Reliability Test of an Established Pedestrian Environment Audit in Rural Settings

Abstract

To test inter-rater reliability and internal consistency of a reduced-item version of an environmental audit tool for physical-activity resources in a rural setting. A cross-sectional design. Five rural school districts in western New York. A random sample of 391 street segments (i.e., the length of a street contained within a census block.) Fifty street segments were observed a second time for a reliability analysis. The Pedestrian Environmental Data Scan (PEDS), a previously tested instrument for conducting audits of the built environment for physical-activity supports, was used. Modifications were made for use in rural settings. Four tests of reliability (Kappa [κ]; prevalence-adjusted, bias-adjusted κ [PABAK]; percent agreement; and Spearman rho [ρ]) and one test of internal consistency (Cronbach α) were used. Nineteen measures were identified for their utility in the rural setting. These measures had an average κ of .43, an average PABAK of .72, and an average percent agreement of 86.2%, indicating good agreement. They were organized into four subscale areas that had Spearman ρ ranging from moderate (ρ = .38) to strong (ρ = .66). Alpha (α) values for the scale overall (α = .60) and subscales (range, .25-.66) suggested that items included were likely to be multi-dimensional. A simpler audit tool has been identified that might be more appropriate for use in the rural setting.