In July 2022, a cardiac screening CT in South Carolina showed an 8 mm pulmonary nodule. The radiologist did exactly what is expected: described the incidental finding in the report and recommended a follow-up CT in 6 to 12 months. That follow-up scan happened in May 2025 — two years and nine months later. By then the nodule was a 3.1 cm mass in the left lung, and the final diagnosis was stage 4 non-small cell lung cancer with bone metastases. The patient is suing Roper St. Francis Hospital and her primary care physician. The radiologist is not a defendant.
The timeline, as alleged in the complaint
Everything below is an allegation in the filing; no ruling has been issued. According to the complaint, the primary care physician never communicated the finding to the patient and never ordered the recommended CT. A year later, at a routine annual visit, the nodule went unmentioned and no follow-up plan was documented. In July 2024 the patient reported new and progressive respiratory symptoms — intermittent wheezing, chronic cough and exertional shortness of breath. The recorded response attributed the picture to allergies and possible asthma, prescribing albuterol and antihistamines.
By early 2025 the dyspnoea appeared on minimal exertion. She sought a pulmonologist at an affiliate of the same hospital. Reviewing her chart before the visit, the specialist found the 2022 finding and ordered a chest CT, performed on May 7, 2025. The radiologist who compared the two studies described changes “highly concerning for metastatic malignancy, likely lung origin,” recommending PET and tissue sampling. Biopsy confirmed it. In June, a second opinion at MD Anderson in Houston kept the diagnosis and the plan — palliative, not curative.
The plaintiff’s theory does not target the report. It targets the absence of a protocol: the allegation is that the hospital and its physicians lacked systems capable of ensuring that a recommendation for imaging follow-up was carried out within the stated timeframe and communicated to the patient.
Why an 8 mm nodule on a cardiac scan is not a detail
The technical context matters. A coronary artery calcium CT covers a significant portion of the lung fields — an incidental pulmonary finding on that exam is expected, not exceptional. And 8 mm sits exactly at the size where management changes.
Under the 2017 Fleischner Society guidelines, the international reference for incidentally detected pulmonary nodules in adults aged 35 and over, a single solid nodule of 6 to 8 mm calls for a follow-up CT at 6 to 12 months, with a possible further scan at 18 to 24 months depending on risk. Above 8 mm the recommendation escalates to CT at 3 months, PET/CT or tissue sampling. The recommendation recorded in the 2022 report was therefore guideline-concordant. The system simply did not execute it.
One important caveat: Fleischner does not apply to nodules found in a lung cancer screening programme — those follow Lung-RADS — nor to patients with known cancer or immunosuppression. The distinction matters because it changes both interval and aggressiveness, and conflating the two systems is a common source of inadequate follow-up.
The link that broke was not the diagnosis
This is the part that concerns everyone working in imaging. The radiologist saw it, described it and recommended action. The failure happened in the next stretch of the loop: communication, scheduling, tracking the open item. In almost every department, that stretch has no formal owner.
Communication of actionable findings has an ACR practice parameter and is covered by a Joint Commission patient safety goal, but the yardstick usually stops at the critical result — the finding that demands immediate contact. An 8 mm nodule is not critical; it is actionable in months. That intermediate category — important, not urgent — is exactly where open items evaporate, because it triggers neither a phone call nor an alarm, just a sentence at the end of a report.
Add to that the variability of the wording itself. Follow-up recommendations written as free text, with no code, no target date and no structured field, are invisible to any automated system — and depend entirely on a human reading, remembering and acting. The same fragility shows up in how departments handle incidental findings generally, which is why tools such as AI systems built to manage incidental radiology findings have found a market.
What a closed loop actually requires
“Closed-loop” is not software; it is a process with four requirements most departments do not fully meet:
Tag it machine-readably. The recommendation has to be a structured field — modality, interval, due date — not a loose sentence. Without that there is no queue, only an archive. Hold the open item in a live list. Someone, or some system, must maintain a register of open recommendations with due dates. Escalate what lapses. When the date passes with no exam performed, the case has to rise to a named person: care navigator, follow-up nurse, clinical coordination. Close the loop with the patient. The patient needs to know an exam is pending and when it is due, without depending on remembering a conversation in a consultation room.
Automated extraction of recommendations from free text exists and works reasonably well, and it is one of the clearer-return uses of AI in radiology. Patient-facing pressure is pushing the same way: as we covered, 44% of reports now reach patients before their doctors, which makes structured, unambiguous recommendations a practical necessity rather than a nicety. But no tool replaces deciding who answers for an open item when it lapses.
The same problem, other health systems
In Brazil and much of Latin America, the radiologist almost never has a direct channel to the patient, and the report travels through an information system that rarely talks to the scheduling system. The result is the same open loop, with two aggravating factors: fragmentation across payer, imaging clinic and attending physician, and no formal obligation to track follow-up recommendations at all.
There is a low-cost path before any platform investment: standardise the recommendation sentence with an explicit interval, tag the report with a retrievable code, pull a monthly list of recommendations issued, and check how many became exams. The first report is usually uncomfortable — which is precisely why it is worth running. The literature already established that incidental findings on lung CT signal real cancer risk; the question was never whether the finding matters, but whether anyone goes back for it.
What this case teaches before any verdict
Whatever the court decides, the operational lesson is already available. A technically correct report protects neither the patient nor the practice if the recommendation dies in the archive. From a risk standpoint, a radiologist who writes “follow-up in 6-12 months” with no way of knowing whether the follow-up occurred is carrying an exposure they do not control — and the patient is carrying one they do not even know exists.
Source: Radiology Business
