What is radiology turnaround time?
Radiology turnaround time is the interval between two recorded points in the imaging pathway — and there are four defensible pairs of points, not one. Departments usually report the gap between images becoming available and the report being signed. Referring clinicians experience the gap between placing the order and knowing the result. Both get called TAT, which is why published benchmarks contradict each other.
When I sit with a radiology operations manager and ask what their turnaround time is, I get a number in about five seconds. When I ask which two timestamps produced that number, the answer takes ten minutes and usually involves opening the RIS and arguing with a colleague. That gap between the confidence and the provenance is the whole subject of this article.
A departmental head defending a four-day MRI figure to an executive is usually being measured on a clock that includes scheduling while reporting on one that excludes it. That conversation cannot be won until the segments are separated.
The four segments — and which one your number is measuring
The imaging pathway splits into four intervals, each owned by a different part of the hospital. Order-to-exam belongs to scheduling. Exam-to-available belongs to the modality and the technologist. Available-to-report belongs to radiology. Report-to-communication belongs to the RIS, the EHR and the critical-results process. A department quoting "our TAT" is nearly always quoting the third one alone.
| Segment | From | To | Who actually controls it |
|---|---|---|---|
| Order-to-exam | Order placed | Patient scanned | Scheduling, transport, patient factors |
| Exam-to-available | Patient scanned | Images complete in PACS | Modality, technologist QC, network |
| Available-to-report | Images available | Report signed | Radiologist supply, worklist order, case complexity |
| Report-to-communication | Report signed | Clinician actually knows | RIS/EHR routing, critical-results process |
The practical consequence is uncomfortable. If your order-to-exam interval for outpatient MRI is eleven days and your available-to-report interval is fifty minutes, then reporting speed accounts for well under one per cent of the total wait. Improving the segment you measure is not the same as improving the experience anybody is complaining about.
Where the timestamps come from, and where they lie
All four segment boundaries are recorded somewhere and can be pulled without new instrumentation. The order timestamp sits in the RIS or EHR order record. The exam timestamp lives in the DICOM StudyDate and StudyTime attributes and in the modality's acquisition tags. Images-complete has to be inferred — from the last stored instance, or from the modality performed procedure step — because DICOM has no "study complete" message. Report-signed and report-verified are RIS or EHR events.
1. Order placed
RIS or EHR order record. Watch for orders entered retrospectively after a verbal request, which compresses the first segment artificially.
2. Exam performed
DICOM StudyDate/StudyTime and modality acquisition tags. Clock skew between a modality and the archive of even a few minutes will show up as negative intervals.
3. Images available
Inferred from the last stored instance, or from the modality performed procedure step. DICOM has no 'study complete' message, so this boundary is always an inference; studies that arrive in batches make the images look complete long before the radiologist can safely read them.
4. Report finalised
RIS/EHR signature events. Decide explicitly whether you are counting the preliminary read or the final, and what an addendum does to the clock.
Every one of those four has a known failure mode, and they do not all push in the same direction — two of them silently shorten your reported number and two inflate it, which is why the figure on the dashboard rarely matches the experience on the ward. The addendum question is the one departments most often leave undefined: if a report is signed at 14:00 and amended at 09:00 the next day, a system that reopens the clock will produce a nineteen-hour outlier that has nothing to do with reporting speed.
Getting these events to line up is an integration problem before it is an analytics problem. If your order and report events do not reconcile against the same accession number, you are measuring two populations and subtracting them. That is the same plumbing described in closing the loop between orders and reports.
Why the median hides your real problem
A mean turnaround time is distorted by a handful of extreme studies; a median discards them entirely. Neither tells you about the cases that actually cause harm. Percentile reporting — P50, P90 and P99 for each segment and each priority class — is the honest form, because it makes the tail visible as a number rather than as an anecdote in a morbidity meeting.
Consider two departments. One reports a forty-minute median with four per cent of studies finishing past twenty-four hours. The other reports a ninety-minute median with no tail at all. On a league table the first looks twice as good. Clinically, the first has a problem and the second does not, because harm does not live in the middle of the distribution.
Expert Insight: report the tail, not the average
If you publish one number, publish P90 by priority class rather than a median across all studies. A median rewards you for the routine outpatient chest X-rays that dominate your volume, and stays perfectly flat while the studies most likely to hurt someone drift further into the tail. In my experience, departments that switch to percentile reporting usually find something in the P99 that nobody had been looking at.
The tail is also where staffing reality shows up. Overnight coverage, subspecialty availability and holiday rotas do not move the median much, because they affect a small share of studies. They move P99 dramatically. This is closely tied to how much unstructured decision-making a reader is carrying, which we cover in radiologist cognitive load in high-volume departments.
Why MRI, CT and X-ray have different expectations
Modality-specific targets are not an excuse, they are a description of the work. MRI protocols run longer, are more likely to need prior comparison, more likely to require a subspecialty reader, and are more often scheduled than urgent. CT sits between. Plain radiography is short to acquire and frequently read without priors. A single departmental target flattens three genuinely different processes into one misleading figure.
Protocol length
MRI sequences run one after another with the patient on the table throughout; a plain film is a single exposure. That difference sits entirely in segment two, before a radiologist has seen anything.
Priors for comparison
Cross-sectional work is far more often reported against a previous study. Fetching that prior, or chasing one held by another hospital, lengthens segment three for reasons that have nothing to do with reading speed.
Subspecialty reading
Some studies need a specific reader who is not on site every day. That concentrates delay in a small number of cases, which is why it barely moves the median and moves P99 sharply.
Scheduled versus urgent mix
MRI is mostly booked work; plain radiography is mostly walk-in and ward work. Different arrival patterns mean a blended figure compares two queueing problems, not two teams.
This is also why MRI turnaround time and CT turnaround time are separate questions rather than one question with two answers. An outpatient MRI ordered for chronic knee pain and a non-contrast head CT ordered from the emergency department share almost nothing except a PACS. Blending them produces a number that describes neither, and it is why emergency pathways are usually measured separately — see door-to-diagnosis time in the ER.
What a turnaround policy should actually say
A usable radiology turnaround time policy names the two timestamps for each measured interval, sets separate targets per modality and priority class, states the percentile it is measured at, and defines what happens when a study breaches. A policy that says only "reports within 24 hours" is unenforceable, because nobody can agree on when the twenty-four hours started.
Definitions before targets
Name the exact start and stop event for each segment, in the source system, using field names an analyst can query. Include the addendum rule.
Targets by class, not in aggregate
Separate rows for stat, urgent, inpatient and routine, and separate rows per modality. One number per class, measured at a stated percentile.
A named exception path
What a reader does when a study cannot meet target — who is told, how it is recorded, and how those cases are reviewed rather than quietly excluded.
Communication counted, not assumed
A signed report is not a delivered result. Measure segment four explicitly, including the critical-results acknowledgement, or it will never improve.
For the professional framing around reporting practice and quality metrics, the American College of Radiology and RSNA are the bodies whose published guidance is worth reading directly rather than in summary — and note the setting each target assumes before importing it.
What AI moves — and what it does not
AI acts on one of the four segments and can shorten a second. It can reorder the available-to-report queue so that studies with suspected significant findings are opened earlier, and it can shorten report-to-communication by triggering a critical-alert path the moment a finding appears rather than after a signature. It does not scan patients faster and it does not create radiologist hours.
That second point deserves to be stated without decoration, because it is where most vendor claims quietly break. Reordering a worklist does not reduce the total work in it. The queue still contains the same studies and the same reading time; what changes is the order in which people wait. A department at capacity that reorders its queue has improved the turnaround time of urgent studies by lengthening the turnaround time of routine ones. That is usually the right trade — it is not a reduction in turnaround time, and calling it one is how credibility gets lost. The case for making that trade deliberately is made in automated worklist prioritisation, and this article does not repeat it.
Fractify enters exactly here. It is a clinical decision support tool for medical imaging — X-ray, CT, MRI and dental — that returns structured reports with findings, confidence scores, urgency levels on a 1–5 scale and treatment guidance, and a qualified clinician reviews every finding. The 1–5 urgency score is a segment-three input; the alert it can raise is a segment-four mechanism. Neither touches segments one or two, and any honest account of hospital deployment says so up front.
The caveat I owe you: an urgency score is only worth the process standing behind it. If nobody is rostered to act on a flagged study at three in the morning, the flag has moved that study to the top of a list nobody is reading, and your P99 will not move at all.
What to do next
Before you adopt any benchmark, spend one afternoon pulling four timestamps for a single month of studies and plotting each segment separately at P50, P90 and P99, split by modality and priority. In most departments that exercise relocates the problem — often out of radiology entirely. Then set your targets against your own distribution, and publish the definitions alongside the numbers so the next person to quote them knows which clock they came from.
Frequently asked questions about radiology turnaround time
These are the questions radiology managers, quality leads and referring clinicians ask most often when a turnaround target is being set, reported or challenged. The short answers below assume the four-segment model above: order-to-exam, exam-to-available, available-to-report and report-to-communication. Which segment you mean changes the answer to almost every one of them.
What is turnaround time in a radiology department?
Turnaround time is the measured interval between two recorded events in the imaging pathway. Most departments define it as the time from images becoming available in PACS to the report being signed. Referring clinicians usually mean something wider — from the order being placed to the result reaching them. Always state which two timestamps your figure uses.
What is a standard turnaround time for radiology reports?
There is no single standard. Published targets differ by setting, priority class and modality, and an emergency department figure does not transfer to outpatient MRI. Rather than adopting a number you cannot trace, pull your own timestamps for one month, plot each segment at P50, P90 and P99 by priority class, and set targets against that distribution.
How is radiology turnaround time calculated?
Subtract a start timestamp from an end timestamp for every study, then report the distribution rather than a single average. The order event comes from the RIS or EHR, the exam time from DICOM StudyDate and StudyTime, images-complete from the last stored instance, and report-signed from RIS signature events. Reconcile all four on the same accession number.
What is a normal MRI results turnaround time?
It depends almost entirely on which segment you count. For scheduled outpatient MRI, waiting for the appointment usually dwarfs the reporting interval, so a patient's experienced wait and the department's reported turnaround time can differ by orders of magnitude. Ask your department for the order-to-communication figure specifically, not the reporting figure.
Why do CT and MRI reports take longer than X-rays?
MRI and CT studies contain far more images, are more likely to require comparison with prior examinations, and more often need a subspecialty reader who may not be on site. Plain radiography is quick to acquire and frequently read without priors. Applying one departmental target across all three flattens genuinely different processes into a misleading number.
Should turnaround time be measured as a mean or a median?
Report percentiles instead — P50, P90 and P99 — for each segment and each priority class. A department with a fast median but a few per cent of studies running past twenty-four hours has a worse problem than one with a slower median and no tail, because harm lives in the tail.
Does AI reduce radiology turnaround time?
It moves one segment and can shorten a second. AI can reorder the reading worklist so suspected significant findings are opened earlier, and can trigger a critical-alert path as soon as a finding appears. It does not scan patients faster and does not create radiologist hours. Reordering a queue changes who waits rather than reducing the total wait.
What should a radiology turnaround time policy include?
Four things: the exact start and stop events for each measured interval, named in the source system; separate targets by modality and priority class; the percentile each target is measured at; and a documented exception path for breaches, including who is informed. Measure report-to-communication explicitly, because a signed report is not a delivered result.
See Fractify working on your own scans — live demo takes 15 minutes.
Request a Free Demo →