Artificial Intelligence and Hand Hygiene Accuracy: A New Era in Infection Control for Dental Practices

Objective

The study aimed to assess the efficacy of an artificial intelligence (AI) model in evaluating hand hygiene (HH) performance compared to infection control auditors in dental clinics.

Material and Method

The AI model utilized a pretrained convolutional neural network (CNN) and was fine-tuned on a custom data set of videos showing dental students performing alcohol-based hand rub (ABHR) procedures. A total of 66 videos were recorded, with 33 used for training and 11 for validating the model. The remaining 22 videos were designated for testing and the AI- infection control auditors comparison experiment. Two infection control auditors assessed the HH performance videos using a standardized checklist. The model's performance was evaluated through precision, recall, and F1 score across various classes. The level of agreement between the auditors and the AI assessments was measured using Cohen's kappa, and the sensitivity and specificity of the AI were compared to those of the infection control auditors.

Results

The AI model has learned to differentiate between classes of hand movement, with an overall F1 score of 0.85. Results showed a 90.91% agreement rate between the AI model and infection control auditors in evaluating HH steps, with a sensitivity of 85.7% and specificity of 100% in identifying acceptable HH practices. Step 3 (back of fingers to opposing palm with fingers interlocked) was consistently identified as the most frequently missed step by both the AI model and the infection control auditors.

Conclusion

The AI model assessment of HH performance closely matched auditors' evaluations, suggesting its reliability as a tool for evaluating and mentoring HH in dental clinics. Future research should explore the application of AI technology in different dental settings to further validate its feasibility and adaptability.