Natural Language Processing of Electronic Dental Records for Population Health, Quality Indicators, and Clinical Decision Making
Access status:
USyd Access
Type
ThesisThesis type
Doctor of PhilosophyAuthor/s
Pethani, FarhanaAbstract
Oral diseases are among the most prevalent non-communicable diseases that disproportionately affect the socioeconomically disadvantaged. Electronic dental records (EDRs) may contain useful information to address these issues at the population, practice, and patient levels, but much ...
See moreOral diseases are among the most prevalent non-communicable diseases that disproportionately affect the socioeconomically disadvantaged. Electronic dental records (EDRs) may contain useful information to address these issues at the population, practice, and patient levels, but much of the useful patient information is stored as clinical notes. While natural language processing (NLP) methods offer potential value for dentistry, tools developed for other healthcare areas may be unsuitable for dental applications due to distinct clinical workflows, vocabulary, and EDR structures. There are, however, major knowledge gaps in the use of NLP methods in dentistry. In this thesis, I aimed to determine whether NLP methods applied to clinical notes in EDRs could be used to support population health, quality indicators, and clinical decision making. The objectives were to identify vulnerable subpopulations based on the prevalence of social determinants of health, and to understand and predict patient returns for complications following a dental extraction. The research comprised a systematic review of the use of NLP methods in dentistry followed by three primary studies: 1. evaluating NLP methods to extract social determinants of health data; 2. classifying reasons for patient returns following dental extraction visits; and 3. evaluating predictive models to estimate the risk of return due to complications following dental extraction visits. Findings suggest that while language models may be helpful, the way patient information is currently captured in EDRs is a limiting factor in the value of NLP methods. Further research is required to understand the barriers to consistent and complete documentation in EDRs. To facilitate high quality documentation without disrupting clinical workflows, I therefore speculated that implementing artificial intelligence scribes, re-designing EDR systems, and facilitating interoperable medical and dental records may be useful future directions.
See less
See moreOral diseases are among the most prevalent non-communicable diseases that disproportionately affect the socioeconomically disadvantaged. Electronic dental records (EDRs) may contain useful information to address these issues at the population, practice, and patient levels, but much of the useful patient information is stored as clinical notes. While natural language processing (NLP) methods offer potential value for dentistry, tools developed for other healthcare areas may be unsuitable for dental applications due to distinct clinical workflows, vocabulary, and EDR structures. There are, however, major knowledge gaps in the use of NLP methods in dentistry. In this thesis, I aimed to determine whether NLP methods applied to clinical notes in EDRs could be used to support population health, quality indicators, and clinical decision making. The objectives were to identify vulnerable subpopulations based on the prevalence of social determinants of health, and to understand and predict patient returns for complications following a dental extraction. The research comprised a systematic review of the use of NLP methods in dentistry followed by three primary studies: 1. evaluating NLP methods to extract social determinants of health data; 2. classifying reasons for patient returns following dental extraction visits; and 3. evaluating predictive models to estimate the risk of return due to complications following dental extraction visits. Findings suggest that while language models may be helpful, the way patient information is currently captured in EDRs is a limiting factor in the value of NLP methods. Further research is required to understand the barriers to consistent and complete documentation in EDRs. To facilitate high quality documentation without disrupting clinical workflows, I therefore speculated that implementing artificial intelligence scribes, re-designing EDR systems, and facilitating interoperable medical and dental records may be useful future directions.
See less
Date
2025Rights statement
The author retains copyright of this thesis. It may only be used for the purposes of research and study. It must not be used for any other purposes and may not be transmitted or shared with others without prior permission.Faculty/School
Faculty of Medicine and Health, The University of Sydney School of Public HealthAwarding institution
The University of SydneyShare