Artificial intelligence (AI) is especially effective at pattern recognition, and its use is growing rapidly.
Outside of healthcare, AI recognizes obstacles in autonomous-driving cars. It can count calories on a dish photographed on a cell phone. It can detect dangers and errors at construction sites. AI is even helping reunite lost pets with their owners. The same pattern-recognition capability is now reshaping medical imaging.
AI applications in dermatology
Within healthcare, AI is being used to perform initial screening of skin lesions, especially valuable given the limited access to dermatology care. AI tools can be used in conjunction with telemedicine, or they can be used by primary care clinicians to triage dermatology referrals.
Changing mammography screening
Artificial intelligence is increasingly being used in mammography to increase sensitivity and reduce dependence on radiologist capacity. The largest radiology group recently purchased an AI company, and many radiology practices are now integrating AI into their operations.
A randomized trial of 105,000 women in Sweden published in The Lancet in January compared women whose mammograms were interpreted by two radiologists with those whose mammograms were interpreted by AI and a single radiologist. More earlier cancers were detected in the AI group without an increase in false positives. The AI group had fewer invasive cancers discovered over a two-year follow-up period.
Billing and access challenges
However, there's currently no universally accepted billing code for AI reading of mammograms. Some practices add AI without charging for it, although many bill patients between $40 and $100 for AI readings. This could create disparities, as those lower-income plan members are less likely to pay extra for AI services. Eventually, when billing codes become available, AI readings would be paid for without patient cost-sharing if AI-enhanced mammograms were recommended by the U.S. Preventive Services Task Force.
Detecting secondary conditions beyond primary diagnoses
Artificial intelligence is also likely to allow radiologists to find diseases other than those being evaluated by the scan. For example, mammography includes images of the blood vessels in the breast. The presence of calcium in these vessels shows an increased risk of cardiovascular disease. This could prompt providers to recommend more intensive blood pressure or cholesterol therapy. CT scans of the trunk ordered to diagnose abdominal complaints could detect osteoporosis, fatty liver disease and muscle and fat composition. CT scans of the lungs could also detect coronary artery disease. An important clinical challenge will be to avoid ordering follow-up tests that may cause more harm than good.
Improving rather than replacing radiologists
Radiology is an example of how AI can enhance jobs rather than eliminate them. Radiologists always do a final reading of any image initially interpreted by AI, and demand for radiologists is growing with the increased use of AI. More imaging capability means more images to read — AI expands the pipeline rather than shrinking it.
Implications for employer-sponsored health plans
Payment for these advances will be a challenge for employer-sponsored health insurance. Be sure to:
- Watch for new CPT codes covering AI-assisted imaging
- Assess whether your plans' preventive care structures are positioned to cover these services
- Consider the equity implications if AI enhancements are billed as patient cost-sharing before standardized coverage rules are established
