Agile Bi


Machine Learning in Public Health

Understand what influences hospital cost

A steadily climbing health expenditure has resulted in a more complex budgeting process. The budget team needed a mechanism to better understand what influences hospital cost and length of stay, two of the most imperative factors when deciding the future budget.

“Managing the scare resources of staffing and facilities is an ongoing challenge for any health service. This data-driven approach enables more effective planning and utilisation so more patients can be supported.”

Too many variables to consider

Too many variables to consider

One of the most challenging tasks the budget team faces is deciding how many factors to consider when building a new budget plan. There are a myriad of variables that can affect budget with a varying degree of influence. The team had to prepare a number of scenarios manually, and choose the ones that appeared most plausible. The traditional approach is lengthy and expensive. Besides, it is prone to bias and less responsive to evolving environmental conditions.

How to better forecast?

The budget team relied on simple forecasting method for creating a budget plan. They assess what occurred in the past, and makes a reasonable extension to estimate what occurs next. Such method did not provide adequate accuracy to respond to a complex and changing environment such as public health system.

Machine Learning helped understand what affects cost  

AgileBI built a predictive model that helped identify what factors most affect the cost, and by how much they have influenced on key factors such as hospital length of stay. We used the latest advancement in the Machine Learning field known as a neural network for developing the model.

We created an interface in Power BI for the budget team to interact with the model. In this interface, the team can configure various factors, such as in which hospital the patient receives treatment, it will perform simulation using the inputs and display the outcome graphically.

ML Evaluation cropped ML Evaluation cropped

Results and future plans

The team now have a comprehensive set of tools for performing budget and forecast. They can now better understand influential factors to the cost and length of hospital stay.

The next step is to incorporate external data such as weather, census and economics. This makes the model more complete when forecasting the cost.

Results and future plans


Let us show you how you can make your data work for you.