Patients may access radiology results before speaking with a clinician, leaving complex imaging language to be interpreted without immediate support. Lung cancer screening CT reports can be especially difficult when incidental findings or suspicious results require follow-up. A 2026 survey in Radiology Advances assessed whether large language model-generated summaries could improve understanding. Among 1815 U.S. adults, the summaries improved comprehension, confidence about next steps and self-rated anxiety, with the strongest benefits among respondents with lower self-rated English and health literacy.
Summaries Improve Understanding
Participants reviewed three types of lung cancer screening result. One was a negative screen, one was negative for lung cancer but included complex incidental findings and one was suspicious for lung malignancy and required action. Each scenario tested understanding of the main finding and the expected next step. The summaries used simpler wording and a consistent structure, while retaining the clinical meaning of the original CT information.
Comprehension improved after participants read the patient-friendly summaries. The improvement appeared across the negative, complex incidental and suspicious scenarios. Understanding of main findings was already high for the negative screen but still increased after the summary was added. Gains were larger when the original information was more complex. The clearest improvement concerned next steps, especially in the more complex and actionable results, where many participants initially struggled to identify what should happen next.
Subjective experience also improved. Participants found the summaries easier to understand than the original CT wording and reported greater confidence about the next step. The change was particularly visible in the technically complex scenario that was negative for lung cancer but included incidental findings. Anxiety scores fell across all three scenarios, including the result suspicious for lung cancer. The findings indicate that clearer wording can reduce uncertainty, even when the clinical content remains potentially concerning.
Must Read: Blockchain for Strengthening EHR Security
Literacy Differences Shape the Impact
The benefits were not evenly distributed across all respondents. Participants with lower self-rated English literacy showed the greatest improvement in objective comprehension after reading the summaries. They also reported a larger reduction in anxiety than respondents who rated their English literacy as middle or high. This pattern suggests that simplified radiology wording may be especially useful for people who are less comfortable with complex written English.
Self-rated health literacy also shaped the effect. Respondents with lower health literacy did not show the same pattern for every comprehension measure, but they did report greater anxiety reduction than those with high health literacy. Participants with middle health literacy also showed more anxiety reduction than the high-literacy group. These results point to a communication need that extends beyond plain English alone. Understanding medical forms and health-related instructions may influence how patients experience imaging results and how much reassurance clearer explanations can provide.
Several other participant characteristics were considered, including age, experience working in health care and whether a person had read one of their own radiology results during the previous year. After adjustment for relevant factors, self-rated English and health literacy remained important predictors of benefit. Most participants rated their English and health literacy as high, and the group with low self-rated health literacy was relatively small. Even so, the strongest measured effects were concentrated among respondents who may be most vulnerable to misunderstanding technical imaging information.
Communication Needs Remain
Patient-friendly summaries also influenced intended communication with clinicians. After reading the summaries, more participants said they would wait for a scheduled appointment rather than try to contact their health care provider on the same day. This shift appeared across all three scenarios. It was most noticeable for the complex result that was negative for lung cancer but included incidental findings, and less pronounced for the result suspicious for malignancy. Some participants selected other communication options, such as using a patient portal, waiting for a clinician to call or arranging follow-up.
The summaries were widely viewed as helpful. Most participants rated the additional summary as at least moderately helpful, and more than half rated it as extremely helpful. This positive response was accompanied by clear requests for more information. Free-text comments most often asked for clearer guidance about next steps and urgency. Participants also wanted more context about risk and prognosis. Other requests concerned clarification of medical findings, treatment or procedure information, lifestyle and prevention guidance, definitions of medical terms and access to images or more data.
These remaining gaps matter because translation into simpler language does not automatically answer all practical questions. Participants still wanted to know what a finding meant for them, how quickly action was needed and what the result might imply. The summaries improved comprehension and anxiety, but some respondents still misidentified next steps in the more actionable scenarios. Direct clinician communication therefore remains important, particularly when results require follow-up or carry potential clinical seriousness.
Large language model-generated patient-friendly summaries improved understanding, ease of reading, confidence about next steps and self-rated anxiety in a vignette-based assessment of lung cancer screening CT results. The greatest benefits appeared among respondents with lower self-rated English and health literacy. The results remain preliminary because participants were online volunteers rather than patients undergoing lung cancer screening. Further validation in real patient populations is needed before such summaries can guide clinical practice. The main unresolved communication needs concern urgency, next steps, risk and prognosis.
Source: Radiology Advances
Image Credit: iStock
References:
Serna JA, Yu Y, Diaz P et al. (2026) Self-reported comprehension of large language model-generated summaries of lung cancer screening reports: a vignette survey. Radiology Advances, 3(2): umag008.
