Laboratory results, surgery reports and other information are continuously collected on treatment situations when patients are treated at different units in the wellbeing services county. By studying this so-called real-world data, i.e. Real World Data (RWD), we can gain a better understanding of the true effectiveness of treatment. This information is helpful when choosing treatments for future patients.

While a patient population that is strictly limited and suitable for a clinical trial based on inclusion criteria, real-life data is available for a much larger number of patients and over a longer period of time. Clinical trials often focus on testing drugs and treatments, and real-life data indicates effectiveness and patient satisfaction, among other things. Simply put, while clinical research has shown that treatment is effective against the disease at least to some extent, it is only in practice that the treatment given to patients tells what the real effects of the treatment are.

Aiming for informed decisions

In cancer treatments, real-world data essentially complements the results of clinical trials and helps to understand which group of patients will benefit most from the treatment. In addition, information is obtained on which patient group the treatment may even significantly impair the quality of life, in which case the disadvantages of the treatment outweigh the potential benefits. This information is helpful for the treating physician, whose goal is to individually select the treatment that benefits the patient in question.

The wellbeing services county of Pirkanmaa has  a team of research informatics services (clinical informatics), whose core task is to help clinicians reach health information relevant to patient care.

“The Clinical Informatics Unit is absolutely necessary, because no one else can utilize the existing data. The team is already paying for itself by being able to reduce the number of treatments that do not actually benefit the patient, says Chief Physician Annika Auranen, Director of the Tays Cancer Center,.

Typically, the idea for research comes from clinicians who consider the health and economic effectiveness of different therapies and what could be done better.

 “We can investigate, for example, whether patients treated with a certain treatment procedure return to working life better than others,” says Data Manager Leena Hakkarainen, who leads the clinical informatics team.

The clinical informatics team is an indispensable link

Real-world data accumulates from several different health information systems. Data from fully electronic systems are available fairly comprehensively from 2007 onwards. The data ends up in so-called data lakes. With the wellbeing services county, material from more and more systems will end up in data lakes.

However, the data ending up in data lakes cannot be utilized by researchers as such, but essential data must be mined for each research and its objectives. Typically, for example, information is retrieved from text entities with the help of certain descriptors and processing codes given to the computer, i.e. algorithms. Data protection and the way the question is posed also play an important role. The clinical informatics team, as part of other research support services, helps with all this.

– We aim to communicate more so that researchers on the Tays campus know about our services and that they are free of charge for our clinicians, Hakkarainen says.

In information retrieval, the informatics team serves researchers both domestically and internationally. There are a lot of requests, so prioritization is necessary. Projects to be  prioritized include  the  ongoing so-called OOO studies in the wellbeing services county of Pirkanmaa, i.e. the right treatment, for the right patient, at the right time. For national surveys coming from Findata, which usually contain data subject from several organizations, the information must be submitted within 30 working days  as required by law. That’s when other projects must wait . There are 5–15 such requests per year. According to Hakkarainen, biobank studies in which certain kinds of variables are searched from genomic data and health data are sought alongside samples are ongoing all the time.

Security as a priority

Data security and the lawful handling of sensitive patient data are essential. Health system data is not usually transferred outside the wellbeing services county, but algorithms are created for different research purposes to retrieve the necessary information. Research algorithms can be sent simultaneously to Tays’ informatics team and around the world.

The algorithm is like a battery of questions from a researcher that has been turned into code. For example, it can be used to study how many cancer patients have been given a certain medicine over a certain period of time and how their laboratory results have changed since then. Hakkarainen’s team modifies the algorithms related to external search requests as needed. After searching, they check that the search result is of high quality and that the code has searched for what it is supposed to search. Only those search results that are considered relevant and to which the research in question is entitled according to the data permit will be disclosed to the research. This also ensures that no individual patient can be identified from the data.

A secure virtual workspace for researchers has been built at Tays that meets data protection laws, where researchers can safely process and  analyze the data of their own research. Only the information that falls within the  scope of the data or research permit in question is available to researchers in the working environment of their research.

Information retrieval is developed in cooperation with university researchers

Retrieving and combining data from different systems does not happen at the click of a finger, so the clinical informatics team also collaborates with researchers at the University of Tampere. Professor Mark van Gils and his group are developing algorithms that utilize multiple data sources, such as images, patient monitoring data, text and biobank data.

In addition to different data sources, the disparity in the data of the same sources also causes gray hairs.

“Different countries have different devices, sources and formats, or different ways of recording things, so combining data is challenging,” van Gils says.

Van Gils says many research groups focus on individual sources, such as magnetic resonance imaging, electrocardiogram (ECG) or genome-related analysis. However, the big challenge is combining data from different sources. The teams of Hakkarainen and Van Gils have created use cases for which researchers may need information from different sources. Hakkarainen and his partners provide data, and van Gils’ group is looking for solutions for combining data to support clinical decision-making. Planning has been carried out during 2023 and the process is ongoing.

Towards harmonization

One step towards producing unified data is the international OMOP data model (Observational Medical Outcomes Partnership). Simply put, the model means an internationally uniform code list of diagnoses, medicines and procedures, among other things, that is used to translate local medical terms into internationally recognizable ones.

“The OMOP model is essential for us to be able to respond quickly to international data requests,” says Hakkarainen.

However, there is still no hierarchy or automatic modelling method for patient texts. Many things that are valuable to medicine are recorded in free-form text, so mining data also requires a lot of manual work and collaboration, as well as clinicians’ knowledge of recording methods and concepts, as well as the expertise of data experts to collect the necessary things for research. In cancer treatment, the hope would be to have a way to record the chosen treatment line and its effectiveness.

“Now the effectiveness is visible in life expectancy data, but the aim is to harmonize the recording of data so that information on quality of life, harms and different treatments given could also be obtained automatically from information systems in simple ways. For this reason, too, the harmonization of clinical data should be developed with particular vigour, Annika Auranen emphasizes.