The Validation of the Actigraph GT3X Step Counter in Youth Who Are Blind

By Elizabeth K. Lenz, Brooke E. Starkoff, Lauren J. Lieberman, and Danny Too

Elizabeth K. Lenz is an associate professor of exercise science in the department of kinesiology, sport studies, and physical education at the SUNY Brockport.

Brooke E. Starkoff is an assistant professor of exercise science in the department of kinesiology at Valparaiso University.

Lauren J. Lieberman is a distinguished service professor in the department of kinesiology, sport studies, and physical education at SUNY Brockport.

Danny Too is an associate professor of biomechanics in the department of kinesiology, sport studies, and physical education at the SUNY Brockport.

Abstract

Background: Identifying valid tools to assess physical activity in youth who are blind is paramount.

Methods: Nine participants who were blind (14.1±2.1 years old) completed two, 400-meter laps around a track. Steps were recorded via accelerometer (ACC) and direct observation (DO). Analyses examined the relationship between number of steps measured by DO and ACC, and between mean number of steps for each.

Results: Average steps by ACC and DO were 664.9±72.7 and 683.4±96.3, respectively. Pearson correlations revealed significant relationships between laps for ACC (r=0.978, p<0.00) and DO (r=0.998, p<0.00), and between ACC and DO (r=0.955, p<0.00). Paired t-tests revealed no differences between trials for ACC, DO, and DO compared to ACC (p=0.146-0.841). Regression analysis revealed no proportional bias between differences in steps taken between trials (p=0.158-0.977), but there was proportional bias (p=0.04) between steps taken with DO and ACC. 

Conclusion: Accelerometry can be used to measure steps for youth who are blind.

Keywords

accelerometer, blind, youth, validation, direct observation

Introduction

One of the most common modes of physical activity (PA) across age groups is walking, and the primary way researchers track this activity is through measuring the number of steps people take via activity monitors. The need for accurate and reliable PA monitors is essential, as they can be utilized to examine the quantity of PA, determine appropriate dose amounts of PA required to alter specific health parameters, and evaluate the effectiveness of programs to increase habitual PA (Trost, 2007). The Actigraph is a well-known accelerometer (ACC) and step counter used by researchers internationally to capture and monitor PA levels as well as determine if PA interventions are effective, but less is known about the validity of the Actigraph in individuals who are blind, especially youth.

Youth with visual impairments (VI) have been shown to have low levels of physical fitness, ultimately affecting their health and quality of life (Haegele et al., 2019; Lieberman et al., 2010). This is further compounded by other factors including the level of VI, such that children and adolescents with greater magnitude of VI (those with no light perception) perform fewer physical activities than sighted youth resulting in more time with sedentary behaviors (Greguol et al., 2014; Kroksmark & Nordell, 2001; Lieberman et al., 2010). Subsequently, researchers and physical educators have become increasingly interested in PA and health promotion in youth who are blind (Lirgg et al., 2017). Yet, tracking the extent and magnitude of PA within this population may be difficult due to differences in movement as a result of the blindness (Haegele & Porretta, 2015). These differences include varying gait patterns, posture, and walking speeds, as they have to utilize more senses to navigate their environment compared to sighted youth (Hallenmas et al., 2010: Silva et al., 2018). Examples of this include forward head with kyphosis, lumbar lordosis, stiff arm swing, and small step length during walking (Chen & Wang, 2009; Gazzellini et al., 2016; Hallemans et al., 2011). Furthermore, individuals with VI exhibit worse balance and greater fear of falling during PA compared to their sighted peers (Silva et al., 2018).  Ultimately, these factors may affect how PA trackers measure step count in youth who are blind.

Over the last five years there has been an increase in the research of PA tracking devices for individuals with VI, from pedometers to ACC, for the accuracy of step count measurement (Beets et al., 2007; Haegele & Poretta, 2015; Haegele et al., 2017). Many different devices have been validated in children and adolescents who are sighted to accurately quantify steps per day (Haegele & Porretta, 2015; Mooses et al., 2018; Trost, 2007). Although research has been conducted utilizing PA devices to assess these habits of individuals with VI, few studies have validated these tools within this specific population of children who are blind and to our knowledge the Actigraph ACC step count feature has not been validated in this population (Haegele & Porretta, 2015; Haegele et al., 2017; Kozub et al., 2005; Willis et al., 2012; Zhu & Haegele, 2019).

Previous studies have shown that ACC are a valid and useful tool to assess steps in adults with VI (Holbrook et al., 2009), but there is limited information on the accuracy of this ACC to count steps in youth who are blind. Examining the validity of this tool in youth who are blind is important because they are less likely to achieve the recommended amounts of PA, which puts them at higher risk of developing poor health conditions (Lobenius- Palmer et al., 2018). Therefore, precise and practical tools to measure PA are needed in order to effectively assess and promote PA programming in this population (Sadowska & Krzepota, 2015).

Several studies have utilized ACC to assess overall PA in adults with VI (Sadowska & Krzepota, 2015; Willis et al., 2012). Therefore, the purpose of this study was to determine the reliability and validity of the Actigraph ACC steps in youth who are blind. We hypothesized that ACC would not differ in steps taken compared to the gold standard of direct observation (DO) and, therefore, would be an accurate tool to assess steps in youth who are blind.

Methods

Participants

Nine youth (age = 14 ± 2.1 years old; height = 152.3 ± 14.5 cm; weight = 49.5 ± 11.4 kg; body mass index = 21.1 ± 3 kg/m2) with a VI of no light perception (B1 classification by the United States Association of Blind Athletes) (IBSA Visual Classifications) completed this study during a one-week sports camp for children and adolescents with VI (Table 1). Steps were measured for each participant during the track and field activity time at camp.

Table 1. Participant characteristics.

Anthropometric Variable

 

     Age (years)

14.4 ± 2.1

     Height (cm)

152.3 ± 14.5

     Weight (kg)

49.5 ± 11.4

     BMI (kg/m2)

21.1 ± 3

cm = centimeters; kg = kilograms; BMI = body mass index; m = meters
Note: Data presented as mean ± SD.

Data collection

Institutional Review Board approval was obtained for this study. Permission was obtained from legal guardians of the youth and from the participants. Height and weight were measured via stadiometer and digital scale (SECA 769, Hamburg, Germany), and level of VI was obtained from participants and verified by their parents or guardians. Only those who had no light perception (B1) participated in this study. Prior to the beginning of data collection, each participant was familiarized with the goals of the study and the ACC. In addition, participants were previously familiarized with the track and their sighted guides.

The Actigraph GT3X ACC was used to record the steps taken. After the participants were familiarized to the ACC and the elastic belt that held the ACC it was secured around the waist. The ACC was aligned with the hip and placed above the knee on the mid-axial line according to manufacturer’s instructions. Since mobility aids, such as human guides and canes, were used by the participant, the device was worn opposite of the mobility aid as recommended by Holbrook et al. (2011).

Once the participants were fitted with the ACC, they were instructed to complete two laps (trials) of walking around a 400-meter track. The participants either walked independently, used a cane, or were led by human guide at their own pace. A cane was permitted if they felt safer or more comfortable. This was considered appropriate as it would mimic typical conditions for each individual participant. Human guide technique was used due to the fact that the track had no tactile markings to guide the participants (Rosen, n.d.).

During each lap the participants were also assessed by utilizing direct observation (DO) via Kodak Zi8 pocket video camera. Separate video recordings were taken of each participant’s feet while walking to obtain an accurate step count for trials one and two. Prior to each trial, researchers recorded the participant identification code, date, and beginning and end time of the trial. After each participant completed the two trials, the ACC data was downloaded and recorded. Step data from the ACC was time matched to when the participant began and ended their walking trials.

Data Analysis

Following data collection, researchers determined the step counts taken by each participant by reviewing the video recorded from each lap walked. Each video was viewed at least twice, and steps were counted and recorded during each viewing session for each participant. If a difference between determined observed steps was identified the video was reviewed a third time. Steps from each trial were entered and averages for each participant’s laps were determined.

Analyses were conducted with SPSS version 22.0 for Windows and alpha level was set a priori at p £ 0.05. Descriptive statistics were reported as means ± standard deviations. Pearson correlations and paired t-tests were conducted to examine the initial relationship between walking trials and ACC compared to DO. Absolute percent error was determined based on the difference in the number of steps between ACC and DO. A 95% confidence interval was determined from a regression equation to predict ACC from DO. Additional Bland-Altman analysis was completed (with a one-sample t-test and regression analysis) to determine whether there was a proportional bias between trials 1 and 2 using ACC and DO.

Table 2. Physical activity steps observed via direct observation and accelerometry.

Physical Activity Measure (Steps)

All (N = 9)

Direct Observation Trial 1

685.3 ± 96.6

Direct Observation Trial 2

681.6 ± 96.4

Direct Observation Average Trials

683.4 ± 96.3

Accelerometry Trial 1

664.3 ± 68.6

Accelerometry Trial 2

665.6 ± 77.6

Accelerometry Average Trials

664.9 ± 72.7

Note: Data presented as mean ± SD.

Results

Accelerometer reliability

The correlation between ACC steps taken in trial 1 and trial 2 was also large and significant (r = 0.978, p = 0.000) and there were no significant differences in the number of steps measured between the two trials (t = - 0.207, p = 0.841). A Bland-Altman analysis was undertaken (with a one-sample t-test and regression analysis) to determine whether there was a proportional bias between trials 1 and 2 for ACC counted steps. A one-sample t-test revealed no significant differences between the differences in number of steps taken between trials 1 and 2 (p = 0.841) for the ACC. Furthermore, regression analysis revealed no proportional bias (p = 0.158) in the differences in number of steps taken between trials 1 and 2.

Comparison of direct observation to accelerometry

Overall, the average number of steps counted with DO was 683.4 ± 96.3 steps compared to the 664.9 ± 72.7 steps counted with ACC, highlighting an average difference of 18.5 fewer steps counted with ACC (Table 1). Absolute percent error was determined to be 2.7%. The correlation for the average steps taken during both trials between DO and ACC was significant (r = 0.955, p = 0.000). A one-sample t-test revealed no significant differences between the number of steps taken between DO and ACC (t = 1.611, p = 0.146). A Bland-Altman plot and analysis (with a one-sample t-test and regression analysis) were used to determine whether there was a proportional bias between DO and ACC. A plot of the mean number of steps (between DO and ACC) versus the difference in the number of steps (between DO and ACC) revealed that eight out of nine data points (participants) were within a 95% confidence interval as depicted in Figure 1. However, regression analysis revealed a proportional bias (p = 0.04) in the differences in number of steps taken between DO and ACC, in favor of DO. These results found that, as the number of steps taken increased, the difference between the number of steps counted with DO and ACC also increased. This can be attributed to one participant being an outlier; they had the largest number of steps counted by both DO and ACC, and the largest difference in step count between the DO and ACC. Since the outlier is an anomaly and would generally be excluded, the ACC is accurate and is a reliable measurement of steps for children who are blind.

Bland-Altman plot of mean steps counted (between direct observation and accelerometry) versus differences in steps counted with a 95% confidence interval.

Figure 1Bland-Altman plot of mean steps counted (between direct observation and accelerometry) versus differences in steps counted with a 95% confidence interval. Notes: Diff = average trial difference in the number of steps between direct observation and accelerometry; Mean = average trial number of steps for direct observation and accelerometry

Discussion

The present study examined the use of the Actigraph ACC as an appropriate tool to measure steps walked around a track by youth who are blind. These results indicate that DO is a reliable measurement of steps counted supporting its use of the gold standard in PA measurement. The results also indicate that, when compared to DO, the use of ACC is a reliable and valid measure of the number of steps counted for youth who are blind. When the number of steps taken in trial 1 was compared to steps taken in trial 2 by ACC the results revealed ACC to be a reliable measure of steps counted. Additionally, the strong, positive correlations between steps taken in trials 1 and 2 found via video recording, highlighted DO as a reliable tool for comparison. The 95% confidence interval, based on a regression analysis to predict ACC from DO, was determined to be ± 46 steps. Although there was a significant correlation between steps counted with DO and ACC (r = 0.955, p < 0.00), and no significant differences in steps counted with DO and ACC (p = 0.146), to determine if there was a bias in the number of steps counted with ACC, a Bland-Altman plot and analysis was undertaken. The plot and analysis revealed that as the number of steps counted increased, there is a bias with an average of 18.5 fewer steps counted with ACC when compared to DO. Although regression analysis revealed that this bias was statistically significant (p = 0.04), in practical applications this is not significant. A difference of 18.5 fewer steps counted with ACC, out of an average of 683.4 steps counted with DO, results in 2.7% fewer steps counted. Although this difference could be deemed as minor, we do not know what this may indicate for longer distances. Therefore, it can be concluded that ACC is a reliable and valid means to measure step counts for short distance movement in youth who are blind, with an understanding that it can result in a lower number of steps when compared to DO.

Having valid and reliable PA monitors is essential when accurately assessing steps per day in youth who are blind. A review of studies examining the validity of step counters in individuals with disabilities found that the presence of a disability hinders the validity of the tool (Kenyon et al., 2013). Therefore, there is a substantial need to ensure that these tools correctly portray PA within these populations. The use of PA tracking devices, such as pedometers and ACC, has been found to be a valid and reliable tool for PA assessment in sighted children (Freedson et al., 2005; Mooses et al., 2018; Trost et al., 2005). In addition, talking pedometers have been validated for use in assessing PA in children with VI (Beets et al., 2007). However, these pedometers only record steps and no other information regarding the intensity and volume of PA like the ACC does. Although we did not specifically review intensity of activity, future research should focus on the validity of ACC collecting PA intensity. Little is known about the accuracy of ACC in counting steps in youth with VI, specifically those who are blind. Therefore, our findings may be among the first to deem the Actigraph ACC an appropriate tool for counting steps within this population, which is especially salient for research purposes examining quantity of PA determined by steps walked.

In summary, the DO method utilized in this study was deemed reliable with significant correlations between the two trials of steps walked and non-significant differences in the number of steps walked between trials. Therefore, it was chosen to be used as the comparison method when examining validity and reliability of steps recorded by ACC. Subsequently, in a comparison of ACC to DO, the ACC was both a valid and reliable measurement of steps taken among youth with VI. Therefore, the ACC can also be used in addition to, or independently from, DO in accurately assessing steps taken in this population.

Limitations

The only major limitation of this study was the use of human guides rather than, or in addition to, the use of personal assistance devices. Future research should examine the validity of the ACC in moving without a guide and with the participant’s typical assistance device.

Conclusion

In order to monitor PA accurately in research and in practice, it is important to determine which devices are valid and reliable among youth who are blind. These results suggest that, for youth who are blind, the ACC is an appropriate and practical tool to use for guided, short distance PA, in place of DO when assessing steps walked. However, because our participants were blind these results are not representative of all youth with VI and more research is needed. Due to the number of participants with varying levels of VI and the health disparities that exist, it is suggested that more research be conducted on youth with a range of VI to ensure that there are accurate and reliable devices that can be used for PA research in this population. Further research should focus on the validity of this tool for measuring energy expenditure and PA intensity.

Implications for Practitioners and Families

Physical activity is a major determinant of health and quality of life. Although significant research is available regarding the physical activity habits of youth, there is a dearth of literature pertaining to those with disabilities and, specifically, those with VI. Yet, we know that youth with VI tend to acquire less PA than their sighted peers. Current PA guidelines do not necessarily consider those with VI, who often engage in more sedentary behavior due to barriers to PA. Identifying the Actigraph ACC as a valid tool to use in quantifying PA will be useful in understanding current behavior and creating policy for improving physical activity. The Actigraph ACC is an appropriate tool to measure the quantity of PA determined by step count.

References

Beets, M. W., Foley, J. T., Tindall, D. W., & Lieberman, L. J. (2007). Accuracy of voice-announcement pedometers for youth with visual impairment. Adapted Physical Activity Quarterly, 24(3), 218-227. https://doi-org.ezproxy.valpo.edu/10.1123/apaq.24.3.218

Chen, S. & Wang, Y. (2009). The impact of different walking environment on the gait and posture pattern of individuals with visual impairment. Asian Journal of Exercise and Sport Science, 6(1),53-59. https://js.sagamorepub.com/ajess/article/view/2471

Freedson, P., Pober, D., & Janz, K. F. (2005). Calibration of accelerometer output for children. Medicine and Science in Sports and Exercise, 37(11), S523-S530. https://doi.org/10.1249/01.mss.0000185658.28284.ba

Gazzellini, S., Lispi, M. L., Castelli, E., Trombetti, A., Carniel, S., Vasco, G., Napolitano, A., & Petrarca, M. (2016). The impact of vision on the dynamic characteristics of the gait: strategies in children with blindness. Experimental Brain Research, 234(9), 2619–2627. https://doi.org/10.1007/s00221-016-4666-9

Greguol, M., Gobbi, E., & Carraro, A. (2014). Physical activity practice among children and adolescents with visual impairment - influence of parental support and perceived barriers. Disability & Rehabilitation, 37(4), 327-330. https://doi.org/10.3109/09638288.2014.918194

Haegele, J. A., & Porretta, D. L. (2015). Validation of a talking pedometer for adolescents with visual impairments in free-living settings. Journal of Visual Impairment & Blindness, 109(3), 219-223. https://doi.org/10.1177/0145482X1510900306

Haegele, J. A., Brian, A. S., & Wolf, D. (2017). Accuracy of the Fitbit Zip for measuring steps for adolescents with visual impairments. Adapted Physical Activity Quarterly, 34(2),195-200. https://doi-org.ezproxy.valpo.edu/10.1123/apaq.2016-0055

Hallemans, A., Ortibus, E., Truijen, S., & Meire, F. (2011). Development of independent locomotion in children with a severe visual impairment. Research in Developmental Disabilities, 32(6), 2069–2074. https://doi.org/10.1016/j.ridd.2011.08.017

Holbrook, E. A., Caputo, J. L., Perry, T. L., Fuller, D. K., & Morgan, D. W. (2009). Physical activity, body composition, and perceived quality of life of adults with visual impairments. Journal of Visual Impairment & Blindness, 103(1), 17-29. https://doi.org/10.1177/0145482X0910300104

Holbrook, E. A., Stevens, S. L., Kang, M., & Morgan, D. W. (2011). Validation of a talking pedometer for adults with visual impairment. Medicine and Science in Sports and Exercise, 43(6), 1094-1099. https://doi.org/10.1249/MSS.0b013e318205e2d6.

IBSA Visual Classifications/USABA Recognized Low Vision Classification. United States Association of Blind Athletes. http://usaba.org/membership/visual-classifications/

Kenyon, A., McEvoy, M., Sprod, J., & Maher, C. (2013). Validity of pedometers in people with physical disabilities: A systematic review. Archives of physical medicine and rehabilitation94(6), 1161-1170. https://doi.org/10.1016/j.apmr.2012.11.030

Kozub, F. M., Oh, H. K., & Rider, R. A. (2005). RT3 accelerometer accuracy in estimating short term physical activity in individuals with visual impairments. Adapted Physical Activity Quarterly22(3), 265-276. https://doi.org/10.1123/apaq.22.3.265

Kroksmark, U., & Nordell, N. (2001). Adolescence: The age of opportunities and obstacles for students with low vision in Sweden. Journal of Visual Impairment & Blindness. 95(4), 213-225. https://doi.org/10.1177/0145482X0109500403

Lieberman, L. J., Byrne, H., Mattern, C. O., Watt, C. A., Fernandez-Vivo, M. (2010). Health-related fitness in youths with visual impairments. Journal of Visual Impairment & Blindness, 104(6), 349-359. https://doi.org/10.1177/0145482X1010400605

Lirgg, C. D., Gorman, D. R., Merrie, M. D., & Shewmake, C. (2017). Exploring challenges in teaching physical education to students with disabilities. Palaestra, 31(2), 13-18.

Lobenius-Palmér, K., Sjöqvist, B., Hurtig-Wennlöf, A., & Lundqvist, L. O. (2018). Accelerometer-assessed physical activity and sedentary time in youth with disabilities. Adapted Physical Activity Quarterly, 35(1), 1-19. https://doi.org/10.1123/apaq.2015-0065

Mooses, K., Oja, M., Reisberg, S., Vilo, J., & Kull, M. (2018). Validating Fitbit Zip for monitoring physical activity of children in school: A cross-sectional study. BMC public health18(1), 858. https://doi-org.ezproxy.valpo.edu/10.1186/s12889-018-5752-7

Rosen, S. (n.d.) Human guide techniques: Study guide. American Printing House for the Blind. https://tech.aph.org/sbs/02_sbs_hg_study.html

Sadowska, D., & Krzepota, J. (2015). Assessment of physical activity of people with visual impairments and individuals who are sighted using the International Physical Activity Questionnaire and Actigraph. Journal of Visual Impairment & Blindness, 109(2), 119-129. https://doi.org/10.1177/0145482X1510900207

Silva, E., Fischer, G., Rosa, R., Schons, P., Teixeira, L., Hoogkamer, W., & Peyré-Tartaruga, L. (2018). Gait and functionality of individuals with visual impairment who participate in sports. Gait & Posture, 62, 355-358. https://doi.org/10.1016/j.gaitpost.2018.03.049

Trost, S. G., McIver, K. L., & Pate, R. R. (2005). Conducting accelerometer-based activity assessments in field-based research. Medicine and Science in Sports and Exercise, 37(11), S531-S543. https://doi.org/10.1249/01.mss.0000185657.86065.98

Trost, S. G. (2007). Measurement of physical activity in children and adolescents. American Journal of Lifestyle Medicine, 1(4), 299-314. https://doi.org/10.1177/1559827607301686

Willis, J. R., Jefferys, J. L., Vitale, S., & Ramulu, P. Y. (2012). Visual impairment,uncorrected refractive error, and accelerometer-defined physical activity in the United States. Archives of ophthalmology, 130(3), 329-335. https://doi.org/10.1001/archopthalmol.2011.1773

Zhu, X., & Haegele, J. A. (2019). Reactivity to accelerometer measurement of children with visual impairments and their family members. Adapted Physical Activity Quarterly, 36(4), 492-500. https://doi.org/10.1123/apaq.2019-0040

The Journal of Blindness Innovation and Research is copyright (c) 2021 to the National Federation of the Blind.