Explainable artificial intelligence for individualised treatment selection using real-world data

Patients with the same disease do not always respond to the same treatment in the same way. Differences in age, disease characteristics, other health conditions, and many other factors can influence how well a treatment works. This creates a need for better ways to support more personalized treatment decisions.

The research investigated how artificial intelligence and routinely collected healthcare data can be used to support more individualized treatment selection. The work was carried out across four publications focusing on chronic diseases. Artificial intelligence models were developed and evaluated to predict health outcomes and compare how different treatment options may work for different patients.

An important part of the research was making the models easier to understand. Instead of only producing predictions, the methods were also used to show which patient characteristics influenced the results. This can help researchers and healthcare professionals understand how the models reach their conclusions and assess whether the model outputs are clinically reasonable.

The results showed that real-world healthcare data can provide valuable information about disease outcomes and variation in treatment response. The findings also demonstrated that artificial intelligence could go beyond predicting what may happen to a patient by comparing possible outcomes under different treatment options. This comparison can help identify which treatment may be better suited to an individual patient.

Overall, the research shows how real-world health data, artificial intelligence, and understandable model explanations can be combined to support more personalized healthcare. By taking individual patient characteristics into account, these approaches can contribute to more informed, transparent, and patient-specific treatment decisions in the future.