Monash University
/Monash Health


Healthy Forecasting:
Length-of-Stay (LOS) Prediction Models Relieve Patient Overflow

No one likes to wait in the emergency department of a hospital, even when they aren’t sick. But when illness is involved, fast accessibility to healthcare can literally be critical—especially in an emergency.

That’s why Klarrio, in partnership with Monash University’s Dr Joanne Enticott and PhD Student Kushan Ranakombu, built length-of-stay (LOS) prediction models for Monash Health District. Models were designed to allow the hospital to accurately forecast Length of Stay in the emergency department, in addition to predicting subsequent hospital admission rates.

Download the Case Study


Build Length-of-Stay Prediction Systems to Improve Efficiency

Patients today are coming to emergency departments and being admitted to hospitals with multiple and increasingly complex healthcare needs. Hospital resources are already stretched beyond their designed capacities, and at the same time, government health expenditures are growing at rates significantly greater than GDP.

Monash Health wanted to address the situation proactively, and opted to improve hospital efficiency through technology. The hospital provided Klarrio and Monash University with de-identifiable data sets that detailed every patient and emergency department admission over a five-year period. To maintain privacy, patient names and street addresses were not provided.

The information covered a diverse population of 2 million and included more than 900 admissions involving an estimated 350,000 unique patients. It also detailed 100s of data fields including demographics, medical records, emergency department diagnoses, and more.

Klarrio was asked to develop innovative real-time applications that used cutting-edge machine-learning techniques to alleviate the stress on the system.


Leveraging numerous machine learning algorithms, including gradient-boosting machines and neural networks, Klarrio built models to:

  • Predict LOS in the emergency department and subsequent hospital admissions
  • Predict the probability of a patient returning to the emergency department within 7 days
  • Forecast the number of patients likely to present to the emergency department in every 4-hour period during the week ahead

Utilising new methods borrowed from the field of game theory, Klarrio was also able to explain the root causes of patient LOS, and to drill into the drivers of potentially preventable hospitalisations (PPH) for specific diseases, such as cardiovascular illness.

The Expertise

  • Machine learning
  • Data analytics
  • Field game theory

The Technology behind

  • Gradient boosting machines
  • Neural networks
  • Data streaming

“Don’t be afraid to give up the good to go for the great.”

— John D. Rockefeller


The Monash Health District models can be leveraged and extended to build real-time dashboards that will give hospital administrators the ability to better plan resources.

Dashboards can provide hospital management with both historical and predictive views on the status of multiple admissions criteria in real time, in addition to accurate length-of-stay predictions to manage patient load more efficiently.

The ultimate goal is to maintain a high quality of care , both in the emergency room and hospital-wide, despite ongoing changes in technology and increasingly complex healthcare needs.

unique patients
of data fields

Ask for a demo!

We’re your one-stop cloud-native partner

Big Data engineering. Data science. Data Analytics. Site Reliability Engineering. Consulting. And customized Open Source projects for companies of all sizes. Learn more about what Klarrio can do for you today.

Contact us

Other Projects

Just a few projects examples.