Determination of Xerostomia with Cutoff Value for Salivary Flow Rate using Machine Learning Algorithm

Abstract

Abstract Aim The purpose of this study was to investigate the objective cut-off values of unstimulated (UFR) and stimulated salivary flow rates (SFR) in patients with xerostomia and to present optimal machine learning model with A Classification and Regression Tree (CART) for all ages. Methods A total of 829 patients with oral diseases were enrolled (591 females; mean age, 59.29 ± 16.40 years; age range, 8–95 years old), 199 patients with xerostomia and 630 patients without xerostomia. Clinical characteristics were collected and analyzed together. To investigate which oral and systemic factors affect the presence of xerostomia and the cutoff value of UFR and SFR, the CART machine learning algorithm was repeatedly performed. Results UFR (0.41 ± 0.24 vs. 0.29 ± 0.22 mL/min, p < 0.001) and SFR (1.39 ± 0.94 vs. 1.12 ± 0.55, p < 0.001) were significantly lower in Xerostomia than in non-Xerostomia. The presence of xerostomia had a significant negative correlation with both UFR (r=-0.603, p-value < 0.01) and SFR (r=-0.301, p-value < 0.05). Considering the magnitude of the correlation coefficient, the presence of xerostomia in the patients with oral diseases showed a stronger correlation with the decrease in UFR than with the decrease in SFR. In the diagnosis of xerostomia based on the CART machine learning algorithm, the presence of stomatitis, candidiasis, halitosis, psychiatric disorder, and hyperlipidemia were significant predictors for xerostomia. According to the type of parameters included in each CART algorithm, the cutoff values of UFR and SFR were different, and the specific ranges with significant results are as follows; the UFR of 0.03 ~ 0.18 mL/min, SFR of 0.85 ~ 1.6 ml/min. Conclusion For the diagnosis of xerostomia, a new comprehensive approach was made using an optimal CART algorithm considering salivary hyposalivation and oral/systemic conditions. Xerostomia was negatively correlated with UFR and SFR values, and cut-off values for salivary flow rates varied depending on the underlying clinical factors of the patients.