What the study found
Researchers at Seoul National University Bundang Hospital analyzed 24,146 outpatient abdominopelvic CT interpretations from 14,871 patients, reported between February 2023 and January 2024. Exams were randomly assigned to either on-site radiologists or teleradiologists, letting the team isolate the effect of interpretation setting from case mix. The results were published Sept. 16, 2026 in Academic Radiology and covered by Radiology Business.
Teleradiologists recommended additional imaging in 5.5% of reports, compared with 1.5% for on-site radiologists — an odds ratio of 3.91, or nearly four times the odds of issuing a follow-up recommendation. But the higher rate did not translate into more disease found. Recommendations from teleradiologists were significantly less likely to be completed (adherence odds ratio 0.42) and, when completed, significantly less likely to yield a positive oncologic finding — what the authors call oncologic clinical effectiveness (odds ratio 0.36). A subgroup analysis limited to liver MRI recommendations showed the same pattern.
As corresponding author Jungheum Cho, MD, PhD, of Seoul National University College of Medicine, and colleagues noted, teleradiology's role has expanded well beyond its original purpose: "Initially developed to provide after-hours coverage and for areas deprived of specialized medical services, teleradiology has evolved into a service model that supports not only emergency department images, but also elective imaging interpretation for outpatients in tertiary care hospitals." That expansion is exactly why the size of this dataset — and the randomized assignment of routine outpatient elective reads — matters: this is not an emergency after-hours coverage gap, it's routine daytime practice.
On-site vs. teleradiology: the three metrics side by side
| Metric | On-site radiologists | Teleradiologists |
|---|---|---|
| Recommendation proportion | 1.5% | 5.5% (OR 3.91) |
| Adherence proportion | Reference | Significantly lower (OR 0.42) |
| Oncologic clinical effectiveness | Reference | Significantly lower (OR 0.36) |
| Inter-radiologist variability (ICC) | 0.16 | 0.24 |
Source: Chee CG, Kim J, Cho J, et al. Academic Radiology, 2026. DOI: 10.1016/j.acra.2026.08.105. On-site figures shown as reference group; odds ratios describe teleradiology relative to on-site.
Why this reads as a variability problem, not a quality problem
The single statistic most easily missed in coverage of this study is the inter-radiologist variability figure: an intraclass correlation coefficient of 0.24 among teleradiologists versus 0.16 among on-site radiologists. In plain terms, which individual radiologist happened to read a given scan explained more of the difference in recommendation behavior among teleradiologists than it did among on-site readers. The setting itself — remote versus on-site — is a proxy for something more specific: inconsistent individual practice patterns that show up more sharply when readers lack shared local context.
The study's authors point to plausible mechanisms rather than a competence gap: teleradiologists commonly cover multiple facilities with different protocols and less direct access to a patient's electronic health record, prior imaging, or the ordering clinician — conditions that push individual readers toward more cautious, and more individually variable, follow-up language. That is a description of workflow conditions, not diagnostic skill. Two well-trained radiologists reading the same ambiguous finding under those conditions can reasonably land on different follow-up wording — one recommends a repeat CT "in 3 months," another suggests "clinical correlation," a third says nothing specific at all. None of those three is necessarily wrong; they are simply inconsistent, and inconsistency compounds when it isn't visible until a study like this one lines up thousands of interpretations side by side.
Reframed this way, the finding is less "teleradiology is worse" and more "practice-pattern variability in follow-up recommendations is larger than most groups can see, and remote reading conditions amplify it." That distinction matters for what a fix looks like — a quality problem is solved by changing who reads; a variability problem is solved by standardizing what gets written, regardless of who or where the reader is.
What actually reduces reader-to-reader variability
A durable fix targets the variance itself, not the location of the reader:
Standardize follow-up language against evidence-based criteria
When follow-up recommendations are drafted from a consistent, guideline-anchored template rather than each reader's individual phrasing, the same finding tends to generate the same recommendation — whether the reader is on-site or a thousand miles away.
Make recommendation rate a tracked quality metric
Most groups can report turnaround time by reader; few track recommendation proportion or adherence proportion by reader. This study shows that gap can hide meaningful variability. Surfacing it, on-site and remote alike, is the first step to closing it.
Treat clinical context gaps as a workflow problem to fix, not accept
If limited access to EHR data and referring-clinician communication is a documented driver of more cautious, inconsistent recommendations, closing that access gap for remote readers reduces the variability directly — independent of any reporting tool.
Where structured, AI-drafted reporting fits
This study is a clean illustration of why structured, template-driven reporting matters beyond formatting. AI CT reporting drafts findings into a structured report with standardized, evidence-based follow-up language proposed for each finding type, so the recommendation a patient receives depends on the finding — not on which individual radiologist, on-site or remote, happened to be reading that day. The reading radiologist still reviews, adjusts, and signs every report; standardization narrows the variance in the draft they start from, which is exactly the lever this study points to.
Frequently asked questions
Do teleradiologists recommend more follow-up imaging than on-site radiologists?
Yes. A 2026 study in Academic Radiology of 24,146 outpatient abdominopelvic CT interpretations found teleradiologists recommended additional imaging in 5.5% of reports, versus 1.5% for on-site radiologists reading the same exam type at the same institution — nearly four times the odds (odds ratio 3.91).
Does a higher recommendation rate mean teleradiologists find more disease?
No — the opposite. Follow-up exams recommended by teleradiologists were significantly less likely to be completed (odds ratio 0.42) and, when completed, significantly less likely to turn up a malignancy (oncologic clinical effectiveness odds ratio 0.36) compared with recommendations from on-site radiologists.
Is the gap a teleradiology quality problem?
The study frames it as a variability problem rather than a quality problem. Inter-radiologist variability in recommendation behavior, measured by intraclass correlation coefficient, was higher among teleradiologists (0.24) than on-site radiologists (0.16) — meaning which individual reader interpreted the scan mattered more for teleradiologists than for on-site readers. The authors attribute this partly to teleradiologists covering multiple sites with different protocols and less access to clinical context, prompting more cautious, inconsistent follow-up language.
How can imaging groups reduce follow-up recommendation variability regardless of where the reader sits?
Standardizing follow-up language against evidence-based criteria — rather than leaving wording to each reader's individual judgment — reduces reader-to-reader variability whether the radiologist is on-site or remote. AI-drafted structured reports that propose standardized, guideline-anchored follow-up recommendations for the radiologist to review and finalize are one way to apply that consistency at scale, without changing who is accountable for the signed report.
Source: Chee CG, Kim J, Cho J, Kim Y, Kim M, Hwang J, Lee Y, Do Y, Lee JW, Kim YH. "Follow-up Examination Recommendation and Clinical Effectiveness of On-Site Versus Teleradiology Interpretation for Outpatient Abdominopelvic CT." Academic Radiology, published online Sept. 16, 2026. DOI: 10.1016/j.acra.2026.08.105. Covered by Radiology Business. Figures rounded as reported.