Public-private consortium aims to develop therapy assistants for physicians in Europe
Berlin, 4 March, 2026
Clinical practice guidelines summarize how people with a specific disease should be treated according to current scientific knowledge. They support physicians in finding the optimal therapy. However, guidelines are becoming increasingly difficult to navigate. A consortium of public and industry partners, led by Charité – Universitätsmedizin Berlin, aims to remedy this: In the GUIDE-AI project, the team is developing AI-based assistants that highlight deviations from the guidelines – and could thus improve the treatment of millions of people. GUIDE-AI is being funded with €9.5 million over four years as part of the EU’s Innovative Health Initiative.
Medical guidelines are intended to help treating physicians determine the best therapy in each individual case. “However, medical knowledge doubles every 73 days, and some guidelines now comprise more than one hundred pages,” explains Dr. Matthias Gröschel, physician and head of a research group in the Department of Infectious Diseases, Pulmonology, and Intensive Care Medicine at Charité. “Furthermore, the recommendations are regularly updated. It’s simply a huge challenge to keep track of all the details and updates for a multitude of diseases.”
Photo: At a glance: The GUIDE-AI navigators are designed to highlight deviations from medical guidelines within patient software. © Charité | Janine Oswald
Take heart failure, specifically heart failure with reduced ejection fraction (HFrEF): If a physician wants to treat it according to guidelines, they have to weigh a whole range of factors on a patient-specific basis – blood pressure, heart rate, kidney function, electrolyte balance, comorbidities, and much more. Because some medications, while helping with heart failure, also put a strain on the kidneys or blood pressure, they also have to consider the nephrological guidelines for people with kidney disease. In short, determining a safe and effective treatment for each individual HFrEF patient has become complex – and therefore requires time, which is often lacking in everyday medical practice.
19 Partners for a European Guideline Navigator
The GUIDE-AI consortium now aims to use large language models (LLMs) to guide physicians through the guideline jungle at a glance. In this project, research institutions are working hand in hand with small and medium-sized enterprises, patient advocacy groups, and partners from the pharmaceutical industry. These partners come from seven European countries and Israel. The guideline assistants are being developed in various European languages and are designed to take local regulations into account. The collaboration is coordinated by Matthias Gröschel, while AstraZeneca is responsible for project management.
The goal: Navigators that—integrated into hospital information systems or practice management software—compare prescribed therapies based on individual patient data with guideline recommendations and suggest changes in case of discrepancies. A hospital information system, similar to practice management software, is a computer program that displays data on individual patients to medical staff, including radiology images, lab results, and lists of prescribed medications.
Returning to the heart failure example: The HFrEF Navigator is designed to alert physicians in their practices or hospitals within the computer program if only three of the four recommended medications have been prescribed, or if a different dosage would be preferable for one of the medications. “Ultimately, the treating physician always decides on the therapy based on all relevant information,” says Matthias Gröschel. “But with the GUIDE AI Navigators, we want to highlight opportunities to optimize therapy according to the guidelines.”
Returning to the heart failure example: The HFrEF Navigator is intended to alert physicians in their practices or hospitals within the computer program if only three of the four recommended medications have been prescribed, or if a different dosage would be preferable for one of the medications. Focus on Four Common Chronic Diseases
The GUIDE AI team is initially focusing on four chronic diseases that are particularly widespread and place a significant burden on health, but which, despite established guidelines, are not always treated according to recommendations: heart failure with reduced ejection fraction (HFrEF), chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), and asthma. An estimated 160 million people across Europe are affected by these diseases. A specific navigator will be developed for each of the four diseases, with plans to expand the scope to include other conditions, such as inflammatory bowel disease.
The first step will be to assess the need for guideline assistants in medical practice. Subsequently, various AI models will be tested, and the most suitable one—or perhaps several—will be selected as the basis for each navigator. “Naturally, we place great importance on protecting sensitive medical data,” emphasizes Matthias Gröschel. “For example, the AI models should ideally be usable on local servers, but at a minimum, they must comply with European data protection regulations.”
Part of the project also includes a study that examines whether physicians, with the support of the navigator, actually prescribe guideline-compliant therapies more frequently. Finally, the program is also intended to provide information in layman’s terms – to guide not only medical staff but also patients in managing their illness.
About GUIDE-AI
The GUIDE-AI consortium unites 19 partner institutions from academia, healthcare, and the pharmaceutical industry, located in Germany, Italy, the Netherlands, Estonia, Slovenia, the United Kingdom, Switzerland, and Israel. It is coordinated by Charité – Universitätsmedizin Berlin. The project is funded by the Innovative Health Initiative (IHI) (Grant Agreement No. 101253015). IHI receives support from the European Union’s Horizon Europe research and innovation program, as well as from COCIR, EFPIA, Europa Bio, MedTech Europe, and Vaccines Europe.