Section: 
Science articles
Science articlesPredicting the paratonsillar abscess development probability using logistic regression method
A. P. Yastremskii (1), A. I. Izvin (1), A. G. Sannikov (1), S. D. Zakharov (1)
(1) Tyumen State Medical University Ministry of Healthcare of the Russian Federation, Tyumen, 625023, Russia
UDK: 616.322-031.64-002.3:519.252
DOI: https://doi.org/10.18692/1810-4800-2019-2-95-102
ABSTRACT
Currently, the new tests, methods, and tools for diagnosing the diseases and forecasting the probability of possible complications are being developed and recommended for application in both medical science and practice. The use of such models allows a specialist to verify a diagnosis of the disease, to determine the prognosis and development of possible complications. The article presents the method of development of the model of paratonsillar abscess development probability. The authors used a retrospective sample set of 359 clinical cases with acute inflammatory pharyngeal diseases. There were two groups of patients: the patients with acute tonsillitis, included in the 1st group, and the patients with paratonsillitis/paratonsillar abscess – in the 2nd group. The authors made a comparative analysis of clinical signs of the above-mentioned diseases using contingency tables. The logistic regression method was used as a tool for development of mathematical and statistical model of predicting the probability of the event relevant for the researcher with two possible outcome variants. Using the presented method, the logit-regression equations were obtained, and one of them – with higher indices of agreement – was chosen, providing the revealing of the disease predictors that form the basis for mathematical model “Predicting the paratonsillar abscess development probability”. To determine the diagnostic sensitivity, specifics, and efficacy of the model, the authors used ROC (Receiver Operating Characteristic) analysis method. ROC-curve provides the assessment of the quality of the model of separating the two classes based on the finding of the cutoff point of the continuous or discrete quantitative sign. The developed model provides predicting the probability of paratonsillar abscess development with the result of diagnostic efficacy up to 97.8%.
Publication date:
28.03.2019
Keywords:
sample of patients, acute tonsillitis, paratonsillitis, logistic regression, disease predictors. For citation:
Yastremskii A. P., Izvin A. I., Sannikov A. G., Zakharov S. D. Predicting the paratonsillar abscess development probability using logistic regression method. Rossiiskaya otorinolaringologiya. 2019;18(2):95–102. https://doi.org/ 10.18692/1810-4800-2019-2-95-102