Concord specialists use AI to save customer budgets

2024: Developing a service to save customer budgets

The technological division of the insurance company "Consent" has developed a service based on artificial intelligence, which allows to increase the efficiency of the processes for checking reports on provided medical services within the framework of voluntary health insurance programs (VHI). The main task of AI is to provide automated verification of treatment methods and prophylaxis assigned to clients by clinics. The company announced this on September 12, 2024.

The VHI policy covers a large amount of medical services provided by partner clinics. But in some cases, the client is assigned procedures that do not have practical benefits in restoring his health. As a result, these appointments increase the cost of the insurance policy from year to year. The service of the company "Consent" allows you to track such situations. This leads not only to a decrease in the loss-making of the insured, but also allows you to protect the rights of patients to receive high-quality and necessary medical care.

The model that we have implemented is comprehensive and consists of three components: firstly, the accumulated experience of specialized specialists, expressed in a set of conditions under which the service cannot be provided to the insured, and secondly, digitized clinical recommendations and their compliance with the diagnosis, and finally, machine-based training that allows you to assess the likelihood of refusal of the service. As a result, the company has repeatedly improved the quality of the services provided and reduced the costs of mainly corporate clients, "said Vladimir Kozhevnikov, head of the analytics and machine learning center.

This AI has been in the company's service since 2019. During this time, he helped save more than 45 million rubles to customers of Concord. In the fifth version of the service (the update took place in April 2024), the company introduced clinical recommendations of the Ministry of Health of the Russian Federation - this made it possible to several times increase the accuracy of determining ineffective prescriptions.

There are several boxed products in the insurance market that allow you to perform related functions, but we decided to focus on our development. Yes, this is a long time, for example, for our system we tried on reference vectors, KNN, random forest bagging and other popular methods, and in the latest iteration we settled, it seems to me, on the most suitable algorithm. But the time spent is leveled by the advantages of our own development: we can conduct a more flexible configuration that fits the requests of our company, introduce new modules, and also regularly improve the rules due to client recommendations, - said Vladimir Kozhevnikov.