Older adults typically spend much of their day sitting down [10
]. Given the growing body of evidence on the ill-effects of too much sitting [3
], it is important to include questions about sitting time in population surveillance of PA and epidemiological studies on sedentary behavior and health [1
]. In this qualitative study, we document the cognitive processes used by a sample of older adults to respond to two widely used sitting-time questions. The results suggest that the ways that individual older adults answer sitting-time questions vary depending on (1) their understanding of the question, particularly the scope of the activities to be included and the timeframe to be considered; and (2) the strategies they employ to compute time spent sitting on a daily basis.
Understanding the scope of sitting activities to report
Participants' perception of the scope of activities to include in their responses was largely guided by the examples provided. Many participants only reported activities indicated in the examples. This tendency was also observed in cognitive testing of another PA questionnaire in middle-aged US adults; some participants in that study considered the list of examples to be exclusive, rather than suggestive of activities to be included [26
]. The complexity of the cognitive processing was decreased in that study by adding more varied examples to some questions and converting questions with many examples into multiple questions with fewer examples [26
]. Although splitting questions could be a good strategy for obtaining more accurate information, short questionnaires are preferred for surveillance [20
]. An option that may improve comprehensibility but would not affect the length of the questionnaire would be to add examples of sitting activities that are more relevant for older people, as was done in a Swedish study [22
]. Additional qualitative research could determine appropriate activities to include.
Participants expressed difficulties conceptualizing 'sitting leisure activities'. This problem became more apparent for PASE than for IPAQ, probably because PASE exclusively asks about leisure-time sitting, whereas IPAQ also assesses other domains. Particularly for PASE, participants were unsure whether to include sitting while eating, resting/napping or driving, as these activities could be considered leisure-time activities, but were not included in the examples. Thus, some participants reported these activities while others did not.
Participants' perceptions of the scope of the activities to include also depended upon the domains from which the examples came. Neither questionnaire explicitly mentions transport or work as domains. Consequently, some participants reported, for example, transport-related sitting because they considered such sitting within the leisure-time domain, while others did not report this activity, as they did not consider it within the domains explicitly stated in the question. These findings suggest that respondents' understanding of the range of activities they are expected to report could be improved if they were given examples that reflect the full range of sitting activities in each domain. The results of this study further indicate that transport-related sitting is not an activity that people naturally consider in responding to questions about sitting activities. If respondents are expected to include transport-related sitting in their answers, they should be informed to do so. Clarifying domains and adding appropriate examples are likely to improve the accuracy of the responses, but further research is needed to evaluate the impact of doing so.
Understanding the timeframe of sitting activities to report
Both IPAQ and PASE ask about sitting on a day in the last 7 days, with IPAQ specifying 'a weekday'. However, the majority of participants reported sitting for a usual day, or averaged sitting time across days in a usual week. Because many participants indicated that sitting time did not vary much across days or weeks, it was likely easier for them to recall their usual activities than to recall specific activities in the past 7 days that may not have been part of their usual week.
The issue of asking about a 'usual' or 'typical' week instead of the 'last 7 days' or 'last week' has been debated in physical activity epidemiology. For example, in the original 12-country validation study of IPAQ with predominately middle-aged adults, participants experienced difficulties with the interpretation of a usual week, and therefore, the use of the 'last 7 days' was recommended [20
]. This is in line with earlier thinking in this field, which suggested that recall of actual PA participation during a set period (e.g. last 7 days) may provide more accurate estimates than recall of usual PA participation [27
]. Our results suggest that this finding may not be applicable to recall of sitting time.
Formulation of a response to sitting duration
Participants did not only have problems with reporting sitting in the past 7 days, but also with selecting a day. In the absence of instructions on how to select 'a day', participants choose one or another of various logically justifiable strategies, including choosing a day at random, selecting the day with the least sitting time, and choosing the day with the most sitting time. Estimates of sitting time, of course, varied, depending on the strategy chosen.
Similarly, since neither questionnaire suggests a strategy for calculating hours spent sitting during a day, participants used a variety of strategies, which also affected the estimates of sitting time. Participants found this step more challenging for IPAQ than for PASE, given its open-ended response format. This was particularly the case if their sitting time varied from day to day. Our overall impression was that it might be easier if participants were asked to report on a specific day, such as 'the total time spent sitting last Wednesday', as is used for the phone-administered IPAQ when respondents cannot answer the sitting question because their sitting times varied between days [20
]. Another option would be to ask about time spent sitting 'yesterday' as respondents may find it easier to recall what they did in the last 24 hours than to recall the activities of any earlier day. These options do not overcome the problem of accurately assessing weekly sitting time if activities vary from day to day, but the suggested strategies may help respondents to focus on sitting in the last 7 days.
In general, participants found PASE's closed format easier to use, and consequently, the strategies used to formulate a response were more consistent across participants for PASE than for IPAQ. It may therefore be beneficial to use a closed format to assess sitting time in older adults, as PASE does, but to provide a greater selection of response options from which to chose, starting with 0-2 hours, and progressing in 2-hour increments, up to a response option of > 12 hours. Such an approach could decrease the time and the cognitive processing needed to complete the sitting question. Regardless of the format used, instructions should be added to give guidance in selecting a day and calculating time spent sitting, in an effort to ensure the biases in the estimates are consistently in the same direction.
Limitations and strengths
It is usually not possible to generalize the findings of cognitive interviews to the general population because of sampling issues [12
]. Although participants for this study were purposefully selected, the generalizability of the findings may be limited, because participants were generally well educated, cognitively healthy, and reported good overall health. Because education levels could influence questionnaire comprehension [14
], future studies may want to target less educated people, in order to get a better understanding of how this population responds to sitting questions. As with all qualitative studies, use of the think-aloud process could have influenced the responses (i.e., reactivity effects). Answering the self-report questions in a face-to-face interview, instead of completing the questions on one's own, may also have influenced the responses. Our approach to minimizing this included not providing guidance in how to interpret and respond to the questions and probing after participants had completed the questions. Responding to questions from different questionnaires in one interview may also have affected participant's responses. The questions were therefore completed in random order. It is not likely that the order influenced the responses, as major themes were present in all transcripts.
The large sample size is a strength of this study. In general, sample sizes for studies using cognitive interviews are small, based on the assumption that problems with questionnaires will become apparent in only a small sample [12
]. However, not all problems with a questionnaire may be uncovered in small samples. The findings of Blair et al. show a relationship between sample size and the number of problems with a questionnaire, with the highest number of uncovered problems for a sample size of 50 [14
]. They therefore recommend larger sample sizes.
Where to from here: continue to use self-report sitting-time measures like the IPAQ and PASE sitting-time questions?
Self-report data are more vulnerable to bias and measurement error than objective data, and quantitative studies generally report only poor to fair agreement between self-reported and objectively-measured physical activities [6
] and sedentary time [28
]. A potential explanation by Bauman et al. (2006) for the low validity of self report against objective measures is that questionnaires and objective measures such as accelerometers measure different aspects of physical activity behavior [6
]. Bauman et al. state that sedentary time is the least well-measured type of activity [6
]. Studies such as the current qualitative study that document how people respond to sitting-time questions are needed to understand what problems are occurring during the process of answering these questions, why these problems are occurring and to offer suggestions for improving these questions.
The findings of the present study may help explain the generally low to moderate agreement between self-reported and objectively-measured sitting time. One explanation could be that objective measures such as accelerometers register all sedentary activities, including brief periods that participants are unlikely to recall for self-report measures. Another explanation could be that while accelerometers register sedentary activities in all domains (e.g., leisure, work, transport), self-report measures often assess sitting time in only some domains (e.g., PASE asks about leisure-time sitting only). Finally, accelerometers register all activities with an intensity under a pre-specified cut-point, which could also include lying down and light intensity standing activities, whereas self-report questions ask about sitting specifically. Combined use of accelerometers and inclinometers could potentially overcome this problem and, if so, might be a way to obtain more valid sitting-time data.
Although some surveillance studies have started to include objective measures of sitting and PA, it is likely that researchers will continue to use self-report sitting questions. Self-report questions are convenient and inexpensive to use and therefore suitable for large population studies [6
]. Furthermore, studies have shown that self-reported sitting time is associated with a range of health outcomes [3
]. A key recommendation from a recent review on the measurement of sedentary time was for population-based studies to include both self-report measures and device-based measures [28
]. Given the frequent use of sitting-time questions in these large studies, it is important to be aware of problems with self-report sitting-time questions and to improve these questions to ensure we collect the most accurate self-report sitting-time data possible. It is critical that researchers be aware that different survey instruments claiming to measure sedentary behavior may not be collecting information on the same activities and in the same domains, which should be taken into account in the interpretation of findings from different studies of sedentary behavior and health.