> ## Content Index
> Fetch the complete content index at: https://www.cpj.fyi/llms.txt
> Use this file to discover other available public pages before exploring further.

# The radiologist will see you now
- URL: https://www.cpj.fyi/radar/the-radiologist-will-see-you-now/
- Published: 2026-11-16T17:50:24.000Z
- Updated: 2026-11-16T17:50:24.000Z
- Description: Who's looking at your X-Ray? Who's interpreting it? Who's doing something about it? Where do we want that work to go?
- Author: Clay Parker Jones
- Tags: Radar, Artificial Intelligence

**AI seems to be making expert work easier to outsource before it makes the experts more productive,** as economists Marco Francesconi and Xiaoyu Xia argue in a [recent analysis](https://cepr.org/voxeu/columns/ai-and-geography-expert-work-evidence-radiology?ref=cpj.fyi). 

The reading moves An illustrative diagram of radiology organization. On the left, three local practices each serve a nearby hospital. On the right, a shared remote radiology group serves the same three hospitals. AI may help verify remote interpretations. The share of Medicare-enrolled radiologists affiliated with groups operating in five or more states rose from 9 percent in 2017 to 47 percent in 2025\. The diagram illustrates a proposed mechanism; the percentages measure group affiliation, not remote work or jobs replaced by AI. Local practices serve nearby hospitals HOSPITAL A HOSPITAL B HOSPITAL C Reading capacity stays local REMOTE RADIOLOGY GROUP AI may help check the reads HOSPITAL A HOSPITAL B HOSPITAL C Reading capacity is shared Hospital Radiologist reading scans Radiologists affiliated with groups operating in five or more states Share of Medicare-enrolled radiologists 9% 2017 47% 2025 DIAGRAM: PROPOSED MECHANISM · DATA: FRANCESCONI & XIA (2026) CPJ.FYI / RADAR 

*In 2017, 9% of Medicare-enrolled radiologists were affiliated with groups operating in five or more states. By 2025, the share was 47%.*

At Radiology Partners, the largest network of radiologists in the US, the average number of markets listed per radiologist **rose from 1.41 to 2.41** between 2017 and 2025, while **Medicare interpretations per reader fell** after 2021 in claims observed through 2024\. Across the study, the authors find **no broad increase in interpretations per radiologist** in Medicare fee-for-service claims. 

The way the authors think this works is through faster, easier verification of findings. Images have for some time been easy, fast and cheap to send for a distant specialist to review, but the buyers (physicians and hospitals) have to believe in that specialist's work. Software (probabilistic or otherwise) that flags findings, supplies measurements or prioritizes scans can provide additional ways to check the work. 

In the authors’ model, local radiology services require some minimum staffing commitment, even when demand is low. A group serving many locations can pool demand and share capacity, spreading the fixed costs of verification. That would make remote expertise economical where maintaining a local team would be expensive. So: tools leave specialist’s tasks largely intact while changing which organization employs/delivers them, which clients they serve and which communities retain their expertise.

The obvious worry here is *enshittification of radiology*, where we make something cheaper to deliver, worse, less plentiful. The [working paper](https://repository.essex.ac.uk/44128/1/cesifo1%5Fwp13086.pdf?ref=cpj.fyi) finds evidence that reading moved away from smaller markets, and residents continued to receive imaging. 

Locally billed *procedures* fell too, including work requiring physical presence. Some of that decline reflects changes in billing location, and the data does not establish whether access to procedures deteriorated. For a hospital buying *remote* readings, that choice includes maintaining local procedural capacity; if reading work helps sustain a local practice, moving that work could affect the economics of keeping specialists nearby for procedures. For now, biopsies still need someone in the room. 

Worth watching.