AIR_PTE

AI-based risk forecasting for assessing responses to treatment

Application industry: Health
Technology area: Machine learning

AIR_PTE is developing artificial intelligence (AI) methods to predict the effects of treatment options, taking into account individual health risks. Health insurance data from Canada and Germany provides the source material.

The baseline
As advances are made in medical research, more and more innovative and specialised treatment methods are becoming available. This positive development, however, presents doctors with the difficulty of deciding on the best option in each individual case. The effectiveness of treatment depends on various factors such as pre-existing diseases and demographic data, so the consequences of a particular treatment may vary greatly from one person to the next. Today, decisions on therapy are usually made on the basis of experience, published studies and guidelines. In practice, however, they do not always apply to the actual patients to be treated.

The project goal
Accounting data from statutory health insurance schemes covering the past ten years provide the opportunity to scrutinise large numbers of cases, including rare constellations, for correlations between disease progression, treatment paths and treatment outcomes. The aim is to apply innovative AI methods to this pseudonymised data in order to obtain statistically significant proof of the effects of different treatment options. Machine learning is used to detect temporal patterns from which new characteristics and reliable forecasting models can be derived. Venous thromboembolism and its various treatment options is being used as the pilot indication. Over the course of the project, the methods will be generalised for use with other diseases. AI-based risk prognosis is intended not only to support decision-making in the choice of therapy but also to help identify target groups for innovative care models so that cost-intensive treatment options can be better targeted.

Application and practical benefits
The project will result in a Rapid Evidence Repository (RER) for use at the point of care in compliance with data protection regulations. The doctor or health care professional will enter the characteristics and treatment goals for the specific patient anonymously into a mask and will then instantly receive information about the potential effects of the treatment options available. To allow the doctor to understand how the individual prognosis was generated, all the characteristics that had a significant influence on the information are displayed. The RER process will be used only with the patient’s consent.

Term: September 2020 to August 2022

Consortium: DCC Riskanalytik GmbH (lead member), InGef Institute for Applied Health Research, McGill University Montreal, Dept. of Medicine (Canada), Macadamian HealthConnect (Canada)

Contact:
Prof. Thomas P. Zahn, DCC Riskanalytik GmbH

Website:
www.risikoanalytik.de 
www.ai-evidence.de