216 - Utilizing Natural Language Processing to Objectively Characterize Dental Behavior

Purpose: The Frankl Behavior Rating Scale (FBRS) is one of the most used behavior scales in pediatric dentistry. The FBRS has general guidelines to assess patient behavior but can be subjective when not applied as intended. Natural language processing (NLP) is an area of research and application that explores how computers can be used to understand and manipulate natural language text to do useful things.​ The purpose of our study aims to evaluate the current method of assigning behavior scores at Nationwide Children’s Hospital (NCH) and to utilize NLP to assign behavior scores more objectively and accurately to our pediatric dental patients.

Methods: The study was completed in 3 parts: (1) a video survey administered to dental faculty & residents, (2) manual annotation of behavior note text, and (3) adjudication and analysis of the annotation protocol.

Results: Dental faculty and residents yield moderate reliability to identify the true FBRS score when shown video footage of the dental encounter. Providers largely do not agree with one another when evaluating dental behavior. Manual annotation of behavior notes suggest strong reliability and substantial agreement among independent annotators.

Conclusion: Limited quantifiable data exists to support the continued use of the FBRS. The issue may not be the fault of the scale itself but rather how providers are currently applying it. This study describes an alternative method of using NLP to supplement the FBRS to create a more objective, accurate, and consistent scale. More research needs to be completed to substantiate this claim.