How to verify teleradiology vendor claims

Your teleradiology vendor's monthly report shows 97% of STAT studies signed inside the hour. The same week, your ED medical director forwards two overnight cases that sat unsigned past the 90 minute mark and asks what the contract is for. Both can be right, because the report measures an interval nobody defined out loud. To verify teleradiology vendor claims, demand four disclosures in writing: median and 90th percentile turnaround by priority tier, the two timestamps the clock runs between, SLA compliance over the trailing 12 months against those tiers, and the written definition of a major discrepancy behind the discrepancy rate. Then check the turnaround half yourself. Your RIS or PACS already holds a received timestamp and a signed timestamp for every study you've ever sent, and one month of that export produces figures to set beside whatever the vendor reported.

If you're still assembling the question list, start with our teleradiology buyer's guide. This page is about grading the answers.

Average turnaround is the most gamed number in imaging

An average is the easiest turnaround figure to flatter, because wildly different distributions can produce the same mean. Picture two vendors, both reporting a 45 minute average STAT turnaround. The first signs nearly every study between 35 and 55 minutes; a physician gets the report in roughly the time it takes to work up the next patient, at every hour of the night. The second signs most studies in 20 minutes and a stubborn overnight subset in three hours, and the fast daytime volume launders the slow nights out of the arithmetic. The mean can't tell these two apart. The median can: it shows what a typical study experiences and shrugs off a handful of outliers. The 90th percentile finishes the job, because it describes how long your worst-served patients wait, and that tail is where clinical risk lives. So ask for both figures, by tier, for the trailing 12 months, and expect them within the week, since a vendor that measures its own distribution has these numbers sitting in its reporting stack. A vendor that offers only an average has either never computed the distribution or would prefer you didn't see it.

Five teleradiology vendor claims to demand in writing

Five written questions separate vendors that measure from vendors that market. Send them verbatim. The written answers are the deliverable, and the tone of the reply is a second data point.

1. Turnaround. "Please provide median and 90th percentile turnaround by priority tier for the trailing 12 months, and state the two timestamps the clock runs between."
2. SLA compliance. "What percentage of studies were delivered within tier over the trailing 12 months, and do missed studies stay in the denominator?"
3. Quality. "What share of final signed reports receive an independent double read, what is the major discrepancy rate, and what written definition of a major discrepancy produces it?"
4. Access. "What is the escalation path to a reading radiologist by phone at 3 a.m., and what is your measured time to reach one?"
5. Coverage. "Are the radiologists reading our studies licensed in the state where our patients are located, and which fellowship subspecialty reads each part of our case mix?"

Every one of these has a checkable answer, and the fifth tracks a federal requirement rather than a preference: 42 CFR 482.22(a)(4)(iii) requires the distant-site physician to hold a license issued or recognized by the state in which the hospital whose patients receive the service is located. Adjectives arriving where numbers should be is itself a finding.

Four ways a turnaround number gets flattered

Vendors flatter turnaround by stopping the clock early, starting it late, trimming the denominator, or leading with a best month. Each is defensible in isolation. Each makes a slow service look fast.

Stopping at a preliminary read. A clock that stops at the preliminary read can show half the time of one that runs to the final signed report. The final signed report is the document your clinicians act on. It is also the one 42 CFR 482.26(d) requires the interpreting practitioner to sign and the hospital to keep for at least five years. Ask which endpoint the vendor's clock uses. The distinction is covered in STAT vs preliminary vs final reads.

Starting at assignment. When the clock starts at radiologist assignment, every minute a study spends waiting in the queue vanishes from the measurement. Queue time is precisely the time your ED spends calling to ask where the report is.

Trimming the denominator. Some vendors drop studies that missed their tier, or re-tier them after the fact, then report compliance on what survives. Ask directly whether misses stay in the denominator permanently.

Leading with a best month. A single quiet month tells you about that month. Require the trailing 12, and require the same figure monthly thereafter, so the number stays honest once the contract starts.

Reference calls that produce signal

A reference call produces signal when you ask for specifics a vendor can't coach: the worst month, the 3 a.m. phone call, the missed finding. Vendors hand you their happiest customers, and that's fine; your job is to learn what happy looks like under load. Ask what the worst month with the vendor was and what the vendor did about it. Ask how long it takes to get a reading radiologist on the phone overnight, and whether the monthly report lists every out-of-tier study individually with a cause. Ask whether they've ever had a clinically significant missed finding and how it was closed. Ask whether turnaround has drifted since signature, what happened during their last volume spike, and what they'd negotiate differently today. Then listen for texture. A reference who can't recall a single rough night has either been briefed or hasn't used the service long enough to hold an opinion, because every working clinical relationship accumulates friction.

Verify teleradiology vendor claims from your own timestamps

The strongest way to verify teleradiology vendor claims runs entirely on your own data, because your RIS or PACS already records when each study arrived and when its report was signed. Export one full month with these fields per study: accession, modality, priority tier, ordering time, study received or last image time, and final signed report time. Most RIS platforms expose them in a turnaround or productivity report; on the PACS side they usually live in the study status audit log. Subtract received from signed for each study, then take the median and the 90th percentile within each tier. Ten minutes in a spreadsheet yields a figure you can hold against the vendor's report. If yours and theirs diverge by more than a few minutes, you're measuring different intervals. The definitions are the finding.

Segment before you conclude

A blended monthly number hides the specific failures that hurt, so split the same export by tier, by hour, by modality and by day of week. The tier cut catches STAT inflation, where so much gets marked STAT that the label stops meaning anything. The hour-of-day cut exposes overnight staffing gaps that a daily average smooths over. The modality cut shows where a vendor's panel runs thin, often MRI or breast. The day-of-week cut tests whether weekend and holiday coverage extends past the brochure.

The overnight cut usually pays for the whole exercise. A vendor whose 2 a.m. to 6 a.m. 90th percentile runs triple its daytime figure doesn't have 24/7 coverage; it has daytime coverage with an overnight queue. See weekend and holiday coverage for what 24/7/365 coverage looks like when it is staffed that way.

Put the scorecard in the contract

Write the monthly performance report into the contract as an obligation with a fixed format. One page, every month, six lines. Median and 90th percentile turnaround by tier; out-of-tier studies counted and individually listed with a cause; SLA compliance for the month and the trailing 12 months, with misses kept in the denominator. Then the double-read sampling rate and the discrepancy rate it produced, escalation events with the measured time to reach a radiologist, and volume read against remaining headroom. Compare it with your own export monthly for the first quarter, then spot-check quarterly. If your site is ACR accredited, the peer review line is your obligation as much as the vendor's: the ACR accreditation requirements for supervising physicians name ensuring that all physicians providing services at the facility are actively participating in a formal peer review program, and studies read remotely are still services provided at your facility.

Run a 30 day pilot with a pass or fail gate

A paid 30 day pilot on live volume, gated by criteria written before day one, tells you more than any reference call. Thirty days covers at least four weekends and enough overnight volume for a stable 90th percentile. A workable gate: STAT median under 45 minutes and STAT 90th percentile under 90 minutes, measured from your timestamps; zero unexplained out-of-tier studies; a radiologist reachable by phone within 15 minutes on every escalation; and the monthly report delivered in the contract format. State plainly that the pilot converts to a term only on a pass; a vendor confident in its numbers takes that deal without flinching.

One caveat: if your facility sends only a handful of studies a month, a formally gated pilot is more process than the decision warrants, and the five written questions alone will separate the field.

AstraRad's answers to the same five questions

We publish the disclosures this page tells you to demand, on our SLA and QA methodology page. The clock runs from last-image arrival to radiologist signature, with no pauses for nights, weekends or holidays. Median STAT turnaround is 30 minutes, against tiers of STAT under one hour, Urgent under four hours and Routine under 24. Over the trailing 12 months, 99.4% of reports came back inside their tier, with every miss listed by study and kept in the denominator. One in 20 final signed reports gets an independent second read from a subspecialist blind to the original. Major discrepancies run under 0.3% of signed reports, each reviewed and closed with the reader. Behind those figures sit 240 board-certified, fellowship-trained subspecialists reading on scheduled shifts, 24/7/365, every one physically located in the United States and licensed in the state where your patients are. STAT reads at a 30-minute median shows that pace from the emergency department side. Grade all of it the way you'd grade anyone else, from your own PACS timestamps.

Which refusals should end the conversation with a teleradiology vendor?

Two refusals end the conversation: declining to name the two clock timestamps, and declining to confirm that misses stay in the compliance denominator. Without the timestamps the turnaround number is unreadable; without the denominator the compliance rate is unverifiable. Below those sit four softer refusals, in descending severity. First, a missing written definition of a major discrepancy, which leaves the discrepancy rate meaning whatever the vendor decides. There is no external number to fall back on either: the ACR RADPEER program states plainly that no scoring benchmarks have been established. Second, an average offered where a median and 90th percentile were requested. Third, a best month presented as a trailing figure, in which case ask again for 12 and watch what arrives. Fourth, a nameless escalation path to a radiologist, which reads fine on paper and hurts at 3 a.m. If you hear "that's proprietary," remember that pricing structure sometimes is, and the location of two timestamps never is. Weigh honest bad news accordingly. A vendor that reports a 90th percentile it's embarrassed by, with the cause attached, has told you more than one that reports only a flattering median.

Run the method on everyone, including us. Learning to verify teleradiology vendor claims is only worth the hour if the incumbent gets graded too. A turnaround number without its two timestamps is an adjective; the number that survives your own RIS export is the performance. Ask us for the trailing 12 month figures alongside per-report pricing, or book a coverage consultation to scope a gated pilot. If you're still narrowing the field, our honest comparison of teleradiology companies covers who does what well.