Project Description




The Emergency Room (ER) at Zuyderland Medical Center receives, examines and treats all patients with emergency conditions. It is the most active department within the hospital with a constant flow of incoming patients suffering from an acute illness or injury. The ER is typically staffed with nurses and physicians who treat a wide spectrum of medical conditions, including cardiac arrests, strokes, serious injuries, pregnancy complications, asthma and various other conditions.

To enhance its patient care, Zuyderland aimed to use AI and machine learning to anticipate patient admissions to the ER. The nature of this ward is often associated with a high-stress medical environment, where every second counts. By accurately predicting patient admissions, Zuyderland wants to allocate resources efficiently and ensure they have the right number of medical staff, beds, equipment and medications available to immediately address incoming cases. During this project, a Data Science expert from Mediaan Conclusion supported Zuyderland with:

  • Providing bi-weekly (short-term) forecasts of patient inflows, taking into account variables such as weather conditions and holidays.

  • Developing a user-friendly dashboard to visualize data and facilitate decision-making.


The main objective was to ensure that ER patients received the necessary care. It was vital to sort patients through triage codes, reflecting the severity of their condition. An “emergency” can be something major, such as a car accident, or something less serious, such as falling down the stairs or suffering minor burns from fireworks on New Year’s Eve. The ER figures out how urgent each case is and treats patients accordingly. In the worst situations, patients go to the Intensive Care Unit (ICU). By knowing how many patients are coming in and what codes they have, Zuyderland can determine who needs immediate attention. They can also decide which medical staff, beds, equipment and medications are needed immediately. So, it made sense to take factors such as weather conditions and holidays into our predictions. For example, does the hospital need more manpower during New Year’s Eve or receive more patients during rainy days?


As we took into account variables like weather conditions and holidays, we had to make short-term bi-weekly forecasts to ensure the accuracy of our predictions. To do this, we needed models that could quickly adapt to this type of data. Another challenge was that our historical data was limited, spanning only a few years. For instance, in a year with no snow, it was impossible to make reliable predictions since we had no relevant samples. This limitation naturally had an impact on our prediction results. Conversely, when we had complete data, our model was able to make accurate predictions. Another challenge was breaking down our forecasts by specialty within the hospital workforce to determine the manpower needed at specific times.


Throughout this project, we maintained a close collaboration with our client. The following techniques and technologies were used:

  • Python

  • Time-Series Forecasting

  • Machine Learning

  • Microsoft SQL

  • Tableau

  • Docker


The information from the predictions gives Zuyderland insight into how to optimize patient care in the emergency room so that patients of varying severity levels receive the proper attention they need. The user-friendly dashboard offered easily understandable information, allowing the client to take preventive measures and manage & allocate their resources efficiently. Additionally, they also gained valuable insights into how variables like weather conditions and holidays affect admissions to the ER.