Development and validation of the Brunel lifestyle physical activity questionnaire

Abstract

The purpose of the present programme was to develop and validate a theoretically-grounded instrument to measure the planned and unplanned dimensions of lifestyle PA (PPA and UPA; Dunn, Andersen, & Jakicic, 1998). In Study 1, two samples of British adults (Internet: N = 742; paper: N = 563) were used to establish the content validity of the Brunel Lifestyle Physical Activity Questionnaire (BLPAQ). Exploratory factor analysis yielded a two-factor model (UPA and PPA) that produced acceptable fit indices using confirmatory factors analyses with both samples. The purpose of Study 2 was to examine the test-retest reliability of the BLPAQ over 5 weeks using a sample of leisure centre users, university staff members, and university students (N = 337). High correlations were observed between the two administrations (range = .93-.98; p < .01). Thereafter, the data were subjected to proportion of agreement (PoA) analysis as advocated by Nevill, Lane, Kilgor, Bowes, and Whyte (2001). Both PPA and UPA demonstrated satisfactorily high internal agreement (PoA > 95%). In Study 3, the BLPAQ was cross-validated using two criterion measures: the Baecke Questionnaire of Habitual Physical Activity (Baecke, Burema, & Frijters, 1982) and the Godin’s Leisure-Time Exercise Questionnaire (Godin & Shephard, 1985). Multiple linear regressions were performed to predict PPA and UPA from the subscales of the two reference measures. The predictive models differed markedly in terms of gender. Subsequently, the sample of 338 British adults was divided into two subsamples, and these were subjected to a cross-validation using the Limits of Agreement (LoA) methodology advocated by Bland and Altman (1986). The agreement plots revealed that both BLPAQ subscales demonstrated acceptable inter-sample agreement when compared to the criterion measures. In Study 4, a series of structural equation models were tested with the aim of predicting PPA and UPA using the variables that constitute the Theory of Planned Behaviour (TPB). The TPB was able to predict PPA but not UPA. The addition of a direct path between past behaviour to UPA did not result in a significant prediction. Further work is required to examine the factorial structure of the PPA subscale and to increase the number of items in the UPA subscale. In sum, the programme has contributed a valid and reliable theory-based measure of PA as well as evidence to support the utility of the TPB in PA research. However, the TPB framework may require the addition of predictors such as past behaviour and actual behavioural control.