Implementation of a structured training program for retrospective video analysis of Parkinson's disease

• Main Author: Eden, Gabrielle Marie
• Other Authors: Thomas, Kevin C.
• Language: en_US
• Published: 2019
• Subjects: Medical imaging; Structured training program; Training methods
• Online Access: https://hdl.handle.net/2144/34883

INTRODUCTION: Retrospective video analysis (RVA) has been a popular method of analysis in many research fields, evidenced by autism behavioral research, child play behavior, and caregiver-resident interactions (Baranek, 1999; Gilchrist et al., 2018; Gilmore-Bykovskyi, 2015). Given the widespread use of RVA, the number of studies using it to augment their study designs provide sparse details about the training methods for this level of analysis, making it difficult to maintain a standard level of rigor across different institutions (Haidet, Tate, Divirgilio-Thomas, Kolanowski, & Happ, 2009). METHODS: A structured training program was developed for naïve coders (n=5). The structured training program was composed of five stages with careful introduction of behaviors and regular checkpoints. Statistical Analysis: The output generated by the naïve trainees was analyzed with paired t-tests, Fisher’s Exact Test, ANOVA, percent agreement, and Cohen’s kappa. RESULTS: No difference was found between the different trainees, demonstrating the trainees were trained to a similar level of expertise. The overall recognition of behaviors increased by 2.1% from the first to last training video analysis. Discrete behaviors had a higher level of agreement. CONCLUSIONS: The structured training program demonstrated a small increase in the recognition of behaviors, with a higher recognition in the derived MDS-UPDRS behaviors.