0.001). The mean bias was -2,242 steps/d, although between-device differences were highly varied (95%LoA of -7,738 to 3,253;
EP
Figure 1). The mean absolute percent error was 28%, which is higher than the ≤10% error that is desired for valid step estimates under free-living conditions.23 The mean bias for daily steps was
AC C
greater in boys (-3,320 steps/d; 95%LoA -7,643 – 1,003; 39% MAP) than girls (-1,389 steps/d; 95%LoA -7,133 – 4,356; 18% MAP; Figure 1). An example day illustrating differences in hourly and cumulative steps between devices is found in Figure 2.
Compared with steps, agreement between devices was notably lower for time spent in physical activity intensities (all ICCs<0.66; online supplement Table S1 and Figure S1). Discrepant results are likely explained by methodological differences. For example, the Fitbit records only
9
ACCEPTED MANUSCRIPT
MVPA that is accrued in bouts of ≥10 min duration,24 which is in line with adult physical activity guidelines25 but not current guidelines for children18 and thus not how we calculated MVPA based on accelerometry. This explains person-days where the accelerometer registered
RI PT
some MVPA but the Fitbit did not. In contrast, the Fitbit registered substantially more MVPA than the accelerometer on other person-days, which was almost exclusively observed in younger children. The proprietary Fitbit algorithm categorises MVPA as ≥3METs, which in turn is
SC
derived from estimated caloric expenditure based on movement, age, sex, height and weight.24 Energy cost of activity is greater in children than in adults,26 and it is unclear if and how the
M AN U
Fitbit algorithm corrects for this in our younger participants. Fitbit also states that the heart rate feature does ‘a better job of recognizing active minutes’,26 but it is not disclosed how this is done. Sedentary time was the only activity intensity metric that yielded agreement between devices that was comparable to steps (ICC=0.845, p<0.001). However, we faced challenges
TE D
regarding data availability and the need for additional data manipulation when the Fitbit was not worn over night (see online supplement for details) which currently limits the usefulness of this metric. All analyses were performed on person-day datasets that were unadjusted for multiple
EP
observations per person; we found similar results using a reduced person-level data-set (not
AC C
shown), confirming that no meaningful within-person effects necessitated adjustments.
Diagnostic values of Fitbit devices to identify adherence to physical activity guidelines A ROC curve was created using Fitbit steps as the classifier and meeting physical activity guidelines (≥60 min MVPA based on accelerometry) as the true-status reference (Figure 3). The ROC area under the curve was 0.82 (95%CI 0.74-0.90). The coordinate with the greatest sum of sensitivity and specificity (cut-point that identifies meeting guidelines) was 12,562 steps/d.
10
ACCEPTED MANUSCRIPT
Device wear time analyses and its effect on validity
RI PT
Out of a possible 20 participants who had their devices managed through Fitabase, high resolution minute-by-minute Fitbit data were available for 15 participants and 55 person-days. Infrequent Fitbit syncing resulted in some loss of high resolution data. Validation results were
SC
similar in this sub-sample compared with the entire sample (ICC=0.908; bias -2,236). On average, Fitbits were worn for 36 minutes longer than ActiGraphs (Fitbit 866 min/d vs.
M AN U
accelerometer 830 min/d), although differences in wear time were highly varied (95%LoA -334 to 261). Girls wore both devices for longer (844 min/d vs. Fitbit 906 min/d) compared with boys (ActiGraph: 813 min/d vs. Fitbit 821 min/d), although neither sex-difference was statistically significant (ActiGraph p=0.288 and Fitbit p=0.060). Differences in wear time between devices
TE D
tended to be greater in girls (-62 min/d, 95%LoA -294 to 170) than in boys, but the variance was also greater in boys(-7.8 min/d, 95%LoA -360 to 344). Within either device, wear time was not related to steps (ActiGraph r2=0.015, p=0.368; Fitbit r2=0.003, p=0.708). Between-device
EP
differences in wear time was a significant correlate of between-device differences in steps, but explained only 7% of the variance (r2=0.072, p=0.048). On days were the two devices had
AC C
similar wear time (n=12; ±30 min) the bias in steps was -1,854 (95%LoA -4,815 to 1,107).
DISCUSSION
In children with CHD, we assessed the validity of the Fitbit Charge HRTM - a wrist-worn commercial tracker - to estimate physical activity. We demonstrated a strong correlation between the Fitbit and the ActiGraph for daily steps, and importantly, we identified a cut-point of ~12,500
11
ACCEPTED MANUSCRIPT
Fitbit steps/d to correspond to meeting physical activity guidelines (≥60 minutes MVPA/d from accelerometry), which will have utility for clinical practice and future research.
RI PT
Physical activity – a cornerstone of health Physical activity during childhood is of paramount importance for optimal cardio-metabolic health27 and mental well-being,28 and active children are more likely stay active as adults.29 We
SC
are facing a physical inactivity crisis where fewer than 1 in 10 Canadian children are meeting physical activity guidelines.21 For children with CHD, who are at high cardiovascular risk as a
M AN U
consequence of their condition,3 there has been an urgent call for action to promote physical activity to optimize long-term cardiovascular health and quality of life in these individuals.1,2 Innovative strategies are needed to engage these children and their families in health promotion.
Utility of commercial activity trackers in children in clinical practice and research
TE D
Exercise training programs are effective at improving physical activity levels in children with CHD,30 but access to such programs is typically limited to those residing near metropolitan areas. Considering the relevance of technology for our young patients, the role of commercial activity
EP
trackers for remote / home-based physical activity promotion is already being explored.4,6,7
AC C
Commercial trackers such as the Fitbit offer several advantages over research-grade activity devices: 1) aesthetically pleasing and user-friendly hardware; 2) appealing user-experience through the online ‘dashboard’; 3) cost-effective and readily available; 4) no specialist software and/or data processing expertise necessary; 5) wrist-based, a preferred device location in children and youth.31 In addition, services such as Fitabase can streamline the remote management of multiple users at once. This is important to effectively facilitate exercise prescription, where
12
ACCEPTED MANUSCRIPT
continuous remote monitoring of participant activity levels are needed in order to provide tailored feedback and recommendations.
RI PT
Validity of Fitbit devices in children The target consumer group for commercial trackers are adults, whose most common activity is walking.32 Fitbit algorithms are “designed to look for motion patterns most indicative of people
SC
walking”.10 No previous research has validated the Fitbit in children, but adult studies
consistently demonstrate good validity for steps under free-living conditions.8,33-35 Children
M AN U
engage in sporadic bouts of high intensity activity,9 and it is unclear to what extent, if any, these differences in activity patterns may impact algorithms to accurately identify physical activity. We found agreement between devices was most consistent for daily steps – a physical activity metric that most closely resembles total activity rather than intensity. The poorer agreement for time spent in physical activity intensities is in part explained by the Fitbit only recording MVPA
TE D
bouts of ≥10 min duration (in line with adult guidelines),24 and the interpretation of Fitbit MVPA being further complicated by undisclosed proprietary algorithms drawing on estimates of caloric
EP
expenditure that may not be valid in children.
AC C
Sex differences in Fitbit validity: wear time or physical activity? Between-device step differences in steps were nearly twice as great in boys as in girls. We investigated whether this was explained by poorer compliance in boys. Girls tended to wear both devices for longer than boys, but it is the girls who were less consistent with wearing both devices equally, and clearly favouring the Fitbit. Regardless, wear time differences overall explained only a small portion of the variance in step differences (7%), and a sub-analyses of person-days with similar device wear time still registered a substantial between-device step
13
ACCEPTED MANUSCRIPT
difference (-1,854). Thus, we do not believe that device wear time is an important factor affecting our validation results.
RI PT
Physical activity volume and patterns could instead play a role in our observed sex-differences. Our physical activity data hints at well-documented sex-differences in physical activity, with boys generally being more active than girls.36 In addition to more sporadic bursts of high
SC
intensity activity in children compared with adults,9 boys commonly play games, whereas girls frequently walk for leisure.37 Walking is an activity that Fitbits are optimized to capture,10 which
M AN U
could explain the more favourable step results in girls. The notion that boys may engage more in higher intensity, sustained play/games is supported by our finding that between-device agreement for MVPA was good for boys (ICC=0.824), but poorer for the sample overall (ICC=0.657; Table S1).
TE D
Strengths, limitations and future directions
This is the first validation study of commercial activity trackers in children. Our analyses identified the daily Fitbit steps as the most suitable metric for use in children, in line what has
EP
been found for adults.8 Our sample was a convenience sample of patients attending pediatric
AC C
cardiology clinics and most had CHD. We previously found that our CHD patients are as active as healthy Canadians of the same age and sex, and thus believe the current findings are of relevance to the wider pediatric community.
We acknowledge that there was considerable variance in our results. For example, we found that while the Fitbit recorded on average 2,242 more steps than accelerometers did, 95%LoA indicated that there was considerable variability. A certain degree of variance is to be expected in
14
ACCEPTED MANUSCRIPT
a context where children are tasked with collecting behavioural data under free living conditions, and are acceptable in light of the strong correlations we observed.
RI PT
It is a limitation that we did not assess the validity of the Fitbit against a direct measure of physical activity (e.g. direct observation) and/or energy expenditure (e.g. indirect calorimetry), but instead against a widely used accelerometer14 and associated MVPA cut-points that are based
SC
on indirect calorimetry in children.15 Adult validation studies reported different device
performance between controlled, laboratory-based treadmill walking and running versus free-
M AN U
living conditions;8,35 thus, validation under free-living conditions is clearly warranted. However, (unknown) differences in device wear time can impact validation studies under such conditions. We used algorithms to identify non-wear from a lack of movement signal, but different devices may have different thresholds as to what acceleration signal constitutes a step, for example. More
TE D
direct measures of wear time are desirable, such as skin sensors in the newer ActiGraph models (optimized for wrist placement). For Fitbits, the continuous heart rate feature has potential to identify wear time, but access to the fee-for-service Fitabase and advanced data processing skill
EP
are required.
We did not assess intra- or inter-device reliability of the Fitbit, but this has been reported
AC C
elsewhere.e.g. 38 There are currently no published studies of the Fitbit Charge HRTM, likely due to the discrepancy between the fast-paced release of new commercial trackers and the lag in research and peer-reviewed scientific publication. Fewer adult studies have assessed the newer wrist-based Fitbit models, and although they performed well, they were typically inferior to the hip-based models.e.g. 35 It is possible that hip-based Fitbit models yield more comparable results
15
ACCEPTED MANUSCRIPT
to the ActiGraph in children. However, children favour wrist-based over hip-based devices,31 and the importance of compliance should be carefully evaluated against the need for accuracy.
RI PT
CONCLUSION In children with CHD, commercial activity trackers can provide general estimates of physical activity using daily step counts, and a cut-point of ~12,500 Fitbit steps/d relates to meeting
ACKNOWLEDGEMENTS
M AN U
prescribe and monitor physical activity in children.
SC
physical activity guidelines (defined as ≥60 min MVPA/d). Fitbits may be useful to remotely
We are grateful to the participants and their families who participated in this research. We thank the nurses, administrative staff and research staff at the Children’s Heart Centre and our
FUNDING SOURCES
TE D
partnership clinics for their support.
EP
We received funding form the Heart and Stroke Foundation of Canada (F1504) and the Child &
AC C
Family Research Institute, University of British Columbia.
REFERENCES
1. Takken T, Giardini A, Reybrouck T, et al. Recommendations for physical activity, recreation sport, and exercise training in paediatric patients with congenital heart disease: a report from the Exercise, Basic & Translational Research Section of the European Association of Cardiovascular Prevention and Rehabilitation, the European Congenital Heart and Lung Exercise Group, and the Association for European Paediatric Cardiology. Eur J Prev Cardiol. 2012; 19:1034-65.
16
ACCEPTED MANUSCRIPT
2. Longmuir PE, Brothers JA, de Ferranti SD, et al. Promotion of physical activity for children and adults with congenital heart disease: a scientific statement from the American Heart Association. Circulation. 2013; 127:2147-59.
RI PT
3. Kavey RE, Allada V, Daniels SR, et al. Cardiovascular risk reduction in high-risk pediatric patients: a scientific statement from the American Heart Association Expert Panel on Population and Prevention Science; the Councils on Cardiovascular Disease in the Young, Epidemiology and Prevention, Nutrition, Physical Activity and Metabolism, High Blood Pressure Research, Cardiovascular Nursing, and the Kidney in Heart Disease; and the Interdisciplinary Working Group on Quality of Care and Outcomes Research: endorsed by the American Academy of Pediatrics. Circulation. 2006; 114:2710-38.
M AN U
SC
4. Gibson C, August K, Greene J, et al. A televideo exercise and nutrition program for children with acute lymphoblastic leukemia in maintenance therapy: design and methods. Open Access Journal of Clinical Trials. 2015; 7:77-84. 5. Hayes LB, Van Camp CM. Increasing physical activity of children during school recess. J Appl Behav Anal. 2015; 48:690-5. 6. Hooke MC, Gilchrist L, Tanner L, Hart N, Withycombe JS. Use of a Fitness Tracker to Promote Physical Activity in Children With Acute Lymphoblastic Leukemia. Pediatr Blood Cancer. 2016; 63:684-9.
TE D
7. Jacobsen RM, Ginde S, Mussatto K, et al. Can a Home-based Cardiac Physical Activity Program Improve the Physical Function Quality of Life in Children with Fontan Circulation? Congenit Heart Dis. 2016; 11:175-82.
EP
8. Evenson KR, Goto MM, Furberg RD. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act. 2015; 12:159.
AC C
9. Welk GJ, Corbin CB, Dale D. Measurement issues in the assessment of physical activity in children. Res Q Exerc Sport. 2000; 71:S59-73. 10. Fitbit Inc. How does my tracker count steps? Available at: https://help.fitbit.com/articles/en_US/Help_article/1143. Accessed on August 9, 2016. 11. Warnes CA, Williams RG, Bashore TM, et al. ACC/AHA 2008 Guidelines for the Management of Adults with Congenital Heart Disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (writing committee to develop guidelines on the management of adults with congenital heart disease). Circulation. 2008; 118:e714-833.
17
ACCEPTED MANUSCRIPT
12. de Onis M, Onyango AW, Borghi E, et al. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ. 2007; 85:660-7.
RI PT
13. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000; 320:1240-3. 14. Sherar LB, Griew P, Esliger DW, et al. International children's accelerometry database (ICAD): design and methods. BMC Public Health. 2011; 11:485.
SC
15. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. Calibration of two objective measures of physical activity for children. J Sports Sci. 2008; 26:1557-65.
M AN U
16. Gabel L, McKay HA, Nettlefold L, Race D, Macdonald HM. Bone architecture and strength in the growing skeleton: the role of sedentary time. Med Sci Sports Exerc. 2015; 47:36372. 17. Trost SG, Pate RR, Freedson PS, Sallis JF, Taylor WC. Using objective physical activity measures with youth: how many days of monitoring are needed? Med Sci Sports Exerc. 2000; 32:426-31.
TE D
18. Canadian Society for Exercise Physiology. 24-Hour Movement Guidelines for Children and Youth. Available at: http://www.csep.ca/CMFiles/Guidelines/24hrGlines/Canadian24HourMovementGuidelines 2016.pdf. Accessed on August 9, 2016. 19. Akobeng AK. Understanding diagnostic tests 3: Receiver operating characteristic curves. Acta Paediatr. 2007; 96:644-7.
EP
20. Sirard JR, Slater ME. Compliance with wearing physical activity accelerometers in high school students. J Phys Act Health. 2009; 6 Suppl 1:S148-55.
AC C
21. Colley RC, Garriguet D, Janssen I, et al. Physical activity of Canadian children and youth: accelerometer results from the 2007 to 2009 Canadian Health Measures Survey. Health Rep. 2011; 22:15-23. 22. Tamayo C, Manlhiot C, Patterson K, Lalani S, McCrindle BW. Longitudinal evaluation of the prevalence of overweight/obesity in children with congenital heart disease. Can J Cardiol. 2015; 31:117-23. 23. Tudor-Locke C, Sisson SB, Lee SM, et al. Evaluation of quality of commercial pedometers. Can J Public Health. 2006; 97 Suppl 1:S10-5, S-6. 24. FitBit. What are active minutes? http://goo.gl/C6RTnZ. 2015; 18
ACCEPTED MANUSCRIPT
25. Canadian Society for Exercise Physiology. Canadian Physical Activity Guidelines for Adults 18-64 years. Available at: http://csep.ca/CMFiles/Guidelines/CSEP_PAGuidelines_adults_en.pdf. Accessed on August 9, 2016.
RI PT
26. Harrell JS, McMurray RG, Baggett CD, et al. Energy costs of physical activities in children and adolescents. Med Sci Sports Exerc. 2005; 37:329-36.
SC
27. Ekelund U, Luan J, Sherar LB, et al. Moderate to vigorous physical activity and sedentary time and cardiometabolic risk factors in children and adolescents. JAMA. 2012; 307:70412. 28. Biddle S, Asare M. Physical activity and mental health in children and adolescents: a review of reviews. British Journal of Sports Medicine. 2011; 45:886-95.
M AN U
29. Malina RM. Tracking of physical activity and physical fitness across the lifespan. Res Q Exerc Sport. 1996; 67:S48-57. 30. Duppen N, Takken T, Hopman MT, et al. Systematic review of the effects of physical exercise training programmes in children and young adults with congenital heart disease. Int J Cardiol. 2013; 168:1779-87.
TE D
31. Fairclough SJ, Noonan R, Rowlands AV, et al. Wear Compliance and Activity in Children Wearing Wrist- and Hip-Mounted Accelerometers. Med Sci Sports Exerc. 2016; 48:24553.
EP
32. Dai S, Carroll DD, Watson KB, et al. Participation in Types of Physical Activities Among US Adults--National Health and Nutrition Examination Survey 1999-2006. J Phys Act Health. 2015; 12 Suppl 1:S128-40.
AC C
33. Tully MA, McBride C, Heron L, Hunter RF. The validation of Fibit Zip physical activity monitor as a measure of free-living physical activity. BMC Res Notes. 2014; 7:952. 34. Ferguson T, Rowlands AV, Olds T, Maher C. The validity of consumer-level, activity monitors in healthy adults worn in free-living conditions: a cross-sectional study. Int J Behav Nutr Phys Act. 2015; 12:42. 35. Kooiman TJ, Dontje ML, Sprenger SR, et al. Reliability and validity of ten consumer activity trackers. BMC Sports Sci Med Rehabil. 2015; 7:24. 36. Nader PR, Bradley RH, Houts RM, McRitchie SL, O'Brien M. Moderate-to-vigorous physical activity from ages 9 to 15 years. JAMA. 2008; 300:295-305.
19
ACCEPTED MANUSCRIPT
37. Fakhouri TH, Hughes JP, Burt VL, et al. Physical activity in U.S. youth aged 12-15 years, 2012. NCHS Data Brief. 2014:1-8.
RI PT
38. Dontje ML, De Groot M, Lengton RR, Van der Schans CP, Krijnen WP. Measuring steps with the Fitbit activity tracker: An inter-device reliability study. J Med Eng Tech. 2015; 39:286-90.
FIGURE LEGENDS
SC
Figure 1. Association between daily steps measured by accelerometer and Fitbit
M AN U
Scatter plot (A) and Bland-Altman plot (B) for the overall sample (n=138 person-days, n=30 individuals). Note the strong correlation between the two devices (A, dotted line), while a significant bias suggests that the Fitbit measured more steps than the accelerometer. There was systematic bias, whereby the differences between devices were greater at higher total steps. Note the outliers likely indicate that the accelerometer was worn for longer than the Fitbit, which we confirmed in wear time analyses.
Figure 2. Example day of a participant’s hourly steps by accelerometer and Fitbit Note how the Fitbit registers more steps at almost every hour of the day, which significantly amplifies the total cumulative steps for the day.
AC C
EP
TE D
Figure 3. Receiver-operating curve identifying a Fitbit step cut-point. Fibit steps ≥12,500 steps/d identify those who meet physical activity guidelines (defined as ≥60 min/d based on accelerometry).
20
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT