Implementing an Artificial Intelligence System in the Work of General Practitioner in the Yamalo-Nenets Autonomous Okrug: Pilot Cross-sectional Screening Observational Study

Background. Early identification of risk factors (RF) associated with cardiovascular diseases (CVD) is essential for the prevention of CVDs and their complications. CVD risk factors can be identified using Artificial Intelligence (AI) systems, which are capable of learning, analyzing and drawing conclusions. The advantage of AI systems consists in their capacity to process large amounts of data over a short period of time and produce ready-made information.
Objectives. Evaluation of the efficiency of implementing an AI software application by a general practitioner for identifying CVD risk factors.
Methods. The study included data from 1778 electronic medical histories of patients aged over 18, assigned to an outpatient and polyclinic department of Muravlenkovskaya Gorodskaya Bolnitsa (Muravlenko municipal hospital), Yamalo-Nenets Autonomous Okrug (Russia). The study was conducted in four stages. The first stage involved a preliminary training of the Artificial Intelligence (AI) system under study using numerous CVD risk assessment scales. The Webiomed predictive analytics and risk management software by K-SkAI, Russia, was selected as a platform for this purpose. The second stage included an analysis of medical data to identify CVD risk factors according to the relative risk scale for patients under 40 and the SCORE scale for patients over 40. At the third stage, a specialist analyzed the previous and new information received about each patient. According to the results of the third stage, four risk groups for CVD (low, medium, high and very high) were formed. At the fourth stage, newly diagnosed patients with a high risk of CVD, who had not been previously subject to regular medical check-up, were directed for additional clinical, laboratory and instrumental follow-up examination and consultations of relevant specialists. Statistical data in absolute terms and as a percentage were obtained. Statistical processing of the results was carried out by a computer program aimed at medical decision support. Content visualization was performed in spreadsheets and charts.
Results. Based on the data obtained, the AI system under study divided all patients into CVD risk groups and identified uncounted factors. The AI system confirmed a high and very high risk of CVD according to SCORE (Systematic COronary Risk Evaluation) in 623 people, who were already receiving appropriate cardiological assistance. The RFs that had not previously been taken into account in the diagnosis were recorded in 41 (11.5%) patients from the very highrisk group and in 37 (12.7%) high-risk patients. The AI system identified a high risk of CVD in 29 people who had not been previously under care of a general practitioner or other specialists due to their infrequent visits to health care facilities. These patients were detected by the AI system following periodic and preliminary medical check-ups (35%), full in-patient treatment for other diseases (31%), when seeking help of other specialists (17%), as well as when obtaining a medical certificate for a driving license (12%), admission to a swimming pool (3%) or possessing a weapon (2%). In a group with the newly diagnosed patients at a high risk of CVD, men dominated (24 persons, 82%) and women comprised only 8% (5 persons). All these people were of working age between 40 and 50. In order to confirm the information received, the supervising physician subsequently referred patients for a follow-up examination, as a result of which only 1 person (3%) was not diagnosed with a somatic pathology.
Background. Early identification of risk factors (RF) associated with cardiovascular diseases (CVD) is essential for the prevention of CVDs and their complications. CVD risk factors can be identified using Artificial Intelligence (AI) systems, which are capable of learning, analyzing and drawing conclusions. The advantage of AI systems consists in their capacity to process large amounts of data over a short period of time and produce ready-made information.
Objectives. Evaluation of the efficiency of implementing an AI software application by a general practitioner for identifying CVD risk factors.
Methods. The study included data from 1778 electronic medical histories of patients aged over 18, assigned to an outpatient and polyclinic department of Muravlenkovskaya Gorodskaya Bolnitsa (Muravlenko municipal hospital), Yamalo-Nenets Autonomous Okrug (Russia). The study was conducted in four stages. The first stage involved a preliminary training of the Artificial Intelligence (AI) system under study using numerous CVD risk assessment scales. The Webiomed predictive analytics and risk management software by K-SkAI, Russia, was selected as a platform for this purpose. The second stage included an analysis of medical data to identify CVD risk factors according to the relative risk scale for patients under 40 and the SCORE scale for patients over 40. At the third stage, a specialist analyzed the previous and new information received about each patient. According to the results of the third stage, four risk groups for CVD (low, medium, high and very high) were formed. At the fourth stage, newly diagnosed patients with a high risk of CVD, who had not been previously subject to regular medical check-up, were directed for additional clinical, laboratory and instrumental follow-up examination and consultations of relevant specialists. Statistical data in absolute terms and as a percentage were obtained. Statistical processing of the results was carried out by a computer program aimed at medical decision support. Content visualization was performed in spreadsheets and charts.
Results. Based on the data obtained, the AI system under study divided all patients into CVD risk groups and identified uncounted factors. The AI system confirmed a high and very high risk of CVD according to SCORE (Systematic COronary Risk Evaluation) in 623 people, who were already receiving appropriate cardiological assistance. The RFs that had not previously been taken into account in the diagnosis were recorded in 41 (11.5%) patients from the very highrisk group and in 37 (12.7%) high-risk patients. The AI system identified a high risk of CVD in 29 people who had not been previously under care of a general practitioner or other specialists due to their infrequent visits to health care facilities. These patients were detected by the AI system following periodic and preliminary medical check-ups (35%), full in-patient treatment for other diseases (31%), when seeking help of other specialists (17%), as well as when obtaining a medical certificate for a driving license (12%), admission to a swimming pool (3%) or possessing a weapon (2%). In a group with the newly diagnosed patients at a high risk of CVD, men dominated (24 persons, 82%) and women comprised only 8% (5 persons). All these people were of working age between 40 and 50. In order to confirm the information received, the supervising physician subsequently referred patients for a follow-up examination, as a result of which only 1 person (3%) was not diagnosed with a somatic pathology.
Conclusion. The efficiency of the AI system under study comprised 97%. Permanent monitoring of all parameters of electronic medical histories and outpatient records is an efficient method for timely identification of RF at any visit of a person to a health care facility (preventive and periodic medical examinations, regular check-ups, specialist consultations, etc.) and their assignment to respective CVD risk groups. Such monitoring ensures an effective medical supervision of able-bodied populations.