The "dying breed" debate
In July 2026, a post from a practicing physician set off a large clinician discussion that spread widely across medical social media, and it was picked up by Radiology Business. His argument: physicians trained to justify every scan — to articulate the clinical question and the pretest probability before ordering — are becoming a "dying breed." In his words, "CT scans have essentially been continually abused as an extension of the physical exam."
The replies split predictably. Some radiologists agreed the ordering discipline has eroded; emergency physicians pushed back that the malpractice environment leaves them little choice. As one put it: "Until someone thanks me for NOT getting a CT scan, I'll continue doing what I think I need to do to keep patients safe." Both sides are describing the same force from different ends — and it is worth separating the anecdote from the data.
The anecdote is a social-media thread. The data underneath it is durable, peer-reviewed, and points one direction: the overutilization of medical imaging is not a talking point, it is a measurable, multi-decade trend that keeps outrunning the visits and the workforce meant to absorb it.
More scans per visit, not more visits
The clearest recent evidence comes from a 2025 Radiology analysis of fee-for-service Medicare that tracked emergency-department imaging from 2013 to 2023. Its finding is striking precisely because visit volume went down while imaging went up.
| Per 100 Medicare beneficiaries | 2013 | 2023 | Change |
|---|---|---|---|
| ED encounters | 65.0 | 54.5 | −16% |
| CT exams | 18.7 | 36.7 | +95.8% |
| CT exams per 100 ED encounters | 28.7 | 67.1 | +133.5% |
Source: Rosenkrantz & Cummings, Radiology (2025). The authors note plainly that "such growth in ED imaging can strain radiology practices given radiologist shortages…" The demand isn't coming from a rising tide of patients — it's coming from more imaging per patient. That is the demand-side signal the "dying breed" thread was gesturing at.
A trend measured in decades, not headlines
The ED numbers are the sharp edge of a long curve. A 2019 JAMA study of more than 135 million exams across seven US health systems and Ontario found that CT use among adults 65 and older roughly doubled — from 204 to 428 exams per 1,000 person-years between 2000 and 2016. MRI use more than doubled over the same window.
Notably, that study found growth had decelerated from its early-2000s pace — but decelerating growth is still growth, layered on top of an already far larger base. See Smith-Bindman et al., JAMA (2019). Two decades of compounding volume don't reverse because ordering habits are debated online.
Why this becomes a capacity problem
Here is where the demand story collides with a supply story. Radiologists cannot be manufactured on the timeline that scans accumulate. Training a radiologist takes over a decade; a CT order takes a click. As far back as 2015, a Mayo Clinic workload analysis calculated that "the average radiologist interpreting CT or MRI examinations must now interpret one image every 3–4 seconds in an 8-hour workday to meet workload demands" — see McDonald et al., Academic Radiology (2015). That was a decade and roughly one ED-CT doubling ago.
A 2025 Radiology special report on the workforce put it flatly: "the increasing number of imaging studies, owing to advancing technology and an aging population, is outgrowing the capacity of radiologists" (Afshari Mirak et al., 2025). Whether or not imaging-cautious ordering physicians are truly disappearing, the arithmetic is the same: volume is a compounding line, and headcount is a nearly flat one. The gap between them is the reporting-capacity squeeze.
Two ways to absorb the overflow
Faced with more studies than radiologists, an imaging operation has essentially two levers. It can add readers — hiring, outsourcing, or paying overtime — which scales linearly with cost and is bounded by a workforce that isn't growing fast enough. Or it can raise the throughput of each read without cutting the human out of it.
Grow headcount proportionally
The intuitive response, but the one the shortage constrains most. Radiologist supply is inelastic on any timeline that matters, and demand for read capacity is spiky and 24/7. Cost rises in step with volume, and locum or overtime coverage is the most expensive form of it.
Raise per-read efficiency with AI drafting
AI can produce a structured, comprehensive draft report for a study that a radiologist then reviews and finalizes. The same radiologist covers more studies per shift because the blank-page work is done — absorbing overflow volume without hiring proportionally more people. The human stays fully in the loop.
The two levers aren't mutually exclusive, but only the second one bends the cost curve away from volume. And crucially, it does so without asking anyone to solve the ordering-behavior debate first — it treats the volume as a given and focuses on the read.
Where AI CT reporting fits
This is the problem AI CT reporting is built for. A foundation-model system drafts a full, structured preliminary report for a study; xAID's in-house radiologist reviews every preliminary; and the report is delivered ready-to-sign so that your reading radiologist signs the final. The overflow is absorbed at the drafting stage, not by cloning radiologists. No AI system is FDA-cleared in the US to file a final read on its own, and this workflow keeps that line bright: in-house review on every preliminary, final signature stays with your reading radiologist. For imaging groups watching volume outrun their schedule — a pressure also visible in radiologist compensation trends — that is the difference between hiring against a shortage and reading through it.
Frequently asked questions
What is the overutilization of medical imaging?
Overutilization of medical imaging refers to ordering scans that are unlikely to change management — imaging driven by defensive medicine, throughput pressure, or habit rather than a clear clinical question. It shows up as imaging volume growing faster than patient visits. In a national Medicare analysis, CT use per 100 beneficiaries in the emergency department rose 95.8% between 2013 and 2023 even as ED encounters per 100 beneficiaries fell 16%, meaning the growth came from more scans per visit, not more patients.
Are imaging-cautious ordering physicians disappearing?
A widely discussed 2026 clinician thread argued that physicians trained to justify every scan by articulating the clinical question and pretest probability are becoming a 'dying breed,' as CT increasingly functions as an extension of the physical exam. Defensive medicine, a fear of missed diagnoses, and payment models that reimburse per scan all push in the same direction. The debate is anecdotal, but the utilization data it points to is real and long-running.
How much has imaging volume grown relative to patient visits?
Long-run data shows sustained growth. In a 2019 JAMA study of more than 135 million exams, CT use among adults 65 and older in US health systems roughly doubled from 204 to 428 exams per 1,000 person-years between 2000 and 2016. More recently, a 2025 Radiology analysis of fee-for-service Medicare found CT use per 100 ED encounters rose 133.5% from 2013 to 2023 — growth concentrated in scans per visit rather than visit counts.
How can radiology handle rising imaging volume without adding headcount?
Because radiologist supply cannot scale as quickly as imaging orders, the practical lever is per-read efficiency. AI can produce a structured, comprehensive draft report for a study, which a radiologist then reviews and finalizes — absorbing overflow volume without hiring proportionally more radiologists. The human stays accountable: no AI system is FDA-cleared in the US for autonomous final reporting, so a radiologist reviews the preliminary and the reading radiologist signs the final read.
Sources: news peg — Radiology Business (2026). Utilization data — Rosenkrantz & Cummings, Radiology (2025); Smith-Bindman et al., JAMA (2019). Workload & workforce — McDonald et al., Academic Radiology (2015); Afshari Mirak et al., Radiology (2025). Figures are as reported.