Prediction of the course of chronic rhinosinusitis using a mathematical model

Purpose of research: development of a mathematical model for predicting the course of chronic rhinosinusitis.
Material and methods: The levels of biomarkers of endothelial dysfunction in the blood (homocysteine, highly sensitive C-reactive protein, cystatin C, D-dimer) were analyzed. The study involved 269 people, including 149 patients with CRS aged 18 to 59 years, who sought surgical help at the otorhinolaryngological department of the Tambov Clinical Hospital. The control group consisted of 120 people without pathology of the paranasal sinuses, comparable in age and gender. The СRS group was divided into 2 subgroups: polypous rhinosinusitis (PolRS) –67 people and chronic purulent rhinosinusitis (CPurRS) -82 people. Patients with diseases or pathological conditions affecting changes in the levels of the studied biomarkers in the blood are excluded from the study. Statistical data processing was carried out using methods of descriptive statistics and correlation analysis. The differences at the p<0.05 level were considered statistically significant. The subject of the study was a predictive model developed on the basis of logistic regression, ROC analysis was used as a tool for assessing the quality of the model.
Results. Analysis of the obtained data showed that the levels of these biomarkers were statistically significantly higher in the groups with pathologies than in the control group. A multifactorial analysis was performed to assess the correlation of independent predictor factors with a group of patients: between patients with CRS (PolRS and CPurRS) and the control group. Correlation analysis showed the presence of a strong direct reliable association of the group code with the indicators «homocysteine» (R=0.728, p=0.001), «cystatin C» (R=0.712, p=0.004) and «D-dimer» (R=0.779, p=0.0021). A correlation of average strength (R=0.376, p=0.0034) was revealed with the parameter «hsCRP». According to the results of the analysis, all 4 parameters are included in the mathematical model: «homocysteine», «cystatin C», «D-dimer» and «hsCRP». The logistic regression method was used to build a predictive mathematical model. In order to assess the diagnostic significance of the obtained model, a ROC analysis was performed. The area under the ROC curve (AUC) is 0.99, which indicates the high efficiency of the model. Using the F1-Score metric, the optimal cut-off threshold of 0.3864 was determined. When the calculation result value is < 0.3864, it means that the subject has no clinical and laboratory manifestations of endothelial dysfunction, and if ≥ 0.3864, it means the presence of clinical and laboratory manifestations of endothelial dysfunction. When evaluating the performance of the model with a cut-off threshold of 0.3864, the following characteristics were obtained: sensitivity – 99.6%, specificity – 100%, accuracy – 100%.
Conclusions. Based on the obtained mathematical model, it is possible not only to diagnose ED in MS, but also to monitor clinical and laboratory manifestations of ED after treatment. Timely diagnosis of ED using mathematical modeling will help optimize treatment regimens and improve the prognosis of the course of CRS. When predicting the course of the disease, an integrated approach should be applied. The method of ED diagnosis using mathematical modeling described in the article will contribute to the optimization of medical measures and the improvement of the prognosis for CRS.
Keywords. Chronic rhinosinusitis; endothelial dysfunction; highly sensitive C-reactive protein; D-dimer; cystatin C; homocysteine; biomarker; mathematical model; logistic regression; forecasting in medicine
Conflict of interest. The author declare that they have no conflict of interest.
Funding. This study required no funding