ABSTRACT
Background Dentists are a high-risk population of work-related musculoskeletal disorders (WMSDs), where the body part with the highest prevalence is the neck. Objective To analyze potential influencing factors of neck pain among dentists, and explore a prediction model of neck pain in dentists. Methods Dentists from different hospitals in Fuzhou were selected as study subjects by stratified cluster sampling according to hospital characteristics (dental hospitals, general hospitals, and dental clinics). The basic information, presentation of WMSDs, and its influencing factors were investigated by using the Chinese version of Musculoskeletal Disorders Questionnaire and the Subjective Workload Assessment Technique. A total of 655 questionnaires were collected, of which 603 were valid, with an effective rate of 92.1%. Multiple logistic regression was used to analyze potential influencing factors of neck pain in dentists. A prediction model of neck pain of dentists was constructed by using neural network model, and the prediction efficiency of the model was evaluated. Results The neck was the body part with the highest prevalence (43.8%, 264/603) of WMSDs among dentists. The results of multiple logistic regression analysis showed that female (OR=2.709, 95%CI: 1.852-3.962, P <0.001), working age of 10-<20 years (versus <10 years, OR=3.836, 95%CI: 2.471-5.957, P<0.001), keeping head up or down for a long time (OR=8.492, 95%CI: 2.203-32.731, P=0.002), holding head sideways for a long time (OR=2.210, 95%CI: 1.376-3.550, P<0.001), maintaining the same sitting spot for a long time (OR=4.336, 95%CI: 2.192-8.579, P<0.001), and psychological load value ≥70 (versus <40, OR=1.901, 95%CI: 1.038-3.480, P=0.037) increased the risk of neck pain in dentists. Sufficient operating space (OR=0.507, 95%CI: 0.302-0.850, P=0.010) and doing some exercise during work break (OR=0.670, 95%CI: 0.453-0.991, P=0.045) reduced the risk of reporting neck pain among dentists. A neural network prediction model of dentists' neck pain was constructed with 1 hidden layer and 6 hidden layer neurons. The percentage of correct prediction of training set was 89.6%, and the percentage of correct prediction of test set was 83.9%. The order of importance of the independent variables included in the model were working age, holding head sideways for a long time, psychological load, etc. The result of neural network model of neck pain among dentists showed that the area under the curve of receiver operator characteristic (ROC) was 0.940 (95%CI: 0.922-0.958, P<0.001). When the maximum diagnostic value was determined by the ROC curve, the sensitivity was 84.8%, the specificity was 91.2%, and the Youden Index was 0.760. Conclusion Neck pain of dentists is affected by many factors, such as individual factors (gender and working age), ergonomic factors (keeping various postures and operations for a long time, operating space, etc.), psychological factors (different levels of psychological load) and so on. The neural network model can be used as a prediction tool to explore the risk of reporting neck pain among dentists.