“With AI, SIDIAP data can be used to validate risk prediction models and thus work on their implementation in the health system.”

We spoke with Maria Aragón, SIDIAP coordinator, about the evolution of the platform, its impact on research and the challenges that lie ahead..

Let's start from the beginning. You trained as a computer engineer. Were you always clear that you wanted to dedicate yourself to it?

When I was little, I was very good at languages and philosophy; I had a natural aptitude for them. However, my real interest lay in understanding things that presented a personal challenge, such as mechanics and mathematics. Additionally, my older brother had a computer when I was still in school. I struggled with tasks like installing programs, playing games, and connecting to the Internet. As I grew older, I found myself using it to chat with neighbours. So, when it was time to choose a field for high school, I was already leaning towards computer engineering.

How was your arrival at IDIAPJGol and your first contact with SIDIAP?

I was working at the UPC when a colleague told me that there was a computer science position at the Institut. I got interested and, since I really liked databases, when they told me what SIDIAP was, I thought it could be a good fit. And said and done, I started to work as a Data Manager and to understand what Primary Care is under the guidance of Dr. Elorza

How has SIDIAP evolved since you started working?

When I arrived, SIDIAP was very young, there were manual processes and it was not yet widely known internationally. In Spain, and in Europe in general, the search was beginning with real data(Real World Data). Now, the number of research projects with Catalan participation, their level and impact has increased greatly. Thanks to this, the team behind it has also grown, we have improved the documentation we send to researchers and also the technology we use. For example, the data update time has been reduced by half, which is now semiannual. We have also enhanced quality indicators, expanded the sources available for research, and increased our participation in international Common Data Models (CDM).

And what do you think has been your personal contribution?

Over the years I consider myself a good translator between research teams and their needs and technicians. On a personal level, I have worked a lot on the automation and optimization of SIDIAP's technological processes, on the national and international visibility of the platform and the team's work, which is much more than just providing data, and on promoting technological projects from home.

Would you highlight any specific study in which SIDIAP has been key to taking great steps forward? SIDIAP data have been used in many different pathology studies that have achieved great impact.

I do especially remember the time-trial projects that were carried out at the beginning and during the COVID-19 pandemic, when we were all confined, where patients were first characterized, then waves and complications were derived and, afterwards, assessed for adverse effects in the vaccines.

However, before and after the pandemic, high-impact studies have been carried out, such as the REGIPREV study, the first project with SIDIAP data in 2011, the APRES, where it was seen that the use of antibiotics in some cases can be counterproductive in terms of resistance, or the 4E, which evaluated the effectiveness of statins in the elderly population.

You can see all the publications with SIDIAP at: https://www.sidiap.org/index.php/ca/activitat.

Are you already applying Artificial Intelligence (AI) in SIDIAP?

Yes, we are currently collaborating with researchers to apply AI algorithms in research projects within Primary Care. We are working on free text (NLP), prediction modeling, and the development of Deep Learning algorithms within databases to ensure pseudo-anonymization.

What challenges do you think AI will allow SIDIAP to achieve? 

I think that with AI, and with explainability techniques (XAI), we will be able to find relationships that are not yet preconceived and this makes it very exciting. SIDIAP data could be used to validate risk prediction models and thus work on their implementation in the health system.

Thanks to AI, it will also be much easier to guarantee that the data requested for research is risk-free and thus simplify its access.