Guidelines summarize how persons with a certain disease should be treated according to current knowledge, thereby supporting doctors in finding optimal therapy. Keeping an overview of guidelines is however becoming increasingly difficult. A consortium comprising public and industry partners and headed by Charité – Universitätsmedizin Berlin is now aiming to remedy the situation: engaging in the GUIDE-AI project, the team is developing AI-based assistants that draw attention to deviations from the guidelines – and opening up potential that could improve the treatment of millions of people. GUIDE-AI is being funded with 9.5 million euros over a four-year period as part of the EU's Innovative Health Initiative.
Medical guidelines are intended to help physicians determine the best therapy for each individual patient. "Medical knowledge doubles every 73 days, however, with some guidelines now comprising more than a hundred pages," explains Dr. Dr. Matthias Gröschel, physician and head of a research group in the Joint Department of Infectious Diseases, Respiratory Medicine and Critical Care Medicine at Charité. "What is more, the recommendations are updated regularly, meaning that it is a challenge to keep track of all the details and updates for a large number of diseases."
Take chronic heart failure, for example, or more precisely heart failure of the HFrEF type: If a doctor wants to treat this condition according to current guidelines, a whole range of factors have to be weighed up for each individual patient – blood pressure, heart rate, kidney function, electrolytes, concomitant illnesses and other factors. Due to the fact that some medications help against heart failure but also put a strain on the kidneys or blood pressure, the physician must also keep an eye on the nephrology guidelines for patients with chronic kidney disease. Put succinctly: Determining safe and effective treatment for individual HFrEF patients has become complex – and requires time, which is often lacking in everyday medical practice.
19 Partners teaming up for a European guideline navigator
The GUIDE-AI consortium now wants to harness large language models (LLMs) to point medical professionals the way in the guideline jungle. Research institutions are working hand in hand with small and medium-sized companies, patient representatives and partners from the pharmaceutical industry. They come from seven European countries and Israel. The guideline assistants will be set up in various European languages, while taking local data privacy and treatment requirements into account. The coordination of the consortium is in the hands of Matthias Gröschel, while AstraZeneca is responsible for project management.
The goal: Navigators that – integrated into the hospital information system or the practice software – compare the prescribed therapy with the guideline recommendations based on individual patient data and recommend changes in the event of relevant deviations. A hospital information system, similar to medical practice software, is a computer program that displays data on individual patients to medical staff, from radiology images and laboratory values through to a list of prescribed medication.
Back to the heart failure example: In the future, the HFrEF-Navigator will alert the doctor in the medical practice or in the clinic within the computer program in the case that only three instead of the four recommended medicines are prescribed or if a different dosage would be preferable for one of the medicines. "In the end, the assigned physician always decides on the therapy in light of all the relevant information," as Matthias Gröschel explains. "But with the help of the GUIDE-AI navigators, we want to point out opportunities to optimize treatment according to the guidelines."
The focus is on four common chronic diseases
Initially, the GUIDE-AI team is focusing on four chronic diseases that are particularly widespread and entail major health burdens but are not always treated according to the recommendations despite established guidelines: HFrEF heart failure, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD) and asthma. It is estimated that more than 160 million people across Europe are impacted by these diseases. A specific navigator is to be developed for each of the four diseases, with plans to extend the system to additional conditions, such as inflammatory bowel disease, in the future.
In the first step, the team aims to survey the barriers and needs in optimizing guideline-directed care. Various AI models will then be tested and the most suitable – perhaps several – will be used to power the respective navigator. "It goes without saying that we attach utmost importance to protecting sensitive medical data," emphasizes Matthias Gröschel. "For example, the AI models should be usable on local servers wherever possible, but at a minimum comply with European data protection regulations."
A randomized clinical trial examining as to whether doctors actually prescribe guideline-compliant therapies more frequently with the support of the Navigator represents an additional part of the project. Finally, the program should also provide information that is readily comprehensible for laypeople – so as to not only guide medical staff but also patients in dealing with their illness.
About GUIDE-AI
The GUIDE-AI consortium brings 19 partner institutions together from science, healthcare and the pharmaceutical industry based in Germany, Italy, the Netherlands, Estonia, Slovenia, the UK, Switzerland and Israel. It is coordinated by Charité – Universitätsmedizin Berlin. This project is supported by the Innovative Health Initiative Joint Undertaking (IHI JU) under grant agreement No 101253015. The JU receives support from the European Union’s Horizon Europe research and innovation program and COCIR, EFPIA, Europa Bío, MedTech Europe, and Vaccines Europe.