In the care sector, medical research relies on anonymous data. This data is also used as statistical input for developing healthcare policies. A typical example of this is the anonymised data that the Belgian healthcare institutions provide to Sciensano/Healthdata for epidemiological monitoring.
When you anonymise data, it is no longer possible to trace the person who is the source of the data. However, there is still some risk of identification. For example, it would not be difficult to use the date of birth of a 100-year-old patient to uncover their identity.
Combining two (or more) data sets may create some risk of identification. It is always good practice to carefully consider, in advance, which and how much data you will exchange, even if it is anonymised.
In certain situations, however, you may need to make data available to a third party, but then want to enable re-identification afterwards under strict conditions. Here, anonymisation is not an option.
One example could be radiology images that you have analysed in the cloud using an AI algorithm for triage or other purposes. Or an external algorithm that uses patient data to calculate which components are needed to perform a knee replacement. Or text and unstructured reports that you make available to a software application that processes and analyses the data in order to optimise care paths.
In order to share medical data safely and responsibly in such situations, you can pseudonymise it. This security measure involves deleting any personal data that can be traced back to the patient (name, date of birth, patient number, etc.) and/or replacing it. For example, you might replace date of birth with age, and municipality with region. The data is then linked to a code that can only be used in the hospital to identify the patient.