CODEX Digest - 1.8.26

1) Confidence and certainty in medical diagnoses within acute healthcare: a scoping review. (subscription required)

Aiyer S, Higham H, Yeung N. BMJ Qual Saf. Epub 2025 Dec 3. 

Overconfidence can lead to diagnostic error. This review analyzes how clinicians' diagnostic confidence relates to accuracy, noting that each is influenced by a range of factors. The review outlines how misplaced confidence affects diagnosis and presents a model to help improve determination of confidence in clinical decision-making and its communication among clinicians.

2) Lung cancer deaths prevented and life-years gained from lung cancer screening. (subscription required) 

Bandi P, Landy R, Star J, et al. JAMA. Epub 2025 Nov 19. 

Annual lung cancer screening (LCS) lowers mortality and is recommended by the US Preventive Services Task Force, yet screening rates are subpar. This research letter reports that nationally only about 9-31% of eligible Americans received LCS in 2024. Raising uptake to 100% could triple the number of deaths prevented and life-years gained. The results underscore the importance of screening in the US population.

3) Frequency, cost, and variation in inpatient diagnostic imaging use in children’s hospitals. (subscription required) 

Collins ME, Stephens JR, Hall M, et al. J Hosp Med. Epub 2025 Oct 9.

Understanding utilization patterns can help curb overuse and improve diagnostic excellence. This study examines imaging use, costs, repeat rates, and hospital variation in hospitalized children. Chest X-rays and cardiovascular ultrasounds were most common, costly, and often repeated. ICU stays and complex chronic conditions led to higher imaging expenses.

4) Patient safety risks from AI scribes: signals from end-user feedback. (This is a preprint that has not gone through the peer review process).

Dai J, Huang A, Nasrallah C, et al. arXiv. Epub 2025 Dec 1.

AI scribes are changing clinical documentation, yet their impact on patient safety remains unclear. The preprint, real-world study reports that that AI scribes may introduce risks such as incomplete or incorrect histories and diagnoses, misrecording of existing diagnoses, and hallucinating conditions.  Further research is needed to evaluate these and treatment safety concerns.

5) Adverse effects of overdiagnosis on our understanding of disease. (subscription required) 

Gram EG, Davies E, Bell K.  BMJ Evid Based Med. Epub 2025 Oct 21.

Accurate understanding of disease relies on appropriate evidence. This commentary discusses how overdiagnosis can cause "knowledge harm" by affecting how medical information is produced, validated, and shared, leading to negative effects for individuals and populations.

6) Measuring what matters in radiology: a guide to selecting, implementing, and interpreting patient-reported outcome measures.

Kamran R, Doria AS, Patlas MN. Can Assoc Radiol J. Epub 2025 Nov 6.   

Reported patient experiences offer insights into clinical practices that inform improvement. This review discusses the value of patient-centered outcome work to radiology practice, highlights evidence that electronic patient-reported outcome measures (ePROMs) improve the patient diagnostic imaging experience, and offers a PROM implementation roadmap drawn from Canadian experience.

7) Fast, accurate assignment of clinical diagnoses from patient notes by a large language model: critical pediatric pneumonia as a use case.

Martin B, Payan M, LaVelle J, et al. Crit Care Explor. 2025;7(12):e1350

Rapid identification of critical illness can mitigate patient harm. This single-center project evaluates GPT-4o's ability to distinguish PICU admissions with or without bacterial pneumonia using clinical notes. The LLM accurately and quickly identified cases, indicating potential for automating tasks that usually require manual chart review.

8)  Artificial intelligence in skin and wound care: enhancing diagnosis and treatment with large language models.(subscription required)

Nelson S, Lay B, Johnson AR Jr. Adv Skin Wound Care. 2025;38(9):457-461.  

AI tools can enhance patient communication and symptom reporting, supporting diagnostic accuracy. This commentary reviews AI tools used in skin and wound care, including machine learning models and patient-facing language bots. The authors emphasize how AI can help clinicians, empower patients, and improve access to health care.

9) A Safety-II approach to learn from practice variation in the diagnostic process in the emergency department: an action research study. (subscription required) 

Roos R, Zwaan L, Maranus M, et al. Diagnosis (Berl). Epub 2025 Oct 13.

Learning from success ("Safety II") rather than just learning from errors is a key route to diagnostic excellence. This single-center study discusses how a Safety-II strategy was applied to identify and learn from diagnostic process variations and best practices in the emergency department, leading to effective adoption rates of the suggested changes.

10) Checking the box: the association between “problem list reviewed” and outdated diagnoses on the list. (subscription required)

Simon J, Panzer J, Ekong A, et al. Appl Clin Inform. 2025;16(5):1779-1786. 

Problem lists are intended to support high-quality care, but are often long and outdated, leaving diagnoses remaining on lists far longer than clinically appropriate. The authors looked at community health center problem lists and if checking the "reviewed" button had any relationship to removal of short-term diagnoses.

11) Accessing AI mammography reports impacts patient follow-up behaviors: the unintended consequences of including AI in patient portals.

Song EC, Bernstein MH, Lay PS, et al. npj Digital Med. Epub 2025 Dec 3.

Direct AI communication with patients raises important issues. This study examines how sharing AI mammogram interpretations through patient portals affects behavior and diagnosis. Patients were randomized to hypothetical mammogram result conditions. Mention of AI in a result increased likelihood of litigation although naming the AI risk of error mitigated this.

12) Perceived racial discrimination in health care in relation to late stage at breast cancer diagnosis.

Xu NN, Barnard ME, Holder EX, et al. Breast Cancer Res. 2025;27(1):208.

Although mammographic screening rates are now similar for Black and white US women, Black women are still more often diagnosed at later stages. This study links perceived racial discrimination to late-stage breast cancer diagnosis, highlighting the need to address inequities in care for Black women in the US.