The PASE total score demonstrated acceptable reliability for men but not for women and low agreement for both men and women. When compared to accelerometry measures, PASE total score had inadequate validity. The total amount of physical activity, as assessed with accelerometry, was greater in women than in men, but this was not confirmed by the PASE total scores.
There was no evidence of gender-related differences for the PASE total score, but men tended to have higher scores than women, mainly because of higher work-related activity scores. In fact, more men than women in our group of patients were still working part- or full-time. The PASE total scores of our TKA patients were comparable to those previously obtained in healthy populations
[15, 16, 21] and in patients with knee pain and disability
. The relatively high PASE scores we observed could be explained, at least in part, by the fact that most TKA patients begin a formal exercise prescription program after surgery and usually maintain it afterwards. In contrast, Tsonga et al. reported a mean PASE score of 68 in Greek women 6 months after TKA. These scores are much lower compared to our patients, likely because the group of women evaluated in this previous study was older (73 years), had a greater BMI (30 kg∙m-2) and was recruited early after surgery compared to the patients considered in our study.
No systematic bias was found for the PASE total score between the two test sessions, despite a substantial reduction of approximately 10% from test to retest. Personality traits, social desirability and social approval were recognized to be possible sources of systematic bias
. Since TKA patients are encouraged by doctors and therapists to exercise, we suppose that patients in this study tended to overestimate their physical activity to attain social approval and desirability of the investigator at the first session
. In contrast, knowing that the PASE total score referred to the week objectively assessed by the accelerometer, patients committed to report their physical activities more precisely and truthfully at the second session. This was particularly reflected in the recreational activity sub-score which is dependent on the reported weekly frequency and daily duration, and this resulted in a significantly lower recreational activity sub-score for women.
The PASE total score showed acceptable reliability for men but not for women according to the predefined threshold proposed by Terwee et al.. These findings are comparable to those obtained in previous studies investigating the reliability of PASE questionnaire in healthy and pathological elderly populations
[14–16, 20] The high mean ICC of 0.91 reported by Dinger et al. in a rural elderly community can be explained by a substantial difference in the study design compared to the present investigation. In fact, questionnaire administrations were separated by only 3 days and referred to the same week in the study of Dinger et al., while in the current study they referred to two consecutive but different weeks. These reliability results indicate an acceptable but moderate ability of the PASE questionnaire to discriminate between male patients following TKA according to their PAL, but not between female patients.
The PASE total score demonstrated low agreement for both men and women due to the large measurement error, with SEM values similar to those recently reported in knee osteoarthritis patients
. The smallest detectable change represents the smallest change in the questionnaire score that can be interpreted as a “real change” above measurement error
. In order to detect real changes in the questionnaire score after an intervention or over time, the smallest detectable change should be smaller than the minimal clinically important change
. Unfortunately, the minimal PASE score change that can be considered clinically relevant in TKA patients is not known. However, the acceptability of PASE agreement results can be interpreted using the noise-to-signal ratio, which can be expressed as an effect size (change in score divided by SEM)
. Tsonga et al. observed an increase of 18% in the PASE total score of female patients between 3 and 6 months after TKA, while an average PASE total score SEM of 34% was observed in the present study. Interpreting the signal as the change in score (18%), the noise we obtained (34%) was almost 2 times larger than the signal. Since the noise has to be smaller than the changes in score, this suggests that the PASE questionnaire has a very limited ability to distinguish measurement error from real changes. Hence, the PASE questionnaire is not suitable for longitudinal monitoring of the PAL in TKA patients of both sexes.
The PASE total score showed inadequate validity in both men and women. Indeed, although the association between the PASE total score and aPAL was significant for men, the correlation coefficients did not reach the defined threshold of r ≥ 0.50 proposed by Terwee et al.. In previous studies, the correlations between the PASE total score and aPAL ranged between 0.43 and 0.52 in healthy elderly people
[15, 17, 21] and was 0.30 in patients with hip osteoarthritis
. There is no consensus on how high correlations should be to demonstrate adequate validity
. Previous studies rated the PASE questionnaire to be valid with significant but lower correlations with respect to those obtained in the present study
[14–16, 18, 20]. We chose to use the more conservative thresholds because the worse the measurement properties are, the higher the risk is for misclassification and biased results
In line with our findings, some studies reported better validity of the PASE total score for healthy elderly men compared to their female counterparts
[16, 17, 19]. This was probably associated to household-related activities, which are generally more frequently performed by women than men
[16, 19]. However, such an assumption is not supported by our observations as household activity sub-scores did not differ significantly between men and women. We nevertheless acknowledge that household activities might be a potential source of misjudgment for PASE questionnaire scoring, since, contrary to recreational and work-related activities, household activities provide a fixed score without considering frequency and duration.
The PASE total score is the standard outcome for this questionnaire. However, for explorative purposes, we determined two PASE intensity sub-scores. Indeed, differentiating patients by the time spent in low and moderate-to-high intensity activities may be important for investigations focusing on the dose–response effects of physical activity. However, unfortunately, we found no significant correlations between the time spent in low or moderate-to-high intensity activities and the respective PASE sub-scores. Therefore, our results, together with those of Svege et al., seem to confirm that the PASE has no ability to differentiate physical activity intensities.
The PASE total score failed to detect the gender-related differences in the total amount of physical activity reported by the accelerometer (women > men). Questionnaires have several limitations associated with recall and reporting bias; they overestimate the time spent on strenuous activities and underestimate activities lasting less than 10 min or with a level of exertion lower than brisk walking
. Therefore, we suppose that the mismatch between accelerometry and PASE total score lies in the ability of the PASE questionnaire to assess low intensity activities, since the reported higher aPAL of women is explained by the fact that they spent significantly more time on low intensity activities than men. Therefore, the lack of gender differences in the PASE total score further weaken its construct validity.
The main limitation of this study is the design used to evaluate the reproducibility of the PASE questionnaire. Since the physical activity level was assessed for reproducibility during two consecutive but different weeks, it cannot be ensured that TKA patients maintained the same physical activity level during this time period. The repeated assessment of the physical activity level referring to the same week would have been methodologically more appropriate. However, a time period of less than a week between two questionnaire administrations would have increased the risk of recall bias
. In addition, accelerometer wearing could have influenced the reproducibility results, by inducing the TKA patients to be more active during the second week. On the other side, we explicitly asked the patients to maintain their usual physical activity habits during this time period. Two patients demonstrated disparate physical activity habits between the two weeks because of holidays and were thus excluded from reproducibility analyses thereby reducing the statistical power.