CODEX Digest - 11.20.25

1) Defining minimum image quality criteria for common diagnostic point-of-care ultrasound images: a position statement of the Society of Hospital Medicine.

Anstey J, Bhasin A, Franco-Sadud R, et al. J Hosp Med. Epub 2025 Sep 22. 

Point-of-care ultrasound (POCUS) is being used for diagnosis across a variety of specialties but no benchmark for image quality exists. This position statement addresses this gap. The framework provides evaluation criteria for use in five common POCUS applications to assess image quality in support of training, quality, and credentialing efforts. 

2) Advancing medical artificial intelligence using a century of cases. (This is a preprint that has not yet gone through the peer review process). 

Buckley TA, Conci R, Brodeur PG, et al. arXiv. Epub 2025 Sep 15.  

This preprint introduces a novel AI discussant (“Dr. CaBot”) and a benchmarking case database (“CPC-Bench”) based on cases from the New England Journal of Medicine (NEJM). These tools were then used to assess LLMs and physician performance on differential diagnosis and medical presentations. It finds LLM results to be superior to physicians, but less so for image assessment and literature identification. The tools have been made publicly available for use in testing diagnostic AI applications.  

3) When caution becomes harm: understanding the psychology of over-investigation. (subscription required) 

Chatterjee S. Am J Med. Epub 2025 Oct 15.

The overuse of diagnostic tests, or over-investigation, is wasteful, burdensome, and harmful to patients. This review looks at cognitive, social, systemic, and relational contributors to excessive testing. The piece calls for adjusting clinician behaviors and patient care efforts to improve testing processes and reduce burnout, patient mistrust, and spending. 

4) Listening to patients: A qualitative study on diagnostic delay, coping strategies and stigma in early-onset colorectal cancer.

Gros HL, Taha-Mehlitz S, Erdem S, et al. Colorectal Dis. 2025;27(11):e70285.  

Patient firsthand accounts of care journeys provide important insights into process gaps that degrade diagnostic timeliness. This study reports on the experiences of Swiss colorectal cancer patients under the age of 50 looking at diagnostic delay, disease management, and post-diagnosis stigma. The results illustrate challenges facing early onset colon cancer patients and highlight the need for improved physician awareness of these challenges to enhance patient care.   

5) Radiologist interaction with AI-generated preliminary reports: a longitudinal multi-reader study. (subscription required) 

Hong EK, Suh C-H, Nukala M, et al. J Am Coll Radiol. Epub 2025 Sep 20. 

Communicating test results effectively is essential for diagnostic excellence, but it takes time and can contribute to clinician workload, delays, and errors. This study examines whether AI could help improve the efficiency of chest X-ray reporting. The results were encouraging, but report quality varied—especially for complex cases—so the findings should be interpreted with caution.   

6) Educational strategies to prepare trainees for clinical uncertainty. (subscription required) 

Ilgen JS, Dhaliwal G. N Engl J Med. 2025;393(16):1624-1632. 

This discussion of clinical uncertainty and how physicians manage it is anchored in the diagnostic process, highlighting its presence across the care continuum. The article calls for educators to help trainees manage uncertainty by understanding how to recognize and communicate uncertainty, while labeling it as an opportunity for learning. The authors suggest that technology will not reduce uncertainty but change the nature of its presence in medical care. 

7) Evaluation of imaging research adherence to the STARD 2015 reporting guideline: update 9 years after implementation and baseline assessment.

Kashif Al-Ghita M, Dawit H, Kazi S, et al. Can Assoc Radiol J. 2025;76(4):631-645. 

High quality research is needed to assess the effectiveness of diagnostic improvement strategies and tools. This review finds that published reporting guidance on high quality imaging research is not routinely followed. Although improvements were realized since a 2016 assessment, gaps in reporting still exist that may undermine the diagnostic imaging accuracy evidence base. 

8) Machine learning-based error detection in the clinical laboratory: a critical review. (subscription required)  

Lin Y, Mensah IK, Doering M, et al. Crit Rev Clin Lab Sci. 2025;62(7):535-547. 

Reliable laboratory testing processes are fundamental for diagnostic excellence. This review summarizes the potential of advanced computing technologies to improve laboratory testing through identifying variations that degrade processes and result in errors. 

9) Screening for asymptomatic tuberculosis among adults with household exposure to pulmonary tuberculosis: a prospective observational cohort study.

Mendelsohn SC, Mulenga H, Tameris M, et al. Lancet Glob Health. 2025;13(11):e1869-e1879.  

Screening for disease is a primary strategy for early diagnosis and disease incidence policy and management, but only if the results are accurate. This South African study finds that chest radiograph screening of asymptomatic individuals missed 40% of tuberculosis cases among household contacts, indicating that estimates of endemic disease may be incorrect.   

10) Grounded large language models for diagnostic prediction in real-world emergency department settings. 

Niset A, Melot I, Pireau M, et al. JAMIA Open. 2025;8(5):ooaf119.  

As emergency departments (ED) face high demand, complex patients, and staffing strain, AI systems could enable faster, more informed care. This Belgian study examines how AI might support diagnostic performance in a real ED. Both open-source and proprietary AI tools performed well in predicting diagnoses, with case-specific factors having the biggest influence on accuracy. Larger, real-time studies tied to patient outcomes and workflow are needed to confirm their full value. 

11) How do health literacy and chronic disease influence the diagnostic evaluation of patients with lung cancer symptoms? 

Sætre LMS, Balasubramaniam K, Wehberg S, et al. Acta Oncol. 2025;64:1455-1464.  .  

An understanding of patient population challenges may help to clarify what distinct characteristics affect participation in diagnostic screening. This Danish study found that former smokers were more likely than current smokers or never smokers to complete lung cancer imaging after seeing a general practitioner. Health literacy did not explain the difference, suggesting the need for different types of initiatives.  

12) Using clinical cases with diagnostic errors and malpractice claims: impact on anxiety and diagnostic performance in GP clinical reasoning education.

van Sassen C, Mamede S, Hooftman J, et al. Adv Health Sci Educ Theory Pract. 2025;30(5):1403-1423.  

The use of error as an educational strategy can be both motivational and stressful for learners. This study investigates how negative case outcome context could impact learning and diagnostic performance. The findings suggest cases involving error or malpractice claims did not induce anxiety or reduce learner confidence, underscoring their potential value as content for clinical reasoning education.