Journal of Pftys/ca/ Activity and Health, 2012, 9, 225-236© 2012 Human Kinetics, Inc,

Reliability and Validity of CHAMPS Self-Reported Sedentary-to-Vigorous Intensity Physical Activity in Older Adults

Eric B. Hekler, Matthew P. Buman, William L. Haskell, Terry L. Conway, Kelli L. Cain,James F. Sallis, Brian E. Saelens, Lawrence D. Frank, Jacqueline Kerr, and Abby C. King

Background: Recent research highlights the potential value of differentiating between categories of physicalactivity intensities as predictors of health and well-being. This study sought to assess reliability and concurrentvalidity of sedentary (ie, I METs), low-light (ie, > I and 2 and .3 METs), and "total activity" (>2METs) from the CHAMPS survey. Further, this study explored over-reporting and double-reporting. Methods:CHAMPS data were gathered from the Seniots Neighborhood Quality of Life Study, an observational study ofadults aged 65+ years conducted in 2 US regions. Results: Participants (N = 870) were 75.3 ± 6.8 years old,with 56% women and 71% white. The CHAMPS sedentary, low-light, high-light, total activity, and MVPAvariables had acceptable test-retest reliability (ICCs 0,56-0.70). Tbe CHAMPS high-light (p = 0,27), totalactivity (p = 0.34), and MVPA (p = 0.37) duration scales were moderately associated with accelerometryminutes of corresponding intensity, and the ,sedentat y scale (p = 0.12) had a lower, but significant correlation.Results suggested that several CHAMPS items may be susceptible to over-reporting (eg, walking, housework).Conclusions: CHAMPS items effectively measured high-light, total activity, and MVPA in seniors, but furtherrefinement is needed for sedentary and low-light activity.

Keywords: exercise, sedentary behavior, measurement, seniors, questionnaire

Recent evidence indicates that light physical activity(defined as >1 and

226 Hekler et al

potentially could be if demonstrated to have sufficientreliability and validity.

Self-report physical activity measures arc comingunder increased scrutiny.'- Self-report measures havemany psychometric problems including response biasissues such as over-reporting or double-reporting'-' aswell as socially desirable responses.'"* Indeed, self-reportphysical activity measures have been shown to be poor ate.stimating absolute levels of physical activity. ' ' '"' Despitethese limitations, self-report physical activity tneasuresare still valuable because they can assess specific behav-iors (eg, walking to run errands vs. walking for exercise)whereas accelerometry and other objective measures suchas doubly-labeled water cannot. Self-report measures arestill valuable for epidemiologic studies in which distri-bution of an accelerometer to all research participantsis impractical. The current physical activity guidelinesare based largely on self-reported estimates of physicalactivity.'" Additional research is still needed to refineself-report physical activity measures with a particularfocus on their utility relative to accelerotnetry.

The primary aim of the current study was to exam-ine the 6-month test-retest reliability and validity of theCHAMPS sedentary and light activity categories. Thesecond aitn was to explore psychometric issues of theCHAMPS with a particular focus on double reportingand over-reporting.

Methods

ProceduresThe Senior Neighborhood Quality of Life Study(SNQLS) was an epidemiological study designed toexamine the relationship of neighborhood "walkability"characteristics to physical activity and other outcomes inadults 66 years of age and older. Study tnethods have beendescribed previously.'^ In brief, the study was conductedin 2 U.S. metropolitan areas (Seattle/King County, WAregion and the Baltimore, MDAVashington DC area), andparticipants were recruited from neighborhoods repre-senting wide variability in neighbothood walkability andincome.''* The study used a 2 x 2 design with a) neigh-borhood "walkability" based on geographic infortnationsystems data related to street connectivity and land-usemix (dichotomized into high walkable and low walkable),and b) neighborhood income (dichototni/.ed into highincome and low income).'"* The pritnary putpose of thestudy was to examine if neighborhood "walkability" andincome were associated with physical activity of olderadults. The protocol was approved by the institutionalreview boards lor the investigators' home institutions.

Participants were recruited via tiiail, followedup with telephone contact, using contact informa-tion obtained from a commercial marketing company.Interested participants completed an infomied consentstatement and were then mailed an accelerometer andsurvey packet. Participants were instructed to wear the

accelerometer for 7 consecutive days, complete thesurvey alter wearing the accelerometer, and mail backthe accelerometer and survey packet to research staff.Six months later the accelerometer and a second surveywere mailed to participants with sitnilar instructions.Participants were recruited and enrolled year-round IVotn2005 through 2007. This 2-phase data collection strategyreduced participant burden, provided 2 7-day periods ofself-report and objective physical activity assesstnent,and controlled for scasonality effects. Upon completionof each assesstnent, participants were given $25.

Participants

Of the 3911 eligible contacts, 25% provided writ-ten informed consent and agreed to have a survey andaccelerometer mailed to them. Of those who completedthe informed consent, 929? returned the survey and worethe accclcrotneter during the first time-point, resulting ina total sample of 896 participants for the first time point(Time 1 ). To answer the validity questions of interest, theanalysis satnplc was limited to only those with completedata on the CHAMPS at Time I. This resulted in a finalsample of 870. Table 1 describes this sample in moredetail. Participants wctc contacted by mail 6 months laterto cotnplctc a second survey and wear an accelerometerfor the second titne point (Time2). Eighty-six percentof the original satnple mailed back the survey and worethe accelerometer at Timc2, resulting in a final sampleof 748 with completed assessments at both time points.These individuals were included in the test-retest reli-ability analyses.

MeasuresDetnographic information was ascertained via sell-report.This included age, gender, race/ethnicity, years of educa-tion, incotnc, and height and body weight to calculateBMI (kg/tn-). Adult self-report of weight and height arecorrelated strongly with objectively measured values.-"

CHAMPS. The original CHAMPS includes 41 itemsrelated to a variety of activities in which midlife andolder adults tend to engage. In the SNQLS,* 11 items,mostly related to transportation, were added to theCHAMPS and I item was deleted (new and deleted itemsare identified in the Online Appendix). Three originalCHAMPS items that were not included in the originalscoring but were at the beginning of the survey concern-ing general activity and flights of stairs climbed per daywere dropped in the interest of parsimony. The CHAMPSu.sed in SNQLS asks participants to think about the past4 weeks and report on the ftcquency and total amount oftime, on average per week, participants engaged in eachactivity. Those participants who reported engaging inthe activity were asked to select 1 of 6 categories reflect-ing the amount of time on average they engaged in theactivity, ranging from less than I hour per week to 9 ormore hours per week. The midpoint of each category's

range was used to provide an estimate of activity time foreach itetTi (eg, I to 2.5 hours = 105 minutes per week).The original CHAMPS scoring includes algorithmsto calculate total caloric energy expenditure per weekbased on "all" activities that would likely impact health(henceforth referred to as "total meaningful activities"),caloric energy expenditure per week based on moderate-to-vigorous intensity activity (ic, physical activity ofgreater than or equal to 3 METs), frequency per weekof cngagctncnt in total tneaningful activities (minutes/week), and frequency per week of engagetnent in mod-erate-to-vigorous intensity physical activities (minutes/week). For this study, we added algorithms to calculatesedentary and light activities. Although we calculatedall of the original CHAMPS scores, we were interestedprimarily in the average frequency per week (ic, minutes/week) participants reported engaging in each physicalactivity intensity level (ie, sedentary, light, modcrate-to-vigorous), as this was most readily cotTiparable to thecoding of the accelerometry data within SNQLS, andsitnilar, datasets.^'

All CHAMPS items were assigned a MET valuebased on Stewart ct al* and the 20(X) Cotnpendium ofphysical activities.' As in the original Stewart et al papet,**MET values were lowered for some CHAMPS itetns ( 18based on the current techniques) based on likely reducedexertion among adults over 65 years old lor these activi-ties relative to the general adult population (see OnlineAppendix for a full rationale of MET changes for eachitem).'* In addition, 4 items received different MET valuesfrom the original Stewart ct al values based on activityintensity classifications frotn the 2000 Compendium.Finally, 13 sedentary and light intensity itctns that werenot originally assigned a MET value and 10 new itemswere assigned a MET value. See Online Appendix formore details on each CHAMPS item.

Based on previous work with accelerometry witholder adults^ and to provide additional discriminationacross the range of light intensity activities, CHAMPSitems were categorized into 2 light intensity levels: low-light (METs > I and 2 and 1 and 3 and 6).

Accelerometry. Participants also wore an Actigraphaccclcromctcr for 7 days that was included with eachquestionnaire packet mailing (mailed 6 months apart).Participants were asked to wear the accelerometer duringwaking hours. A follow-up phone call was scheduledwith the participants for the day after they received theaccelerometer to review the key points of the writteninstructions and to answer any questions. In addition,research personnel contacted pat ticipants at least 1 other

CHAMPS Sedentary-to-Vigorous Activity 227

time by phone during the week to answer any questionsthat may have arisen. Participants were also provideda toll-free nutnber to use if any other questions aroseduring the week that they wore the accelerometer. Par-ticipants were asked to wear the accelerometer beforefilling out the questionnaires; therefore, the 1-weekaccelerometry data collection occurred during the4-week time-span being captured by the CHAMPS.

The Actigraph (Actigraph, Manufacturing Technol-ogy Incorporated, model 7164 and 71256, Fort Walton,FL) is a small, electronic, uni-axial device that is wornon the waist and measures activity counts (epoch set at 1minute for this study). It has been extensively validatedin a variety of populations, including older adults.'̂ '̂̂ 'Data compliance and cleaning procedures were consis-tent with other large-scale cross-sectional accelerometerstudies,''* such that a) valid hours of data consisted ofno more than 45 consecutive 'zero' values (interpretedas nonwear time), b) a valid day was defined as at least8 valid h/day, and c) participants with less than 5 validdays (using a 30 consecutive 'zero' and 10 valid h/dayrule) or less than 66 valid hours across 7 days were askedto tewear the accelerometer.

To date there are no widely accepted accelerometer"cut points" to differentiate activity intensity levels spe-cifically for older adults.-^ Minutes of moderate-vigorous(>I952 counts/min) and sedentary ( 100 and < 1041 counts/min) were identified as low-lightactivity. Activity minutes that were equal to or above theCopeland value but below the traditional adult value of1952 (ie, >1041 and 104I) was also calculated for concurrent validationof the total meaningful activity CHAMPS variable. Thisaccelerometry value roughly corresponds with all of theactivity classified as either high-light or moderate-to-vlgorous intensity activity and is referred to subsequentlyas high-light/moderate-to-vigorous intensity activity.

Statistical Analyses

All analyses were completed using SAS Enterprise Guide4.1 (SAS Institute Inc., Cary, NC, USA). Summary sta-tistics wete calculated for all variables of interest, with

228 Hekler et al

means and standard deviations reported for continu-ous variables and percentages reported for categoricalvariables.

Test-retest reliability of the CHAMPS was assessedby calculating the 1 -way single-measure intraclass corre-lation (ICC) between study Time 1 and Time 2. The ICCis a better indicator of stability over time than Pearsonproduct-moment correlations as it measures both thepercent of variance explained by each individual and themean differences over time.

To assess the validity of the CHAMPS, scales fromthe CHAMPS and accelerometry were compared usingSpearman rank-order correlations because of nonnormaldistributions for the majority of the CHAMPS and accel-erometry variables. To explore possible over-reportingon the CHAMPS, Bland-Altman Plots comparing accel-erometry and CHAMPS were used.̂ * A Bland-Altmanplot displays the difference between 2 scales on the samemetric (eg, min/wk for this study) compared with theaverage value of the 2 scales. Accelerometry has beenshown to under-represent total energy expenditure,-^which is partially attributable to misclassification of lowimpact moderate-to-vigorous intensity activities (eg,bicycling) as sedentary or light activity.-"^ As neither theCHAMPS nor the accelerometer are "gold standards"for sedentary and light intensity activity, these plots arebest interpreted as giving evidence for concurrent ratherthan criterion validity.

The highest response option for the CHAMPS is 9or more hours per week. This may limit the CHAMPS inassessing sedentary and low-light activities that often areengaged in for more than 9 hours per week (eg, televisionviewing). Based on this, the percent of participants whoreported the highest possible option for the CHAMPSwas reported as an indicator of the number of participantswithin the current sample who chose that category forsedentary and low-light activity.

To examine possible double reporting, CHAMPSitems that appeared su.sceptible to potential double report-ing (eg, walking fast/briskly versus walking up hill; heavyhousework versus light housework) were paired, result-ing in 10 item-pairs (see Table 3). Data were coded toreflect different response options (see Table 3). We wereprimarily interested in the number of participants whoreported engagement in 2 similar activities for both thesame category of minutes and an equal number of timesper week. Although not an ideal indicator of double-reporting, we hypothesized this could be indicative ofpotential double reporting. T-tests comparing participantswbo may bave double-reported based on our coding tothose who did not double report on accelerometry-basedphysical activity was done to test the validity of thedouble-reporting algorithms.

ResultsTable 1 reports basic demographics of the sample. Test-retest reliability estimates of the CHAMPS indicatedacceptable 6-month stability for the low-light (ICC =

0.70), high-light (ICC = 0.68), moderate-to-vigorousduration (ICC = 0.66), total meaningful activity duration(ICC - 0.69), moderate-to-vigorous physical activitycaloric expenditure (ICC = 0.61), and total meaningfulactivity caloric expenditure (ICC = 0.64) variables andsomewhat lower 6-month stability for the sedentary vari-able (ICC = 0.56).

Table 2 reports Spearman rank-order correlationsbetween CHAMPS and accelerometer variables. Resultsindicated the strongest correlation between the CHAMPSmoderate-to-vigorous caloric expenditure (p - 0.40, P <.001) followed closely by the CHAMPS total meaning-ful activity caloric expenditure (p = 0.39, P < .001),moderate-to-vigorous intensity duration (p = 0.39, P <.(K) 1 ), and total meaningful activity duration (p = 0.38, P< .001 ) variables when correlated with the accelerometry-based high-light/moderate-to-vigorous intensity activityvariable. A moderate and significant correlation betweenthe CHAMPS and accelerometer moderate-to-vigorousvariables (p = 0.37, P < .(K)01 ) was observed, consistentwitb previous research.--' The CHAMPS and acceler-ometry high-light variables had a moderate but slightlyweaker correlation (p = 0.27, P < .0001). Correlationbetween the CHAMPS and accelerometer values forsedentary was significant but low (p = 0.12, P < .001).Finally, there was no significant correlation between theCHAMPS low-light variable and tbe accelerometry-derived low-light variable (p = 0.06, P = . 10).

Figure 1 shows Bland-Altman Plots comparing theCHAMPS and accelerometry. The Bland-Altman Plotsshowed wide differences in estimates between acceler-ometry and CHAMPS activity time on all variables. Inparticular, tbe CHAMPS indicated fewer minutes/weekof sedentary {Mj¡ff= -2841.6 min/wk; 95% CI = -4476.7to -1206.5 min/wk) and low-light activity iM,iiff= -472.7min/wk; 95th CI = -1937.2 to 991.9 min/wk) and moreminutes of high-light (A/j,^= 395.5 min/wk; 95% CI =-346.2 to 1137.2 min/wk), moderate-to-vigorous {Mjiff=222.4 min/wk; 95% CI = -^02.9 to 847.5 min/wk), andtotal activity (Mj,y= 617.8 min/wk; 95% CI = -504.1 to1739.7min/wk) activity relative to accelerometry.

Examination of the plots suggests that the dis-crepancy between CHAMPS and accelerometery onthe moderate-to-vigorous (Panel D), high-light activity(Panel C), and total activity (Panel E) variables wasmore pronounced among persons who reported beingthe most active (in the direction of over-reporting on theCHAMPS). No noticeable discrepancy between the 2assessment modes was observed between measures byactivity level for low-light activity (Panel B). Further,the discrepancy between the CHAMPS sedentary itemsrelative to accelerometery (Panel A) appeared to be morepronounced as participants reported being more sedentary{\n the direction of underreporting on CHAMPS).

Table 3 reports descriptive statistics on individualitems of the CHAMPS. The highest possible responseoption on the CHAMPS (ie, 9 or more hours) was reportedfor 37.2% of participants on the watching television item,and 28.9% of participants on the reading item. Over half

CHAMPS Sedentary-to-Vigorous Activity 229

Table 1 Demographic Characteristics (N = 870)

NRace

White/Caucasian (%)

Black/African-American (%)

Other (%)

Education

1000

(A)

-8000

1000 1500 2000 2500 3000 3500 4000

Average (CHAMPS-)-Accel/2) min/wk

5000

(B) Average (CHAMPS+Accel/2) min/wk

(C)

-1000400 600 800 1000

Average (CHAMPS+Accel/2) min/wk

Figure 1 — Bland Altman plots: A) Sedentary behavior: B) Low-light activity: C) High-light activity.

(continued)

230

CHAMPS Sedentary-to-Vigorous Activity 231

V)CL

2000

1500

1000

(D)

< S

^ ÊSig -"^Q

-1000

«•1

^ * * • • * * •

r • • *

* •

Jupper tx>und=847.5

*

mean diff=222.4i

lower bourxí=-402.9

200 400 600 800

Average (CHAMPS+Accel/2) min/wk

1000 laao

O)uu<Q.

S:«< 5

IIa>u

Í0)it:Ö

4500

4000

3500

3000

2500

-1000

(E)

500 1000 1500

Average (CHAMPS+Accel/2) mmiviV,

2000 «9W

Figure 1 (continued) — Bland Altman plots: D) Moderate-to-vigorous activity; E) Total meaningful activity (ie, high-light andtnoderate-to-vigorous activity).

and the same number of times per week for 2 similaractivities across several of the item sets, including golfingwith a cart versus golfing while carrying clubs (13.2%reported the 2 items similarly), walking fast versus walk-ing leisurely (8.6%), heavy versus light gardening (8.2% ),heavy versus light house work (6.1%), walking leisurelyversus walking for errands (5.4%), and aerobic machinesversus stair machines (5.4%).

Aggregation of these data revealed that 21.3% ofparticipants reported the satne category of tninutes perweek and the same number of times per week on at least1 item-pair. Individuals who tnay have double-reportedalso had significantly tnore tninutes of mtxlcratc-vigorousactivity based on accclcrotnctcty compared with thoseindividuals that did not double-report on any items [/(848)= -2.45, A/"" = 105.0 ± 132.4 vs. M'̂ '""« = 82.1 ± 105.1 ].Further, younger participants (ie, those between the agesof 66 and 69 years) were tnore likely to respond similarlyto related items compared with older participants (ie.

30.3% of 66-69 year olds reported similar values on 1or more items relative to 19.7% of 70-79 year olds and15.7% of 80+ year olds, %' = 15.5, P < .001). In addi-tion, men were tnore likely to report similar values thanwomen (ie, 27.1% and 16.8%> reported similarly on 1 ormore item pairs, respectively, x^ = 1-3.3, P< .(X)l). Nosignificant differences in the tendency toward similarreporting were found based on education {P = .89) orrace/ethnicity {P= .07).

Discussion

The main contribution of the current study was to developand evaluate new measures of sedentary behavior andlight intensity activity from the CHAMPS survey thatare tailored for older adults. Results generally indicatedacceptable reliability lor the CHAMPS sedentary,low-light, high-light, moderate-to-vigorous, and total

Table 3 CHAMPS Item Responses

Sedentary» (MET = 1)Total sedentaiy behavior

Watch television

Read

Sit and talk with friends

Attend an event

Ride in a car

Travel by bus

Travel by subway or train

Use a dial a ride service

Low-light" (MET >1 AND

CHAMPS Sedentary-to-Vigorous Activity 233

Table 3 (continued)

Moderate-to-vigorous" (MET ¿3) Mean± SD% contributionto sedentary

% never doingactivity

% report 9+hours/wk

Aerobic machine w/ arms & legs

Golf, carrying/pulling equipment

Dance

Do stair or step machine

Do water exercises

Do aerobics or aerobic dancing

Swim gently

Play doubles tennis

Swim moderately or last

Jog or run

Bicycle to do errands

Basketball, soccer, or racquetball

Play singles tennis

Skate (ice. roller, in-line)

13.2+8.8±

8.0±

7.5±

7.2+

7.1±

5.9±

3.7±

3.4±

2.6±

2.3±

0.9±

0.5±

O.l±

45.254.841.130.535.132.5 '.36.3

30.8

26.2

20.5

24.3

12.4

6.4

3.6

4.2

X8 .2.52.42.3121.9U1.1480.80.80.80.8

86.3

96.693J -̂

•f 9Ö.1 '

93.8 '93.595.598^97,797.097.899.499.399.9

0.1

0.3

m0.0ai0.50.20.0OX)0:00.00.0

Note. Item listings above do not include the ful

•' Test-Retest Reí: ICC = 0.56; "Test-Retest Reí

wording for each item. For the full

ICC = 0.70;'Test-Relest Rel: ICC

wording, please see the Online Appendix.

= 0.68; ""Test-Retest Rel: ICC = 0.66.

Table 4 Comparison of Responses to CHAMPS Items Suspected of Double-Reporting

J .

CHAMPS items compared

Talk vs visit with Iriends

Talk vs play a game

Walk fast vs walk leisurely

Walk fast vs walking uphill

Walk leisurely vs errand walking

Golf w/cart vs golf carry clubs

Aerobic mach, vs stair machine

Aerobic machine vs aerobics

Heavy vs light gardening

Heavy vs light housework

Full sample

Equal min,unequal

# times/wk

22.3

6.2

4.7

0.0

9.2

0.1

0.5

0.5

6.4

6.3

Only participants that respondedto at least 1 of the 2 items in

affirmativelythe pair

Equal min. Equal min, Equal min,equal unequal equal

# times/wk # times/wk # times/wk

1.5

0.2

6.8

0.0

4.3

0.8

1.2

0.8

5.4

5.2

22.4

6.4

6.0

0.0

J: 11.7

1.9

2.2

2.6

9.8

7.4

1.5

0.28.6

0.0

5.4

13.2

5.44.58.26.1

Note. All above values are percents of the given sample (ie, either the full sample or only a subsample of participants that responded afñrmativelyto at least I of the 2 items in the item-pair).

meaningful activity scales. Comparisons with acceler-ometer data indicated acceptable concurrent validity forthe previously validated moderate-to-vigorous scalesand total meaningful activity scales that were similar toprevious studies.-" Results suggested that tbe new high-light duration scale was significantly correlated withaccelerometry, albeit with a slightly weaker correlation.There was poor concurrent validity for sedentary andlow-light scales.

These results suggest that the CHAMPS moderate-to-vigorous duration and caloric expenditure, total

meaningful activity duration and caloric expenditure, andhigh-light variables are all reliable and valid methods forscoring the CHAMPS. The addition of the "high-light"scoring method to the CHAMPS could prove usefulfor future epidemiologic studies and intervention trialsamong older adults. As indicated by previous research,high-light intensity activity appears to have positivehealth benefits on physical health and psychosocial well-being after controlling for moderate-intensity physicalactivity and sedentary behaviors.^ This finding suggeststhat interventions among older adults may not need to

234 Hekler et al

promote moderate-to-vigorous intensity activity butinstead could gain similar health benefits by targetingonly high-light intensity activity. This might be a particu-larly fruitful area for intervention research atnong olderadults as this group already spends a significant portionof time in light intensity activity and thus an interventionto promote increa.sed high-light activity may be easier forolder adults to adopt and maintain.

The weak concurrent validity results for the sed-entary and low-light measures tnay be at least partiallyattributable to the ceiling effect ob.served for severalsedentary and low-light activities (eg, respondents couldnot report doing television viewing, reading, or drivingfor more than 9 hours per week due to the formattingof the CHAMPS). Approximately half of participantsreported the highest possible response option on at least 1sedentary item and one-third of participants reported thehighest possible response option for low-light activities.

Bland-Altman plots showed that participantstended to report more moderate-to-vigorous and high-light activity and less low-light and sedentary activitythan that measured by accelerometry. These resultsare consistent with other studies suggesting self-reportmeasures may misrepresent absolute levels of activity,'''including over-reporting of moderate activity comparedwith accelerometry.'^ One explanation for the absolutedifferences in sedentary minutes between the CHAMPSand accelerometry measures is that the CHAMPSunder-represents the activities individuals do whilesedentary. As the CHAMPS was not originally intendedto assess sedentary behavior systematically, more itemsmay need to be added to the measure to fully assesssedentary behavior (eg, relaxing, or listening to music).More importantly, the observed ceiling effects observedwithin the sedentary scale should be addressed. Potentialimprovements would be to expand the re.sponse optionsto include higher values or rewriting the CHAMPS withan open-ended response option, as is currently availablefor other physical activity self-reports (eg, the Seven-dayPhysical Activity Recall,-*- International Physical ActivityQuestionnaire").

To examine the potential impact of using an open-ended response option instead of the categorical option,we completed a small pilot study with 14 community-dwelling adults ages 50 years and older (mean age = 63.4± 7.4). Participants completed 2 versions of the CHAMPSon 2 consecutive days with the order of administrationrandomized. One version was the traditional CHAMPSusing the categorical response option and the other ver-sion asked participants to state "how many hours perweek" they participated in an activity (eg, walking fast/briskly) with an open-ended response option. Participantsreported significantly more sedentary and low-lightactivity on the open-ended version compared with thecategorical version (Sedentary: Open-ended = 2413.1 ±1501.1 min/wk vs. Categorical = 1393.9 ± 328.4 min/wk. Low-light: Open-ended = 2683.2 ± 1577.7 min/wkvs. Categorical = 1714.3 ± 497.5 min/wk). Although itdid not reach statistical significance in this small pilot.

the open-ended response option produced lower reportsof moderate-to-vigorous activity (Open-ended = 427.8 ±334.6 min/wk vs. Categorical = 548.6 ± 436.9 min/wk).

Pilot study results, coupled with results from thelarger investigation indicating the CHAMPS may under-estimate sedentary and low-light activity and overestimatemoderate-to-vigorous activity relative to accelerometry,suggest that changing the CHAMPS to an open-endedresponse option may make the CHAMPS more useful inassessing absolute levels of activity. As this specific rec-ommendation is based on a small pilot study, additionalempirical validation and reliability testing using a larger,more diverse sample is required before any conclusionscan be drawn. Sotne groups of older adults may find itdifficult to fill out the open-ended option, although nodifferences in perceived difficulty or the amount of timeto complete the survey were found in the pilot study.

Another explanation for the apparent overestitna-tion of moderate-to-vigorous intensity activity andunderestimation of sedentary behavior is that self-reportmeasures are poor indicators of absolute levels of physicalactivity.""" Indeed, results from the Bland-Altman plotssuggest that the CHAMPS is poor at estimating absolutelevels of physical activity at all intensities, assuming thataccelerometry is a better indicator of absolute physicalactivity. Therefore, the CHAMPS should likely not beused to estimate absolute levels of physical activity.Previous research on the CHAMPS, however, indicatesthat the CHAMPS does consistently correlate withother measures of physical activity including a Mini-Logger 2000 (a device that measures both activity andheart rate),^' VO-max,-*'* fitness tests (eg, chair stands,6-min walk),-"'̂ ''- '̂' accelerometer,•^•' other self-reportedphysical activity measures,̂ '••'̂ and self-reported healthmeasures.̂ '• '̂* In addition, the CHAMPS has been usedsuccessfully within clinical trials and thus is sensitive tosmall but meaningful changes.*-^'•'''Taken together withthe present findings, these results suggest that, althoughthe CHAMPS should likely not be used as an absoluteindicator of physical activity, it does function well as anindicator of relative physical activity.

Both gardening and household activities were com-monly reported in this cohort, and other studies suggestedthese kinds of routine or instrumental forms of activity aresubject to over-reporting.'^ As part of tbe pilot activitiesreferred to above, 16 participants were asked to describein more detail their activities during gardening and house-hold activities. Re.spon.ses suggested these items tended tobe over-reported unless tbllow-up questions were asked.Many participants revised their estimations of time spentin these activities downward in response to probes by tbeinterviewer. If these pilot results are representative, it maybe preferable to reword the gardening and houseworkitems. For example, a caveat such as, "please count eventsin which you arc engaging in the activity for at least 10minutes at a time," or "please include only time when youare actively engaging in this specific activity" may help toreduce over-reporting on these items (such probes wereused by the interviewer to reduce over-reporting in the

CHAMPS Sedentary-to-Vigorous Activity 235

pilot). Other options would be to offer a more inclusivelist of heavy household and heavy gardening activities soas to reduce estimations about activities that a participantmay perceive as strenuous but are not listed as sueh inthe 2000 Compendium. These suggestions require furtheretTipirical examination.

Analysis of selected tnatched-pairs of items for pos-sible double reporting suggested that 1 in 5 participantsin the larger investigation tnay have double reported butonly on 1 of the selected item pairs. Participants whomay have double-reported were also more active basedon accelerometry and were younger men. These resultssuggest that participants may simply have been moreactive, indicating that double-reporting may not be aserious problem within the CHAMPS, Nonetheless, morerefined methods for examining double-reporting (eg,detailed querying following completion of the survey) areinquired before any definitive conclusions can be drawn.

Limitations and Strengths

There are several limitations to this study. One concernsthe use of existing data to investigate attributes of tbeCHAMPS, as opposed to a study specifically designedto evaluate new CHAMPS scores and improve themethod in a tnore controlled manner. There are knownlimitations to the use of accelerometers to assess lightintensity activity,^' Therefore, accelerometry is not a true"gold standard" indicator of light activity and thus thecorrelations between accelerotnetry and the CHAMPSshould be interpreted with caution. Although the presentdistinction between low-light and high-light activity wasbased on previous work by Copeland et al,-' the Cope-land cut-point, which is approximately half of tbe con-ventional moderate-to-vigorous "Freedson" cut-point,-'was developed as an indicator of moderate-to-vigorousactivity, not light intensity activity. Nonetheless, previ-ous research that linked light intensity activity to healthoutcomes defined light activity as activity that wasgteater than sedentary and less than the Freedson countpoint,- Therefore, the Copeland cut-point, although notoriginally defined as identifying "high-light" activity,may nonetheless be a good indicator for differentiatinglow-light and high-light as it is consistent with priordefinitions of light activity,- This conclusion is furthersupported by a recent study that examined the relationof different activity intensities to health outcomes, whichused the Copeland cut-point to differentiate between"low-light" and "high-light" activity,'

The current study also has several strengths. Thelarge sample size was drawn from 2 regions of the coun-try, thus increasing generalizability of results. The sampleincluded a large number of 80+ year old participants (n= 235), an age group that is often difficult to study. Asis evident in Table 1, participants had a wide range ofincome and educational backgrounds, further improvinggeneralizability. CHAMPS and accelerometer data werecarefully matched by intensity, enhancing the validityanalyses.

ConclusionsResults from the current study suggest the CHAMPS canreliably assess sedentary to moderate-to-vigorous activityin a diverse sample of older adults, and the moderate-to-vigorous activity and total meaningful activity variableswere validated by accelerometery at the levels typicallyobserved in other investigations. A new finding was theCHAMPS "high-light" scoring procedure appears to bea reliable and valid way of scoring the CHAMPS thatmay prove useful for future intervention trials that focusmore exclusively on "high-light" physical activity andepidemiologic studies examining health impacts of activi-ties of various intensities. CHAMPS and aecelerometryvariables were found to be only weakly associated forthe sedentary and low-light variables, suggesting that,in its current form, tbe CHAMPS should not be used toassess sedentary and low-light activity. This finding is notunexpected, given that the CHAMPS was not originallydeveloped to measure these 2 categories of activity behav-ior. It is possible that more precise estimates of sedentarybehavior and physical activity could be captured usingan open-ended response format, as well as additionalitems for sedentary behavior, but this would need to beempirically evaluated. Results further indicated that theCHAMPS may significantly overestimate moderate-to-vigorous intensity activity and underestimate sedentarybehaviors. Based on this, the CHAMPS is likely bestused as a relative indicator of physical activity ratherthan an absolute indicator. Overall, results from this studyconfirmed the validity of the CHAMPS for moderate-to-vigorous and total meaningful activity, as well as high-light activities-the latter consisting of activities in whicholder adults regularly engage, making them an importanttarget for future study.

Acknowledgments

This study was funded by the Public Health Service GrantHL07714I from the National Heart, Lung, and Blood Institute.Hekler was supported by Public Health Service Training Grant5 T32 HL 007034 from the National Heart, Lung, and BloodInstitute.

References1. Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium

of physical activities: an update of activity codes and METintensities, Med Sei Sports Exerc, 2000:32(9):S498-S516.

2. Healy GN. Dunstan DW, Salmon J, et al. Objectivelymeasured light-intensity physical activity is independentlyassociated with 2-h plasma glucose. Diabetes Care.2007:.30(6):I384-1389.

3. Hu FB, Li TY, Colditz GA, Willett WC, Manson JE. Tele-vision watching and other sedentary behaviros in relationto risk of obesity and type 2 diabetes millitus in women.y/\A//l. 2003:289:1785-1791,

4. Levine JA, Vander Weg MW. Hill JO, Klesges RC, Non-exercise activity thermogenesis: the crouching tiger hiddendragon of societal weight gain, Arterioscler Thromh VaseBiol. 2006:26(4):729-736.

236 Hekler et al

5. Vandelanotte C, Sugiyama T, Cardiner P, Owen N. Asso-ciations of leisure-time intemet and computer use withoverweight and obesity, physical activity and sedentarybehaviors: cross-sectional stud. J Med Internet Res,9(3):e26. 20O9;l I(3):e28.

6. Owen N, Bauman A, Brown W. Too much sitting: a noveland important predictor of chronic disease risk? BrJ SportsA/É>Í/. 20G9;43(2):8l-82.

7. Buman MP Hekler EB, Haskell WL, et al. Objective lightintensity physical activity associations with rated healthin older adults. Am J Epidetniol, 2010:172:1 155-1165.

8. Stewart AL, Mills KM, King AC, Haskell WL. Gillis D,Ritter PL. CHAMPS physical activity questionnaire forolder adults: outcomes for interventions. Med Sei Sports&i'/-c.2001;33(7):l 126-1141.

9. Troiano RP, Berrigan D. Physical activity in the UnitedStates measured by accelerometer: comment - response.Med Sei Sports Exerc, 2008;40(6): 1189.

10. Physical Activity Guidelines Advisory Committee. P/iy.$i-cal Activity Guidelines Advisory Cotntttittee Report, 2008,Washignton, DC: U.S. Department of Health and HumanServices; 2008.

11. Courneya KS, Jones LW, Rhodes RE, Blanchard CM.Effects of different combinations of intensity categories onself-reported exercise. Res Q E.xerc Sport. 2004;75(4):429-433.

12. Westerterp KR. Assessment of physical activity: a criticalappraisal. EurJ Appl Physiol. 2009;105:823-828.

13. Altschuler A, Picchi T, Nelson M, Rogets JD, Hart J,Stemfeld B. Physical activity questionnaire comprehen-sion: lessons from cognitive interviews. Med Sei Sports&e/-c. 2009:41 (2):336-.343.

14. Sallis JF, Saelens BE. Assessment of physical activity byself-report: status, limitations, and future directions. ResQ Exerc Sport. 2OOO;71 (2):S 1 -S14.

15. Neilson HK, Robson PJ. Friedenreich CM. CsizmadiI. Estimating activity energy expenditure: how validare physical activity questionnaires? Atn J Clin Ntitr.2OO8;87(2):279-29I.

16. Soundy A, Taylor A, Faulkner C, Rowlands A. Psycho-metric properties of the 7-day physical activity recallquestionnaire in individuals with severe mental illness.Arch Psychiatr Nttrs, 2007;21 (6):309-316.

17. Shigematsu R, Sallis JF, Conway TL, et al. Age differencesin the relation of perceived neighborhood environment towalking. Med Sei Sports Exerc, 2009;41 (2):314-321.

18. Sallis JF. Saelens BE, Frank LD, et al. Neighborhoodbuilt environment and income: examining multiple healthoutcomes. Soc Sei Med, 2009;68(7): 128.5-1293.

19. Frank LD, Sallis JF. Saelens BE, et al. The development of awalkability itidex: application to the Neighborhood Qualityof Life Study. Br J Sports Med, 2009:bjsm.20O9.O5870l.

20. Stewart AL. The reliability and validity of sell-reportedweight and height. J Chronic Dis, 1982;35:295-309.

21. Troiano RP, Berrigan D, Dodd KW, Masse LC, TilertT, McDowell M. Physical activity in the United Statesmeasured by accelerometer. Med Sei Sports Exerc,2008;40(1):181-I88.

22. Welk GJ. Use of accelerometry-based activity monitors toassess physical activity. In: Welk GJ, ed. Physical activityassesstnents for health-rektted resectrch. Champaign, IL:Human Kinetics, Inc.; 2002:124-141.

23. Pruitt LA, Glynn NW, King AC, et al. Use of accelerom-etry to measure physical activity in older adults at risk formobility disability. y/\gmg PhysAct, 2008; 16(4):416-4.34.

24. WelkGJ. Principlesofdesignandanalysesforthecalibra-tion of accelerometry-based activity monitors. Med SeiSports Exerc, 2005;37(11. Suppl):S5OI-S511.

25. Freedson PS, Melanson E, Sirard J. Calibration of theComputer Science and Applications Inc. Accelerometer.Med Sei Sports Exerc, 1998;30(5):777-781.

26. Matthews CE, Chen KY, Freedson PS. et al. Amount oftime spent in sedentary behaviors in the united states,2003-2004. v4m y Ê/)if/em/o/. 2008; 167(7):875-881.

27. Copeland JL, Esliger DW. Accelerometer assessment ofphysical activity in active, healthy older adults. J AgingPhys Act, 2009;\l{\)'.\l-?:0,

28. Bland JM, Altman DG. Statistical tnethods for assessingagreement between two methods of clinical measurement.Lattcet, 1986;l(8476):.307-310.

29. Starling RD, Matthews DE, Ades PA, Poehlman ET.Assessment of physical activity in older individuals: adoubly labeled water study, [erratum appears in J ApplPhysiol 2001 May;90(5):following table of contents] JAppl Physiol. 1999;86(6):2090-2096.

30. Yokoyama Y, Kawamura T, Tamakoshi A, et al. Cotnpari-son of accelerometry and oxymetry lor measuring dailyphysical activity. CircJ. 2002;66(8):751-754.

31. Harada ND, Chiu V, King AC, Stewart AL. An evaluationof three self-report physical activity instruments for olderadults. Med Sei Sports E.xerc. 2001 ;33(6):962-970.

32. Sallis JF, Haskell WL, Wood PD, et al. Physical activityassessment methodology in the Five-City Project. Atn JEpidetniol. 1985;I21:91-1O6.

33. Craig CL, Marshall AL, Sjostrom M, et al. Internationalphysical activity questionnaire: 12-country reliability andvalidity. Med Sei Sports Exerc, 2OO3;35(8): 1381-1395.

34. Resnicow K, McCarty F Blissett D, Wang T, Heitzler C,Lee RE. Validity of a modified CHAMPS physical activityquestionnaire among African-Americans. Med Sei SportsExerc, 2OO3;35(9):I537-154.S.

35. Cyarto EV, Marshall AL, Dickinson RK, Brown WJ.Measurement properties of the CHAMPS physical activityquestionnaire in a sample of older Australians. J Sei MedSport. 2006,9(4):?,\9-n6.

36. Moore DS, Ellis R, Allen PD, et al. Construct validation ofphysical activity surveys in culturally diverse older adults:a comparison of four commonly used questionnaires. ResQ Exerc Sport. 2008;79( 1 ):42-50.

37. Resnick B, King A, Riebe D, Ory M. Measuring physi-cal activity in older adults—use of the community healthactivities model program for seniors physical activityquestionnaire and the Yale Physical Activity Survey inthree behavior change consortium studies. West J NursRes, 2008;30(6):673-689.

38. Godard MP, Standley CM. Relationship betweenCHAMPS physical activity questionnaire and functionalfitness outcomes in older adults. Act Adaptation Aging.

39. Wilcox S, Dowda M, Levitón LC, et al. Active for lifefinal results from the translation of two physical activityprograms. Atn J Prev Med. 20O8;35(4):.'í40-351.

40. King AC, Ahn DK, Oliveira BM, Atienza AA, CastroCM, Gardner CD. Protnoting physical activity throughhand-held computer technology. Am J Ptev Med.2008;.34(2): 138-142.

41. Crouter SE, Bassett DR, Jr. A new 2-regressionmodel for the Actical accelerometer. Br J Sports Med,2008;42(3):217-224.

Copyright of Journal of Physical Activity & Health is the property of Human Kinetics Publishers, Inc. and its

content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.