Frequent physical activity has been related with a reduction in risk for hospitalization and mortality in patients with chronic obstructive pulmonary disease (COPD).1–3 Consequently, there is a growing interest in measuring physical activity in these patients in order to study its determinants and effects as well as its clinical evaluation. There are several tools used to measure physical activity, and research on their ease of use, validity and reliability for monitoring physical activity in patients with COPD has grown exponentially. Similarly, the use of these monitors in clinical research of physical activity and COPD has increased,4 despite some disagreements about their interpretation.5 Nevertheless, there is a lack of research in COPD about the ease of use, validity or reliability of questionnaires,4 which are widely used instruments for measuring physical activity in other fields of research. Physical activity questionnaires have identified the health benefits of physical activity in numerous diseases and chronic processes. The questionnaires have a generalized applicability, are low-cost, easy to use and are the method of choice for measuring physical activity in large-scale epidemiological studies.6–8 The specific sociodemographic characteristics of COPD patients (seniors, retirees) and their functional limitations impede the use of any questionnaire that has not been previously validated in this population. We selected the Yale Physical Activity Survey (YPAS)9 because it is one of the most detailed, available for senior subjects and whose validity, reliability and sensitivity to change have been previously published.9–11 The YPAS reflects the volume, frequency and intensity of physical activity, expressed as metabolic equivalents of task (MET), meaning MET-h/week, which are able to estimate the effects of physical activity as a continuous parameter even in the lowest levels of activity (predictable in COPD).8

The objective of the present study is to validate the YPAS in COPD patients in order to quantify the levels of physical activity and classify the patients according to these levels using a physical activity monitor (accelerometer) as a reference.

Table 1 demonstrates the characteristics of the sample. The patients used the accelerometer a mean of 6 days and registered a mean of 95% of the daytime hours (13.5h out of a maximum of 14h). A total of 79 (48%) patients did a minimum of 30min of at least moderate physical activity (56%, 45%, 49%, and 40%, in mild, moderate, severe and very severe COPD, respectively). All the physical activity parameters of the questionnaire positively correlated with the parameters of the accelerometer, with Spearman's coefficients that varied from 0.29 to 0.52 (all P<.001; Table 2).

The intraclass correlation coefficients for the time spent performing activity, intensity (MET) and energy expenditure were 0.397, 0.360 and 0.339, respectively, showing that less than half of the variance of the concordance was attributable to the variability of the individuals. The corresponding Bland–Altman plots showed that the mean values were significantly different between the questionnaire and the accelerometer (Fig. 1). There was a wide variability in the agreement; in 45%, 56%, and 57% of the cases it was higher than the previously defined limit for agreement for the time spent in each activity, the intensity and the energy expenditure, respectively.

Figure 1.

Intraclass correlation coefficients (ICC) and Bland–Altman plots of the physical activity parameters of the monitor and the survey in 172 patients with chronic obstructive pulmonary disease (COPD) (the ICC is a proper statistic for evaluating the degree of agreement between measurements of the same construct, and can be conceived as the ratio of the intra-individual variance over total variance).

(0.28MB).

The mean values of the physical activity variables recorded by the monitor increased in accordance with the tertiles of the corresponding parameters in the questionnaire (all the P<0.001) (Fig. 2). Fig. 3 shows that the area under the ROC curve between the summary of the activity according to the questionnaire and sedentarism was 0.71 (95% CI: 0.63–0.79). Table 3 shows three different cut-points with high sensitivity and specificity for identifying the sedentary individuals based on the questionnaire.

Figure 2.

Distribution (diagram) of the physical activity parameters of the accelerometer in accordance with the tertiles of the corresponding parameters of the questionnaire in 172 patients with chronic obstructive pulmonary disease (COPD).

(0.28MB).

Figure 3.

Receiver operating characteristic (ROC) curve of the summary index of activity, determined from the questionnaire, compared with sedentarism, defined as <30min/day of moderate physical activity based on the monitor, in 172 patients with chronic obstructive pulmonary disease (COPD). An area under the curve of 0.71 means that, if two individuals are selected randomly (one above and another under the cut-points of the questionnaire), 71% of the time they will be correctly classified as sedentary/non-sedentary, in accordance with the monitor.

(0.06MB).

Table 4 shows that: (a) being a current smoker and being actively employed led to higher values in the questionnaire than in the accelerometer; and (b) growing levels of activity as defined by the accelerometer were related with a decrease in the differences between instruments. There was no statistically significant interaction or modification of the effect between being a smoker and active worker. Other potential determinants showed no association in the bivariate or multivariate models.

The stratified validation analyses showed very similar results for the ex-smokers and/or retired individuals, while in current smokers (n=47 [28%]) and/or in those actively employed (n=12 [7%]) we observed lower correlations and a poorer classification of the physical activity parameters. The sensitivity analysis, detailed in Section “Methods”, showed very similar levels for all the analyses (data not shown).

The present study evaluated the validity of a self-administered physical activity questionnaire in a COPD population by means of the comparison of the YPAS questionnaire with the SenseWear® Pro2 Armband accelerometer. In order to evaluate the valid properties of the survey, several strategies were used that demonstrated the following: (a) positive correlations that were statistically significant between both tools; (b) a reduced agreement between instruments and a substantial overestimation of physical activity by the questionnaire; (c) statistically significant differences in the parameters of the monitor in accordance with the tertiles of the questionnaire; and (d) satisfactory performance for the identification of sedentary patients.

The correlations between the questionnaire and the monitoring system varied from low to moderate (0.30–0.48) in the sample of patients with COPD in the present study, which corresponds with the values from 0.35 to 0.45 that are predictable based on the studies published in healthy individuals of older age.9,10,23 In addition, it is much greater than the value of 0.14 observed in a small-scale study done in COPD patients which compared the units of movement of the vector of a tri-axial accelerometer and a questionnaire of the activity over the previous 4 days.26 The greatest correlation coefficient was obtained between the survey summary index and the steps taken as measured by the monitor, which are the most widely accepted parameters of each tool.4 As was expected, the 6-min walk distance (an objective parameter for exercise capacity) exhibited slightly higher correlation coefficients with the accelerometer (an objective parameter for physical activity) than with the questionnaire (a subjective parameter of physical activity). Nevertheless, it is interesting to comment that the questionnaire correlated better with the monitor than with the distance walked in 6min, which supports the opinion that physical activity is a concept different from that of capacity for exercise.27

Until now, no studies have ever examined the agreement between the quantitative parameters of the YPAS questionnaire and the parameters of the accelerometer. The data of the present study show statistically significant differences between instruments in the means of the physical activity parameters, all of which are higher in the questionnaire. The ICC estimations imply that half of the variance in the physical activity parameters can be attributed to the tool more than to the individual. In addition, the prevalence of cases that surpassed the previously defined limits of agreement (±30% of the mean estimation of the accelerometer) was around 50%. This great variability could be explained in part by the weakness of the questionnaire. The effect of social desireability28 when responding to a questionnaire (a self-administered instrument) could entail exaggerating socially accepted activities, which would lead to an overestimation of the time dedicated to the physical activity in the survey. Regardless of the cause, these data suggest that the YPAS should not be used for individual quantitative estimations of physical activity parameters.

It is worth mentioning that the analysis of the possible determinants for the differences between the accelerometer and the questionnaire only identified two variables in addition to physical activity. Being a current smoker or being actively employed increased the difference between the two instruments. One possible explanation of the first is that current smokers may over-report their physical activity due to the effect of what is socially desirable. As for the patients who were actively employed, it is possible that the questionnaire requires more direct, clear questions about the degree of physical activity done during work instead of the individual question that we added about activity at work. Lastly, it is reasonable that the levels of physical activity are also independently related with the differences between the questionnaire and the monitor, with fewer differences in the lower levels of physical activity. As a whole, and, although the variance in the differences of physical activity explained by these variables is lower than 30%, this analysis suggests that, in future research on the Yale survey in COPD patients, adjustments could be necessary for these specific populations.

The present study is the first that examines the capacity of the YPAS for discriminating physical activity among groups and for identifying sedentary individuals. As the health benefits of physical activity have been evaluated in general with physical activity in ordinal categories more than as a continuous variable,20 the results of the present study regarding the increase in mean accelerometer values in accordance with the tertiles of the questionnaire is exceedingly important. It is not known to what point the selection of cut-points (either statistically or clinically pertinent) could influence in these results. However, and given that the degrees of physical activity change through studies and questionnaires, the use of categories based on percentiles is a common practice. For the identification of sedentarism, the authors used an accepted cut-point,20 and the questionnaire seemed useful for identifying this circumstance. As long as COPD patients have a lower degree of physical activity than healthy adults of the same age29—related with their poorer state of health and poorer outcome30—the questionnaire could contribute to the early detection and treatment of sedentarism in clinical practice. From the several cut-points obtained in the analysis of the ROC curve with high sensitivity and specificity, it is proposed to use a cut-point score of 51 in the summary index of the YPAS questionnaire for the identification of sedentary patients. The reason for this choice is that the authors gave priority to the sensitivity over the specificity in the identification of sedentary individuals given the absence of harmful psychological effects associated with the diagnosis of sedentarism and the lack of adverse effects related with the recommendation of performing physical activity.

One limitation of the present study is that it does not evaluate reliability. Previous studies have demonstrated that the YPAS has an appropriate reliability in healthy individuals of older age (ICC=0.65), which could be very similar in the case of patients with COPD.10 Second, the calculation of the energy expenditure in the questionnaire could have led to an incorrect estimation of the energy expenditure in the COPD population due to the use of absolute intensities (MET values) derived from healthy individuals.22 In addition, the use of a physical activity monitor as a reference for energy expenditure could be debatable because the application of the reference algorithm of the monitor could translate into an erroneous estimation of physical activity.31 Other parameters, like the doubly labeled water method (DLW), provide more valid information about energy expenditure.7 Unfortunately, the DLW is limited to experimental studies due to its complexity, the material necessary and cost. The authors considered that the SenseWear® accelerometer is an appropriate instrument for validating the questionnaire because previous studies have demonstrated an appropriate agreement (ICC 0.54–0.73) in the physical activity parameters of its measurements and those using indirect calorimeter.32 It has been previously validated for low-intensity activities in COPD pacients,18 and it has been demonstrated that, for senior patients with low cultural levels, it is inoffensive and appropriate.15 Lastly, the sample of the present study included mostly men, which represents the actual distribution of sexes in COPD in Spain. Although it is probable that the degree of physical activity may be different in the sample of the present study than in others with a different male:female proportion, it is highly unlikely that this were to cause bias in evaluating the validity of the questionnaire.

It could be debated whether a validated questionnaire for physical activity in COPD patients is necessary or not, given the recent technological progress made in monitoring the activity of this population. Several authors coincide in affirming that the selection of the tool for measuring physical activity in COPD patients depends on the characteristics and the objectives of the research.4,5,19 Without belittling the relevance of monitoring activity, questionnaires (if these are validated and reliable) are still the most cost-efficient instruments for epidemiological research and control.33 This could also be the reason for discussion if the instruments that monitor physical activity provide data comparable to the self-evaluated experience of physical activity in COPD patients, which is unknown due to the lack of a conceptual physical activity framework in this population. One of the main advantages is the use of agreement and discrimination parameters to quantify and classify physical activity in COPD patients in addition to the generally used exclusive correlation analysis. An additional advantage comes from the use of the accelerometer for a mean of 6 days as a reference because recent studies have demonstrated systematic changes during the week19 that otherwise would have only been covered by the questionnaire. A final, novel advantage is that, in order to improve the results obtained with the questionnaire, we included patient body weight in the calculation of energy expenditure based on physical activity instead of assuming that all the subjects weighed 60kg, which is the usual strategy.9

We conclude that the Yale Physical Activity Survey is a valid tool for classifying the physical activity in patients with COPD, but not for quantifying it. Its use in large-scale epidemiological studies, including either COPD patients or patients with similar chronic processes, will provide an adequate and cost-efficient estimation of the effects of physical activity levels in these patients. Given that the summary index of the questionnaire demonstrated adequate capacity for discriminating sedentarism, the use of the series of questions that make up this index (which is a very small part of the questionnaire) should be considered a screening instrument for identifying patients who run the risk of sedentarism.

The present study was funded by the Health-Care Research Fund [FIS PI052292], the Spanish Society of Pulmonology and Thoracic Surgery [SEPAR 2004/136] and the College of Physiotherapists of Catalonia [R02/08-09]. Judith Garcia-Aymerich has a research grant from the Instituto de Salud Carlos III [CP05/00118], Ministry of Health, Spain. The PAC-COPD study is funded with grants from: the Health-Care Research Fund [FIS PI020541], Ministry of Health, Spain; the Agència d’Avaluació de Tecnologia i Recerca Mèdiques [AATRM 035/20/02], Catalonian regional government; Spanish Society of Pulmonology and Thoracic Surgery [SEPAR 2002/137]; Catalonian Pulmonology Foundation [FUCAP 2003 Beca Marià Ravà]; RESPIRA Network [RTIC C03/11]; RCESP Network[RTIC C03/09]; Health-Care Research Fund [PI052486]; Health-Care Research Fund [PI052302]; La Marató de TV3 Foundation [num. 041110]; DURSI [2005SGR00392]; Novartis Pharmaceuticals, Spain; AstraZeneca Pharmaceuticals, Spain. CIBERESP and CIBERES receive funding from the Instituto de Salud Carlos III, Ministry of Health, Spain. The sources of funding did not participate in either the study design, data collection, data analysis or interpretation, the drafting of the manuscript or the decision to present the article for publication. The researchers are independent from the funding sources.