EASON, KAREN E.; MÂSSE, LOUISE C.; KELDER, STEVEN H.; TORTOLERO, SUSAN R.
Research has shown that physical activity serves a preventive function against the development of several major chronic diseases, such as noninsulin-dependent-diabetes mellitus, osteoporosis, stroke, hypertension, colon cancer, obesity, depression, and coronary heart disease (20). However, studying physical activity and its benefits is difficult due to the complex nature of physical activity and the associated difficulty in measuring physical activity. One of the difficulties in measuring physical activity is that activity type, intensity, frequency, and duration can vary considerably both within and among individuals and populations. Furthermore, the extent to which physical activity patterns vary within and among individuals and over time may have important implications in measurement of activity. For example, investigators have suggested the need for multiple days of activity measurement, but it is uncertain how many days are necessary to capture an individual’s typical physical activity pattern (13). Understanding the variations in physical activity patterns will help identify how many days of reporting activity are needed.
Recent studies have reported varying outcomes related to the number of recording days needed to reliably estimate physical activity. The disagreement among studies may be due to the methods used to measure activity and the sample being studied. Studies using an accelerometer report the recommended number of days needed to reliably estimate activity as 4–5 for children (19), 8–9 for adolescents (19), and 3–6 for adults (6,10). With the diary method, 8–14 d of reporting are recommended (3,6). The varying methods of measuring activity, the varying populations, and the varying number of days reported as necessary to reliably estimate activity presented in these studies indicate a need for further research in this area to more clearly identify measurement days needed. Furthermore, none of the current studies examine activity by the type of activity, only total activity. Understanding the variability of activity by type of activity such as work, household chores, or exercise may be useful in determining the number of days necessary to reliably estimate physical activity. It may be the case that differing number of days of recording activity are required for different types of activity, or that some types of activity are reliably measured with fewer days. Similarly, patterns of activity may differ for individuals who work and those that are unemployed. Some studies have shown that working status affects levels of physical activity (5,9,11,15). Further, patterns of activity may differ by ethnicity. Therefore, the purpose of this study was to estimate variation in energy expenditure among older African-American and Hispanic female workers and nonworkers, and identify how many days of diary self-report are needed to reliably estimate activity for total activity and type of activity. This study focused on understanding variability by activity type as this information can be used to guide the development of measurement protocols and instruments to measure physical activity.
Participants for this analysis were obtained from the Women On The Move [WOTM] study, a project aimed at developing and validating physical activity surveys for minority women. WOTM involved a sample of 260 African-American and Hispanic women aged 40–70 yr old residing in Houston, TX. To participate in the study, women had to meet the following criteria: 1) no health limitations that prevented them from being physically active, 2) not pregnant or planning on becoming pregnant during the study, and 3) no plans to move out of the geographical area within the next year. Women were recruited to participate in the WOTM study through the media (print, television, and radio), community presentations, and posting of flyers. After recruitment, a total of 656 women (311 African-American and 340 Hispanic) expressed interest in the WOTM study (i.e., left a telephone message), of which 590 women were screened by telephone for eligibility. Of the 590 women screened by telephone, 386 women volunteered to attend the in-person screening. After the in-person screening, 306 met the study eligibility criteria. Of the 306 eligible women, 260 enrolled in the WOTM study, 130 African-American, and 130 Hispanic women. A total of 227 of these women completed a 7-d diary and were used in the analysis for this study. Women who accounted for 75% or more minutes of each day (not including sleep) for 6 d were included in the analyses.
Approval for the study was received from the human subjects committees at the University of Texas-Houston Health Science Center and the Centers for Disease Control and Prevention. During the in-person screening, details of the study were explained to the participant, and written informed consent was signed. After informed consent was given, weight was measured, and a demographic questionnaire was completed. Participants were also trained at this time on recording information into the activity diary. A trained interviewer instructed the participant how to complete the diary, and then had the participant complete a practice diary. To evaluate the participants’ ability to complete an accurate diary, the participant was asked to complete a 1-d diary the following day and return it in a pre-addressed and stamped envelope. Returned diaries were reviewed for completeness, readability, accuracy, and correct use of format and codes. Those women who returned acceptable diaries were eligible to participate in the study. Additional diary training was conducted later in the protocol.
The WOTM study included three 2-h interviews (T1, T2, and T3) exactly 1 wk apart and a 1-h follow-up interview (T4) conducted 6 wk after the third interview (T3). Trained interviewers scheduled and conducted the interviews. Continuity with the subject was established by having the same interviewer conduct all four interviews. Physical activity questionnaires were administered at each of these interviews. Between the second and third interviews (T2 and T3), the 7-d activity diary was completed. Participants were thoroughly trained on the procedure for completing the diary. In addition to the training participants received at the screening session (T0), they were again given detailed instruction at the T2 interview on how to complete the diary. The diary completed by the participant for screening purposes was reviewed with the individual and used to further explain correct completion of the diary. Participants completed a second practice diary at the T2 interview by recording activities they performed in the previous 4 h (or the prior evening if the interview was held in the morning). The interviewer reviewed this diary with the individual as it was completed and answered any questions the individual had regarding completion of the diary. During the 7 d following, the interviewer contacted the participant to make sure she was completing the diary and to answer any questions she had. The diary was reviewed by the interviewer for completeness and clarity when it was returned at the third interview. Subjects who completed all study components received compensation, and those who partially completed the study received a prorated compensation.
Participants completed an eight-item demographic questionnaire providing their age, race, primary language, occupation, income, education, and health conditions that would prevent activity. This questionnaire was completed at the in-person screening.
The diary used for analysis in this study was a 7-d diary developed in both English and Spanish for the WOTM study. Instructions, a completed sample day, and practice pages were included in the front of the diary. Five blank pages were provided for each day of recording. Each day, subjects recorded the date, the time they got out of bed and the time they went to bed, and whether it was a workday or nonworkday. Subjects were instructed to record all activities lasting 10 min or longer. For each activity, subjects recorded the start and end time for the activity, a description of the activity, the type of activity (W for paid work; H for house and yard activities; C for child, family, and pet care; E for exercise, dance, and sports; V for volunteer and/or community work; T for transportation; P for personal care; Walk for walking; and O for other activity), the intensity or effort the activity required (1 = activity performed in a sitting, standing, or lying position; 2 = activity with intensity comparable to strolling or walking slow; 3 = activity like brisk walking; and 4 = activity more intense than brisk walking), and how much, if any, additional weight was lifted or carried during the activity. The diary used in this study was similar to those that have been published, except that participants had to record more information. Diaries have been shown to provide valid and reliable assessment of physical activity. These studies have validated the diary against doubly labeled water, dietary intake, and room calorimetry (1,4,7,8,12,16,17). The diary used in this study was validated against an accelerometer (Computer Science Application, model 7164, Shalimar, FL), and the correlation between the diary and the accelerometer was 0.41.
Two research staff independently coded each diary entry with an activity type code, metabolic equivalent (MET) code, and a MET value. The MET code and corresponding MET value were assigned to each entry using the Ainsworth et al. Compendium of Physical Activities (2). Activity type codes (paid work, exercise, household activities, yard work, self care, transportation, volunteer work, child and family care, and pet care) were identified and used for coding each entry. The activity type codes were usually coded according to the participants’ recorded activity type. One exception to this rule was walking activity, which was listed as an activity type on the diary and used by the participants as an activity type. Walking was not used as an activity type in coding because walking is performed as part of other activities and was coded into the activity type appropriate for the activity. For example, walking as part of work was coded as activity type work, walking the dog was coded as pet care, walking to the store was coded as transportation, and walking for exercise was coded as exercise. A second exception occurred when participants combined activities from two or more activity types in one entry and used a single activity type. Coders were instructed to split activities when the combined activities could be assigned different activity type codes and if the time appeared to be equally divided between the two activities. For example, eating breakfast and doing the dishes would be split into activity types personal care and household activities and half of the time would be given to each activity.
After both coders completed each diary, a third independent research assistant then reconciled any differences. Reconciling rules were established to maintain consistency. The reconciler generally reviewed all activity type codes and MET codes to ensure correctness and agreement. Additionally, all walking activities, all exercise activities, all activities with a MET value above 3, and all discrepancies between coders were examined.
For each subject, total activity kcal per day and kcal by activity type per day were used in the analyses. Standardized kcal (not weight dependent) were derived by multiplying the MET value from the compendium (2) times the minutes for each reported activity. Standardized kcal assume a standard weight of 60 kg. Total activity kcal were computed by summing all activities reported in the diary (not including sleep). Aggregated kcal were computed for the following activity types: paid work, exercise, household activities, yard work, self care, transportation, walking, volunteer work, child and family care, and pet care.
Because it has been shown that activity patterns vary between individuals who work and those who do not (5,9,11,15), the analysis was completed separately for workers and nonworkers. Number of work minutes per day was summed for a weekly total. Those who were employed 1800 min (30 h) or more per week were considered as working full time and were classified in this paper as workers. Those who were not employed and worked less than 1800 min were not considered to be full time workers and were classified as nonworkers. Although the nonworker group included a portion of women who had casual work hours, the majority of the women worked less than 10 h·wk−1 or not at all. Only paid work (in the home or outside the home) was included in the analysis for the activity type work. Volunteer work and household work (unpaid) were analyzed separately.
Forty-two single-facet, subject-by-day design generalizability studies (G-studies) were completed ([9 activity types and 1 total activity for nonworkers, 10 activity types and 1 total activity for workers]*2 for each ethnic group). Generalizability theory (G-theory) is a statistical theory about the dependability of behavioral measurements (18). It could be considered an extension of the intraclass correlation, the commonly used statistical procedure for assessing reliability or consistency in measurement over time (also known as classical test theory). Unlike classical test theory, which looks at the relative standing of individuals to each other and assumes that the systematic error variance associated with different days of activity measurement is equal, G-theory provides an estimate of the individual’s variation across time as well as multiple sources of error in measurement (18). In other words, G-theory not only provides an estimate of the individual’s variation across time, it can also indicate the error variance associated with conditions of measurement such as days of measurement, test forms, or administrators, and the measurement error due to unsystematic sources or random events, in a single analysis. Consequently, G-theory can be used to determine how many measurement occasions (or test forms or administrators) are needed to obtain dependable scores. G-theory provides a summary coefficient (G-coefficient) that is analogous to classical test theory’s intraclass correlation. Shavelson and Webb (18) define an acceptable level of generalizability as a G-coefficient of at least 0.80. Independent analysis was conducted for total activity and 9 of the 10 activity types for African-American and Hispanic workers and nonworkers (40 analyses), and the work activity type for African-American and Hispanic workers (2 analyses). Variance components for 40 of the single-facet designs were computed in the PROC VARCOMP statistical subroutine in SAS (14). Type I sums of squares were computed given that all designs were balanced. The PROC GLM statistical subroutine in SAS (14) was used for the two single-facet designs for the work activity type because of its unbalanced design (number of workdays varied by individual). In this case, Type III sums of squares were computed because the two designs were unbalanced. Variance components were calculated for subjects (ς2s = [mean squares for subject-mean squares for residual]/number of days), days (ς2d = [mean squares for days-mean squares for residual]/number of persons), and residual (ς2s x d,e = mean squares for residual). If a near zero negative variance arose, it was assumed to be caused by sampling error (a small sample being drawn from an indefinitely large universe). The approach taken in this analysis was to set the negative estimate to zero (18). Work was the only activity type analyzed as an unbalanced design because typically individuals have set days that they work, generally 5 d, but this may vary. The unbalanced design was used so that only days where work activity was recorded were used in the analysis. Activity in all of the other activity types could be done by choice on a day-to-day basis and could be done on any day or all days of the week so all 7 d were used in the analysis, creating a balanced design.
Percentage of the total variance for each component (subjects, days, and residual) was calculated by dividing each variance estimate by the sum of the variance component estimates. Absolute variance estimates and generalizability coefficients (G-coefficients) were then computed. The estimated absolute variance was calculated using Shavelson and Webb’s (18) formula ς2Abs = ς2d/nd + ς2s x d,e/nd, where ς2d is the day facet variance component, ς2s x d,e is the residual variance component, and nd is the number of days used for the investigation. The decision to use absolute values rather than relative values was made because each of the subject’s activities were independent of the others; in other words, the subject’s scores were not relative to one another. G-coefficients were calculated using Shavelson and Webb’s (18) formula GAbs = ς2s/(ς2s + ς2Abs), where the G-study subject variance component is divided by the sum of the subject variance component and the estimated absolute error variance calculated for the desired number of days.
The results of the G-study were then used for decision studies (D-studies). Estimated absolute variance and G-coefficients were calculated for the D-studies by using the same formulas as the G-study and varying the number of days under investigation.
A total of 227 women (111 African-American and 116 Hispanic) from the WOTM study (N = 260) had a complete 7-d diary and were used in the analysis for this study. The 33 participants excluded from this study because they did not complete a diary were not significantly different in age (P = 0.802), height (P = 0.561), weight (P = 0.719), BMI (P = 0.712), race (P = 0.352), income (P = 0.822), or education (P = 0.162) than the 227 women used for the analysis. Demographic characteristics of the study sample are presented in Table 1. A total of 105 (67 African-American and 38 Hispanic) worked 30 or more h·wk−1 and were considered workers, and 122 women (44 African-American and 78 Hispanic) were classified as nonworkers.
Estimated variance components, absolute error variance, and G-coefficients for activity kcal and kcal by activity type are shown in Table 2. For all items, 7 d represent the G-study. Of the 42 analyses, 4 had G-coefficients ≥ 0.80, 16 had G-coefficients < 0.80 but ≥ 0.70, 15 had G-coefficients < 0.70, and for 7 of the analyses, G-coefficient could not be computed because there was no interday variability. Exercise for African-American workers, household for African-American nonworkers, transportation for Hispanic workers, and pet care for Hispanic workers reached an acceptable level of generalizability with 7 d of reporting (G ≥ 0.80).
Decision studies use the results of a G-study to determine how the reliability would improve by varying the conditions of measurement. For this study, the number of days was varied in the D-study to determine how many diary-recording days would be required to obtain a G-coefficient of 0.80. Results of the D-studies are shown in Table 3. For the yard work activity type for African-American workers and nonworkers, the volunteer work activity type for African-American and Hispanic workers and African-American nonworkers, the child and family care activity type for African-American workers, and the pet care activity type for African-American nonworkers, only 1 d of activity was recorded for the majority of subjects reporting activity in these activity types. Therefore, with no interday variability, valid G-coefficients could not be calculated for these activity types.
The purpose of this study was to estimate variation in energy expenditure for African-American and Hispanic workers and nonworkers, and identify how many days of diary self-report are needed to reliably estimate total activity and activity by type using generalizability analyses. It was found that 8–14 d were needed to reliably estimate total activity. By activity type, exercise for African-American workers, household activities for African-American nonworkers, and transportation and pet care for Hispanic workers could be reliably estimated with 7 d reporting or less. For African-American workers, days of diary self-report required to reliably estimate actiity by type ranged from 6 to 30. For Hispanic workers, days of diary self-report required to reliably estimate activity by type ranged from 7 to 42. For African-American nonworkers, days of diary self-report required to reliably estimate activity by type ranged from 6 to 48, and for Hispanic nonworkers, days of diary self-report required to reliably estimate activity by type ranged from 8 to 111.
For this population, it appears that there is more consistency in daily activity for both African-American and Hispanic nonworkers compared with workers. Fewer reporting days of total activity were required for nonworkers compared with workers, and only three activity types (exercise, transportation, and walking) for African-Americans and four activity types (transportation, walking, pet care, and yard work) for Hispanics could be reliably estimated with fewer reporting days for workers compared with nonworkers. This may be due to the older female minority population. Perhaps there is more routinization of activity among older minority women who don’t work or the type of work that older minority women engage in may be inconsistent in the daily activity performed.
It appears that high intrasubject variability (inconsistent pattern of activity) and low interday variability (very few days the activity is performed) result in a very high number of reporting days to achieve reasonable reliability. The high number of reporting days indicates difficulty in measuring the behavior due to the inconsistency of the behavior.
This study may be limited in its findings due to the population of older minority women. The activities performed may not be representative of other populations. However, the study does provide important information for a population not well studied. A second limitation is the diary method of recording activity, which may not be representative of actual physical behavior. Even though thorough training was conducted with review by a trained interviewer, the accuracy of the data is relative to the participants reporting. However, the diary method still represents the most cost effective, efficient, and practical method of collecting a comprehensive picture of physical activity in free-living populations. The findings were also limited by the number of subjects who reported activity in each of the activity types and how many days they performed the activity. All participants did not report activity in all activity types on every day, which should be considered when evaluating the outcomes.
The results of this study indicate the need for more than 1 wk of diary self-report to achieve reliable estimates of activity. In this population, at least 8 d of diary self-report are needed to reliably estimate total activity for Hispanic nonworkers, 11 d for African-American nonworkers, 12 d of diary self-report are needed to reliably estimate total activity for African-American workers and 14 d for Hispanic workers. These findings support previous studies that report 8–14 d of diary self-report are required (3,6). Whereas previous studies have not looked at required reporting days by activity type, this study found that two activity types (exercise for African-American working subjects and household activity for African-American nonworking subjects) could be reliably estimated in less than 7 d. Other activity types for African-American nonworkers required up to 48 reporting days, for Hispanic nonworkers up to 111 d, for African-American workers up to 30 reporting days, and for Hispanic workers up to 42 d. Even though measurement of activity by activity type provides more comprehensive information about a person’s activity patterns, future studies should consider the feasibility of this method, expecting that a low frequency of activity in the activity type, either in the number of subjects reporting activity in the activity type and/or the number of days the activity is performed, will result in a high number of reporting days required to achieve a reasonable level of reliability. It appears that measurement by activity type may be useful when looking at particular types of activity in specific populations, as observed in this study for the work activity for workers and the household activity for nonworkers. Being aware of the frequency of activity in specific activity types for a study population and being selective in the activity type measured will make this a useful and more comprehensive method of analysis compared with looking only at total activity.
This work was supported by “Behavioral Science Education Cancer Prevention and Control”, National Cancer Institute/NIH Grant no. 2R25CA57712-06.
Data for this study was collected for the Women On The Move study funded by the Women’s Health Initiative through the Centers for Disease Control and Prevention (CDC U48/CC609653).
Address for correspondence: Karen E. Eason, Dr.P.H., University of Texas-Houston, Center for Health Promotion and Prevention Research, 7000 Fannin St. #2624C, Houston, TX 77030; E-mail: firstname.lastname@example.org.
1. Acheson, K. J., I. T. Campell, O. G. Edholm, D. S. Miller, and M. J. Stock. The measurement of daily energy expenditure: an evaluation of some techniques. Am. J. Clin. Nutr 33: 1155–1164, 1980.
2. Ainsworth, B. E., W. L. Haskell, M. C. Whitt, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med. Sci. Sports Exerc 32 (Suppl. 9): S498–S516, 2000.
3. Baranowski, T., M. Smith, W. O. Thompson, J. Baranowski, D. Hebert, and C. deMoor. Intraindividual variability and reliability in a 7-day exercise record. Med. Sci. Sports Exerc 31: 1616–1622, 1999.
4. Borel, M. J., R. E. Riley, and J. T. Snook. Estimation of energy expenditure and maintenance energy requirements of college-age men and women. Am. J. Clin. Nutr 40: 1264–1272, 1984.
5. Clark, A., and K. Haag. Exercise participation among working women. N. Z. J. Health Phys. Educ. Recreat. 21: 5–8, 1988.
6. Coleman, K. J., and L. H. Epstein. Application of generalizability theory to measurement of activity in males who are not regularly active: a preliminary report. Res. Q. Exerc. Sport 69: 58–63, 1998.
7. Edholm, O. C., and J. G. Fletcher. The energy expenditure and food intake of individual men. Br. J. Nutr. 9: 286–300, 1955.
8. Geissler, C. A., T. M. O. Dzumbira, and M. I. Noor. Validation of a field technique for the measurement of energy expenditure: factorial method versus continuous respirometry. Am. J. Clin. Nutr. 44: 596–602, 1986.
9. Green, E., S. Hebron, and D. Woodward. Work and leisure. In: Women’s Leisure, What Leisure? London: Macmillan, 1990, pp. 85–112.
10. Gretebeck, R. J., and H. J. Montoye. Variability of some objective measures of physical activity. Med. Sci. Sports Exerc 24: 1167–1172, 1992.
11. Harrington, M., and D. Dawson. Who has it best? Women’s labor force participation, perceptions of leisure and constraints to enjoyment of leisure. J. Leisure Res. 27: 4–24, 1995.
12. Kalkwarf, H. J., J. D. Haas, A. Z. Belko, R. C. Roach, and D. A. Roe. Accuracy of heart-rate monitoring and activity diaries for estimating energy expenditure. Am. J. Clin. Nutr. 49: 37–43, 1989.
13. Laporte, R. E., H. J. Montoye, and C. J. Caspersen. Assessment of physical activity in epidemiologic research: problems and prospects. Public Health Rep. 100: 131–146, 1985.
14. Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. SAS®
System for Mixed Models. Cary, NC: SAS Institute Inc, 1996, pp. 18–23, 140.
15. Nelson, M. A. Health practices and role involvement among low-income working women. Health Care Women Int. 18: 195–205, 1997.
16. Riumallo, J. A., D. Schoeller, G. Barrera, V. Gattas, and R. Vauy. Energy expenditure in underweight free-living adults: impact of energy supplementation as determined by doubly labeled water and indirect calorimetry. Am. J. Clin. Nutr. 49: 239–246, 1989.
17. 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.
18. Shavelson, R. J., and N. M. Webb. Generalizability Theory: A Primer. Newbury Park, CA: Sage, 1991, pp. 1–26, 83–115.
19. Trost, S. G., R. R. Pate, P. S. Freedson, J. F. Sallis, and W. C. Taylor. Using objective physical activity measures with youth: how many days of monitoring are needed? Med. Sci. Sports Exerc. 32: 426–431, 2000.
20. U.S. Department of Health and Human Services (1996). Physical activity and health: a report of the Surgeon General (Publication No. 017-023-00196-5). Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996, pp. 81–172.