Intra and interobserver agreement of the Dynamic Imaging Grade of Swallowing Toxicity Scale (DIGEST) in fiberoptic endoscopic evaluation of swallowing (FEES): the importance of observer-tailored training

Abstract

Purpose The Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) is a scale to quantify the severity of pharyngeal dysphagia in head and neck cancer (HNC) patients. This study (1) described the training process of the observers for DIGEST in fiberoptic endoscopic evaluation of swallowing (FEES), (2) determined observer agreement on the DIGEST in FEES, (3) explored the effect of bolus consistency on observer agreement, and 4) explored criterion validity of the DIGEST in FEES. Methods Twenty-seven dysphagic HNC patients were enrolled. Two observers completed a training program for DIGEST in FEES. Observer agreement on the Penetration-Aspiration Scale (PAS), percentage of pharyngeal residue (PPR), and DIGEST grades was determined using linearly weighted Cohen’s kappa coefficient (κ). Results Due to insufficient observer agreement after the first measurement attempt, additional training was organized using an elaborated manual with descriptions of the visuoperceptual variables, thereby improving observer agreement. Intraobserver agreement was almost perfect on the PAS (κ = 0.86–0.88) and PPR (κ = 0.84–0.86). Interobserver agreement was substantial on the PAS (κ = 0.78), almost perfect on the PPR (κ = 0.82), substantial on the safety grade (κ = 0.64), almost perfect on the efficiency grade (κ = 0.85), and substantial on the summary grade (κ = 0.71). Bolus consistency had an effect on observer agreement. A significant correlation was found between DIGEST efficiency grade and EAT-10. Conclusion The DIGEST showed to be a reproducible measurement for FEES in terms of observer agreement. However, agreement between novice observers on the DIGEST was only reached after specific observer-tailored training. Observer agreement should be analyzed by taking bolus consistency into account during training, as this might affect the interpretation of the outcome. A manual with well-defined descriptions can optimize the reproducibility of DIGEST measurements.