Using AI in processing claims

[iris lilies]

I hope Rob sees this thread. Just a little bit of Googling shows that our federal government is using AI to process all kinds of benefit claims, especially in the SSI disability world.

here is one such article but they’re all over the place:

https://www.govexec.com/oversight/20...i-work/398640/

[iris lilies]

Curious about how European countries with their magical public healthcare systems use AI, I googled that question and this helpful summary, no doubt generated by AI, appeared on my screen. My guess is Rob will be most terrified of the red highlighted ideas:


Several European healthcare systems are starting to utilize AI for claims processing, with countries like Germany, the Netherlands, and the UK leading the way, utilizing machine learning algorithms to analyze large datasets of claims data, identifying patterns and anomalies to streamline the process, detect potential fraud, and improve the accuracy of claim approvals, all while adhering to strict data privacy regulations like GDPR; however, the adoption of AI in claims processing is still evolving across the continent due to the complexity of regulations and the need for robust data infrastructure.

• Processing:
  
  • Fraud Detection: Identifying suspicious patterns in claims data to flag potential fraudulent activities.
  • Claim Review and Eligibility Checks: Automating the initial review of claims to identify those that require further investigation.
  • Predictive Analytics: Using historical data to predict future claim costs and identify high-risk patients.
  • Clinical Decision Support: Providing insights to healthcare providers regarding appropriate treatment options based on claim data.
    

Examples of European countries actively exploring AI in claims processing:

• Germany:
  Considered a frontrunner in digital health innovation, with initiatives to integrate AI in claims processing and reimbursement for digital health applications.
  
• Netherlands:
  Actively developing AI-based solutions for claims review and fraud detection within their healthcare system.
• UK:
  Utilizing AI to analyze large datasets of claims data to improve efficiency and identify potential cost-saving opportunities.

Challenges and Considerations:

• Data Quality: The accuracy and completeness of healthcare data is crucial for the effectiveness of AI algorithms.
• Ethical Considerations: Ensuring AI systems are used fairly and do not perpetuate biases.
• Transparency and Explainability: The ability to understand how AI algorithms arrive at decisions is critical in healthcare settings.

[Tybee]

I imagine AI for this will work better in theory than in practice. I am convinced that the radiologist used AI to create my last knee report, in that the result was absolutely bizarre and made no sense, but the word kind of sounded like the word it should have been. The word it substituted for the word it should have been is a much more common finding, so that is what it said I had, even though they were monitoring a much rarer bone tumor for growth, to see if it had the growth pattern of a malignant tumor. This was a followup xray of the pre-existing bone tumor for size only. The report said something about an endograft, which had no bearing on what was in my knee, but what was in my knee was much less common than the endograft word.

So I think it looks for common things and throws in things with a higher probability, even when it makes no sense. The original tumor had already been visualized. Or it is what LDAHL talked of, the hallucinations that it comes up with.

The problem is the information is not always trustworthy. And you need a human to interpret it.

It took five weeks and multiple calls from the doctor, including escalating it to supervisors, to get the thing rewritten correctly.

[bae]

I have been evaluating a commercial AI system for processing public records requests for a governmental agency, and performing any necessary redactions on the requested documents. It seems to perform better than our expensively-trained staff on many sorts of requests, and avoids expensive legal consultations on the trickier documents.

I suspect we'll budget shortly to use it for real. This will reduce our expenses considerably, provide better service to the public in getting their requests delivered in a timely fashion, and will allow our overworked staff to focus on the boring parts of their job: firefighting, emergency medical services, and rescue services.

[Tybee]

I have been evaluating a commercial AI system for processing public records requests for a governmental agency, and performing any necessary redactions on the requested documents. It seems to perform better than our expensively-trained staff on many sorts of requests, and avoids expensive legal consultations on the trickier documents.

I suspect we'll budget shortly to use it for real. This will reduce our expenses considerably, provide better service to the public in getting their requests delivered in a timely fashion, and will allow our overworked staff to focus on the boring parts of their job: firefighting, emergency medical services, and rescue services.

That is very encouraging to hear!

[LDAHL]

More than one person has warned me to expect a wave of AI systems to start displacing mid- level professionals in any number of fields in the next 10-20 years. Anything where somebody needs to evaluate information and make decisions and monitor and report results. Hopefully they will keep a few humans onboard for when scapegoats are required. It could be that people adept at training AI systems could become the new elite profession. At least for a while.

And if Rob really wants something to worry about, the DoD has 7-800 projects ongoing to develop autonomous weapons systems like the USAF’s “Loyal Wingman” program. And you can probably assume that more bright Chinese engineers are being funneled into that field than game design.