Prior authorization is the “single highest cost for the healthcare industry” in the U.S., totaling some $767 million a year, according to the CAQH index.
Physicians spend nearly two full business days per week on prior authorization requests, and payers devote thousands of manhours reviewing and approving them in an antiquated, manual process involving phone calls and faxes.
The arduous task often delays necessary treatment and sometimes results in treatment abandonment — patients just get tired of waiting, so they give up — both of which hurt patient outcomes and ultimately raise costs in the long run.
Prior authorization has been identified as one of the biggest opportunities for applying artificial intelligence (AI) to help lower the administrative burden and cost. Considering that 82% of healthcare leaders want to see their organizations be more aggressive in adopting AI technology, now may be the perfect time to take the leap toward applying AI to solve the prior authorization problem. Here’s how it can work:
1. Establish parameters for automatic approval.
A machine learning platform can look at previous prior authorization requests and identify the conditions most likely to result in an approval. For example, to be approved for an MRI of the knee, a patient must have already attempted anti-inflammatory medication and physical therapy and had an X-ray. Based on such parameters, payers can build a system for automatically greenlighting incoming requests that meet those conditions, greatly reducing the workload. Those that don’t meet the criteria would get flagged for manual review.
2. Create a standard for prior authorization submission.
One of the biggest issues with prior authorization is that every payer has different requirements. Requirements can differ even within the payer’s own system based on variables like geography, provider group and more. That means providers must figure out the process each time, binding them to a manual, labor-intensive system. An automated system would establish a baseline protocol for submissions. For example, providers could see a checklist of what’s required in the submission workflow based on the payers’ specific parameters, drastically reducing the back-and-forth that frequently ensues when providers fail to initially submit the required documentation.
3. Enable system and data interoperability.
Lack of interoperability has prevented automation of the prior authorization process. While the data required to easily approve prior authorization requests is very often contained in a provider’s electronic health record (EHR), the provider can’t easily share it with the payer for review. The data has to be relayed via fax. Interoperability is essential for the application of AI in prior authorization, and the right platform must have widespread interoperability with every major EHR to enable automated, electronic review of records.
4. Digitize unstructured data.
Some 80% of the data contained in roughly 1.2 billion clinical documents created every year is unstructured, in the form of handwritten medical charts, physician notes, forms and scanned documents. But most payers’ systems can’t read and analyze this data, even though it contains vital details required for prior authorization review. The use of AI would require a system that could digitize and analyze this unstructured data to read and identify the requisite parameters for automated approval. This capability would also have additional data analytics benefits for overall population health and care planning, such as spotting trends, correlations and effective new treatments.
5. Consider social determinants of health (SDoH).
These contextual factors, like socioeconomic status, education and access to care, can play a significant role in patient outcomes. But most EHRs and payers’ systems don’t consider these factors, preventing providers and payers from making the most informed care decisions. By integrating SDoH data from established sources, AI-based prior authorization systems can consider these factors as part of the approval process and flag requests for manual follow-up that meet certain conditions.
See also: The Intersection of IoT and Ecosystems
6. Put submission and approval at the point of care.
By streamlining and accelerating the prior authorization process, AI can slash time spent per transaction from 20 minutes down to just six and in some cases deliver near-instant approvals. This acceleration means the request for authorization can happen at the point of care within the patient visit workflow, reducing time to treatment and treatment abandonment.
Deploying AI to solve the prior authorization problem would dramatically reduce the time and cost associated with this necessary, but cumbersome, part of healthcare. In fact, studies show that automation could cut the cost by nearly 73%, from nearly $10 per transaction to under $3 — a huge impact on lowering the most costly healthcare expense.
With the push to implement AI reaching a crescendo, now is the time for organizations to act or risk getting left behind. We owe it to providers and patients to take advantage of every opportunity to reduce their burden and deliver better care with a better experience at a lower cost.