Accuracy of Predicting Treatment Adjuncts for Patients Requiring Special Care Dentistry

Abstract

Background: Predicting the correct treatment adjuncts in special care dentistry (SCD), such as non-pharmacological behavioural support, inhalation sedation, intravenous (IV) sedation, or general anaesthesia (GA), can be complex. A number of factors could influence the decision-making process on the selection. Inaccurate prediction often leads to treatment failure, disappointment of patient and their care teams, increased waiting times, resource misallocation, negative environmental impacts, and unnecessary restrictive practices. This study aims to measure and optimise accuracy in treatment adjunct selection to facilitate better clinical decision-making.

Methods: A retrospective chart review design received ethical approval (DSREC01/2023) to be performed in the Dublin Dental University Hospital (DDUH), Ireland. All consecutive electronic records of newly assessed patients who required treatment over 24 months (2020-2021) were included. A comparison was made between the treatment adjunct chosen at assessment (predicted treatment adjunct or Pre-TA) and the actual treatment adjunct ultimately provided for successful treatment (actual treatment adjunct or Act-TA). A confusion matrix was generated to establish accuracy, precision, sensitivity, and specificity when predicting each Pre-TA. Multinomial logistic regression was used to find out significant variables that impact the accuracy of treatment adjunct prediction. A prediction model was plotted to identify which features are most relevant when predicting successful Act-TA using Posit Cloud by R Studio.

Results: The records of 106 patients met inclusion criteria. Overall, the accuracy of treatment adjunct selection was 91.5%. Nine cases were deemed inaccurate predictions. Seven of these were planned for IV sedation, resulting in a low precision rate for IV sedation compared to other options. Three major reasons for inaccurate predictions were sedation failure, service availability, and medical complexity. The presence of challenging behaviour and the urgency of treatment significantly impacted the accuracy of treatment adjunct prediction (p<0.05). A decision tree illustrating the probability of each Act-TA was generated. The most important factors in predicting successful adjunct selection were previous adjunct use for dental care, capacity to consent, need for multiple restorations or any extractions, and treatment plan complexity.

Conclusion: This is the first study to assess the accuracy of adjunct selection in SCD, which has revealed a remarkably high accuracy in adjunct prediction. The instances of inaccurate predictions, primarily in cases predicting IV sedation, highlight the need for improved sedation assessment and thorough medical assessment and communication. The development of a supportive decision-making tool for future treatment adjunct selection, which considers past dental history, capacity evaluation, and a clear dental treatment plan, is a significant step forward. The proposed model warrants further validation and generalisation. The study also underscores the importance of education and training in non-pharmacological behavioural support techniques. It also highlights the benefit of teaching decision science to clinicians, specifically regarding decision making in uncertainty and uncertainty tolerance. These findings suggest a roadmap for future research and practice in the field of special care dentistry.