IKNL uses artificial intelligence to determine the chance that a patient has metastatic breast cancer. The model has been trained on the basis of data about procedures, medicines, diagnosis and admissions from the National Register of Hospital Care (LBZ), according to a message on the IKNL website.
Breast cancer is registered at diagnosis within the Dutch Cancer Registry (NKR), but metastases that occur years later are not recorded as standard. Time is the biggest challenge, writes IKNL. Consult a data manager with an EPD to clarify the situation. That takes a few minutes per patient. Since more than 18,000 people are diagnosed with breast cancer every year, it is a time-consuming task.
Data
Application of AI should solve this problem. Data from the national basic register of hospital care (LBZ) of DHD (Dutch Hospital Data) for 2019 and 2020 are used for this. The model has been trained with data on procedures, medicines, diagnosis and admissions from the LBZ in combination with a previously developed ‘gold standard’.
Not black and white
The algorithm determines the chance that a patient has metastatic breast cancer. “That gives an estimate of how useful it is to look up additional data in the patient files of the hospitals,” explains researcher Linda de Munck. She points out that it does not give a black-and-white result, but helps to find a good balance between on the one hand finding out as much as possible about metastases and on the other not having to search for EPDs unnecessarily.