Radiology departments are facing mounting strain.
Imaging volumes continue to rise, with demand expected to increase by 17% for MRI and 25% for CT scans over the next three decades. Even if the radiologist workforce grows by more than 20% over that timeframe, the supply will barely keep pace with imaging demand. That may be an optimistic scenario, considering that subspecialists are substantially more likely to leave the workforce than generalists, and attrition among radiologists in the U.S. has more than doubled since 2014. Given today’s dynamics, “radiologist burnout” is real and a considerable concern within healthcare organizations.
Most healthcare leaders likely already understand the opportunity to use AI to improve speed and accuracy in diagnostic imaging. With AI tools, clinicians can detect subtle abnormalities, prioritize urgent scans, and identify patterns that the human eye alone might miss. **But many leaders are overlooking the opportunity to use AI to improve billing and coding, which can improve both patient and financial outcomes. **
Currently, fewer than 5% of radiologists review their own billing codes, compared to more than 90% of primary care physicians. This disparity may be one reason that radiology reimbursements have dropped by nearly half over the past decade, with rising scrutiny leading to more medical necessity denials. This dynamic leads to preventable revenue losses for healthcare systems. Incomplete documentation also impacts patients, who may miss opportunities for timely follow-ups.
The good news? In addition to improving diagnostics, AI can assist in ICD-10 code identification right within radiologists’ workflows, improving clinical outcomes, helping to recapture lost revenue, and ensuring patients receive the follow-up attention that leads to early intervention and preventative care.
Recapturing Revenue
Radiologists operate in extremely high-volume environments, naturally prioritizing interpreting scans over the manual task of medical code identification. For this reason, in most health systems, administrative staff further downstream in the process have been responsible for this task.
With the deployment of AI within the radiologists’ workflow, radiologists are assisted in code identification and attestation much earlier than in today’s typical workflow, ensuring more accurate and accelerated patient diagnosis and fewer claim denials from payers.
Soon, automating the transfer of key EHR data, such as a clearly stated reason for the exam, into the required documentation will create a timelier and more compliant process for both patients and providers. Even a small increase in approved scans can prevent substantial lost revenue.
‘No Findings Left Behind’
Even more importantly, current coding practices often result in missed opportunities for critical follow-ups that could lead to early detection. For example, an abdominal scan might reveal an incidental lung nodule, but that sort of finding often remains buried in an unstructured report instead of triggering a follow-up. As a result, the patient may never be referred to the appropriate specialist or receive further diagnostic testing.
Similarly, a routine mammogram might reveal arterial calcifications associated with cardiovascular risk, creating an opportunity for further evaluation beyond the original reason for the scan. But again, radiologists are so focused on the immediate diagnostic need that they don’t always have time to make sure incidental findings are tracked and acted upon. AI can automatically flag those results and initiate a follow-up workflow.
The University of Miami Health System (UHealth), which performs more than 1 million imaging procedures each year, partnered with Quantiphi and AWS to address both the clinical and financial sides of this problem. Under its “No Findings Left Behind™” initiative, Quantiphi worked with UHealth and AWS to design and deploy “Hurricode™”, a generative AI application that identifies relevant codes within radiologists’ existing workflows and converts report information into structured data that supports billing and patient follow-up.
Results in the Real World
During a recent webinar, now available on demand, Dr. Alexander McKinney,
chair of the radiology department at the University of Miami, talked about the benefits of this purpose-built AI solution. The effort resulted in a diagnostic data accuracy rate of 90% (up from 58% to 77% previously), reduced manual effort by 40%, and increased forecasted revenue by 34%.
“To get people in a few years before they have pancreatic cancer, or five years before they develop dementia…those are the types of things that we’re able to do now that we weren’t able to do just a few years ago,” McKinney said during the conversation. “It’s not just the revenue, and the ‘money under the couch,’ but also real, meaningful change for patients earlier.”
Our webinar conversation covered:
- How today’s inefficient radiology workflows can cause clinician burnout, negatively impact patient outcomes, and lead to lost revenue from denied claims.
- How Quantiphi leveraged Amazon Bedrock to accelerate code identification and improve accuracy.
- How UHealth achieved measurable results.