Background
Sedentary behaviors, defined prolonged sitting and low levels (1.0-1.5 METs) of energy expenditure [
1] are associated with morbidity and premature mortality, additional to the influence of moderate-to-vigorous physical activity [
2‐
8]. Older adults (≥65 years) are the most sedentary age group with average levels of objectively assessed sedentary time reaching 540 minutes/day and more [
9]. Total sedentary time can encompass several specific behaviors (e.g., television viewing, motorized transport, reading, computer use). Different sedentary behaviors have been linked to different health outcomes [
10,
11] and reducing them will require particular intervention strategies in different contexts [
12]. While accelerometer measurement provides an objective assessment of older adults’ total sedentary time, questionnaires are needed to assess engagement in specific sedentary behaviors [
13].
In their review on measures of sedentary behaviors in adults (≥18 years), Clark et al. [
14] concluded that there is a lack of reliable and valid questionnaires covering a wide range of specific sedentary behaviors. In older adults, only two studies have examined the measurement properties of a sedentary behavior questionnaire that included multiple questions on engagement in multiple specific sedentary behaviors rather than one question targeting participants’ overall sitting time [
15,
16]. In both studies, time spent in the specific sedentary behaviors was summed to obtain a measure of total sitting time, which was then validated against accelerometer-derived sedentary time. These self-report measures did not exhibit strong validity, with self-reported total sitting time underestimating accelerometer-derived sedentary time on average by 216 minutes/day [
15] and 406 minutes/day [
16]. The authors concluded that this underestimation might have partially resulted from the questionnaires not including some specific sedentary behaviors, such as time spent eating, sitting while telephoning, and sitting during household chores. Only one of those studies [
15] reported separately on the reliability of multiple specific sedentary behaviors.
Previous studies have reported differences between men and women in validity findings for a sedentary behavior questionnaire, in US overweight adults [
17] and in European adolescents [
18]. Next to gender, other demographic factors, such as age and educational level, may also influence validity. Except for one study testing the reliability and validity of a question targeting overall sitting time among residents of low versus high socio-economic neighborhoods of Hong Kong [
19], differences in validity results of a sedentary behavior questionnaire between demographic subgroups of the older population remain unexplored.
There is a need for the development and examination of the measurement properties of self-report instruments to address the broader range of older adults’ sedentary behaviors. We examined the criterion validity and test-retest reliability of a new questionnaire assessing a comprehensive set of specific sedentary behaviors in older adults. Additionally, we examined whether the criterion validity of the questionnaire differed according to age, gender and educational level.
Discussion and conclusions
We examined the validity of self-reported total sitting time relative to accelerometer-derived sedentary time among older adults and in different demographic subgroups. Validity was not strong, with a Spearman correlation of 0.30 in the total sample and wide limits of agreement. However, these relationships were stronger than those reported by Hekler et al. [
16], who found a correlation of 0.12 among US older adults, and are comparable to the findings of Gardiner et al. [
15], who reported a correlation of 0.30 among Australian older adults. Furthermore, with 82 minutes/day for a mean average of self-reported and accelerometer-derived sedentary time, our self-report measure of total sitting time underestimated accelerometer-derived sedentary time substantially less than the self-report measures used in the previous studies [
15,
16]. This apparently lower level of underestimation for a mean average of self-reported and accelerometer-derived sedentary time might be explained by the current questionnaire including specific sedentary behaviors that were not employed in previous questionnaires. Several explanations might account for the remaining average underestimation of 82 minutes/day. Firstly, participants might simply be unable to accurately recall and estimate durations of engagement in certain sedentary behaviors. Secondly, social desirability might have resulted in underreporting of certain sedentary behaviors (i.e. television viewing). Thirdly, accelerometers are not the ideal criterion measure to assess sedentary behavior as they cannot distinguish between different postures. Hence, accelerometers might have overestimated older adults’ sedentary time by classifying standing activities at very light intensities as sedentary. However, Healy et al. [
9] identified accelerometer-derived sedentary time as having relatively minimal bias compared to activPAL-derived sedentary time and accelerometers could both over- and under-estimate activPAL-derived sedentary time. For higher levels of average self-reported and accelerometer-derived sedentary time, self-reported total sitting time overestimated the accelerometer-derived measurement. This might be explained by those with high levels of sitting time also engaging in longer bouts of sitting time for which durations might be more difficult to estimate and more likely to be rounded up. Furthermore, these longer bouts might have been interrupted by non-sedentary activities which are registered by the accelerometers as non-sedentary, but which are included in the sedentary time reported by the participants.
Our validity findings differed between demographic subgroups. We observed stronger correlations and narrower limits of agreement for 65- to 74-year-old, male and tertiary educated participants compared to their counterparts. A first explanation for these differences in validity might be that the younger age group, men and those with tertiary education engage more frequently in sedentary behaviors that are easier to recall and report, such as car driving and computer use, compared to their respective counterparts. Furthermore, better cognitive functioning and capacity to recall and report past sedentary behaviors among 65–74 year old compared to 75+ year old participants might explain the better validity results in the younger age group. As mentioned, our validity results for men (rho = 0.35) were better compared to women (rho = 0.24). However, España-Romero et al. [
37] reported similar correlations between self-reported sitting time and sedentary time measured by a combined heart rate and movement sensor among men (rho = 0.17) and women (rho = 0.18) in a sample of British 60- to 65-year-olds. The sample of España-Romero et al. [
37] was younger than our sample and included many non-retired participants, which possibly means that these men and women were more likely to engage in similar sedentary behaviors (i.e. occupational sitting) and, hence, similar validity results. Our most substantial difference in correlation was found between non-tertiary (rho = 0.25) and tertiary educated participants (rho = 0.39). This is in line with findings by Sabia et al. [
38] on the validity of self-reported physical activity among 60- to 83-year-old British participants. They found lower correlations between self-reported and accelerometer-derived PA among those with lower education or occupational position compared to participants with higher education or higher occupational position. Cerin et al. [
19] reported better reliability of a question assessing overall sitting time among Hong Kong older adults living in high compared to low socio-economic neighborhoods. However, their validity results did not differ between high and low socio-economic neighborhoods. More research is needed to further examine demographic differences in validity and to increase the specificity of questionnaires for groups in which lower validity has been observed.
Overall, our findings on validity were not ideal; the correlation coefficients did not reach 0.50 (which was defined as good validity for physical activity questionnaires) [
34] and the 95% limits of agreement were wide. In addition to the reasons described above, another possible explanation for this absence of strong validity is that the target period of our sedentary behavior questionnaire did not overlap with the period the accelerometer was worn. However, in our subsample for the reliability analysis both periods did overlap, but a correlation analyses between their self-reported total sitting time and accelerometer-derived sedentary time did result in a similar correlation (ρ = 0.32) and similar width of the 95% limits of agreement. Additionally, although we followed standard procedures for accelerometer initialization and processing, there is not yet a consensus about many of these procedures [
27]. Although our findings showed only modest validity, they were no worse than those reported for previous questionnaires targeting older adults’ (specific) sedentary behavior(s) [
15,
16]. Moreover, our questionnaire targeted a wide range of specific sedentary behaviors which may have resulted in a lower level of underestimation of total sitting time compared to previous questionnaires [
15,
16]. Additionally, our questionnaire’s format is similar to the IPAQ, which might facilitate administration in studies that assess both physical activity and sedentary behaviors. Given the high prevalence of sedentary behaviors among older adults and the associated health risks, researchers should not delay studies on the health risks, prevalence and correlates of sedentary behaviors and could use the new questionnaire to assess older adults’ sedentary behaviors. Objective measures of sedentary time should be preferred, complemented with the questionnaire to get context-specific information. Our questionnaire might be especially useful for the specific sedentary behaviors for which we found acceptable reliability; TV viewing time, computer use, and car driving. In the meantime, more research is needed to develop questionnaires and objective criterion measures that measure older adults’ engagement in sedentary behaviors more accurately. These validity studies should use different criterion measures (e.g. Actigraph accelerometers and activPALs) and could include log books to examine the validity of specific sedentary behaviors.
Our newly-developed questionnaire for assessing multiple specific sedentary behaviors in older adults was found to be reasonably reliable for total sitting time. In contrast, only three of the twelve specific sedentary behaviors appeared to have acceptable test-retest reliability (i.e. TV viewing time, computer use, and car driving). Test-retest reliability for self-reported total sitting time in the current questionnaire appeared to be as good as or better than previous studies in older adults [
15] and adults [
9] using a sum of multiple specific sedentary behaviors and similar periods between test and retest. Similar to these previous studies, good reliability was found for TV viewing time. This might be explained by TV viewing being easy to recall accurately since it occurs on a regular basis, for prolonged periods, and at specific time points. The same explanation might be true for computer use, which was also found to have acceptable reliability in the current and previous studies [
15,
16]. However, in contrast to poor reliability found in previous studies [
15,
16], we found acceptable reliability for driving a car. Possibly, in our sample, being a car driver is connected to a distinct activity that is performed regularly by the older adults (e.g. going to the supermarket, visiting family) and is, therefore, more-readily recalled.
The remaining questions addressing specific sedentary behaviors did not demonstrate acceptable test-retest reliability, although they could be expected to occur at regular time points for relatively constant durations (e.g. sitting during meals). There might be several reasons for this absence of acceptable reliability. First, it might actually be difficult for older adults to recall and accurately estimate the duration of specific sedentary behaviors. Questionnaires assessing sedentary behavior(s) performed during the past day might offer a solution to this issue [
39], however, these might be less accurate in capturing usual engagement in sedentary behavior(s). Secondly, our test and retest assessment of the specific sedentary behaviors did not target the same seven days. Consequently, the absence of acceptable reliability might simply reflect between-week variability in the sedentary behaviors. Despite the absence of acceptable reliability for the majority of the specific sedentary behavior items, we did find acceptable reliability for total sitting time. This might indicate that the total amount of sitting time does not vary substantially from week to week, but that how it is accumulated changes (one specific sedentary behavior might be replaced by another). Hence, the inclusion of all relevant sedentary behaviors might explain the good reliability results for our measure of self-reported total sitting time.
It should be noted that we tested an interview-based version of our sedentary behavior questionnaire and that our results may not be applicable to self-completion of the questionnaire. For a questionnaire assessing older adults’ physical activity, Dinger et al. [
40] concluded that their observations of very good test-retest reliability (ICC = 0.91) might have resulted from the use of interviews rather than self-completion. Washburn et al. [
41] found better validity results for a telephone-based physical activity questionnaire compared to a self-completion version, but the latter resulted in better test-retest reliability results. Furthermore, we used ‘the last seven days’ as the time frame to report sedentary behaviors. Among adults, similar reliability and validity results for self-reported total sitting time have been observed for ‘the last seven days’ and ‘the usual week’ time frame [
23]. However, it has been argued that older adults might consider a usual rather than the last week when reporting their engagement in physical activity behaviors although they were asked to consider only the last week [
42]. Therefore, in the current study, researchers responsible for data collection were explicitly trained to ensure that participants’ self-reports reflected engagement in sedentary behaviors during the last seven days. To our knowledge, no studies have investigated the influence of administration mode or time frame on the psychometrics of a sedentary behavior questionnaire among older adults. More research is necessary to determine the optimal mode of administration and time frame.
A first strength of the current study is the examination of a sedentary behavior questionnaire that included an extensive list of specific sedentary behaviors. Secondly, our questionnaire had a similar format as the IPAQ, which we used to increase the ease of administration (since the participants were acquainted with the format by previously completing the IPAQ). Thirdly, we investigated differences in validity according to age, gender and education. Our study has limitations, however. First is the use of accelerometers as criterion measure to assess sedentary behavior. Secondly, we only examined the validity of self-reported total sitting time and not the validity of self-reported specific sedentary behaviors. Future studies could include sedentary behavior log books to assess the validity of self-reported specific sedentary behaviors. Thirdly, older adults who were limited by their health to walk a couple of 100 meters were excluded from the current study. Therefore, our findings could not be generalized to older adults with such mobility impairment. Since mobility-impaired older adults may be at increased risk for high levels of sedentary time, future research should investigate the measurement properties of sedentary behavior questionnaires in this subgroup. Fourthly, the size of the subsample for our reliability analysis was not sufficient to perform reliability analyses in different demographic subgroups. The subsample for our reliability analysis was also rather highly educated which might have led to better reliability results.
To conclude, we examined criterion validity and test-retest reliability of a newly-developed sedentary behavior questionnaire, but our findings did not exhibit ideal validity for self-reported total sitting and test-retest reliability for most of the specific sedentary behaviors. However, our findings were comparable to what has been reported by previous studies. Furthermore, our questionnaire tended to result in a lower level of underestimation of sedentary time compared to other questionnaires, possibly explained by the inclusion of additional specific sedentary behaviors. We also observed better validity results for 65- to 74-year-old, male and tertiary educated participants compared to their counterparts. Further research is needed to develop self-report tools and objective criterion measures that accurately measure older adults’ engagement in specific sedentary behaviors and total sitting time.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
JVC, VVH, IDB and BD designed the study. JVC and VVH led the data collection. JVC performed the statistical analyses and drafted the manuscript. All authors critically revised the content of the manuscript and have given final approval for publication.