With the population rapidly declining in Japan, there is a shortage of medical, nursing, and care workers. Could we use the power of AI to turn the knowledge accumulated by each specialist into a common language?Social implementation of a domestic medical-specific language AI model optimized for Japanese clinical practiceWe spoke to Saiki Dai, a researcher at the Center for Emergence Strategy of the Japan Research Institute and co-leader of the "" project, about his research to date and the future he envisions.
In a society with a declining population, everyone must work together to fill the gaps
Japan's population began to decline around 2011, and by 2024 it was down by approximately 900,000 people per year, the largest decline ever recorded. Population estimates predict the decline will accelerate further in the 2030s, reaching a rate of decline of 1 million people. We are approaching a time when we will have to face rapid change unprecedented anywhere in the world.
With the rapid increase in elderly patients, it has become an urgent issue to make the entire medical care process, from the acute phase where life-saving care is provided to the lifestyle phase where long-term rehabilitation is provided, sustainable.
Until now, medical care and nursing for elderly patients have been handled by medical professionals and nursing care professionals, respectively. When the aging rate was not so high, it was fine to divide up the roles, as each person could deepen their expertise. However, as society shrinks, there will no longer be enough demand to share the responsibilities, or it will no longer be possible to allocate personnel in each field, and the system will no longer function. In order to overcome this, each person will need to expand the areas they handle. In other words, they will need to go beyond the division of labor that has existed until now and have a variety of professionals "work together" to fill in the gaps between areas.
In this project, we will collect and organize information from clinical settings based on the knowledge accumulated by experts, and develop a language AI model that will semi-automatically create medical documents while acquiring knowledge through dialogue with doctors.
We want to prevent readmission by making a breakthrough in care at the time of discharge
When elderly patients are hospitalized, they are placed in a supervised environment by many specialists, including doctors, nurses, and pharmacists. As they approach discharge, the specialists gradually begin to remove their support. If the patient does not continue rehabilitation and medication after discharge, the chances of relapse increase, and they will end up returning to the hospital.
If medical professionals' limited resources are not diverted to solving essential problems, situations will occur where patients' lives cannot be saved in time (preventable deaths). Preventing recurrence is important both for patients to be able to continue living at home and for ensuring the sustainability of a society with a shortage of medical and nursing personnel. For this reason, we want to prevent readmission by using care after discharge as a breakthrough.
Hospitals prepare a variety of medical documents, such as detailed symptom descriptions and referral letters. When a patient is discharged, they also prepare a document to be passed on to the community. Many documents simply state, "This treatment was given at the hospital," but if they also include information that connects hospital treatment to care in everyday life, such as, "This type of care and support would be good to have. Here are some specific methods," it will be easier for caregivers in the community to think, "I can do this," or "I should ask for that person's help."
However, it takes time for a human to carefully write such documents. For this reason, this project aims to use the power of AI to create documents that can be easily used by various professionals.
I want to create a common language for professionals to work together
One of the research projects I have conducted so far is the development of "appropriate care management methods" (*1). Care management involves organizing the entire life of the person receiving care from a variety of fields, including medicine, nursing, medication, rehabilitation, mental health, and nutrition. To facilitate smooth collaboration between multiple professions, I wanted to systematize the expected support and organize assessment and monitoring items to consider necessity and implementation. This is how I created the "appropriate care management method."
We created this guide based on the guidelines of each specialized field, and received feedback from care managers who said, "If we had something like this, it would make communication with other professions easier." We hope it will become a common language for professionals to work together.
This project will be based on this "appropriate care management method." When considering "what information should be extracted for Mr. A and what kind of care should be provided," AI will be used to select and discard information. Based on this, documents will be created.
Until now, medical documents have been thought of as something that is passed from one expert to another, but I think it would be better if they were made easier to understand and available to patients themselves.
This time, I will be introducing a doctoral student at the Graduate School of Informatics at Kyoto University, who is also an emergency physician.Naoki OkadaWe are working on a project together with Okada. When I spoke to him, I realized that the volume of documents that needed to be prepared was much greater than I had imagined. I want to ease the burden on doctors, and also contribute to effectively connecting the hospital and the local community. If emergency doctors can properly communicate to the local community what they want patients to be careful about, it will help prevent recurrence. I hope that we can play a part in reducing readmissions after discharge.
(*1) Regarding "appropriate care management methods"HereSee.
(Redirects to the Japan Research Institute website)
