Applications of artificial intelligence to improve revenue cycle management in healthcare hold promise, but executives are concerned about the technology’s accuracy and reliability.
These were some of the results of an Inovalon questionnaire of more than 400 revenue cycle and finance executives and managers, 84% of whom said they were optimistic about AI-driven RCM in hospitals.
However, a third of respondents indicated they were concerned or skeptical about using AI in RCM. The biggest concerns were concerns about accuracy and reliability (31%), lack of awareness/understanding (17%) and AI being too new/untested (15%).
Humans better than AI
Twenty percent of respondents said they are confident that human performance is – at least currently – better than AI.
Julie Lambert, president and general manager of the provider at Inovalon, said Healthcare IT News That applies to RCM, but there are certainly areas that can benefit more from AI.
“Rather than seeing this as an either/or scenario, I challenge us to think about this more, as expertise is a critical foundation for creating AI/ML models that perform and are continually refined,” she said. “When technology and expertise are combined, the potential for the best outcomes is there.”
From her perspective, the areas where AI can be most effective in RCM are the areas that cause the most pain and are currently the most manual for healthcare providers.
Of these areas, denials, prior authorization, and eligibility probably rank near the top for all providers. She said it’s no coincidence that all of these items are interrelated.
“Errors in the registration process mean that applications are subsequently rejected,” Lambert said.
What causes rejections?
If you know which scenarios lead to rejections and how to detect or predict these rejections before they happen, you have the perfect opportunity to deploy AI with expertise and use the acquired data to build and train a model.
“There are opportunities, both within these processes themselves and in the overarching connectivity, to use ML and AI to improve service delivery to providers,” she said.
Lambert added that AI is not static and should never be treated as such.
“Designing a model that continuously learns is a core element of AI: models continuously learn from the data and the feedback loop that naturally arises from the results,” she said.
External factors
It is also crucial that knowledge of external factors that may affect a model is known and taken into account. This may mean regulatory changes that affect the structure of the data, the data elements in the responses, or other factors that may cause anomalies in the data.
“Make sure everyone is aware of any changes that affect the model so that interpretation of the results does not lead to incorrect assumptions or correlations,” Lambert advises.
She added that it is important to make sure that people understand that AI is not just for the top or just for data scientists. Everyone needs to be a part of it and that is what will make AI successful.
“AI needs input from those who are on the ground and managing the data, doing the operations, and managing the workflow to help create models,” she said.
Nathan Eddy is a healthcare and technology freelancer based in Berlin.
Send an email to the writer: nathaneddy@gmail.com
Twitter: @dropdeaded209