Call Check bills telephone consultations with AI at Amsterdam UMC

Since the corona pandemic, telephone consultations have rapidly become commonplace in hospitals. But how do you ensure that these consultations are also billed correctly? Amsterdam UMC recently started working with Bel Check. Project leader Annika Rausch talks about the need and how the system works.

A telephone consultation may only be billed by a hospital if it meets the conditions set by the Dutch Healthcare Authority (NZa). For example, the call must be medically substantive and offer a full alternative to a physical consultation. It is up to the physician, when reporting such a conversation, to assess whether these conditions have been met. And that is where things still too often go wrong, Annika Rausch observes. "That is also logical. Doctors don't know the NZa rules by heart and they don't have to."

Samples

Before the corona pandemic, the number of telephone consultations was up to 50,000 per year. Today, that number is around 300,000. A huge increase that the existing systems in most hospitals are not up to. The workload in finance departments has increased accordingly. Employees now decide on the basis of random samples whether calls may be declared, simply because it is impossible to check all reports manually.

And so technological support was sought. Simple, rule-based algorithms without AI were developed, but they turned out to have too many errors. Says Rausch, "If a report says 'no answer,' the existing algorithms don't bill that call because a call may not have taken place. And rightfully so. But if a physician copies that term in a subsequent report, for example as a reference, that conversation is also not billed. And then the conversation cannot be declared. Wrongly so."

Call Check as a solution

The AI team at Amsterdam UMC developed a smarter solution with the healthcare administration department: Call Check. This AI application reads medical records differently. A system with AI better understands what a report is about. It not only looks at individual words or set lines, but also at the meaning and context of the report. Says Rausch, "The system appears to be just as capable as our staff of correctly assessing calls and assigning them to the correct claim category. Much more accurate than the solutions without AI."

According to Rausch, the gain is not directly in staff savings, but mainly in quality and completeness. All calls can now be reviewed and staff members are left with time to focus on the complex cases for which even the AI model is uncertain.

Validation and scaling

Currently, Bel Check is being validated in several departments within the hospital. Rausch: "Especially in the ENT department, a lot of gains are being made. There, the algorithms were not working well all the time because many reports said 'no hearing.' Bel Check understands the context and doesn't make that mistake."

Accurate claiming is essential, Rausch stresses. "You can compare it to your tax return. If you make mistakes or forget things, you can lose money or have to pay back later. A good system helps you declare everything correctly and completely, so you get what you're entitled to and don't get any surprises."

Open for collaboration

Amsterdam UMC likes to share knowledge gained with other hospitals "During a recent meeting of a knowledge network, I presented our AI module. The interest was great. Hospitals recognize the challenge and immediately see how valuable a solution like this can be."