Closing the Gap: Enhancing Model Use in Healthcare

Prediction models provide estimates, such as breast cancer recurrence risk. These models combine large amounts of scientific evidence to calculate these estimates. Using these models can help clinicians and patients make better informed decisions and to tailor care better to individual patients’ risk profile and preferences. Long-term clinical utility requires 1) regular updates and sometimes model extension to include new markers, 2) thorough evaluation, and 3) integration in clinical workflows. The PREDICT-NL project addressed these three key areas and our efforts have yielded several key outcomes. Firstly, we conducted an exhaustive review of scientific literature to evaluate and expand prediction models, ensuring their ongoing relevance and reliability. This process resulted in the development of a practical guide to assist model developers in selecting the most effective methods for updating and expanding their models. In the context of breast cancer, we applied these methods for updating and expanding prediction models. Our proposed model updating and extension framework will enhance the accuracy and applicability of the models in real-world clinical scenarios in the long-term. Additionally, our research identified crucial factors influencing clinicians' decisions regarding the adoption of prediction models in their daily practice. Understanding these factors is essential for developing effective implementation strategies. We also identified existing barriers and opportunities for leveraging IT infrastructure to streamline the integration of predictive models into clinical workflows. This exploration provided valuable insights into optimizations required to maximize the utility of prediction models in healthcare settings. Lastly, our investigation of patients' preferences regarding outcomes to discuss with healthcare providers and their understanding of different presentation methods for prediction model estimates is essential for promoting patient engagement and shared decision-making in clinical interactions. Overall, the PREDICT-NL project has contributed important insights that will help us to advance the effective utilization of predictive models in clinical practice.