Facebook IconFacebook IconFacebook IconFacebook Icon

How AI can reduce hospital revenue leakage

Sagacify Logo

Challenge

Each time healthcare providers administer care services or medications, they have to manually encode those, so that they can be billed to patients. Occasionally, certain medical procedures or medications are forgotten: for instance, when an infusion is prescribed, a blood bag should always be prescribed as well, but in practice it is not always the case. These omissions result in a loss of revenue for the hospital, which becomes more significant the more frequently prescriptions are forgotten. 

Saint-Luc, the largest hospital in Brussels, asked Sagacify for a solution to avoid potential revenue leakage. 

Key elements  

  • Designed and trained custom AI models to find all invoicing errors, from the combination of  the inami codes, and invoice by invoice
  • Co-creation of an app allowing the financial team to drill down on the data and find where the invoicing errors are. 

Solution

To solve the problem of spotting invoicing anomalies, we split the anomalies detection into two different models: 

  • The first model checks each invoice individually to validate the invoiced prestations against a predicted “patient journey”
  • The second model monitors the time series of all prestation types (INAMI codes) and a validates that the observed variations are normal

Since the integration of the models is of crucial importance in this project, we also developed a user interface in co-creation with the client, in order to let the financial team explore different dimensions of the data and find the reasons behind the spotted loss of revenues.

Results

In production, around 70% of the anomalies pointed out in the top 50 turned out to be real errors, which greatly facilitates the work of the financial team, and helps them spot errors that would have gone unnoticed otherwise. 

Thanks to our collaboration, Saint-Luc is now able to detect anomalies in invoice and can rectify the problems with a significantly reduced effort.

Related cases