Over the past few decades, there has been increased interest and research in the area of physical activity assessment and its relationship to health outcomes. Physical activity has been defined as “any bodily movement produced by skeletal muscles that results in energy expenditure” (6). The majority of epidemiological studies conclude that physical inactivity is positively associated with the incidence rates of morbidity and mortality from a number of chronic diseases (10,11,13–15,21). Furthermore, it has been determined that this is a causal relationship (2,15,20). To study this relationship more closely, epidemiologists and other researchers have developed a number of measurement tools to assess the quantity of physical activity to establish the levels needed to provide a protective effect from chronic disease. The Centers for Disease Control and Prevention and the American College of Sports Medicine recently adopted a new position stance on the amount of recommended exercise for the adult population of the United States (21). The position of these organizations is that every U.S. adult should accumulate 30 min or more of moderate intensity physical activity on most, preferably all, days of the week (21). This statement provides a generalized quantification for the total amount of physical activity recommended.
There has also been increased attention directed to the quantification of physical activity by utilizing mechanical or electronic monitors. These devices provide the necessary objectivity that paper and pencil techniques lack and are not as expensive as the doubly labeled water (DLW) technique. Also, DLW only allows for measurement of total energy expenditure. In certain applications where dose-response relationships are of interest, patterns and intensity levels of physical activity are needed. A recent addition to the activity monitor market, and the subject of this investigation, is the CSA model 7164 (Computer Science and Applications, Inc., Shalimar, FL) accelerometer (8). The CSA model 7164 utilizes the same technology as the previous model (model 5130), but the size of the unit has been reduced making the newer model less obtrusive. The CSA model 5130 has been validated in a laboratory setting during treadmill locomotion in male and female adults (17). The purpose of this investigation was to examine the CSA monitor (model 7164) in a field setting, where it could be used to characterize total as well as daily patterns of physical activity.
Ten men and 10 women were recruited from the university community (25.0, 3.6 yr). Subjects provided written informed consent in accordance with university regulations on the use of human subjects.
The CSA monitor (model 7164) weighs 56.7 g, is 5.0 × 5.0 × 1.5 cm (8), which is approximately the size of a large sport wristwatch. This monitor measures integrated accelerations in the vertical plane via a piezoelectric plate. To begin data collection, the monitor is initialized using an IBM compatible computer. A real-time internal clock allows the researcher to begin collecting at a desired time. Data are accumulated over a user-defined interval or epoch and stored in the internal memory. Collection is ended when the unit is downloaded to the computer. For this study, a 1-min sampling interval was used and data were collected over a 7-d period. After downloading, it is possible to view activity patterns or the amount of physical activity over a specific time frame. The output from the CSA monitor is in counts for each epoch. Counts represent the summed amount and magnitude of accelerations during each epoch.
A 3-d activity diary was the criterion measure of physical activity for this study (5). It divided the day into 24 one-hour segments. For each 15-min block of time, subjects reported the amount of time spent in various activity levels. These levels were determined based on the compendium of activities by Ainsworth et al. (1). The subjects were given a sample list of activities and the corresponding intensity level to categorize their activities accurately. Self-reported energy expenditure was calculated from the 3-d activity diary data. The number of hours spent at each intensity level was recorded for each day. The hours were then multiplied by the appropriate MET value (1 MET = 1 kcal·kg−1·h−1). The products obtained were summed and multiplied by the subject’s body mass to provide estimated total kcal·d−1 (Dkcaltot). Times spent in swimming and showering/bathing were not included in any analyses because the CSA was not worn during these times.
The Stanford Seven Day Recall (SDR) (4) was used for additional comparison with the CSA monitor. The SDR was set up in an interviewer-administered format. The same individual performed all of the SDR interviews. Subjects were asked about their sleep activities and then prompted to recall how many hours each day they spent in moderate, hard, and very hard activities. Light-intensity activity was obtained by subtracting the number of hours spent in sleep, moderate, hard, and very hard activities from the total number of hours in a week (168 h). These procedures have been previously described by Blair (4). The number of hours obtained for each day was multiplied by the appropriate MET value. The values were summed and multiplied by the subject’s body mass to obtain a value of energy expenditure (Rkcaltot).
During the first laboratory visit, subjects completed the informed consent document and any questions were answered. Descriptive characteristics were obtained including age, height, and body mass. During this visit, subjects were familiarized with how to wear the CSA monitor and record their activity in the diary. The CSA monitors were attached firmly to the waist over the right hip (anterior superior iliac spine) using an elastic strap with an adjustable buckle. Subjects were instructed to wear the CSA monitor during all waking hours, except when bathing or swimming. Subjects recorded the times that the CSA monitor was removed and provided information about the activity done during that time. To provide practice at recalling physical activity, they were asked to complete an SDR. Subjects were then given the CSA monitor, the 3-d activity diary, and a list of activities and their corresponding intensity values as guidelines for assessing their own activity (1). The 3-d activity diary was completed on the Thursday, Friday, and Saturday of the 7-d collection period for half of the subjects and Sunday, Monday, and Tuesday for the other half. This provided a more complete comparison between weekday and weekend activity by including all days of the week except Wednesday. During the second laboratory visit, 7 d later, the subject returned the CSA monitor and 3-d activity diary. The subjects then completed the SDR and body mass was measured to ensure the subjects had been weight stable for the preceding week.
The CSA monitor data were reduced with custom software developed for this project. The software categorized each count·min−1 value into light, moderate, hard, or very hard activity. This software also calculated 15- and 30-min mean count·min−1 values and categorized these means into one of the four intensities. Intensity categories based upon CSA monitor data, established in a previous calibration study (9), are presented in Table 1. Because subjects completed the diary by evaluating their activity every 15-min, the CSA monitor data were also divided into 15-min blocks of time. An average count·min−1 value was obtained for each 15-min time period. This average was categorized into one of four intensities (light, moderate, hard, or very hard).
One of the limitations of the CSA monitor is that it is relatively insensitive to physical activities that require little vertical movement. Because of this, the time spent cycling and weight lifting determined from the diary were used to adjust the monitor data. This was accomplished by first eliminating these times from the analyses and then replacing the CSA counts with count values that would represent the typical intensity of such activities based on the compendium of physical activities (1) and the calibration study (9). These data adjustments were not used in any of the analyses after it was determined that the adjusted values did not affect the results. For the purpose of comparing days, day 1 was considered the first weekday that the diary was completed, day 2 was the second weekday, and day 3 was the weekend day for all subjects.
To determine total activity changes and patterns of change across days, separate one-way repeated measures ANOVA (day) were completed for the diary (Dkcaltot) and CSA (cnttot). The Kappa statistic was used to determine the percent agreement between the diary and the CSA monitor (7). To perform this test, subjects were grouped into low, medium, and high active groups based on total counts·.d−1 from the CSA monitor and total kcal·d−1 from the diary. To compare the CSA monitor with the 3-d diary, only complete 15-min blocks from both instruments were used. A variable called “activity minutes” was calculated for each instrument as the sum of minutes spent in moderate, hard, and very hard activities. To determine differences in minutes spent in each activity level, a two-way repeated measures ANOVA (method × day) was performed for each intensity. The location of significant pairwise differences was determined using Tukey’s post hoc procedure.
For the SDR, subjects were asked to recall time spent in various activity intensities to the nearest half-hour. Thus, monitor data were divided into 30-min blocks of time and an average count·min−1 value was categorized as one of the four intensity groups. The mean number of minutes spent in each intensity category for the weekdays, and two weekend days were compared between the SDR and the CSA monitor. A two-way repeated measures ANOVA (method × day) was completed for each intensity category.
Pearson product-moment correlations were computed to compare the CSA monitor with the diary and the CSA monitor with the SDR. Correlations were calculated between the CSA monitor and diary for the number of min spent in each activity category and for total activity per day. Correlations between the CSA monitor and SDR were calculated for total activity on weekdays and weekends.
One subject was eliminated from the analyses due to computer downloading error. The male (N = 9) and female subjects (N = 10) were 25.0 ± 2.4 and 26.0 ± 4.6 yr (P = 0.61), 176.7 ± 13.1 and 167.8 ± 5.0 cm (P = 0.07), and 74.9 ± 10.4 and 66.6 ± 8.0 kg (P = 0.06) respectively. Subjects were weight stable for the week of activity monitoring (P for dependent t-test = 0.98).
Gender and group differences.
Results of a three-way ANOVA (gender × method × day) indicate there were no significant gender by method or gender by day interactions for the number of minutes spent in light, moderate, hard, or very hard activities for either the diary or the CSA monitor data. Therefore, data were pooled for subsequent analysis. Also, there were no significant differences between the group that completed the diary on Thursday, Friday, and Saturday and the group that completed the diary on Sunday, Monday, and Tuesday. Therefore, both groups were pooled for data analysis.
Figure 1 shows the patterns of change across the three days for the CSA monitor and the diary. For both instruments, day 3 total activity was significantly less than day 2 (P < 0.01) but not different from day 1 (P = 0.20). There was a 32% decrease in total counts from the monitor and a 20% drop in total kcal from the diary (approximately 520 kcal) from day 2 to day 3.
Subjects were classified as low, moderate, or highly active based on the tertile of mean counts·d−1 from the monitor and mean kcal·d−1 from the diary. The Kappa of k = 0.53 is significant at P < 0.01, indicating that the CSA categorizes subjects reasonably well into either low, medium, or high active groups (Table 2).
CSA monitor and diary.
Light activity on the weekend day decreased 15% and 13% based on the CSA monitor and the diary, respectively (P < 0.01). A day effect was not detected for moderate, hard, or very hard activities. On average, the CSA monitor recorded 5.9% greater time spent in light activity across days (P < 0.01) in comparison with the diary. For the moderate intensity category, no significant differences between the CSA monitor and the diary were detected (P = 0.68). The CSA monitor recorded 81% less hard activity (P < 0.01) and also recorded 84% less very hard activity across days compared with the diary. Relative to the diary, the CSA monitor recorded 50% fewer minutes spent in the activity minutes category across days (P < 0.01) (Fig. 2).
CSA monitor and SDR.
The significant day main effect indicates that light activity recorded on the 5 weekdays by the CSA monitor and SDR was 9.5% greater than on the weekend days (P = 0.02). For the other intensities (moderate, hard, and very hard), the weekdays were similar to the weekend days for both instruments. There was no significant difference between the number of minutes spent in light activity between the CSA monitor and the SDR (P = 0.26). However, the CSA monitor recorded 43.8% less time in moderate activity (P = 0.06), 94.3% less time in hard activity (P < 0.01), 93.8% less time in very hard activity (P < 0.01), and 67.3% less time in the “activity minutes” category (P < 0.01) (Fig. 3).
Overall activity for the CSA and the SDR differ across days. Although there was a strong trend (P = 0.06) for cnttot to decrease on the weekend according to the CSA monitor (461 ± 130.1 to 360 ± 256.8 counts−100·d−1), the SDR suggests there was no difference between weekday and weekend Rkcaltot (2223 ± 417.2 to 2224 ± 492.5 kcal·d−1;P = 0.99).
Table 3 presents the correlations between the CSA monitor and the diary for each day and for all 3 days pooled for the number of minutes spent in each activity category. The activity minute values were correlated on each day and for the 3 days pooled. One of the subjects was considered an outlier due to employment that required a great deal of moderate and fast walking. The subject worked a total of 13 h during the 3-d diary period. Because of this, the correlations are presented with and without this individual. Removing this subject, most of the correlations, with the exception of the light activity category, are no longer significant. An example of the effect of the outlier is illustrated in Figures 4 and 5.
Table 4 presents the correlations between the CSA monitor (total counts) and the diary (total kcal) and Table 5 shows the correlations between the CSA and the SDR for total counts from the CSA monitor and total kcal from the SDR. The correlations are presented with and without the outlier. The CSA correlated more highly with the diary and the SDR when total counts and total kcal were compared (Tables 4 and 5) as opposed to min in specific activity categories (Table 3).
The major finding of this study is the similar pattern of activity for the CSA monitor and diary as measured with counts and kcal, respectively. Due to the different units of measurement, it is not possible to say if one instrument was consistently higher or lower. The correlations presented here are slightly lower or similar in comparison to previous studies (16,18,23), but the pattern of change over time is the same for both instruments. The similar pattern of activity of the CSA monitor compared to the diary may indicate a potential roll in evaluating changes in activity over time and physical activity’s effect on health.
There was a 20% (520 kcal) decrease in total kcal from day 2 to day 3 calculated from the diary and a 32% decrease from day 2 to day 3 in total counts from the CSA (Fig. 1). It would seem that the decrease in total activity seen on the weekend is partly due to the decrease in light activity. However, there was also a 25% (114 min) increase in the amount of time spent sleeping on the weekends as calculated from the diary. This may account for some of the decrease in light activity and total activity seen on the weekend day.
The Kappa value (Table 2) indicates that the CSA monitor can classify people reasonably well into the appropriate activity level using the diary as the criterion measure. This evidence along with the similar patterns of activity presented in Figure 1 demonstrates that the CSA monitor may be of use in future studies where a change in physical activity over time is of interest.
The CSA monitor provides a higher mean estimate of time spent in light activity by 6% (50 min) and a lower estimate of time in hard and very hard activities by 81–84% (13–14 min) when compared with an activity diary. One explanation for these results is a tendency for people to overreport their time spent in vigorous activity and underreport time spent in light activity. This phenomenon has been documented in other studies using self-report along with direct observation and activity monitoring as the methods of assessing physical activity (3,12,16). Although a list of activities and their corresponding intensities were given, people may have different perceptions of the intensity of their activity based on their current fitness level. This study was conducted with the diary serving as the criterion measure. This method has been reported as valid in several studies (16,18,22), yet it remains a subjective measure of physical activity which is prone to error due to inaccurate recall.
Although there was no significant difference between the CSA monitor and the SDR for the light activity category, there were significant instrument differences in the moderate, hard, and very hard activity categories. The CSA monitor recorded less time spent in these activities. This pattern of differences is similar to that seen between the CSA monitor and the diary and may be due to similar recall errors. The Rkcaltot reported from the SDR were very similar for the weekday and weekend days, which is in contrast to the decreases reported by the CSA monitor and the diary. This stable total activity level (Rkcaltot) coincides with no difference in the amount of sleep recalled by subjects for weekdays and weekends. In contrast, the diary indicates a 25% increase in the amount of time spent sleeping on the weekend day. If the diary is assumed to be the criterion measure, the amount of light activity reported for the SDR may be overestimated and the time spent sleeping on the weekend may be underestimated for this sample (Fig. 3).
The CSA monitor recorded a similar number of minutes spent in light activity and underestimated time spent in moderate and vigorous activities compared with the diary and the SDR. The correlations between the CSA monitor and the diary for the number of min spent in the moderate, hard, and very hard activity categories are, for the most part, nonsignificant with the outlier removed. The correlations for the light activity were significant even with the outlier removed probably due to the larger range of values in this category compared to the moderate and vigorous categories. This may play an important role in determining activity levels in the majority of the population that is predominantly sedentary. Knowledge of the amount of time spent in light activity should reveal the amount of time a person is active and inactive. If a sedentary person gains the most health benefits from leaving the sedentary or lowest active category (19,21), knowing the amount of time spent in sedentary activities may be just as beneficial as knowing how much time was spent in moderate or greater intensity activity.
Tables 4 and 5 present the correlations between the diary, SDR, and CSA monitor for Dkcaltot, Rkcaltot, and cnttot. These somewhat modest, yet significant, correlations are similar to those reported in other studies that have compared activity monitors with either a diary or SDR (18,23). However, using the TriTrac triaxial activity monitor, Matthews and Freedson (16) reported correlations of r = 0.76–0.83 when the TriTrac was compared with a diary and r = 0.77 for the TriTrac correlated to the SDR. These higher correlations may be due to the three axes of the TriTrac providing a more accurate assessment of activity. The higher correlations found by Matthews and Freedson (16) may be due to a greater range of activity levels in that subject pool. However, comparing the kcal levels from the diaries and recalls for both studies revealed similar ranges of values for the present study and the Matthews and Freedson study (1774–3500 vs 1806–3760, respectively).
In conclusion, the CSA monitor overestimated time spent in light activity and underestimated time spent in vigorous activity compared to the diary. However, because the diary method is not a true gold standard, the diary may be underestimating light activities and overestimating vigorous activities. Although correlations between the CSA monitor, the diary and the SDR are lower than those using other monitors (Caltrac, TriTrac), there was a similar pattern of total activity observed between the CSA monitor and the diary. The significant decrease in activity on the weekend reported by the CSA monitor parallels the change observed with the diary indicating that the CSA monitor is sensitive to changes in activity. The significant Kappa statistic indicates that the CSA monitor is able to categorize people into the appropriate tertile of activity. These results indicate that the CSA monitor (7164) can be used in field situations to unobtrusively assess physical activity in young adults.
1. Ainsworth, B. E., W. L. Haskell, A. S. Leon, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med. Sci. Sports Exerc. 25:71–80, 1993.
2. Blair, S. N., H. W. Kohl III, C. E. Barlow, R. S. Paffenbarger, Jr., L. W. Gibbons, and C. A. Macera. Changes in physical fitness and all-cause mortality: a prospective study of healthy and unhealthy men. JAMA 273:1093–1098, 1995.
3. Blair, S. N., M. Dowda, R. R. Pate, et al. Reliability of long-term recall of participation in physical activity
by middle-aged men and women. Am. J. Epidemiol. 133:266–275, 1991.
4. Blair, S. N. How to assess exercise habits and physical fitness. In: Behavior Health: A Handbook of Health Enhancement and Disease Prevention, J. D. Matarazzo, S. M. Weiss, J. A. Herd, N. E. Miller, and W. M. Weiss (Eds.). New York: John Wiley, 1984, pp. 424–447.
5. Bouchard, C., A. Trembley, C. Leblanc, G. Lortie, R. Savard, and G. Theriault. A method to assess energy expenditure
in children and adults. Am. J. Clin. Nutr. 37:461–467, 1983.
6. Casperson, C. J., K. E. Powell, and G. M. Christenson. Physical activity
, exercise and physical fitness. Public Health Rep. 100:125–131, 1985.
7. Cohen, J. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20 (1):37–46, 1960.
8. CSA. (Model 7164) Reference Manual
. Shalimar, FL: Computer Science and Applications, Inc., 1996.
9. Freedson, P. F., E. Melanson, and J. Sirard. Calibration of the Computer Science and Applications, Inc. accelerometer. Med. Sci. Sports Exerc. 30:777–781, 1998.
10. Helmrich, S. P., D. R. Ragland, R. W. Leung, and R. S. Paffenbarger, Jr. Physical activity
and reduced occurrence of non-insulin-dependent diabetes mellitus. N. Engl. J. Med. 325:147–152, 1991.
11. Helmrich, S. P., D. R. Ragland, and R. S. Paffenbarger, Jr. Prevention of non-insulin-dependent diabetes mellitus with physical activity
. Med. Sci. Sports Exerc. 26:824–830, 1994.
12. Klesges, R. C., L. M. Klesges, A. M. Swenson, and A. M. Pheley. A validation of two motion sensors in the prediction of child and adult physical activity
levels. Am. J. Epidemiol. 122:400–410, 1985.
13. Lee, I-M., R. S. Paffenbarger, and C.-C. Hsieh. Physical activity
and risk of prostatic cancer among college alumni. Am. J. Epidemiol. 135:169–179, 1992.
14. Lee, I-M., and R. S. Paffenbarger. Quetlet’s index and risk of colon cancer in college alumni. Natl. Cancer Inst. 84:1326–1331, 1992.
15. Lee, I-M., and R. S. Paffenbarger. Physical activity
and its relation to cancer risk: a prospective study of college alumni. Med. Sci. Sports Exerc. 26:831–837, 1994.
16. Matthews, C. E., and P. S. Freedson. Field trial of a three-dimensional activity monitor
: comparison with self report. Med. Sci. Sports Exerc. 27:1071–1078, 1995.
17. Melanson, E. L., Jr., and P. S. Freedson. Validity of the Computer Science and Applications, Inc. (CSA) activity monitor
. Med. Sci. Sports Exerc. 27:934–940, 1995.
18. Miller, D. J., P. S. Freedson, and G. M. Kline. Comparison of the activity levels using the Caltrac accelerometer and five questionnaires. Med. Sci. Sports Exerc. 26:376–382, 1994.
19. Paffenbarger, R. S., Jr., A. L. Wing, and R. T. Hyde. Physical activity
as an index of heart attack risk in college alumni. Am. J. Epidemiol. 108:161–175, 1978.
20. Paffenbarger, R. S., Jr., and R. T. Hyde. Exercise in the prevention of coronary heart disease. Preventive Med. 13:3–22, 1984.
21. Pate, R. R., M. Pratt, S. N. Blair, et al. Physical activity
and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 273:402–407, 1995.
22. Schulz, S., K. R. Westerterp, and K. Bruck. Comparison of energy expenditure
by the doubly labeled water technique with energy intake, heart rate, and activity recording in man. Am. J. Clin. Nutr. 49:1146–1154, 1989.
23. Williams, E., R. C. Klesges, C. L. Hanson, and L. H. Eck. A prospective study of the reliability and convergent validity of three physical activity
measures in a field research trial. J. Clin. Epidemiol. 42:1161–1170, 1989.