2.2. OHAIRE coding procedure

Five sessions (sessions 1,7,12,18, and 23) were video recorded to capture behavior observation data. These sessions were selected to maximize the total number of participants present during the video recorded sessions and to represent sessions throughout the entirety of the study. Data extraction replicated the OHAIRE Coding System (25) where 10-minute video segments were divided into thirds and 1-minute segments were randomly selected from each of those segments. Therefore, three minutes from each of the selected sessions were randomly selected for behavioral coding for each participant. The OHAIRE coding system was specifically designed for human-animal interaction research projects and demonstrates good reliability and validity (21).

Two research assistants coded the behavior of children with ADHD and their peers. Coders were blinded to the aims and hypotheses of the study, but due to the nature of the study (presence of the dog vs. no dog) raters were not blind to the condition. Coder 1 coded 98% of the data (the 3 minutes that were not coded were dropped from the analysis) and coder 2 coded 23% of the data to establish interrater reliability. Interrater reliability was calculated using Cohen’s Kappa. The overall agreement among raters was 86.5% (k = .865, p<.001). Interrater reliability was also calculated for specific categories of interactions (k = .736, p<.001), facial emotional displays (k = .756, p<0.001), verbal valence (k = 0.98, p<0.001), social communication (k = .655, p<0.001), and problem behaviors (k = .894, p<0.001).

Across the study period’s five assessment sessions, a total of 322 minutes were selected for coding (intervention group: 173 minutes, control group: 149 minutes). The average per child was 9.2 minutes (SD: 2.66, Range: 3-13 minutes). Following the OHAIRE Coding System, coders rated the absence or presence of a behavior in 10 second intervals. The score for each behavior code is the number of 10-second intervals it was present within a minute (i.e., a count ranging from 0-6). Summary scores were created for the following domains: Animal Social Interaction and Human Social Interaction (further separated into Human-Adult Social Interaction and Human-Peer Social Interaction). Each summary score was the sum of the six “interactions” items towards the relevant target, resulting in variables with a possible range of 0-36.