How artificial intelligence helps us check bills

Humans and machines team up

Artificial intelligence (AI) is changing the way we work – including at CSS. In 2021, we checked 23.6 million bills for basic and supplementary insurance. Such a huge volume cannot be tackled by manual checks alone. Our innovative system for verifying benefit entitlement processes over 85% of the bills automatically. We are able to avoid more than CHF 660 million in benefit costs each year that are wasteful and do not reflect the true cost to our insured persons, partly thanks to the increased use of AI.

Hybrid intelligence in checking bills

When it comes to managing benefit costs, we are constantly expanding our expertise and adopting technological advances. This is key to keeping pace with the increasing complexity of the healthcare market. By combining the abilities of humans and machines, we are able to draw on the respective strengths of both. "This 'hybrid intelligence' is crucial in managing benefit costs efficiently and effectively," says Karin Tremp, our Head of Benefit Costs and Rules Management. "Only by continually developing our manual and automated verification and analysis processes can we successfully cope with the volume of bills, which is increasing year by year, and keep premiums as low as possible for our clients."

High healthcare costs due to inefficiency

Healthcare spending in Switzerland totalled CHF 82.5 billion in 2019¹. According to international studies², more than 10% of this amount is attributable to inefficiency. This includes, for example, undetected billing errors, unnecessary administration costs and deliberate abuse of the system by service providers and clients. These wasteful costs are partly responsible for the continuing rise in healthcare expenditure and health insurance premiums in Switzerland. Our benefit costs management aims to identify these inefficiencies and apply countermeasures.

Minimising benefit costs with the help of artificial intelligence

When it comes to checking benefits, we are making more and more use of AI algorithms. One way in which they help is by preventing us from mistakenly making multiple payments for benefit statements that have been submitted twice. AI-based control logics can simultaneously draw on various sources of information to assess the facts and process them in real time. Compared with humans, they are capable, for example, of reaching a decision recommendation within a fraction of a second while taking several billing documents into consideration. We rely on AI-based control logics of this kind for checking accident-related claims for instance, and for recognising billing items and quantities that are noticeably high.

Recognising discrepancies in billing information

In addition to the AI algorithms, we also use established data analysis methods to avoid unnecessary healthcare costs. These analyses enable us to spot discrepancies in benefit statements retroactively. For example, they let us see if the costs billed by a service provider are conspicuously high. In this case, we investigate further and intervene if our suspicions are confirmed.

An example

After family doctor C handed over the practice to his successor, the gross benefit costs billed per patient rose by 195%, from CHF 380 to CHF 1,120. The costs were also no longer comparable with those of other service providers in the same region with a similar group of patients (family doctors A and B). Thorough analyses provided us with the data we needed to prove that this behaviour is fraudulent and to take appropriate steps.

Use of AI in our clients' interests

We encourage our employees to develop their professional, methodological and cultural skills in order to improve the effectiveness of our benefit costs management. We achieve this through increased data literacy and interdisciplinarity, and by developing a digital workplace culture. "The trust of our clients and employees is the key success factor here," emphasises Karin Tremp. "We make sure that progressive automation and the use of AI is in their interests. Furthermore, our clients can trust us to act in accordance with the statutory basis for checking our obligation to pay benefits and to observe data privacy when using their billing data."

Sources:

¹ Statistical Information of the Federal Office of Public Health, 28.10.2021

² Trageser & Vettori et al., 2012, Akademie der Wissenschaften Schweiz;

² European Healthcare Fraud and Corruption Network (EHFCN), 2016.