Optimization of System of Early Detection of Oncological Diseases in Outpatient Medical Organizations

Abstract

INTRODUCTION: Increased attention to the early detection and prevention of malignant neoplasms (MNP) is conditioned by their high medical and social significance.
 AIM: To develop a methodology of early defection and stratification of risk for development of MNP at the outpatient stage using modern information technologies.
 MATERIALS AND METHODS: The data of the official statistics of Rosstat and Health Ministry of Russia for the Voronezh region (VR) were used. To collect the primary information about the level of oncological alertness (OA) of primary care workers, a survey was conducted using a specially designed questionnaire (Oncological Alertness on an Outpatient Visit) which contained 10 questions reflecting the frequency of MNP and precancer detection, doctors’ knowledge, methods of early diagnosis and patient routing in case this pathology is identified. In the survey, 112 medical workers participated. To identify MNP at early stages, methods of evaluation and stratification of the risk of MNP development were elaborated on the basis of a multi-stage analysis of significance of the identified risk factors (‘danger signals’) with the use of artificial intelligence. The method was tried on a test sample (100 patients, MNP in 55).
 RESULTS: A ‘rough’ increase in the incidence rate for 2013–2022 was 11.4%. The mortality rate from MNPs in 2022 was 170.5 per 100 thousand cases, which is 0.2 higher than in the previous year. About 60% of the newly identified MNPs were diagnosed at III–IV stage. A comprehensive study of the causative factors of advanced cases permitted to identify the most significant ones: late seeking medical care, latent asymptomatic course of MNP and insufficient OA level of primary care physicians. When testing the developed technique, the probability for the development of the disease was estimated as high in 41 (82%) patients with MNP, medium in 7 (14%), and in 2 (4%) patients the prognosis was erroneous — a low probability was predicted. Of the 50 patients who did not have MNP at the time of examination, 23 (46%) were referred to the group of low, 21 (42%) — of medium, 6 (12%) — of high risk of having a MNP. After the introduction of the developed technique, the detectability of gastric cancer increased by 3%, of colon cancer — by 2%, of tracheal, bronchial and lung cancer — by 6%, of breast, cervix and prostate cancer — by 1%, 8%, and 2%, respectively.
 CONCLUSION: The developed method permits to identify and exclude unreliable data, to select the optimal feature space characterized by the minimal dimension with sufficient informational value. This permits identification of precancerous conditions at the preclinical stage and facilitates timely detection of MNP at early stages.