Radiology Report Turnaround Time: Benchmarks and Fixes

The short answer: Radiology report turnaround time is the interval between a study being available and the report being signed. In US practice the common targets are under 30 to 60 minutes for emergency department studies, same day for inpatient work, and 24 to 48 hours for routine outpatient imaging, with tighter internal targets on stroke, trauma and suspected pulmonary embolism protocols. Most groups that miss those targets do not have a reading-speed problem. They have a queue problem, a drafting problem and a measurement problem, in that order.

Turnaround time is the metric radiology gets judged on by everyone who is not a radiologist. The ED measures it, referrers complain about it, hospital contracts are written around it, and it shows up in quality reporting. It is also one of the few numbers a group can genuinely move without hiring, which is why it deserves a more careful look than the usual instruction to read faster.

What is a good radiology report turnaround time?

For emergency department imaging, most US hospitals set a target of under 30 to 60 minutes from image available to final report, and the strongest performers hold tighter windows on specific protocols. Inpatient studies are usually expected the same day. Routine outpatient imaging sits at 24 to 48 hours, sometimes stretching further for subspecialty reads. Anything labeled STAT should be measured separately, because averaging it into the general pool hides the failures that actually generate phone calls.

Those are conventions rather than a single national standard. The ACR quality measures used in MIPS reporting track report turnaround for CT and radiography, and individual health systems layer their own service level agreements on top. If you are negotiating a hospital contract, the number in that contract is your benchmark and nobody else's.

How do you calculate radiology turnaround time?

Pick two timestamps and be consistent. The four that matter are order placed, exam completed, image available in PACS, and report signed. Most groups report image available to final signature, because that is the window radiology actually controls. Order to final is the number the ED cares about, and it includes transport and scanner time that is not yours.

Two measurement habits are worth adopting. Report the median rather than the mean, because a handful of overnight outliers will drag an average into fiction. And report a high percentile alongside it, usually the 90th, because the studies that generate complaints are never the median ones. A group with a 22 minute median and a 4 hour 90th percentile has a very different problem from a group with a 45 minute median and a 70 minute 90th percentile, and a single average number makes those two look similar.

Where the minutes actually go

Break a slow report into its parts and the same pattern shows up in most practices.

Waiting in the queue. This is almost always the largest single block, and it is the one least visible in the data. A study that takes eight minutes to read can sit for ninety before anyone opens it, because a worklist sorted by arrival time has no idea which study is the emergency. That is dead time nobody is working on.

Assembling context. Finding the prior study, the relevant clinical note, the protocol document, the previous measurement. In a practice running multiple systems this is a real tax on every read, and it is the kind of friction that a decent search layer over your internal systems can cut more effectively than any change to how radiologists read.

Dictating and correcting. Speech recognition is fast until it is not. The time cost is less the dictation than the editing, and it climbs with fatigue across a shift.

Structuring and formatting. Populating the template, transcribing measurements, writing the follow-up language your group has agreed on. Mechanical work, repeated dozens of times a day.

Communication and documentation. Calling a result, documenting the call, closing the loop. Necessary, and rarely counted in anyone's turnaround metric.

The levers that actually move the number

1. Fix the reading order first

If urgent studies are opened in the order they arrived, no amount of reading speed will fix your emergency turnaround. Reordering the queue so suspected-urgent studies surface first is the highest-leverage change available to most groups, and it costs no additional radiologist minutes. This is what radiology worklist prioritization software is for, and it is why groups that adopt AI triage in radiology usually see the emergency numbers move before anything else does.

Published experience with flagged versus unflagged studies gives a sense of the size of the effect. In one intracranial hemorrhage series, flagged cases were read at roughly 73 minutes against roughly 132 minutes for unflagged ones. The reading itself was not faster. The study simply stopped waiting.

2. Draft the report before the radiologist opens it

The second largest block is the mechanical construction of the report. If the structured template is already populated when the study opens, the radiologist starts from an edit rather than a blank page. That is the argument for AI radiology reporting and for structured radiology reporting generally: the saving is in the typing and the formatting, not in the interpretation, which stays entirely with the radiologist who reviews, edits and signs.

3. Route studies to the right reader

A neuro study sitting in a general queue while the neuroradiologist reads chest films is a turnaround problem disguised as a staffing problem. Subspecialty routing tends to improve both quality and speed, and it is mostly a configuration exercise in your worklist rather than a purchase.

4. Extend coverage rather than compressing reads

Overnight and weekend gaps produce the long tail that ruins a 90th percentile. That is a coverage decision, usually answered with internal call schedules or teleradiology software, not with process improvement.

5. Attack the specific protocols with the tightest clocks

Stroke, trauma and suspected pulmonary embolism carry their own internal targets and their own consequences for missing them. Groups running a AI stroke triage pathway or flagging suspected pulmonary embolism on CTPA are usually solving a turnaround problem on those protocols specifically, not across the whole list.

Does AI reduce radiology report turnaround time?

It reduces the waiting, not the reading. Software that flags suspected urgent findings and pushes those studies up the queue shortens the interval between a study arriving and someone opening it, which is where most of the delay lives. Software that drafts the structured report shortens the interval between opening and signing. Neither makes interpretation faster, and neither should. The radiologist reviews every flag, edits every draft and signs every report.

Be skeptical of vendor turnaround claims in one specific way. A tool that reorders your queue improves the flagged studies by definition, and it does so partly at the expense of everything behind them. Ask what happened to the median across the whole list, not just to the flagged subset, and ask whether routine studies got slower. A group that halved its stroke turnaround while its outpatient backlog grew by two days has moved a problem, not solved one.

Why is my radiology turnaround time getting worse?

Usually volume grew and the queue logic did not change. Imaging volume per radiologist has climbed steadily in US practice, and a first-in-first-out worklist degrades non-linearly under load: at low volume the wait is short regardless of order, and at high volume the ordering decision becomes the dominant factor. Groups often notice the metric deteriorating without any individual working slower, which is exactly what a queueing problem looks like from the inside.

The other common cause is measurement drift. Someone changed which timestamps are compared, or a new modality started reporting into the same pool, or STAT studies got folded into the general average. Before you diagnose a performance problem, confirm you are still measuring the same thing you measured last year.

A practical 90 day plan

The discipline that matters here is changing one thing at a time. Groups that turn on triage, drafting and a new worklist configuration in the same month end up with a better number and no idea which change produced it, which makes the next decision guesswork.

What to ask a vendor about turnaround time

• Which two timestamps does your reported improvement use, and are they the ones we already measure?
• What happened to the median across the entire worklist, not just the flagged studies?
• Did routine and outpatient turnaround get worse while urgent improved?
• How long was the shadow period before the tool changed the live queue?
• What does the radiologist have to do differently, and how many extra clicks is that per study?

Those five questions separate a real workflow effect from a reporting artifact. If you want the full list for a wider evaluation, our radiology AI vendor evaluation questions cover scope, integration, pilot design and contract terms as well.

The honest summary

Turnaround time is mostly a queueing metric wearing a clinical costume. The read is not usually the bottleneck; the wait before the read is, followed by the mechanical work of building the report. Fix the order studies are opened in, remove the typing from the reporting step, measure the whole distribution rather than an average, and the number moves without anyone being asked to work faster. Radiological.ai is built around those first two levers: it flags suspected findings for review, prioritizes the worklist so urgent studies surface first, and drafts the structured report in your template, with the radiologist reviewing, editing and signing every study.