How machine learning keeps ambulance transport on schedule

By and Michele Coiutti | 2 minute read | June 29, 2021

Ambulance

Each year Croce Rossa Italiana Comitato di Udine (the Italian Red Cross of Udine) transports thousands of patients to and from medical appointments by ambulance. We call this service secondary ambulance transport to distinguish it from emergency response. The trips can be between the patient’s home and a healthcare facility or from facility to facility.

Scheduling such trips is a complex challenge requiring excellent logistics skills. If planners inaccurately predict the travel times, CRI Udine cannot provide quality patient service. And too many ambulances tied up in secondary transport can inhibit our emergency response.

Ambulance transport times predicted with 98% accuracy

Travel time on the road is easy to predict with mapping apps. Much harder is estimating the time it takes to get patients in and out of the ambulance. Planners must consider the patient’s condition, age and weight, what floor the patient is on and must be delivered to, the necessity of IV fluids or oxygen, whether the patient needs a stretcher or wheelchair, and much more.

Today, we’ve digitized ambulance scheduling thanks to a first-of-a-kind expert planning system developed with IBM Business Partner IT’S…B2B. Called SoTras, the system uses IBM predictive analytics to analyze records of 200,000 prior ambulance trips. The insights help schedulers predict transport times with 98% accuracy, far more precisely than before. As a result, CRI Udine can better serve our patients and the community.

A hybrid cloud supports patient booking and transport

Developed as a hybrid system integrating cloud, on-premises, web, mobile and human elements, SoTras is a complete solution for transport logistics and patient booking.

For the planning interface, IBM SPSS Modeler applies machine learning to the prior-trip database to predict the time to get patients in and out of the ambulance. The AI engine assesses variables about patient demographics, medical conditions, home and medical facility infrastructure and the impact of needed patient support. Google Maps adds the ability to estimate on-the-road travel time.

SoTras assists with booking by collecting patient transport requests from a public web link to the healthcare system of Udine Province. It also includes a mobile app for drivers to report delays and the ongoing trip status. Integration with Italy’s national healthcare system offers data collection, driver shift management and invoicing of completed trips. A private cloud interface offers system administration.

Better transport scheduling delivers quantifiable benefits

With SoTras increasing planning accuracy, CRI Udine has fewer ambulances on the road at any given time. Plus, optimizing the secondary service frees ambulances for emergency response.

The benefits are quantifiable, as seen from a system audit performed by the Udine Chamber of Commerce. In a single day with 101 secondary transports, the service needed four fewer ambulances and crews than usual due to planning efficiency. The 18 ambulances used drove 118 km less than is typical for the same trips.

Analyzing these and other metrics, we expect to reduce the fleet’s travel by 46,000 km annually, lessening diesel fuel consumption by 7,077 liters and CO2 emissions by 18 tons. And fewer transport delays deliver a crucial measure of success: a superior experience for the patients of CRI Udine.