The new assistant for doctors: the AI ​​project of Assuta and Microsoft

At the center of the move announced by Assuta and Microsoft is the development of a personal artificial intelligence agent for each doctor. Unlike normal medical computer systems, where the doctor is required to go through screens, tests, hospitalization summaries and referral letters, the agent is supposed to read the material in the medical file and present the important information for the meeting.

Thus, before the patient enters the room, the doctor will be able to receive a summary of the previous diagnoses, the medications, the latest tests, treatments that have already been tried and changes that have occurred in the medical condition. The goal is to reduce the time spent searching for information in large amounts of documents and reduce the risk that a significant detail will disappear within the file.

During the visit, the system will be able to listen to the conversation between the doctor and the patient, after receiving the required approvals and in accordance with the privacy rules, and produce an orderly transcript. She will then process the information and draw up a draft of a complete medical document, which will include the patient’s complaints, the medical background, the test findings, the doctor’s impression and recommendations for further action.

The doctor will be required to go over the document, correct it and approve it before entering the medical file. Basuta emphasizes that the system will serve as a decision support tool only and will not replace medical judgment. The responsibility for the diagnosis, choice of tests and treatment will remain with the doctor.

The agent should also assist in the steps after the visit. According to the plan, he will be able to identify that the patient needs an examination or another meeting, locate a suitable date and assist in coordinating a follow-up appointment according to the medical need and availability in the system. In the future, additional operations may be added along the treatment continuum, subject to medical control and information security rules.

One of the main problems in the health system is the large amount of time that doctors devote to documentation, typing summaries, going over documents and performing administrative actions. Sometimes the work continues even after the end of the clinic day. According to the goals defined for the project, the use of the agent may cumulatively save hundreds of hours of doctor’s work every month.

This saving may allow shortening the duration of the visit without reducing the time spent talking with the patient. Instead of the doctor looking at the screen for a significant part of the meeting and typing the details of the conversation, the system will prepare the documentation for him. This way it will be possible to receive additional patients during the working day, expand the availability of appointments and shorten the waiting times in clinics and institutes.

Developing such a system requires more than connecting transcription software to the doctor’s computer. In order for the agent to be able to understand medical documents, he has to go through a long process of study, practice and training. As part of the process, models of medical documents, visit summaries, test results and professional terms are entered into the systems, while removing identifying details and maintaining patient privacy.

The system practices distinguishing between central information and marginal details, identifying connections between diagnoses, medications and tests, and formulating a summary that will be short but will not miss essential information. Medical teams check the products, mark mistakes and teach the system how to present the information in a way that will help the doctor during the work.

One of the challenges will be to prevent errors typical of artificial intelligence systems, in which details are omitted, confusion between patients or the creation of information that did not appear in the original documents. For this reason, the products should not automatically enter the medical file without a human examination. The development and implementation will be carried out gradually, while examining the accuracy, safety and the effect of the system on the work of the teams.

The first pilot is expected to start within a year. Initially, the system will probably be operated on a limited scale, to check how it integrates into the daily work, how much time it saves, what the quality of the summaries is, and how comfortable the doctors and patients feel with automatic documentation of the meeting.

Alongside the development of the personal agent, Assuta and Microsoft agreed on a gradual transition to the company’s cloud infrastructure and on the establishment of an organizational platform for the development and operation of additional artificial intelligence agents. The infrastructure may be used in the future for solutions in the fields of medicine, service and operation, while maintaining strict rules of privacy, information security and protection of medical information.