Reliability and Validity of Environmental Audits Using GigaPan Technology in Parks
Published in: Preventive Medicine Reports, Volume 13, pages 293-297 (March 2019). doi: 10.1016/j.pmedr.2019.01.011
Posted on RAND.org on June 10, 2019
Park quality and features can contribute to more engaging places for play and recreation. However, assessing park characteristics remains a challenge. This study measured the reliability of GigaPan as a method for assessing park characteristics as well as the validity of GigaPan compared to Google Street View (GSV) and direct observation (DO). A total of 65 target areas (16 parks total) in Pittsburgh, PA were assessed using GigaPan, GSV, and DO from July 2015-January 2016. For reliability and validity, 14 and 28 variables were examined, respectively. Cohen's kappa was used to assess inter-rater reliability. Sensitivity and specificity were used to measure validity. Of the 14 variables included in the inter-rater reliability analysis, five variables had almost perfect reliability (kappa > 0.80) and three variables had substantial reliability (kappa > 0.60). Of the 28 variables included in the validity analysis, GigaPan was able to correctly classify 17 of the 28 variables and GSV was able to correctly classify 15 of the 28 variables with a sensitivity >80%. There were no significant differences between sensitivity and specificity between GSV and GigaPan. GigaPan performed similarly to GSV with DO being used as the gold standard. Further, GigaPan overall had high reliability among the features measured. A strength of GigaPan is the ability to be implemented quickly in the field, making it a viable alternative to GSV particularly when temporality is an important factor.