What the study did
Researchers led by Babita Panigrahi, MD, of Johns Hopkins University School of Medicine, studied screening mammograms performed between November 2024 and January 2026, as reported by Radiology Business on a study published in the Journal of the American College of Radiology. Out of 36,384 screening mammograms, 4,647 (12.8%) were recalled for additional workup (BI-RADS 0) — the routine mammogram recall rate most imaging centers already live with.
The intervention wasn't aimed at that whole recalled population. At the moment of interpretation, radiologists flagged only the subset of BI-RADS 0 findings they judged highly suspicious. Staff then arranged the diagnostic appointment directly by phone, using reserved slots, targeting evaluation within about a week. The team built this on top of their existing Epic EHR — an electronic flag, reserved appointment slots, and staff education, with no new infrastructure required, according to the researchers.
Only 57 patients — about 1.2% of all recalls — were routed onto the expedited pathway. That's the point: the pathway is narrow by design, reserved for the cases where speed matters most, not a blanket promise to see every recalled patient faster.
The timeline effect
For patients on the expedited pathway, the median screening-to-diagnostic interval was 7 days, compared with 25 days for the standard pathway. The diagnostic-to-biopsy interval was 14 days versus 19 days. Combined, the median screening-to-biopsy interval was 24 days on the expedited pathway versus 45 days on the standard one — close to half.
| Interval | Expedited pathway | Standard pathway |
|---|---|---|
| Screening to diagnostic evaluation | 7 days (median) | 25 days (median) |
| Diagnostic to biopsy | 14 days (median) | 19 days (median) |
| Screening to biopsy (overall) | 24 days (median) | 45 days (median) |
| PPV of abnormal screening | 91.2% | 5.2% |
| PPV among biopsied patients | 94.5% | 18.2% |
Figures as reported by Radiology Business on the Johns Hopkins study published in the Journal of the American College of Radiology; see the journal record. PPV = positive predictive value.
The accuracy gap explains why the narrow flag works: among expedited cases, 91.2% of abnormal screening findings and 94.5% of biopsies were positive for cancer, versus 5.2% and 18.2% on the standard pathway. Radiologists weren't fast-tracking noise — they were fast-tracking the cases most likely to be cancer, and the system protected capacity for exactly those cases. The clinical rationale is straightforward: delay to diagnostic evaluation is associated with later-stage disease, a lower chance of breast-conserving surgery, decreased survival, and more patients lost to follow-up.
Why this is a triage story, not a mammography story
Strip away the modality and the mechanism is generic: out of a large volume of studies, a reviewer separates a small, high-suspicion subset at the moment of read, and routes only that subset onto a protected fast track with reserved capacity. The 1.2% figure matters as much as the 24-versus-45-day result — it shows that halving wait time for the cases that matter didn't require rebuilding the whole scheduling system, only correctly identifying the minority that needed it.
CT has the same structural problem, at larger scale. Incidental findings and high-suspicion pulmonary nodules sit in exactly this position: a large population of scans, a small subset that is genuinely urgent, and a routine scheduling process that treats both the same way. The evidence on what happens without a triage mechanism is discouraging — one cohort study of patients with incidental solitary pulmonary nodules found that 52% did not receive appropriate follow-up per guideline recommendations, as reported in The American Journal of Medicine Open. A mechanism that reliably separates the high-suspicion 1–2% from routine volume — whether a radiologist's flag or a structured AI draft that surfaces the finding explicitly in the report — is the lever, not a modality-specific fix.
Where AI-assisted reporting fits
The Johns Hopkins pathway relied on a radiologist's eye plus an EHR flag and reserved scheduling slots — no AI involved. The reason it worked is instructive for AI-assisted CT reporting all the same: speed only helps the patients who need it if the system first identifies which studies need it. A structured, comprehensive report draft that consistently surfaces high-suspicion findings — rather than relying on a busy radiologist to remember to flag them during a dictation queue — extends the same logic to CT. xAID's draft is reviewed by an in-house radiologist on every study and delivered ready-to-sign, so the triage benefit shows up as a faster, more consistent draft rather than a new scheduling workflow layered on top of an already-stretched one.
Frequently asked questions
What is the expedited mammogram recall pathway?
It is a workflow studied at Johns Hopkins in which a radiologist flags only the most suspicious BI-RADS 0 screening mammograms at the moment of interpretation. Staff arrange the diagnostic follow-up appointment by phone using reserved slots, targeting evaluation within about a week, instead of the patient waiting for a routine scheduling call.
How much faster was the expedited pathway?
The median time from screening to diagnostic evaluation was 7 days on the expedited pathway versus 25 days on the standard pathway. The median time from screening all the way to biopsy was 24 days versus 45 days — close to half.
How many patients were actually fast-tracked?
Out of 36,384 screening mammograms performed between November 2024 and January 2026, 4,647 (12.8%) were recalled for additional workup, and only 57 of those recalls (1.2% of all recalls) were routed through the expedited pathway — the subset the radiologist judged highly suspicious.
Does this triage logic apply to CT, not just mammography?
The underlying mechanism is modality-agnostic: a human or AI reviewer separates a small, high-suspicion subset from a large volume of studies at the moment of read, and routes only that subset onto a fast, protected track. CT programs face the same shape of problem with incidental findings and high-suspicion nodules, where published adherence to recommended follow-up has been reported as low as 48%.
Source: Radiology Business, reporting on a study by Babita Panigrahi, MD, and colleagues (Johns Hopkins University School of Medicine) published in the Journal of the American College of Radiology (journal record); and Kattih Z, et al., "Rate of Incidental Lung Nodule Follow-Up," The American Journal of Medicine Open (2025). Figures are rounded as reported.