In polyclinics, making more accurate diagnosis with the help of artificial intelligence

July 17

In the capital's polyclinics, there is a system for supporting medical decision-making. The accuracy of making a preliminary diagnosis using artificial intelligence in Moscow polyclinics increased by 7.4 percent. Now, in 46 percent of cases, physicians choose the preliminary diagnosis proposed by the neural network. This was told by Anastasia Rakova, Deputy Moscow Mayor for Social Development.

"The system of medical decision-making support became an integral part of the Moscow physicians work 10 months ago. During this time, the neural network has been trained on 12 million visits, and an improved version of the service has been launched in all Moscow polyclinics since the beginning of May. It has become more accurate by 7.4 percent and, as a result, more convenient for the physician's work, that has a positive effect on the quality of services for patients. In general, the development of digital services in healthcare contributes to reducing the probability of errors in making a preliminary diagnosis and increasing the detection of diseases at the primary healthcare. The latest evidence suggests that when choosing one or three diagnoses, the physician agrees with the proposals of the neural network in 46 percent and 68 percent of cases, respectively," the deputy mayor said.

How artificial intelligence works in polyclinics

The system for supporting medical decision-making began to be used on June 1, 2019. Since September 2020, it has been working in all adult polyclinics in Moscow. The system consists of several modules and completely repeats the patient's clinical path: from collecting anamnesis to making a diagnosis and assigning а therapy. According to the UN, the model of artificial intelligence underlying the module for making a preliminary diagnosis is consistent with the goals of sustainable development.

The service analyzes patient complaints and offers the physician preliminary diagnoses according to the codes of the international classification of diseases (ICD-10). The neural network processes the complaint text entered by the physician, compares it with 12 million impersonal patient records in EMIAS, analyzes the patterns and offers the physician the three most likely diagnoses. The physician can choose one of them or make an independent decision. Today, the accuracy of the model reaches 68 percent when choosing from three diagnoses and 46 percent when choosing one.

For creating the system, one of the first largest studies in Russia and in the world was conducted. The developers summarized the statistics of all diagnoses accepted in city polyclinics for one year - and this is more than two million visits - and ranked them according to the probability of diseases according to ICD-10. 265 categories were selected from the list, with which in 95 percent of cases patients applied to Moscow's polyclinics at initial consultation.

In addition to making a preliminary diagnosis, package appointments are available for a physician. They offer the physician a one-page reference book, which indicates the necessary examinations and consultations in this case. The physician can use them to send the patient to additional laboratory and instrumental examinations or to consult with specialty doctors to confirm the diagnosis. From the list, the physician will also see whether the patient has recently visited a specialty doctor, whether he has passed the recommended examinations. The results are immediately reflected here, and the doctor does not need to look for them in the menu sections. At his discretion, the doctor may agree with the recommendations of the service or appoint additional examinations or referrals for consultations, as well as remove from the list those that the patient has recently passed. Using the system allows you to reduce the average time of appointment of diagnostic testing by 10 times.

The development of a system for supporting medical decision-making minimizes routine in the actions of a doctor and allows him to pay attention to details that could slip under the radar. It also allows the doctor to avoid repeated appointments for the same examination, which means that the burden on the healthcare system as a whole is reduced.

The implementation of such solutions has become possible due to a single digital platform, which is being developed jointly with the Moscow Social Development Complex and the Department of Information Technology.


Latest Events

Osnovnoy instinkt

July 29August 4

Unesyonnye prizrakami

July 22August 4
If you continue to use our website, you are agreeing to accept the use of cookies on your device. Cookie files ensure the website’s efficiency and help us provide you with the most interesting and relevant information. Read more about cookie files.
Accept ccokies