Validity and Reliability of Short Physical Activity History: Cardia and the Minnesota Heart Health Program

Abstract

Validity and reliability of a short physical activity history were assessed in two studies. Validity was studied in 2766 women and 2303 men, participants in CARDIA, a biracial study. Ages ranged from 18 to 30 years. The activities performed in the past 12 months by ≥ 50 percent of participants were walking/hiking, nonstrenuous sports, shoveling/lifting during leisure, running/jogging and home maintenance/gardening. Validity was indirectly assessed by studying the relationships of total activity to skinfold thickness, total caloric intake, duration on a self-limited maximal exercise test, and high density lipoprotein cholesterol. Less than perfect correlation are expected since physical activity is not the only factor affecting the validation criteria and since physical activity patterns change over time within each person. Comparing the highest physical activity quartile to the lowest physical activity quartile, mean level of sum of three skinfolds was 10.7 mm less for women (correlation coefficient (r) = −0.15, P < 0.001) and 6.9 mm less for men (r = −0.12, P < 0.001); mean level of caloric intake was 158 kcal morefor women (r = 0.07, P < 0.001) and 875 kcal morefor men (r = 0.21, P < 0.001); mean level of duration on treadmill was 132 seconds more for women (r = 0.36, P < 0.001) and 95 seconds more for women (r = 0.25, P < 0.001); and mean level of high density lipoprotein cholesterol was 4.8 mg/dL more for women (r = 0.13, P < 0.001) and 3.2 mg/dL more for men (r = 0.11, P < 0.001). Reliability was studied in a separate population by comparing questionnaire results in an initial telephone administration with results obtained two weeks later (N = 129). Similar types and amounts of activity were reported in this group as in the group studied for validity. Test-retest correlation coefficients for three summary scores ranged from 0.77 to 0.84, and were at least 0.57 for each of the 13 activity groupings queried. This questionnaire typically takes 5–10 minutes to administer. It yields moderately detailed information about type and amount of usual leisure time physical activity.

INTRODUCTION

There is an extensive history of study of physical activity and its role in the development of coronary heart and other diseases.¹ Blackburn and Jacobs² have observed that of the 43 studies reviewed by Powell et al.,¹ there were 19 in which physical activity was self-reported by individuals. In all 19 of those, either leisure or work physical activity was inversely related to either coronary or all causes death. The observed relationship between physical activity and disease outcome was more varied in the remaining 24 studies in which physical activity was not self-reported. To this list may be added two recent reports^3,4 which show an inverse relation of leisure physical activity to coronary heart disease and all causes death. One recent study⁵ found no relation between “regular strenuous exercise or hard physical labor” and coronary heart disease incidence; however it was based on only a single question about physical activity. Thus, accurate measurement of physical activity is a continuing concern. Methodological concerns related to the measurement of physical activity have been reviewed.^6,7

Although a number of brief questionnaires for self reporting of physical activity have been formulated, none provide the combination of high reliability with information about frequency of participation in a broad range of specific activities. In order to have a brief and reliable questionnaire which asked in some detail about typical rather than recent activity, the Physical Activity History (PAH) was developed for use in the study titled Coronary Artery Risk Development in Young Adults (CARDIA). The development of the PAH was built on the experience of previous investigators studying the measurement of physical activity. It yields scores related to energy expenditure during the past year for total activity as well as for heavy and moderate intensity activities. It also provides information about frequency of participation in specific categories of activity.

The most direct validation of a physical activity questionnaire would be long-term direct observation of usual behaviors. Such observation is logistically difficult to obtain, and is not available here. The PAH is considered to meet an indirect measure of validity if known effects of physical activity are seen in independent measures of caloric intake, fatness, fitness and high density lipoprotein cholesterol (HDL-C). Relationships are expected between a true measure of physical activity and the indirect validation criteria, but the correlation is expected to be less than perfect because other factors besides physical activity contribute to the level of each variable. In addition, the pattern of physical activity changes across time within person, and the resulting within person variation in the physical activity measure attenuates the correlation coefficients with the criteria variables.

No reliability study for the PAH was available in the CARDIA population. A separate reliability study was performed under the auspices of the Minnesota Heart Health Program.

Validity of the PAH is indirectly assessed here in approximately 5,000 participants of CARDIA, a biracial group of men and women, aged 18–30 years. Reliability of the PAH is presented, based on telephone administration of the questionnaire initially and 2 weeks later to 129 men and women aged 18–74 years in a separate study.

MATERIALS AND METHODS

The Physical Activity History was designed by two of the authors (DJ and WH), and formatted for telephone use by another author (PP). Its design was based on concepts and findings from two other questionnaires. First, the Stanford Heart Disease Prevention Program asked 11 questions about usual participation in six moderate activities and frequent participation in five vigorous activities during the previous 3 months⁸. These indices (number of positive answers for moderate and for vigorous activities) had test-retest Pearson correlation coefficients of 0.75 and 0.83, respectively.⁸ We thought specific querying of well defined patterns of behavior, such as the Stanford item “jog or run at least 10 miles per week,” would be useful in a new questionnaire.

Secondly, the specific types of physical activities queried were constructed on the basis of results from the Minnesota Leisure Time Physical Activity Questionnaire⁹ administered to 3615 men and women aged 25–74 years as part of the baseline assessment of the Minnesota Heart Health Program.¹⁰ Grouping of activities was based on similarities in the intensity of the activity. In addition, one question was included about vigorous activity on the job. The categories of activity represented in the PAH include the spectrum of activities found in this Midwestern population.

Folsom et al.¹¹ observed of the Minnesota Leisure Time Physical Activity Questionnaire that “the largest components of unreliability were inaccurate recall of the number of months, the times per months and/or the time per occasion that an activity was performed.” In light of these observations, we theorized that reliability would increase if the participant had to recall less detail. We attempted to achieve this by incorporating in the questions themselves the specific aspects of greatest interest to us: type, frequency, and duration of activity. Type of activity was queried by grouping activities with similar characteristics. The total number of months of performance per year was asked rather than the specific months in which the activity was performed. Duration of activity was approached by asking how many months each class of activity had been performed for at least one hour during the month, and also how many months the activity had been performed for at least a specific number of hours per week during the month. This specific number (which separated the less frequent from the more frequent participants) ranged from 2 to 5 hours per week and depended on what was considered by us to be substantial participation in each activity category (see Appendix). For instance, we considered four 30-minute jogging sessions per week to be frequent, thus frequent jogging or running was designated to be 2 or more hours per week. No attempt was made to determine the actual amount of time spent each month performing each specific activity.

The PAH was pretested for appropriateness of wording, timing, and ranking against known activity patterns in 42 pretest participants. The questionnaire (Appendix) is typically completed in 5–10 minutes.

Scoring of PAH

Scores related to energy expenditure are computed from intensity levels and the months of less frequent and more frequent performance of each activity (see Table I).

TABLE I.

INTENSITY LEVELS FOR ACTIVITIES IN THE PHYSICAL ACTIVITY HISTORY AND ILLUSTRATION OF SCORING ALGORITHM

		Intensity (mets/minute)†	Cutpoint for frequent participation (hours/week)
Heavy Intensity Activities
A. Jog or run		8	2
B. Vigorous racket sports		8	3
C. Bicycle faster than 10 miles per hour		6	2
D. Swimming		6	2
E. Vigorous exercise class or vigorous dancing		6	3
F. Non-job activity such as shoveling, weight-lifting and moving heavy objects		6	3
G. Vigorous job activity such as lifting, carrying or digging		5	5
H. Other strenuous sports such as basketball, football, skating, skiing		8	3
Moderate Intensity Activities
I. Other non-strenuous sports such as softball, shooting baskets, volleyball, ping-pong		4	3
J. Take walks or hikes or walk to work		4	4
K. Bowling or golf		3	3
L. Home exercises, calisthenics		4	3
M. Home maintenance and gardening, including carpentry, painting, raking, mowing		4	5
Activities of a hypothetical individual
Weight lifting	<3 hours/week for 2 months and ≥3 hours/week for 1 month
Golf	<3 hours/week for 3 months
Take walks	<4 hours/week for 6 months and ≥4 hours/week for 6 months
Score = Sum over (moderate, heavy or all) activities of: intensity × (number of months of infrequent activity +3 × number of months of frequent activity)
Moderate score: 4 × (6 + 3 × 6) for walks + 3 × (3 + 0) for golf = 105
Heavy score: 6 × (2 + 3 × 1) for weight lifting = 30
Total score: 105 + 30 = 135 Exercise Units

^†Approximately equal to kilocalories for a 70 kg man.

The intensity level assigned to each activity is roughly the number of kilocalories expended in one minute of activity. Intensities were patterned after those used by Taylor et al.⁹ and Folsom et al.¹¹ which in turn are based on formulation by Reiff et al.¹² Exceptions are bicycling, which was set at an intensity code of 6 to reflect the wording “faster than 10 miles per hour”; and vigorous job activity (not included in Taylor’s or Folsom’s work), which was set at an intensity code of 5, somewhat less than that for weight lifting.

Amount of activity performed is based on months of participation and an assumption about relative duration of activity for frequent versus infrequent participators. The duration of activity in more frequent participation months (in which the activity was performed more than the activity-specific cutpoint of 2–5 hours per week) is assumed somewhat arbitrarily to be three times as great as the duration of activity in less frequent participation months. The sum of the products of Intensity times Amount is computed for each activity class A through M (“running/jogging” through “home maintenance/gardening”), i.e., bi × (mi + 3 × ni), where i ranges through the activity classes, bi is the intensity, mi is the number of months of less frequent participation and ni is the number of months of more frequent participation. The Heavy Intensity Score sums these computed values over the class of 8 vigorous intensity activities (A–H), the Moderate Intensity Score is restricted to the class of the 5 remaining moderate intensity activities, and the Total Score is the sum of these two subscores. Since the number of minutes of performance in any month is unknown and likely to vary substantially across activities, no interpretation as caloric expenditure per week is available, and the score is left unaltered in “Exercise Units.”

Validity study: CARDIA Description and Data Collection

The inital observation of 5115 CARDIA participants occurred between March 1985 and June 1986 in Minneapolis, Minnesota, Oakland, California, Birmingham, Alabama, and Chicago, Illinois. Recruitment was done primarily by telephone. The telephone interview was followed by an extensive clinic visit. Participants were aged 18–30 years, with 45 percent aged 18–24, 52 percent black, 54 percent female and 40 percent having completed ≤ 12 years of education.

The CARDIA examination included a variety of questionnaires and physical measurements.¹³ The PAH was administered by interviewer. Age, sex, education and race (black or white) were self-reported. Triceps, subscapular and suprailiac skinfolds were measured in duplicate. The maximum value which could be accurately measured by Harpenden skinfold calipers was 50 mm. Values recorded larger than 50 mm and missing values in all 122 participants who were too obese for measurement were set to 50 mm. The sums of the three different skinfold values were used in these analyses. Venous blood was drawn in a reclining position with 97% of participants fasting at least 8 hours to measure HDL-C, other lipids and other blood chemistries. HDL-C was determined from frozen plasma, after precipitation by heparin manganese, using an enzymatic method in the North-west Lipid Research Clinics laboratory at the University of Washington.^14–16 Usual dietary caloric intake over the past month was obtained from the specially designed CARDIA quantitative food frequency questionnaire. Very large values for estimated caloric intake (≥10,000 calories per day) were encountered in 23 individuals who were deleted from these analyses. Weight was measured to the nearest 0.2 pounds with participants wearing only a light shirt, shorts and socks. Caloric intake per kilogram of body weight was computed. The measure of fitness was duration until self-limited maximum exertion on a treadmill exercise test with 9 two-minute stages (speed and grade 3.0 miles per hour, 2 percent grade; 3.4, 6 percent; 3.4, 10 percent; 3.4, 14 percent; 3.4, 18 percent; 3.4, 22 percent; 4.2, 22 percent; 4.2, 25 percent; 5.6, 25 percent) and 3 1-minute recovery periods (the first two walking 2.0 miles per hour at 0 percent grade and the third sitting). Due to a variety of minor medical problems, 85 people did not do the treadmill test, and were excluded from the analysis. After further exclusion for missing data, ca. 5000 men and women enter these analyses.

Statistical Methods for Validity Study

The means and standard deviations of Exercise Units spent in Total activity, Heavy Intensity activity, Moderate Intensity activity, and in each of the 13 separate activities (A–M) were computed for CARDIA participants. Percent of participation for the summary activities and each specific activity were also computed. The means and standard deviations of four objective physiologic measures, sum of skinfolds, caloric intake, exercise test duration, and HDL concentration were also computed for each quartile of total, heavy, and moderate activity for men and women as indirect measures of validity. Pearson correlation coefficients were computed between the three activity measures (in continuous form) and the four validation measures and between the three activity measures for males and females. Linear regression of the four validation measures on heavy/moderate and total activity were also computed to assess further the statistical significance of these relationships, to adjust the relationships for the sociodemographic factors age, race, sex and education, and to test for sex interactions.

Reliability study: Telephone Survey

The PAH was administered in a study separate from CARDIA during the fall of 1984 to men and women aged 18–74 years twice over the telephone, generally by different interviewers, with a 2-week interval between calls (±2 days). For purposes of this study, reliability was defined by the extent of agreement between these two interviews. One participant per household was selected at random from randomly selected households with listed telephone numbers in several suburban areas around Minneapolis, Minnesota. Response rate was 82 percent for the initial phone call, with 153 interviews completed. Of the 153 initial interviews, two had incomplete information at retest while 129 were reinterviewed with complete data generating an 86 percent response rate at retest. Of the remaining 22 participants, five refused the reinterview, one had no phone at the time of reinterview, and 16 could not be reached within the narrow time window. The interviewed group was 46 percent male; seven percent ages 18–34 years, 54 percent ages 35–54, 39 percent ages 55–74; 80 percent employed, 19 percent homemaker, retired/disabled or student, and 1 percent unemployed; 32 percent had professional/managerial occupation, 50 percent sales/clerical and 18 percent blue collar; 37 percent were graduates of high school or had less education, while 13 percent had a degree beyond college; 69 percent were married, 15 percent separated or divorced and 16 percent widowed or never married; 67 percent had children living at home.

Statistical Methods for Reliability Study

The means and standard deviations of number of Exercise Units were computed for each class of activity, for the Heavy and Moderate Intensity Subscores, and for the Total Score. Differences between time 1 and time 2 were assessed using paired t-tests, and within person variability was assessed by test-retest Pearson correlation coefficient, r, and by the ratio of the within person to the between person variances, computed as (1-r)/r.¹⁷ Percent of participation and percent agreement of level of participation (to within 1 month of stated number) was computed for each activity; this was done for both total months of activity and for months of more frequent participation. Reliability of the Total Score was also assessed by considering at time 2 the stability of classification by time 1 quintiles. Variations in reliability across categories of the demographic variables listed above were assessed by multiple linear regression, where dependent variables were differences and differences squared of the Moderate Intensity, Heavy Intensity and Total Scores (six regressions), and independent variables were the demographic characteristics discussed above, simultaneously entered as indicator variables for each category. Specifically, regression of the differences between time 1 and time 2 mean levels of Exercise Units tests differences in reporting bias among demographic categories, while regression of the squared differences between time 1 and time 2 mean levels of Exercise Units tests differences in total variability (test-retest variance plus bias squared) among demographic categories.

RESULTS

CARDIA: Extent of Participation

Table II provides mean and standard deviation of Exercise Units spent in each of the 13 separate activities and in the three summary activities, and gives the percent participation in each activity class for all CARDIA participants. The mean (± S.D.) total activity score in Exercise Units was 522 for men and 335 for women (Table II). Exercise Units expended in heavy intensity activity were higher than in moderate intensity activity. For men, strenuous sports, running/jogging, lifting/carrying/digging on the job, walking/hiking, and shoveling/lifting during leisure contributed the most Exercise Units to total activity. For women, walking/hiking, running/jogging, and exercise/dance classes contributed the most Exercise Units to total activity. Total participation ranged from 80 percent for walking/hiking or nonstrenuous sports (women) to 19 percent for racket sports (women). Sex-specific quartile cutpoints of the total activity score are provided in Table III.

TABLE II.

PARTICIPATION IN VARIOUS ACTIVITIES AMONG 5069 CARDIA SUBJECTS PHYSICAL ACTIVITY HISTORY

	Men (N = 2303)			Women (N = 2766)
Activity	Score* ± SD	Percent Participation		Score* ± SD	Percent Participation
		Any	Frequent		Any	Frequent
A. Running/Jogging	71 ± 91	63	51	43 ± 73	48	36
B. Racquet Sports	24 ± 55	32	21	10 ± 32	19	11
C. Biking	34 ± 53	51	41	26 ± 46	44	35
D. Swimming	15 ± 34	34	23	15 ± 36	30	22
E. Exercise/Dance class	31 ± 60	35	27	47 ± 64	60	47
F. Job lifting/carrying/digging	59 ± 78	56	47	31 ± 64	31	24
G. Shoveling/lifting during leisure	48 ± 59	71	56	21 ± 40	47	31
H. Strenuous sports	78 ± 98	66	54	17 ± 47	29	19
I. Nonstrenuous sports	35 ± 46	69	55	22 ± 34	80	62
J. Walking/hiking	53 ± 56	72	56	57 ± 54	80	62
K. Golfing/bowling	10 ± 21	37	23	5 ± 16	26	13
L. Home exercises/calisthenics	28 ± 40	62	41	14 ± 29	40	23
M. Home maintenence/gardening	28 ± 40	62	41	14 ± 29	40	23
Heavy intensity score	359 ± 251	—	—	211 ± 192	—	—
Moderate intensity score	163 ± 116	—	—	125 ± 98	—	—
Total score	522 ± 324	—	—	335 ± 250	—	—

^*Exercise Units. Since minutes per month of activity is unknown, direct interpretation as caloric expenditure is not available. For participation, Any = at least 1 hour duration in at least 1 month, while Frequent = at least 2–5 hours duration per week (depending on activity) in at least 1 month.

TABLE III.

MEAN LEVELS OF FOUR OBJECTIVE MEASURES BY QUARTILE OF TOTAL PHYSICAL ACTIVITY EXPENDITURE AS DETERMINED FROM THE CARDIA PHYSICAL ACTIVITY HISTORY

Total Activity Exercise Units	Sum of Skinfolds (mm)			Caloric Intake (kcal)			Duration on Treadmill (seconds)			High density lipoprotein Cholesterol (mg/dl)
	N	Mean	S.D.	N	Mean	S.D.	N	Mean	S.D.	N	Mean	S.D.
Females:
≤ 149 (Low)	687	64.9	33.0	689	2307	1097	671	432	117	686	53.6	12.4
150–284	691	62.2	31.6	693	2349	1106	670	471	121	689	55.1	12.9
285–467	689	57.7	27.7	688	2352	1162	675	501	124	679	55.6	12.5
≥ 468 (High)	687	54.2	27.7	690	2465	1262	662	564	144	684	58.4	13.5
Twice the S.D. of the difference*		3.2			125			14			1.4
Correlation with total physical activity		−0.15			0.07			0.36			0.13
Males:
≥ 278 (Low)	575	47.6	26.1	574	3131	1390	555	647	137	573	49.1	13.4
279–468	577	44.8	23.0	575	3353	1492	568	686	131	577	49.7	13.2
469–699	569	41.7	20.5	565	3576	1451	557	721	133	565	49.7	12.6
≥ 700 (High)	572	40.7	20.6	552	4006	1779	566	742	140	567	52.3	12.4
Twice the S.D. of the difference*		2.7			185			16			1.5
Correlation with total physical activity		−0.12			0.21			0.25			0.11

SD = Standard deviation.

^*Approximately 2 standard deviations of the difference in mean objective measure between any two quartiles. (The difference is computed as 2 × pooled SD × √ (2/n_min), where n_min is the minimum sample size among the four quartiles.) If the difference between 2 quartile means is greater than or equal to this number, the test of hypothesis that the two means are different carries P < 0.05.

Validity

Linear regressions of the four validation measures on total activity and on heavy/moderate intensity activity were all significant when adjusted for age, race, education, and sex as seen in Table IV. The relationships of the validation measures to total activity differed somewhat by sex, as reflected in the significant sex by total activity interactions. All relationships with total activity remained significant within each sex. Further analyses (data not shown) indicate that these relationships are not different for blacks vs whites, except that the increase in mean caloric intake across total activity quartiles is less in whites than in blacks.

TABLE IV.

LINEAR REGRESSIONS OF FOUR OBJECTIVE MEASURES ON TOTAL PHYSICAL ACTIVITY EXPENDITURE AS DETERMINED FROM THE CARDIA PHYSICAL ACTIVITY HISTORY

	Sum of Skinfolds (mm) N = 5047		Caloric Intake (kcal) N = 5026		Duration on Treadmill (seconds) N = 4924		High Density Lipoprotein Cholesterol (mg/dl) N = 5020
	b	t*	b	T	b	t	b	t
Adjusted for age, race, sex, education	−0.012	−8.8	0.876	13.5	0.125	20.2	0.006	9.4
Is there a sex interaction? (t-value)	3.2		3.0		−2.8		−3.2
Sex specific regressions adjusted for age, race, education
Females	−0.012	−5.6	0.633	7.2	0.135	14.5	0.006	6.4
Males	−0.008	−5.5	1.051	10.8	0.110	12.9	0.004	5.1

^*Degrees of freedom are approximately 5,000; t = 1.96 corresponds to a two-sided significance level of 0.05; t = 2.58 corresponds to a two-sided significance level of 0.01.

Goodness of fit of these regressions, and an increased understanding of the pattern of the relationship between the physical activity measure and the four validation measures is gained by examining Table III, where cross-tabulations of the means and standard deviations of the four validation measures by total activity quartile and overall correlations with total activity are shown. In both males and females, for each successive quartile of total activity, means either increased or decreased monotonically. Specifically, the estimated mean number of calories increased at each successive level of total activity as did mean exercise test duration and HDL-C. As seen in Table IV, correlation coefficients between total activity and the four objective validation measures were all positive except for the negative value reflecting the inverse relationship between sum of skinfolds and total activity. Treadmill duration was the measure most correlated to total activity in both males and females. Caloric intake per kg of body weight and caloric intake itself were highly correlated (r > 0.9 in both sexes), and the two representations of caloric intake had virtually the same relationship to physical activity. For the most part, the mean level of each validation variable in one physical activity quartile differs significantly from the mean level in the adjacent physical activity quartile.

The correlation coefficients in CARDIA participants between total and heavy intensity activity (not shown) were greater than 0.9 for both men and women while between total and moderate activity they were greater than 0.7 in both sexes. The correlations between heavy and moderate intensity activity were 0.49 for men and 0.43 for women. Due to the high correlation between total activity and heavy intensity activity, the relationships between the validation measures and total activity may be expected to be the same as those for heavy intensity activity, but to be somewhat different for moderate activity. Statements about the relationships of the validation measures to moderate activity should take into account its correlation with heavy intensity activity. Additional regressions were done replacing total activity as an independent variable by heavy and moderate intensity activity simultaneously. Heavy intensity activity was related significantly to all four validation measures, holding moderate intensity activity constant. Moderate intensity activity was related only to caloric intake, holding heavy intensity activity constant.

Telephone Survey: Extent of Participation

Table V gives mean and standard deviation of Exercise Units spent in Total activity, Heavy Intensity activity, Moderate Intensity activity and in each of the 13 separate activities (A–M) at time 1 and time 2. The greatest contributions to the mean 424 Exercise Units spent in total activity arise from vigorous job lifting/carrying/digging, shoveling/lifting during leisure, walking/hiking, running/jogging, and home maintenance/gardening. Three of these activities were also the highest contributors to Exercise Units in the CARDIA participants. The range of participation is from 84 percent for shoveling/lifting during leisure, walking/hiking and home maintenance/gardening to 32 percent for racket sports and exercise/dance class, which is comparable to the range of activities for total participation in CARDIA. The number of Exercise Units in each activity and in the three summary scores was similar in the two groups of participants. The younger age distribution in CARDIA was reflected in the higher number of Exercise Units in the more vigorous and “youthful” activities, such as running/jogging, biking, exercise/dance class, and strenuous and nonstrenuous sports. Likewise, the reliability study participants had a higher number of Exercise Units in less strenuous activities such as shoveling/lifting during leisure, golfing/bowling, and home maintenance/gardening. More frequent participation (>2–5 hours duration per week in at least 1 month) ranges from 66 percent for home maintenance/gardening to 20 percent for racket sports. Table VII shows quartile cutpoints of the Total Score at time 1. Thus 25 percent scored less than or equal to 216 Exercise Units and 75 percent scored less than or equal to 576 Exercise Units.

TABLE V.

TWO-WEEK TEST-RETEST RELIABILITY OF INDIVIDUAL PHYSICAL ACTIVITY HISTORY QUESTIONS, MINNESOTA HEART HEALTH PROGRAM TELEPHONE SURVEY (N = 129)

Activity (N = 129)	Time 1 (Mean* ± SD	Time 2 (Mean ± SD)	Difference Mean ± SD	Test-Retest Correlation (r)	Variance ratio (%)(VR†)
A. Running/jogging	46 ± 90	49 ± 91	− 2.4 ± 4.6	0.82	22
B. Racket sports	25 ± 62	24 ± 59	1.8 ± 2.6	0.88	14
C. Biking	27 ± 52	25 ± 49	1.5 ± 3.5	0.73	37
D. Swimming	19 ± 41	24 ± 44	− 4.7 ± 2.4	0.79	27
E. Exercise/dance class	18 ± 47	23 ± 51	− 4.9 ± 3.8	0.70	43
F. Job lifting/carrying/digging	66 ± 91	68 ± 86	− 1.8 ± 5.8	0.72	39
G. Shoveling/lifting during leisure	51 ± 62	49 ± 60	1.6 ± 4.9	0.58	72
H. Strenuous sports	24 ± 62	26 ± 58	− 2.0 ± 3.6	0.75	33
I. Nonstrenuous sports	21 ± 33	16 ± 28	5.3 ± 2.5	0.64	56
J. Walking/hiking	47 ± 51	44 ± 48	2.1 ± 3.2	0.74	35
K. Golfing/bowling	9 ± 18	9 ± 19	− 0.2 ± 1.0	0.81	23
L. Home exercises/calisthenics	26 ± 42	27 ± 44	− 0.9 ± 3.0	0.69	45
M. Home maintenance/gardening	45 ± 43	41 ± 45	3.2 ± 3.6	0.57	75
Heavy intensity score	276 ± 234	287 ± 231	− 10.9 ± 13.1	0.79	27
Moderate intensity score	147 ± 105	138 ± 104	9.5 ± 6.6	0.77	30
Total score	424 ± 289	425 ± 291	− 1.4 ± 14.5	0.84	19

^*Exercise Units. Since minutes per month of activity is unknown, direct interpretation as caloric expenditure is not available.

^†100 × within person variance/between person variance, computed as (1-r)/r.

TABLE VII.

DISTRIBUTION OF SUBJECTS BY QUARTILE, TEST (TIME 1) VS RETEST (TIME 2) IN THE POPULATION-BASED SAMPLE (N = 129)

Time 2	Time 1				Number of Persons in Row
	≤216* (%)	217–363 (%)	364–576 (%)	≥577 (%)
≤ 216	72	16	3	0	29
217–363	19	66	22	3	35
364–576	6	13	53	27	32
≥ 577	3	6	22	70	32
Number of persons in column	32	32	32	33	129

^*Cutpoints based on time 1 quartiles of total activity Exercise Units.

Reliability

The mean and standard error of the difference in Exercise Units between time 1 and time 2 is listed in Table V. Only nonstrenuous sports has a statistically significant mean difference between the two administrations. Overall test-retest correlation is high, 0.77 to 0.84 for the three scores. For individual activities test-retest correlation ranges from 0.57 to 0.88, with a median of 0.73. Activities with test-retest correlation less than 0.7 are shoveling/lifting during leisure, nonstrenuous sports, home exercises/calisthenics and home maintenance/gardening. The ratio of within person variance to between person variance is about 25 percent for the three summary scores, higher for many of the individual activities.

Reliability of months of participation is presented in Table VI. Agreement of reported total activity as presented in the second column of the table occurs when the number of months reported at time 2 and the number of months reported at time 1 agree exactly or differ by one month. Such agreement occurs in at least 49 percent of cases per activity, and is over 72 percent for most activities, except vigorous job lifting/carrying/digging. shoveling/lifting during leisure, nonstrenuous sports, walking/hiking, home exercises/calisthenics and home maintenance/gardening. Agreement to within 1 month of frequent participation exceeds agreement to within 1 month of total participation in every activity except strenuous sports and home maintenance/gardening. Agreement to within 1 month of participation, computed only among participators (not shown), ranged from 39 to 70 percent for total months of participation and from 33 to 60 percent for more frequent participation.

TABLE VI.

RELIABILITY OF PERCENT REPORTED TOTAL PARTICIPATION AND FREQUENT PARTICIPATION N = 129

Activity	Any		Frequent
	Percent* Participation	Reliability of months of Participation	Percent Participation	Reliability of months of Participation
A. Running/jogging	41	84	32	87
B. Racket sports	32	84	20	90
C. Biking	45	78	33	86
D. Swimming	49	80	35	83
E. Exercise/dance class	32	85	22	88
F. Job lifting/carrying/digging	65	67	53	69
G. Shoveling/lifting during leisure	84	59	61	66
H. Strenuous sports	40	88	27	86
I. Nonstrenuous sports	59	65	41	75
J. Walking/hiking	84	49	57	65
K. Golfing/bowling	36	87	25	89
L. Home exercises/calisthenics	54	72	40	75
M. Home maintenance/gardening	84	57	66	56

^*100 = The percent who reported participation at either occasion.

Agreement occurs when the number of months reported at time 2 and the number of months reported at time 1 agree or differ by 1 month. At most, five persons in any activity reported no participation at time 1 and 1 month of participation at time 2; these individuals are counted under agreement. Nonparticipation is counted as 0 months. For participation, Any = at least 1 hour duration in at least 1 month, while Frequent = at least 2–5 hours duration per week (depending on activity) in at least 1 month.

Table VII shows reasonably close agreement of the Total Score at time 1 and time 2 in quartiles for which cutpoints are derived from time 1. The dependence of test-retest reliability on the demographic factors was studied by regressing the difference in scores, (time 1–time 2), and the squares of these differences, on the demographic factors sex, age, employment status, educational attainment, marital status and presence of children in the home. No significant dependencies were identified. There were no significant differences in reporting bias or in total variability among demographic classes (analyses are not shown).

DISCUSSION

Though many questionnaires exist to measure physical activity, the CARDIA investigators in 1984 found none which was brief and reliable, and at the same time comprehensive. Consequently, the CARDIA investigators developed the PAH. To put this new questionnaire in perspective, we briefly review general directions in the development of physical activity questionnaires, as investigators built on others’ experience.

Initial efforts in the measurement of physical activity assigned typical energy expenditure to occupational job titles.¹⁸ This method provided a relatively coarse measure of physical activity which misses important differences between people in leisure time activity, a problem of greater magnitude today. Because of a need for more accuracy, Montoye and his colleagues developed a reliable, but lengthy method in which the participant was required to list activities performed, then be interviewed in a partly open-ended format to elicit details of performance.¹⁹ A few years later Morris and his collegues developed an intensive diary method, whereby respondents recalled details of activity performed in 5-minute blocks throughout 2 days.^20–22 Montoye assumed that there was a typical pattern of activities performed over a long time period, while Morris opted for increased accuracy by asking participants to recall details of recent activity. Using the Montoye questionnaire as a model, Taylor chose to obtain details of activity performed over a long time period, but in a simple, systematic and standard format.⁹

Following these efforts, a number of simpler questionnaires were developed. Paffenbarger et al.²³ devised a brief mail questionnaire, which guided the participant through questions about stair use, walking and sports activities typically performed. Depending on the activity, time frames of 1 week, 1 year, or usual activity were used. Likewise, a variety of simple questionnaires were developed, attempting to classify individuals in broad categories of activity level, typically by asking less than ten categorical questions. Such an approach was taken by the Lipid Research Clinics,^5,24 Godin,²⁵ the 11 questions asked in the Stanford Heart Disease Prevention Program,⁸ and Baecke.²⁶

The investigators of the Stanford Heart Disease Prevention Program developed a brief questionnaire which builds on both Montoye’s and Morris’ approach. It systematically elicits details of physical activity for the past week. Their Seven Day Recall⁸ presents respondents with a detailed list of potential activities and then elicits a recall of hours spent in the previous seven days performing activities at several intensity levels. All these questionnaires have been indirectly validated against fitness, fatness, treadmill performance, HDL-C, all-cause mortality, and/or development of CHD.^9,19–35

The PAH is a brief questionnaire which builds on all of this experience. Specifically, it incorporates three important considerations: 1) a design based on prior data from Taylor’s questionnaire about the distribution of participation in physical activity; 2) concepts from the Leisure Time Physical Activity Questionnaire (LTPA) about respondents’ abilities to recall and synthesize information reliably and validly; and 3) the idea that highly variable behaviors such as physical activity and diet are better ascertained by asking about the representation of activities performed over a long period of time rather than one that asks about accurate recall of recent activity.^36,37 It provides a physical activity profile which covers the activities most commonly performed during the past year, allowing estimates of participation in 13 categories of activity. These activity categories were selected to represent the bulk of activity reported in the Minnesota Leisure Time Physical Activity Questionnaire, as part of the Minnesota Heart Health Program. Summary scoring was devised for energy expenditure (measured in arbitrary “Exercise Units,” since no direct estimate of duration of activity is obtained) in heavy intensity and moderate intensity activities, separately and combined. Levels of activity based on the PAH in CARDIA and in a small, older population-based sample are given in this paper. These levels are in general agreement with each other (compare Table II with Tables V and VI) and with observations using more extensive questionnaires, i.e., Folsom et al.¹¹. The PAH takes less than 10 minutes to administer. It has acceptable reliability in a general population, aged 18–74 years. Its validity is indirectly established in a biracial population ranging in age from 18 to 30 years by its relationships with several objective measures of fatness, fitness and lipid status. As is the case with more extensive questionnaires, the PAH has relatively low correlations with validation criteria. Despite the ability of the PAH to differentiate people according to fitness, fatness and lipid status, the relatively low correlations between physical activity and the validation criteria serve as a caution that physical activity patterns vary widely within individuals across time, and that it remains desirable to formulate more precise physical activity questionnaires.

Investigators wishing to assess level of physical activity may select from among a variety of existing instruments. The choice should be based on precision desired (that is, reliability and validity), feasible mode of administration (through the mail, self-administered with interviewer aid, interviewer administered in person or on the telephone) and whether usual or recent physical activity is of interest. Finally, administrative costs may be a concern. The PAH is limited in that it provides coarse information about usual physical activity and does not provide a measure of caloric expenditure. Nevertheless, it has several advantages. It is brief, has a broad scope, and has a low cost while maintaining a high level of reliability and validity.

APPENDIX PHYSICAL ACTIVITY HISTORY QUESTIONNAIRE

Instructions

1. Did you (specify activity A–M) in the past 12 months for at least 1 hour total time in any month? (For instance you might have done three 20-minute sessions in the month.)

2. If yes, how many months did you do this activity?

3. How many of these months did you do this activity tor at least_(hours) per week?

Ask questions 1–3 separately for each activity A–M.

1. Jog or run (2 hours)

2. Vigorous racket sports (3 hours)

3. Bicycle faster than 10 mph or exercise hard on exercise bicycle (2 hours)

4. Swimming (2 hours)

5. Vigorous exercise class or vigorous dancing (3 hours)

6. Home or leisure activity (snow shoveling, moving, lifting) (3 hours)

7. Vigorous job activity (lifting, carrying, digging) (5 hours)

8. Strenuous sports (basketball, football, skating, skiing) (3 hours)

9. Nonstrenuous sports (softball, shooting baskets, volleyball, ping-pong, leisurely jogging, Swimming or biking) (3 hours)

10. Walls or hikes (4 hours)

11. Bowling, golf (3 hours)

12. Home exercises, calisthenics (3 hours)

13. Home maintenance (gardening, carpentry, painting, raking, mowing) (5 hours)