The Follow-Up Ledger
Radiology Partners says its AI drafts flagged 57.5% more lung nodules or masses on chest X-ray than its radiologists had. A nodule on a chest film often means a chest CT. Here is every state as one dot: how much chest CT it bills today, and how much the draft’s extra nodules would add on top. Move the sliders and watch the dots climb.
Read the Radiology Partners release →
Or read today’s newsletter →
Primary source: Radiology Partners, “Vision Language Model-Powered Radiology Reporting Shows Gains in Detection and Efficiency in Research Study of Over 237,000 Exams,” Nov. 11, 2026 (company release; internal, preliminary study; “Investigational device. Not for sale in the United States”). Data via MIMI Labs: CMS Medicare Physician & Other Practitioners by Geography & Service (CY2024); CMS Medicare Geographic Variation (2024, Original Medicare beneficiaries).
The release reports a relative lift, not a rate. It says the drafts found 57.5% more nodules than the original reports did, but not how many nodules the original reports had per 1,000 films. That missing number decides how much work this is, so it’s a slider here. Everything else on the chart is 2024 Medicare claims.
Extra nodules flagged
Extra chest CTs
Added to chest CT load
Range across states
Each dot is a state, sized by the chest X-rays it billed in 2024 (CPT 71045–71048). X: diagnostic chest CTs billed (71250/71260/71270) per 1,000 Original Medicare beneficiaries. Y: extra chest CTs (chest X-rays × baseline × lift × CT share) as a share of that 2024 CT volume. Dashed line: the pooled figure for the states shown. Readouts sum the states shown. Hover a dot for its numbers.
80/20 · what the dots say
Medicare billed chest X-rays and diagnostic chest CTs across the 50 states and DC in 2024, about films for every CT. That ratio is the lever. At the page’s starting assumptions (15 nodules per 1,000 films, 70% scanned), a 57.5% lift adds chest CTs, or on top of today’s volume. Where films outnumber CTs most, the jump is steepest: . Halve the baseline and every number halves. The release doesn’t publish the baseline, so the honest headline is the ratio, not the total.
Read this before quoting a number
- The baseline and the CT share are yours, not the study’s. The release gives relative lifts only. Starting values here are placeholders to move, not estimates from the literature.
- “More found” isn’t “more real disease.” The release says blinded panels confirmed the findings, but the full papers aren’t public yet. Some extra nodules will be small or benign and need no CT under follow-up guidelines; the share slider is where that judgment goes.
- Services, not people. Counts are billed services by the state where the clinician practices; beneficiaries are Original Medicare enrollees by state of residence. Referral hubs (Minnesota, Maryland, DC) bill far more imaging per local beneficiary, which is why the stress test drops them. Professional and technical components can both appear, so treat volumes as billed reads.
- Medicare Advantage is invisible. Fee-for-service claims only. In states where most enrollees are in MA, the real film and CT volume is larger than what’s drawn.
- A company study of an uncleared product. Internal, retrospective and preliminary. Nothing here says patients did better.