Application of Binary Response Regression Models to Detect Factors Influencing the Occurrence of Infection in Dental Surgeries

Abstract

Postoperative infection is common in dental surgery, for example, in the removal of the third molar. To control these and other postoperative complications, various studies have reported the use of some drug protocols, namely the prophylact and the preemptive ones, using drugs such as dexamethasone and betamethasone. In this work, we used the generalized linear model via logistic regression to verify whether, in addition to the medicaments mentioned, some covariates that are frequently used in dental surgeries influence the occurrence, or not, of postoperative infection in surgeries for removal of the third molar. One of the main reasons that led us to employ such a model is because the response variable (having or not having infection) presents values of the binary type, in addition to being one of the most applied models in the area of health, among them, the dentistry area. The application of descriptive methods and analysis of association via statistical tests were also used to choose other factors that influence the response variable infection in addition to medications. The AIC (Akaike Information Criterion) selection criterion, analysis of the difeerence in the deviations, and the analysis of residual using the half-normal plot for selection and the assumption of the proposed model were employed. The data set under analysis consists of 113 patients submitted to dental surgery in a specialized clinic in the city of Piracicaba, SP-Brazil between 2003 and 2018. Through the proposed model, some important information the covariates in relation to the patients submitted to dental surgery. One of the key information is that characteristics as Age and Dental extractions are associated with the inflammatory processes after surgery. This relationship indicates that the older the patient, the chance of having an infection after surgery increases. The analysis is similar to the Number of extractions.