Comparison of IPAQ‐SF Self‐Reported and Fitbit‐Measured Physical Activity in Young People With Haemophilia

Dear Editor: It is well known that physical activity (PA) can provide several health benefits to people with haemophilia (PWH), and regular PA is therefore encouraged [1]. Largely, the same PA recommendations as for the general population apply also to this patient group [2, 3]. In order to obtain information about PA levels, such as mapping, tracking of changes, and classifying people as being sufficiently physically active or not, some sort of measurement is required. Self-reported measurements by questionnaires are commonly used for this purpose, and a systematic review from 2021 highlighted that, up until then, the majority of investigations into PA among PWH had utilised self-report methods [4]. However, self-reported PA is subject to recall- and potential social desirability bias, as well as measurement imprecision compared to more objective assessments [5]. The International Physical Activity Questionnaire Short Form (IPAQ-SF) is a frequently utilised questionnaire, even though its correlation with objective measures of PA are mostly below the acceptable threshold [6]. Objective measurements (e.g., via use of validated wearable devices/activity trackers paired with standardised protocols) are potentially more accurate, and therefore recommended for more precise PA estimates [7]. Nonetheless, we chose to include IPAQ-SF as an assessment method in our study to enable comparison to an activity tracker, and to allow comparison to results from other studies where this questionnaire has been utilised. In the current study, we aimed to compare IPAQ-SF self-reported and Fitbit-measured PA in PWH. For this purpose, we used data collected through the ‘HemFitbit study’ (ClinicalTrials.gov NCT04181697), where PA data was captured using both self-report and activity trackers. Ethical approval was obtained from the Regional Committee for Medical and Health Research Ethics South East and the Oslo University Hospital Data Protection Officer. All participants (and/or guardians) provided informed consent to participate. Forty PWH A, aged 13–30 years, with moderate or severe haemophilia A on continuous prophylaxis and without inhibitors were included. PA outcomes were average daily minutes of; light (LPA), moderate (MPA), vigorous (VPA), moderate-to-vigorous (MVPA = MPA + VPA) and total (PA = LPA + MPA + VPA). For self-reported PA, the Norwegian version of the IPAQ-SF was used. As prompted in the questionnaire, participants reported time spent in VPA, MPA, and walking, of at least 10-minutes duration, for the last seven days. Because the IPAQ-SF does not have a specific question for LPA, we defined walking as LPA for our comparison. Participants responding ‘don't know/unsure’ to questions were excluded. Objective PA outcomes were collected using the activity tracker Fitbit Charge 3 (Fitbit Inc, San Francisco, California, USA) over a 12-week measurement period for each participant. Invalid days, defined as ≤1,000 steps/day, were excluded (more details about the Fitbit measurements can be found in a previous publication: [8]). One overall daily average per PA outcome was computed per participant for IPAQ-SF and Fitbit based on valid responses/days, in order to get an estimate of habitual daily PA. Statistical analyses were conducted using Stata version 18.0 (StataCorp LLC, College Station, TX, USA). Except for the age variable, data were non-normally distributed. For continuous variables, we used descriptive statistics (mean and standard deviation (SD) for normally distributed variables and medians and interquartile ranges (IQR) for non-normally distributed variables), Spearman's correlation and Bland-Altman plots with 95% limits of agreement (LoA) (2.fifth centile and 97.fifth centile) to compare IPAQ-SF self-reported and Fitbit-measured PA. If there were complete agreement between the two methods, all data points would lie at a horizontal line at zero. Linear regression models were used to test whether the slope of the Bland-Altman regression was significantly different from zero. A p-value of <0.05 was considered statistically significant. We also investigated agreement between the two PA measurement methods in classifying the participants as meeting the World Health Organization (WHO) PA recommendations for MVPA per week (yes/no), with the criteria being 420 min per week (60 min per day) for teenagers and 150 min per week for adults [3]. Weekly minutes of MVPA was calculated by multiplying the daily average of minutes MVPA by seven, to obtain an estimate reflecting habitual time per week. We summarised the results in a 2 × 2 table reporting numbers and percentages of agreement (yes/yes in both measures or no/no in both measures) and disagreement (yes/no or no/yes in the two measures). We obtained valid IPAQ-SF and Fitbit results for 33 participants, whereof 17 were teenagers (age 13–17) and 16 adults (age 18–30). The mean age was 19.5 (SD 6.0) years. All participants were on continuous prophylaxis with a factor replacement product, and most had low joint scores (limited joint problems). Results for time spent in different intensities in median (IQR) minutes per day for the PA variables for the two measurement methods and median differences are presented in Table 1. The differences were particularly large for LPA and total PA. Correlation analyses showed that the correlations between the continuous PA measurements for the two methods were low for all variables (Figure 1A). Bland-Altman plots showed a measurement bias between the two methods for all parameters with wide LoA and an increasing underestimation by IPAQ-SF compared to Fitbit with higher volumes of PA (Figure 1B). IPAQ-SF and Fitbit agreed in their allocation as meeting or not meeting the WHO PA recommendations for 18 (54%) participants (8 agreements on not meeting, 10 agreements on meeting; Table 2). Our results show lack of correlations and large biases in measurement agreement, suggesting that PA assessed by IPAQ-SF and Fitbit are not comparable, with consistent lower estimates of PA by questionnaire compared to Fitbit. There is also a large discrepancy between the methods in the classification of whether participants meet WHO recommendations for weekly time spent in MVPA. Given that this study was designed to address PA assessment in a modern and well-managed haemophilia population, for whom prophylaxis is standard of care, our findings are therefore most applicable to this population, and extrapolation to those with greater joint- and functional impairment should be made with caution. Since IPAQ-SF does not have a category for LPA, but only for walking, we classified time reported walking as minutes of LPA. It can, however, be argued that this is not comparable to Fitbit-measured LPA. Nonetheless, walking is the only estimate available from the IPAQ-SF that could serve as an estimate of LPA. This important difference is a likely explanation for the discrepant results for this outcome, which also affects total PA. Another probable explanation for the discrepant findings is the previously mentioned well documented limited agreement between self-reported and device-based PA measures [6]. Another potential source of ‘measurement error’ and explanation of the differences between the measurement methods, with a seemingly overall underestimation of PA from the IPAQ-SF, is that the questionnaire requests the participant to report only PA ‘sessions’ of at least 10-minute duration. This does naturally not give an actual representation of all PA, which includes also short-duration tasks such as walking around at home, climbing the stairs, doing household-chores, running a short distance to catch the bus, etc. The Fitbit on the other hand records any movement, even very short bouts. The formulation of the questions are likely based on the previous WHO PA recommendations (from 2010), which required 10-minute bouts of PA for it to ‘count’ and being considered relevant to health [9]. However, based on recent research which shows that all PA matters, the bout requirement was removed from the latest (2020) guidelines [3]. Hence, one can argue that this makes the questionnaire outdated. Another factor which may impact the PA estimates of the two methods we have investigated here is that Fitbit measurements were collected over a 12-week period, whilst IPAQ-SF only asks about the last seven days. As such, the habitual daily PA estimates were not based on corresponding days nor the same number of measurement days. With the calculation of one overall daily average for each PA parameter per method, we tried to make the comparison as valid as possible. For Fitbit, the 12-week period was chosen to capture PA habits as well as possible (primarily measured for another part of this study, where we found that PA levels over the 12 weeks were stable [8]). For IPAQ-SF, however, the questionnaire only evaluates one week, and this is not necessarily representative for an individual's habitual PA. Still, the intention of the questionnaire, based on its wording (including for example the phrase “physical activities that people do as part of their everyday lives” and questions with the word “usually”), seems to be to assess PA habits [10]. Hence, we found it relevant to compare estimated daily habitual PA in our study, via the outlined approach. Marked discrepancies were observed between the two measurement instruments in classifying participants as meeting or not meeting the recommended weekly levels of PA. The overall discordance (lack of agreement between methods) of 46% is substantial and raises concerns regarding the accuracy of both instruments. Such misclassification may have important implications, as it could result in individuals erroneously perceiving themselves as compliant with PA recommendations when they are not, or, conversely, to underestimate their adherence. Of note is, however, that dichotomising a continuous variable reduces the richness of the data and does not present nuances, which could be particularly important for individuals near the cut-off. Nonetheless, this dichotomisation is the most commonly utilised approach in the literature and in public health surveillance, and we therefore adopted this method to maximise comparability with existing studies and to facilitate interpretation relative to WHO guidelines. Finally, we do acknowledge that Fitbit activity trackers are not a ‘gold standard’ for PA measurements, and among PWH, we recently demonstrated that the Fitbit Charge 3 overestimated steps and minutes of LPA, VPA and MVPA compared to the research-grade accelerometer ActiGraph GT3X-BT [11]. However, the Fitbit still provides more objective measurements than self-report assessment of PA, alike several other activity trackers, whether research- or consumer-grade devices. In conclusion, we find that it is important for researchers to keep in mind that self-reported PA is often not ideal. Currently, several more accurate measurement options are available, which are convenient to use in smaller populations, as is often the case in research involving PWH. Our findings presented here emphasise the recommendations of using objective measurement methods such as activity trackers to better estimate PA levels, which is important to form a basis for targeted PA advice for PWH, as well as others. Which type of device to use is, however, a remaining question, and the answer will depend on factors such as purpose, setting and population. REDM, PAH, HG, CSR and MG designed the study. REDM collected the clinical data. REDM analysed the data. REDM drafted the manuscript. All authors contributed to data interpretation and manuscript revision. All authors approved the final version. The HemFitbit-study is financially supported by an unrestricted research grant from Bayer HealthCare. We thank the participants for their time and effort and all clinicians who aided data collection. PAH and REDM have received funding to institution from Bayer HealthCare for research carried out in this work. HG, CSR and MG stated that they had no interests which may be perceived as posing a conflict or bias. This study is part of the ‘HemFitbit study’ and has received ethical approval from the Regional Committee for Medical and Health Research Ethics South East (2019/549). The study was conducted in accordance with the Declaration of Helsinki. The authors declare no conflicts of interest. The data supporting the study findings are not publicly available due to privacy or ethical restrictions. Data may be made available upon reasonable request to the corresponding author.