Purpose: Identification of risk factors that influence the outcome of the patient, their ranking on the contribution to the outcome of treatment, as well as determining the possibility of their additional diagnostic evaluation and correction in the deviation at the preoperative preparation stage with the subsequent construction of a prognostic model. Material and methods: The study included patients who received treatment in the surgical department in A. I. Burnasyan Federal Medical Biophysical Center from January 2009 to July 2017, including workers of nuclear facilities that are exposed to ionizing radiation in professional conditions. The study was conducted in 112 patients, 42 of whom (37.5 %) were men and 70 (62.5 %) women aged 25 to 85 years (59.6 ± 13.2). Among the persons included in the study, 25 men and 26 women were exposed to long-term exposure to ionizing radiation from external sources under production conditions during labor activity within the limits of annual maximum permissible doses, averaged 124.6 ± 10.7 mSv. The work experience under conditions of exposure to ionizing radiation ranged from 5 to 35 years, an average of 24 years. The mean age was 59.1 ± 13.4 years. At the end of hospitalization after surgical treatment, 51 patients were discharged (45.5 %), and 61 (54.5 %) died. In all patients, the parameters of the functioning of various organs and systems were collected, including taking into account the anamnestic data of oncological patients, with differentiation in the final outcome of surgical treatment. To determine the leading risk factors for the lethal outcome of the oncosurgical patient, the Fisher criterion χ2 was used. Based on the leading risk factors for constructing mathematical models, the logistic regression equation was used. The mathematical models were analyzed by researching the area under the ROC curves. Results: Using the Fisher criterion χ2, factors were determined by which the groups of survivors and died patients differ: patient age, body mass index, history of heart rhythm disorders, fraction of cardiac output, Hb level in the blood, presence of protein in urine, INR indicator in coagulograms. Based on the identified factors, twelve mathematical models were constructed using the binary logistic regression method, allowing patients to be divided into groups with the outcomes of hospitalization died / survived after surgery. A mathematical model with the best discriminating ability was chosen. Based on the prognostic model, a decision rule was designed that allows to rank patients into three groups: green (patients with a minimal risk of death), yellow (patients who need preoperative correction), red (patients with the maximum risk of death, decision about surgery is necessary to be solved on a consultation).
Read full abstract