Applying Process-Oriented Data Science to Dentistry

Background: Healthcare services now often follow evidence-based principles, so technologies such as process and data mining will help inform their drive towards optimal service delivery. Process mining (PM) can help the monitoring and reporting of this service delivery, measure compliance with guidelines, and assess effectiveness. In this research, PM extracts information about clinical activity recorded in dental electronic health records (EHRs) converts this into process-models providing stakeholders with unique insights to the dental treatment process. This thesis addresses a gap in prior research by demonstrating how process analytics can enhance our understanding of these processes and the effects of changes in strategy and policy over time. It also emphasises the importance of a rigorous and documented methodological approach often missing from the published literature. Aim: Apply the emerging technology of PM to an oral health dataset, illustrating the value of the data in the dental repository, and demonstrating how it can be presented in a useful and actionable manner to address public health questions. A subsidiary aim is to present the methodology used in this research in a way that provides useful guidance to future applications of dental PM. Objectives: Review dental and healthcare PM literature establishing state-of-the-art. Evaluate existing PM methods and their applicability to this research’s dataset. Extend existing PM methods achieving the aims of this research. Apply PM methods to the research dataset addressing public health questions. Document and present this research’s methodology. Apply data-mining, PM, and data-visualisation to provide insights into the variable pathways leading to different outcomes. Identify the data needed for PM of a dental EHR. Identify challenges to PM of dental EHR data. Methods: Extend existing PM methods to facilitate PM research in public health by detailing how data extracts from a dental EHR can be effectively managed, prepared, and used for PM. Use existing dental EHR and PM standards to generate a data reference model for effective PM. Develop a data-quality management framework. Results: Comparing the outputs of PM to established care-pathways showed that the dataset facilitated generation of high-level pathways but was less suitable for detailed guidelines. Used PM to identify the care pathway preceding a dental extraction under general anaesthetic and provided unique insights into this and the effects of policy decisions around school dental screenings. Conclusions: Research showed that PM and data-mining techniques can be applied to dental EHR data leading to fresh insights about dental treatment processes. This emerging technology along with established data mining techniques, should provide valuable insights to policy makers such as principal and chief dental officers to inform care pathways and policy decisions.