Using Informatics and Natural Language Processing to Develop a Clinical Decision Support Tool to Improve Intimate Partner Violence Diagnosis by Dentists and Assess its Feasibility and Acceptability

Dentists may be the first and only care providers for intimate partner violence (IPV) survivors, as a significant percentage of IPV incidents involve injuries to the face/mouth/teeth. IPV has severe consequences for survivors, impacting their physical, mental, and oral health, and leading to long-term societal and economic repercussions. Despite recommendations from the American Dental Association, dentists face barriers such as limited time, training hindering their ability to effectively screen for IPV. Evidence shows that 2-50% of care providers screen for IPV. To streamline IPV screening for dentists, Dr. Banava will employ state-of-the-art machine learning techniques to develop a clinical decision support to deliver the right information, to the right person, in the right intervention format, through the right channel, and at the right time within the workflow provide ("five rights" approach). Using an integrated data-driven IPV-specific clinical decision support tool will improve patient outcomes, and reduce missed cases. The specific aims are: Aim 1: Identify Prevalence, Patterns and Correlates of IPV-related Orofacial Injuries and Conditions in Electronic Health Records. Dr. Banava will leverage standard IPV-related orofacial injury and condition terms to conduct a comprehensive search within the UCSF structured electronic health record (EHR) database Epic (APeX), which encompasses electronic dental records as well. Next, Dr. Banava will extract embedded IPV-related data from notes, manually annotate, and validate with patient chart review and actual diagnosis (gold standard). Later, Dr. Banava will apply NLP techniques, such as Named Entity Recognition, to annotate the same set of notes and evaluate their precision, recall, and F1 score. Additionally, Dr. Banava will leverage advanced Large Language Models (LLMs) within the NLP framework to further enhance the annotation process. Aim 2: Develop an IPV-Specific Clinical Decision Support Tool and Assess its Feasibility and Acceptability. Dr. Banava will collaborate with the UCSF APeX Enabled Research team to develop a prototype of an IPV-specific clinical decision support tool, leveraging generated data. Adopting a human-centered design approach, Dr. Banava will engage domain experts and conduct iterative focus groups to ensure the tool's user-friendliness, alignment with clinical workflows, and effectiveness in decision-making and case management. The focus groups, following a mixed methods research design, will take place during both the development and post-development phases of the project. Additional groups may be included if information saturation is not achieved. To evaluate the feasibility and acceptability of the developed tool, Dr. Banava will use the Technology Acceptance Model and incorporate prompts accordingly. A comparison will be made between the tool and a standard IPV screening approach. The primary objective of the K23 Award is to provide Dr. Banava with structured training in advanced quantitative statistical analyses and modeling techniques, health informatics, natural language processing, and mixed methods research. This award will empower Dr. Banava to bridge knowledge gaps and streamline IPV screening for dentists, ultimately leading to improved patient outcomes.