Background Subjective instruments for physical activity assessment in children have particularities, and to assess the reliability of these instruments is crucial. The aim of this study was to determine the validity of the Netherlands Physical Activity Questionnaire (NPAQ).

Methods Population under study was Brazilian children aged 4 to 11 years old. Data collection took place by application of the NPAQ through face-to-face interviews with mothers` children and utilisation of GT1M Actigraph accelerometers by five consecutive days as the reference method. Validity analyses were performed by correlation coefficient, sensibility and specificity, as well ROC curve.

Results Two hundred and thirty nine children participated of the study. The prevalence of physical activity was 73.2%. The mean and median of the NPAQ score were 25.5 and 26, respectively. Based on the area under the ROC curve, the median value presented the best indicators of sensibility (59.4%) and specificity (60.9%), and the area under curve was 0.63. The predictive capacity of the NPAQ to identify active children was high regardless the cut-off point chosen. This capacity was even higher from the score of 30 points on. The correlation coefficients between the last question of the instrument and the accelerometer variables were better than those from the rest of the NPAQ.

Conclusions Based on sensibility and specificity values, the NPAQ did not show satisfactory validity. More results of the predictive capacity of the last question will be better explored. Physical activity level of children estimated from questionnaires must be interpreted with cautions.