Radiology Pre-Reads: What the Turnaround-Time Data Actually Shows in 2026
National data shows imaging turnaround has doubled since 2014 and 42% of ED CTs take 3+ hours. What radiology pre-reads change — and how to measure it.
Radiology pre-reads are usually discussed as a staffing answer. The more useful way to think about them in 2026 is as a measurement problem: the national turnaround-time data has been moving in one direction for a decade, and a pre-read step changes which number a treating clinician is actually waiting on. This piece stays with the data — what the published studies show, what an emergency department is supposed to be achieving, and how to measure whether adding preliminary read coverage changed anything at your own site.
⚠️ Important scope note
For human patients at USA facilities, RAD365 provides preliminary reads only. Your facility's own licensed radiologist always completes and signs the final interpretation. RAD365 does not provide final reads for human patients.
For veterinary and pet imaging there are no preliminary reads — veterinary studies are read as complete final reports by specialist veterinary radiology groups.
increase in time to imaging results since 2014, per a JACR study of a 5% Medicare fee-for-service sample, 2014–2024 (reported by U.S. News)
of emergency CT scans took more than three hours to reach a report, in a study of 10,000+ ED CTs (Becker's Hospital Review / KevinMD)
image-available to final-report standard cited in EMRA's radiology turnaround-time literature review, with tighter targets for stroke and trauma
open radiologist positions listed on the ACR Career Center; some counts put it above 1,800
The decade-long trend, in one study
The most consequential turnaround-time finding of 2026 was published online in the Journal of the American College of Radiology in August. Using a 5% sample of Medicare fee-for-service beneficiaries from 2014 through 2024, the authors tracked report turnaround across modalities and found that it continued to increase throughout the period. U.S. News, covering the paper, reduced it to the sentence that circulated: time to results has doubled since 2014.
What makes that finding difficult to dismiss is the method. This is not a satisfaction survey or a single-institution audit; it is a decade of national claims data, and the trend line does not reverse anywhere in it. Scanner counts grew over the same period. Cross-sectional protocols got faster to acquire. Throughput improved. And reports still took longer to arrive, because acquisition capacity and interpretation capacity are separate constraints, and only one of them expanded.
What that looks like in an emergency department
The national average is an abstraction; the emergency department is where it becomes a clinical decision. A study of more than 10,000 emergency department CT scans — covered in the radiologist-shortage reporting carried by Becker's Hospital Review and KevinMD — found that 42% of the time, more than three hours passed before a report was available. The authors attributed the delay to radiologist workload rather than to acquisition or transport. The images were sitting there, complete and available, waiting for someone to read them.
Now set that against the benchmark. EMRA's literature review on radiology turnaround time cites an emergency standard of image-available to final report in under 30 to 60 minutes, with materially tighter expectations for stroke CT and CTA, trauma CT and suspected pulmonary embolism — studies where the report changes management inside the hour, sometimes inside ten minutes.
Stat callout
The published emergency benchmark is under 30–60 minutes. The measured reality in a 10,000-study ED CT sample was more than three hours in 42% of cases. That is not a small variance around a target — it is a different order of magnitude, and hiring alone cannot close it while the ACR job board carries 1,400+ unfilled radiologist posts.
Which clock a pre-read actually moves
A preliminary read does not replace the final report and does not shorten it. What it does is insert an earlier documented checkpoint into the same timeline. The study becomes available; a structured preliminary impression is delivered within a contracted window; the treating clinician has something to act on. The facility's own licensed radiologist then completes and signs the final interpretation on the normal workflow, with critical findings already flagged and communicated.
Two clocks, therefore, not one:
| Measure | Without pre-read coverage | With pre-read coverage |
|---|---|---|
| First clinically actionable impression | Same as final report | Contracted window from study-available |
| Final signed report | Facility radiologist | Facility radiologist — unchanged |
| Critical-result communication | Begins at first read | Begins at preliminary impression |
| Overnight 95th percentile | Typically the worst segment | The segment coverage is bought for |
| Who signs for human patients | Your radiologist | Your radiologist — always |
The measurement trap: medians hide the problem
Turnaround-time distributions have long right tails. A department can hold a perfectly respectable median while running a 95th percentile of four or five hours overnight — and every complaint, every incident report and every clinical risk lives in that tail, not in the median. If you manage to the median you will improve the number that nobody was complaining about.
The comparison problem is worse. There is no enforced definition of "turnaround time", so the start timestamp is a choice: order placed, acquisition complete, study available to the reader, or study assigned. Vendors choose differently, some exclude breached studies from the denominator entirely, and two quoted averages are frequently two different measurements. Before comparing partners, get the definition in writing. This is the same discipline that makes peer review and QA reporting meaningful rather than decorative: the metric only means something if everyone agrees what is being counted.
How to baseline before you contract anything
- Pull ninety days of study-level timestamps from RIS and PACS — raw rows, not a dashboard summary.
- Fix the start timestamp and write it down. Study-available to the reader is the most defensible choice for measuring reading capacity.
- Segment by shift, modality and priority class. Overnight and weekend are where the distribution goes bad, and a blended figure conceals it.
- Report median and 95th percentile side by side. If the gap is large, the problem is the tail, not the average.
- Separate reading delay from routing delay. Prior-fetch failures, reconciliation errors and worklist misprioritisation are cheaper to fix than reading capacity, and fixing them makes any added capacity more effective.
- Agree the same definition with the partner before signing, and require raw study-level data in reporting so you can audit it.
Why hiring alone is not the plan
The ACR Career Center currently carries more than 1,400 open radiologist positions, with some counts running above 1,800. Residency output is not close to that number, and the pipeline lag is measured in years, not budget cycles. That does not make recruitment pointless; it makes recruitment insufficient as the sole strategy for a department whose 95th percentile is already unacceptable this winter. The operational lever available now is coverage design — where the first look happens, how fast, and how it is measured — which is the layer that RAD365's human radiology operations support is built around.
About the author
Trisha Seal writes on radiology operations and imaging workflow for RAD365. RAD365 operates preliminary read coverage for USA facilities — structured preliminary impressions delivered against contracted, auditable turnaround windows, with critical-finding communication protocols and per-account percentile reporting. For human patients, the facility's own licensed radiologist always completes and signs the final interpretation.
Baseline your turnaround-time performance
Talk to RAD365 about what your own 95th-percentile data shows overnight, and what preliminary read coverage would change against it.
Discuss pre-read coverage →Radiology pre-reads and turnaround time: frequently asked questions
The Turnaround-Time Data
What does the national data actually show about radiology turnaround times getting worse?
A Journal of the American College of Radiology study published online in August 2026, built on a 5% sample of Medicare fee-for-service beneficiaries between 2014 and 2024, found that imaging report turnaround times continued to rise across modalities through the end of the study period. U.S. News summarised the headline finding plainly: time to results has roughly doubled since 2014. That is a decade-long directional trend measured on national claims data, not a survey of impressions, which is what makes it hard to argue with.
Why did imaging turnaround times increase through 2024 even as imaging volume and scanner counts grew?
Because capacity to acquire images and capacity to report them are different constraints. Scanner counts, throughput and cross-sectional protocols all expanded, and each additional CT or MRI produces far more images per study than the plain films they displaced. Interpretation time per study rose accordingly while the number of radiologists available to do the interpreting grew far more slowly. More scanning capacity with static reporting capacity produces exactly the queue behaviour the JACR data captured.
What percentage of emergency CT studies take more than three hours to reach a final report, and why?
A study of more than 10,000 emergency department CT scans, covered in the radiologist-shortage reporting carried by Becker's Hospital Review and KevinMD, found that 42% of the time it took more than three hours to get a report back. The authors pointed to radiologist workload as the bottleneck rather than acquisition or transport. In other words, the images existed and were available long before the report did — the delay sat in the reading queue.
Is the 'turnaround time has doubled since 2014' finding true across all modalities, or concentrated in a few?
The JACR analysis reported increases across modalities rather than a single outlier driving the average, though the absolute magnitude differs by modality and by setting. Cross-sectional studies with high image counts and complex reporting requirements sit at the slower end; simpler radiographic studies remain faster in absolute terms but still moved in the same direction. The practical takeaway for an imaging director is that this is not a CT problem to be solved in isolation.
What's considered an acceptable radiology report turnaround time in an emergency department?
The Emergency Medicine Residents' Association literature review on radiology turnaround time cites a working standard of image-available to final-report in under 30 to 60 minutes for emergency studies, with materially tighter targets for stroke CT and CTA, trauma CT and suspected pulmonary embolism, where the report changes management within minutes. Set that against the finding that 42% of ED CTs took more than three hours and the size of the gap is obvious.
What a Pre-Read Actually Changes
How does adding a preliminary read change the turnaround-time picture without changing the final-report timeline?
It changes which clock the treating clinician is waiting on. The final report still belongs to your own licensed radiologist and is produced on your own workflow. What a pre-read inserts is an earlier, documented, clinically actionable checkpoint — a structured preliminary impression delivered within a defined window of the study becoming available. The emergency physician gets a finding to act on while the final interpretation proceeds normally behind it.
Does adding a pre-read step slow down the eventual final report from the facility's own radiologist?
It should not, and when it is set up properly it usually helps. The preliminary impression arrives in the record as a starting point rather than a blank study, and critical findings have already been flagged and communicated, which removes the scramble that otherwise interrupts the reading list. What does slow finals down is a badly integrated pre-read process that forces manual reconciliation — which is why routing and worklist design matter as much as the read itself.
Can pre-reads reduce ED boarding time for patients waiting on imaging results?
For the subset of patients whose disposition genuinely depends on the imaging result, yes — that is the mechanism by which pre-read coverage produces measurable operational benefit. If a patient is held solely because nobody has yet looked at the scan, moving the first look forward by two hours moves the disposition decision forward with it. It does nothing for boarding driven by bed availability or admitting-service delays, so measure the imaging-dependent cohort separately rather than expecting a single global number to move.
Is the turnaround-time problem mainly a staffing shortage or a workflow/routing problem?
Both, and the proportions differ by site, which is why baselining matters before contracting anything. The national picture — the ACR Career Center currently listing well over 1,400 open radiologist positions, with some counts above 1,800 — is unambiguously a supply constraint. But at an individual hospital a meaningful share of elapsed time is often routing, prior-fetch failures, worklist prioritisation and study reconciliation rather than reading capacity. Fix the routing first; it is cheaper, and it makes any added reading capacity more effective.
Measuring It At Your Facility
What operational metrics should a hospital track before and after adding pre-read coverage?
Track image-available to preliminary-impression, image-available to final-report, and critical-result communication time, each split by priority class, by modality and by shift. Add the percentage of studies breaching your own ED target and the volume of clinician calls chasing results. Those five together tell you whether coverage changed clinical experience or only changed an average.
How should a hospital baseline its own turnaround-time performance before contracting a pre-read partner?
Pull at least ninety days of study-level timestamps from the RIS and PACS rather than relying on a dashboard summary, and be explicit about which timestamp starts the clock — order, acquisition complete, or study available to the reader. Segment by shift, because overnight and weekend performance is usually where the distribution goes bad. Then agree with the prospective partner that the post-contract measurement will use the identical definition. Most disappointing outcomes are definition mismatches, not performance failures.
Why do different vendors measure 'turnaround time' differently, and why does that matter when comparing partners?
Because there is no enforced definition, and the choice of start timestamp alone can change a reported number by hours. Some measure from study-available, some from assignment to a reader, some exclude studies that were re-routed, and some report only completed studies while excluding those that breached and were escalated. When you compare two quoted averages you are frequently comparing two different measurements. Ask for the definition in writing before comparing the numbers.
What's the difference between median and 95th-percentile turnaround time, and which one should hospitals actually manage to?
The median describes the typical study; the 95th percentile describes the study that generates the complaint, the incident report and the clinical risk. Turnaround distributions have long right tails, so a healthy median coexists comfortably with a tail of multi-hour outliers. Manage to the 95th percentile and report both. A partner unwilling to be measured at the 95th percentile is telling you where their tail sits.
What happens to turnaround-time performance during a volume surge like flu season or a mass-casualty event?
Surges are where fixed-capacity arrangements fail, because the queue lengthens faster than any single team can absorb. The relevant contractual questions are whether the partner has surge capacity across time zones, whether targets are suspended or maintained during declared surges, and how triage priority is re-ordered when everything is urgent. Ask for last winter's actual percentile performance during the peak weeks rather than the annual average, which flattens exactly the period you care about.
Does RAD365 publish its own turnaround-time percentiles?
Reporting is provided per account against the definitions agreed at contract, at median and 95th percentile, split by priority class and shift, because a single blended figure hides the behaviour that matters. Prospective clients can ask for representative performance data for comparable volume and case mix during evaluation. What we do not do is quote a single marketing average detached from a measurement definition — for the reasons set out above, that number is not comparable to anything.
Scope & Compliance
Is the deliverable the same for human and veterinary patients?
No, and the distinction is absolute. For human patients at USA facilities RAD365 provides preliminary reads only; your facility's own licensed radiologist always completes and signs the final interpretation, and RAD365 does not issue final reports for human patients under any arrangement. For veterinary and pet imaging there are no preliminary reads; veterinary studies are read as complete final reports by specialist veterinary radiology groups. Any proposal you receive from us states which of the two applies.
How quickly does RAD365 typically deliver a preliminary read once a study is available?
Target windows are set per contract and per priority class, with the tightest windows reserved for the study types where minutes change management — stroke CT and CTA, trauma, suspected pulmonary embolism. The clock starts at study-available to the reader, and performance is reported against that definition at both median and 95th percentile. Any partner quoting one blanket number across all priorities is describing an average rather than a commitment.
What should a hospital ask about turnaround-time reporting before signing with any pre-read partner?
Six things: which timestamp starts the clock, whether performance is reported at median and 95th percentile or median only, whether breached and escalated studies are included in the denominator, how performance is reported during declared surges, what the remedy is for sustained misses, and whether you receive the raw study-level data or only a summary. If you cannot audit the number independently, you are not measuring the partner — you are reading their marketing.