-
Out of 120 recruited children, 99 (82.5%) were included. The data from 21 children were excluded due to technical problems with the device and failure in performing the task on 5 required valid days. As shown in Table 1, the gender distribution was approximately equal, with 48 boys (BMI: 21.1 ± 5.3) and 51 girls (BMI: 19.2 ± 3.2), a mean age of 13.0 (s: 2.5) years, and 48.5% overweight or obese. There were no significant differences in mean age, weight, height, BMI, and household education between students included and excluded from the analyses.
Participant characteristics All Boys Girls P value* Sample number 99 48 51 − Age1 mean ± SD 13.0 ± 2.5 13.3 ± 2.5 12.8 ± 2.4 0.265 Height (m)a mean ± SD 1.6 ± 0.1 1.6 ± 0.2 1.5 ± 0.1 < 0.001 Weight (kg)a mean ± SD 49.8 ± 17.2 55.9 ± 19.9 44.0 ± 11.7 < 0.001 BMI (kg/m2)a mean ± SD 20.1 ± 4.5 21.1 ± 5.3 19.2 ± 3.2 0.025 BMI categories n (%) Overweightb 29 (29.3%) 18 (37.5%) 11 (21.6%) 0.083 Obeseb 19 (19.2%) 13 (27.1%) 6 (11.8%) 0.054 Ethnicb n (%) Majority 91 (91.9%) 53 (93%) 57 (90.5%) 0.517 Minority 8 (8.1%) 4 (7%) 6 (9.5%) Father's educationb n (%) None, primary or unknown 8 (8.1%) 5 (10.4%) 3 (5.9%) 0.479 Secondary 53 (53.5%) 28 (58.3%) 25 (49.0%) college, university of trade 38 (38.4%) 15 (31.3) 23 (45.1%) Mother's educationb n (%) None, primary or unknown 14 (14.2%) 8 (16.7%) 6 (11.8%) 0.245 Secondary 43 (43.4%) 25 (52.0%) 18 (35.3%) college, university of trade 42 (42.4%) 15 (31.3%) 27 (52.9%) Note. BMI = body mass index. *P values, which were calculated by univariate ANOVA procedures or χ2 test of variances between genders. aDescribed as x ± s because of the normal distribution. bDescribed as absolute number (percentage) because of categorical data. Table 1. Characteristics of the study population
-
The median (25th, 75th quartile) values of each physical activity variable and ICC across 7 days from the wristband and the accelerometer are shown in Table 2. The test for normality revealed that physical activity variables were non-normally distributed. The ICC for median values of physical activity variables across the 7-day wristband measurements ranged from 0.71 to 0.82, less than that of accelerometer measurements (ICC: 0.72–0.91). As measured by the wristband, there was no difference across days for steps, time spent on total physical activity and LPA, whereas only accelerometer measurements of LPA and MPA showed no significant difference.
Physical activity variables Wristband activity monitor Accelerometer Average per daya ICCb 95% CIb P valuec Average per daya ICCb 95% CIb P valuec Steps 7,244.0 (4,590.5–10,205.5) 0.71 0.62–0.79 0.320 8,331.0 (6,002.5–10,737.5) 0.72 0.62–0.80 0.019 PA (min) 166.0 (102.5–233.5) 0.76 0.68–0.83 0.723 146.0 (106.3–196.8) 0.87 0.82–0.90 0.032 MVPA 89.0 (54.0–137.0) 0.81 0.75–0.86 0.015 54.5 (38.1–74.5) 0.76 0.68–0.83 0.005 VPA 34.0 (14.0–58.0) 0.71 0.61–0.80 0.023 15.3 (6.7–23.5) 0.75 0.66–0.82 < 0.001 MPA 52.0 (27.0–85.0) 0.82 0.75–0.87 0.003 38.5 (25–52.5) 0.83 0.77–0.87 0.075 LPA 66.0 (34.5–101.5) 0.81 0.75–0.86 0.368 91.0 (59.8–126.5) 0.91 0.88–0.93 0.052 PAEE (kcal) 112.0 (71.0–162.5) 0.76 0.68–0.83 0.004 327.9 (217.1–485.7) 0.86 0.82–0.90 < 0.001 Note. PA = physical activity; MVPA = moderate-to-vigorous physical activity; VPA = vigorous-intensity physical activity; MPA = moderate-intensity physical activity; LPA = light-intensity physical activity; PAEE = physical activity energy expenditure; ICC = Intraclass Correlation Coefficients; CI = Confidence Interval. aData were described as median (25th quartile, 75th quartile). bThis is the intraclass correlation coefficient and 95% confidence interval across the 7 days. cThis is the P value for between-day differences using repeated measures analysis. Table 2. Reliability and summary statistics for the wristband activity monitor and accelerometer
The highest level of physical activity occurred on day 1, and the lowest level was on day 4 (Figure 2). The daily patterns of steps, physical activity energy expenditure, and time spent on physical activity by intensity were similar between the two monitors. By using the Spearman-Brown prophecy formula to determine the optimal number of days for activity data collection, it was suggested that a minimum of 3 days of wristband measurement would be required to achieve an ICC of 0.8.
-
The correlation between 7-day averages for the wristband and the accelerometer was strong for steps (rho: 0.72). For time spent on total physical activity, MVPA, VPA, and physical activity energy expenditure correlations were moderate (rho: 0.51 to 0.63), while for MPA (rho: 0.38) and LPA (rho: 0.45), they were week (Table 3). When each day was examined separately, the correlation coefficients were significant, and the lowest correlation coefficients appeared on Sunday or Day 1. Results from the analyses of MAPE demonstrated that the errors of steps and time spent on total physical activity measurement by the wristband were low (MAPE < 15%), but time spent on different intensities of physical activity and physical activity energy expenditure were relatively high (range from 29.1% to 95.9%).
Physical activity variables Spearman’s correlations between Wristband activity monitor and Accelerometer* MAPE (%) between the two monitors Day 1 Day 2 Day 3 Day 4 Day 5 Saturday Sunday 7-day average Steps 0.58 0.74 0.78 0.76 0.79 0.64 0.54 0.72 −14.5 PA (min) 0.51 0.66 0.74 0.68 0.74 0.55 0.34 0.63 8.0 MVPA 0.30 0.40 0.67 0.66 0.65 0.41 0.41 0.55 61.2 VPA 0.20 0.36 0.60 0.55 0.52 0.39 0.40 0.51 95.9 MPA 0.15 0.33 0.49 0.56 0.42 0.38 0.09 0.38 36.1 LPA 0.35 0.39 0.63 0.51 0.52 0.41 0.20 0.45 −29.1 PAEE (kcal) 0.51 0.64 0.66 0.56 0.49 0.53 0.46 0.57 −68.1 Note. PA = physical activity; MVPA = moderate-to-vigorous physical activity; VPA = vigorous-intensity physical activity; MPA = moderate-intensity physical activity; LPA = light-intensity physical activity; PAEE = physical activity energy expenditure; MAPE = median of absolute percentage error. *All P < 0.001. Table 3. Validity and MPAE of wristband activity monitor with accelerometer
Bland-Altman plots were constructed to give a visual representation of the agreement between the two monitors for physical activity variables (Figure 3). Specifically, analyses identified a mean underestimation of 633.5 steps and 244.1 kilocalories, with a mean overestimation of 21.4 minutes for time spent on total physical activity and 42.6 min for MVPA by the wristband. There did appear to be an obvious trend in the data points on the plots except for a slight trend for steps. More time spent on total physical activity and MVPA were more likely to be overestimated, and the differences between the two methods were much higher, whereas a higher level of physical activity energy expenditure was more likely to be underestimated, and the difference between the two methods were much higher.
Validity and Reliability of the Wristband Activity Monitor in Free-living Children Aged 10−17 Years
doi: 10.3967/bes2019.103
- Received Date: 2019-07-11
- Accepted Date: 2019-09-23
-
Key words:
- Physical activity /
- Measurement /
- Children /
- Wristband activity monitor
Abstract:
Citation: | YANG Xi, JAGO Russell, ZHANG Qian, WANG Yu Ying, ZHANG Jian, ZHAO Wen Hua. Validity and Reliability of the Wristband Activity Monitor in Free-living Children Aged 10−17 Years[J]. Biomedical and Environmental Sciences, 2019, 32(11): 812-822. doi: 10.3967/bes2019.103 |