Deciding priority and how urgent a patient’s need for medical care is, is of utmost importance for any healthcare facility. It is not just a question of current symptoms but also a decision that should not be taken without access to the medical history. This is an area where the error margins must be minimized while delivering high quality care and maintaining efficient use of resources. Studies have shown that people feel more comfortable chatting with a bot, than calling official help.

 

Purpose

For Swedish healthcare, it is important to decide how urgent a patient’s need for care is. Tenfifty has together with our client Visiba Care developed an advanced bot, based on Bayesian statistics and other AI techniques. 

Data sources

Initially, automatic conversion from official source text took place. Next, we developed a sophisticated Bayesian belief network tree, in close cooperation with medical doctors.

Technologies

Bayesian belief network, text vectorization, various NLP solutions.

Outcome

Using data encoded together with medical professionals, the bot can decide what question to ask next and perform a diagnosis, together with probability, based on noisy and incomplete data. Since January 2020, this solution is used live by patients.

Author

David Fendrich
CTO
David Fendrich is our Chief Technical Officer and one of the founders of Tenfifty. He has been working with AI and data science for almost two decades, helping companies put cutting edge solutions into practical use.