Bronchopulmonary dysplasia (BPD) is a common and serious lung condition that affects premature babies, especially those born before 33 weeks of pregnancy. It is the most common health issue among newborns in the neonatal intensive care unit (NICU). Although efforts have been made to predict BPD, these attempts often do not reflect current patient populations and lack validation in real-world settings. There is not yet a comprehensive tool to help doctors identify and manage these babies effectively.

A better predictive system could help doctors identify babies at risk earlier, leading to more timely interventions and better care. It could also make the entire healthcare process more cost-effective and efficient. However, developing a tool that can be used in practice requires overcoming some challenges. The predictive model must work well with the data available at the time of diagnosis, which must be easy to collect, reliable, and consistent across different patients.

We are creating a simple, machine learning-based tool that helps doctors predict the risks and outcomes of premature babies, ultimately improving decision-making and patient care.

The development of a decision support tool in healthcare faces several challenges, particularly when it comes to ensuring its adoption in critical care environments. For such a tool to be effective, it must be intuitive and easy to use, especially under pressure. It should also provide timely predictions based on data that is readily available in the moment.

The transition from a research-based model to a fully operational, widely accessible service requires careful planning and significant resources. The research and design process for this tool focuses on several key steps. First, we explored the ideal timing for utilizing the tool to make predictions, ensuring that it can be used effectively when decisions are most critical. Now, we are experimenting with different machine learning models, carefully evaluating their predictive accuracy to identify the best approach. Once we have identified the most promising model, we will design and implement a prototype, testing its real-world effectiveness in healthcare settings to ensure it meets the practical needs of clinicians.

Throughout this process, the goal is to create a decision support tool that not only improves prediction accuracy but is also seamlessly integrated into everyday healthcare practices, helping providers make better, faster decisions for their patients.