Project Update CARRIER – preventing cardiovascular disease by early detection and personalized intervention

Cardiovascular diseases are a major cause of death in the Netherlands and other western countries. The most common cardiovascular diseases is coronary heart disease (CHD). Lifestyle plays an important role in the development of CHD. The aim of CARRIER is to use big data to detect CHD at an early stage and to prevent it by means of personalized lifestyle interventions. This ambitious program started in March 2020: we spoke to clinical data scientist Iñigo Bermejo for an update.

Branching out from oncology

Dr. Iñigo Bermejo is a member of the research group of professor André Dekker at Maastro, a renowned center for radiotherapy. The CARRIER project builds on expertise that was developed in projects related to oncology. “The theory behind prediction models is similar across disease areas” says Bermejo. “The experience we have from oncology-related projects has proven very useful in CARRIER.”

Using a federated learning framework

The prediction model in CARRIER works with data from CBS (Central Bureau voor de Statistiek), general practitioners and hospitals. Combining data from these three sources can be complex. “In the project, we work on the legal and ethical aspects of accessing and using this data in a responsible and secure manner,” explains Bermejo. “We use the Personal Health Train, which is a federated learning framework. Essentially, we bring our algorithms to the data instead of the other way around.” Personal Health Train is designed to give controlled access to data, while ensuring privacy protection.

Other data challenges

A challenge lies in correlating data from different sources. To do that, the CARRIER team is working on the application of secure multiparty computation. Dr. Bermejo: “By using an existing method, we can perform computations on encrypted data without first decrypting it. As the data stays encrypted, this greatly reduces privacy concerns.”

Hospital data comes with particular challenges, he says. “It often comes in the form of free text. To be able to extract relevant data such as a family history of disease, we need natural language processing (NLP) techniques. We use a deep-learning based tool developed by CTCue. I expect similar tools to be used in research more and more, because a lot of relevant data is not stored in a structured form.”

A highly personalized approach

The CARRIER team is developing personalized lifestyle interventions to reduce the risk of cardiovascular disease onset (primary prevention) or to prevent it from getting worse (secondary prevention). “We aim to tailor lifestyle interventions to the patients’ preference, as customized advice has a higher chance of being effective in changing behavior. Compliance with the lifestyle intervention will be promoted and monitored by means of a digital multimedia lifestyle coach (eCoach), which is developed by our partner Sananet.”

Two models are being developed in CARRIER: a prognostic model that aims to identify patients at risk, and a predictive model aimed at predicting the effect of a given lifestyle intervention. The latter is particularly useful in supporting the decision-making process, says Bermejo: “Communicating the effect of a given lifestyle is of great importance, especially in primary prevention. In the project, we research different risk communication techniques to motivate change. We may have to employ different techniques according to level of education. This highly personalized approach is ambitious, but we need to be bold if we really want to change lifestyle attitudes.”

Current state & future outlook

The project is still in its early days, dr. Bermejo explains. “We are working on the data infrastructure, designing the intervention and patient participation workflow et cetera. We have also made a start with the prediction models and the customization of the eCoach. A lot will happen in the coming six months.”

An important question in the design of the study is how to measure success. Ideally, researchers would do a 20 year follow-up study to see if and how cardiovascular disease develops in patients. As that is not within scope of the project, different ways of measuring the effect of lifestyle changes need to be found.

Bermejo believes the CARRIER project will be helpful in making the shift from the traditional approach to cardiovascular disease management (trying to minimize the effects of the disease once it is present) to an approach of primary prevention aided by digital technologies. “This shift will happen in the next 20 to 50 years. I hope the CARRIER project can show people what can happen if we improve interconnectedness and safe access to different data sources.”

More information

14 June 2021


Sign up for our newsletter and stay informed of the latest news Commit2data