CODEX Digest - 7.17.25

Symposium in Diagnostic Reasoning. (subscription required) 

Benbenek MM. AACN Adv Crit Care. 2025;36(2):106-186. 

Nurses are vital partners in diagnosis. This special issue features articles on diagnostic reasoning for nurse practitioners covering topics such as educational strategies and competency assessment tools.  

Reshaping organizational culture in pathology. (subscription required) 

Elzie D. Clin Lab Med. Epub 2025 May 24. 

Safe and collaborative clinical environments are essential for effective care. This article discusses improving the pathology work culture to support accurate diagnosis by flattening hierarchy, encouraging inclusivity, and accepting new technologies such as artificial intelligence. 

Dedicated AI expert system vs generative AI with large language model for clinical diagnoses. 

Feldman MJ, Hoffer EP, Conley JJ, et al. JAMA Netw Open. 2025;8(5):e2512994.

Decision support systems (DDSS) are valuable for diagnosis, but their performance compared to large language models is unclear. This study assessed the performance of both types of decision support systems in three organizations with and without laboratory results. The work found that DDSS outperformed the AI systems in diagnostic accuracy, though the AI results were still good. The study recommends using both types of systems in consort to enhance clinician decision-making. 

**This was a recent CODEX Editor’s Pick.  

A large health system quality improvement intervention providing training and tools to improve detection of cognitive impairment in primary care. (subscription required) 

Gaster B, Zigman Suchsland M, Liao JM, et al. J Am Geriatr Soc. Epub 2025 Jun 5. 

Early diagnosis of cognitive disease is important for successful treatment and patient quality of life. This article describes an approach taken by one health system to improve diagnostic excellence in 14 primary care clinics. The intervention centered on education and tool accessibility in exam rooms to evaluate cognition. The actions increased cognitive assessment completion and newly diagnosed Alzheimer’s disease and related conditions.

Trends of diagnostic adverse events in hospital deaths: longitudinal analyses of four retrospective record review studies. (subscription required)  

Hooftman J, Zwaan L, Sikkens JJ, et al. Diagnosis (Berl). 2025;12(2):201-207.    

Data analysis over time informs assessments of factors contributing to the state of a service or activity. This study reviewed Dutch adverse event data where the diagnostic process was found to be a primary contributor to patient harm. Results show a decline in diagnostic related deaths over four years which plateaued, similarly to trends for general adverse events for the last decade in Dutch hospitals. Many diagnostic adverse events were linked to human actions and organizational factors and were found to be preventable, suggesting the need for ongoing improvements in diagnostic processes. 

Measuring the association between diagnostic errors and neighborhood disadvantage.

Kaiksow FA, Brendel M, Hubbard CC, et al. J Hosp Med. 2025;20(7):727-730.  

Socioeconomic challenges detract from population health and diagnostic excellence. This study examined acutely ill in-patients experiencing diagnostic error who were moved to the ICU or died while hospitalized. The authors found no alignment between patient deaths and their living environments. The study didn't examine long-term community health impacts but suggests social and clinician factors that could affect diagnosis before hospitalization. 

​Artificial intelligence in digital self-diagnosis tools: a narrative overview of reviews. 

Mentzou A, Rogers A, Carvalho E, et al. Mayo Clinic Proc: Digit Health. 2025;3(3):100242   

Digital self-diagnosis tools or symptom checkers are popular with patients, but more research is needed to confirm their usefulness and accuracy. This review identifies gaps in the understanding of symptom checker development, use, effect, and audit. The authors call for multidisciplinary efforts to improve the reliability and safety of these patient-facing tools.  

Sequential diagnosis with language models.

Nori H, Daswani M, Kelly C, et al. arXiv. Epub 2025 Jul 2.

The diagnostic process is multilayered and involves integrating information and human knowledge to arrive at an accurate decision. This developer-authored pre-peer review study introduces a product framework and model that mirrors human decision-making to assess accuracy and the cost effectiveness of diagnostic processes. While the system showed better accuracy than physicians in these tests, its performance in real-world conditions may differ. 

AI and XAI second opinion: the danger of false confirmation in human-AI collaboration. (subscription required) 

Rosenbacke R, Melhus Å, McKee M, et al.  J Med Ethics. 2025;51(6):396-399. 

Second opinions serve as valuable strategies for clinicians and patients to ensure a diagnosis is correct and gain additional treatment perspectives. This article examines the use of AI as a second opinion provider and raises the prevalence of ‘false confirmation’ - when AI validates an incorrect human decision. It calls for physicians to make decisions prior to consulting AI tools and increase their awareness of this type of error when using AI to support decision making.

Misdiagnosis in breast imaging: a statement paper from European Society Breast Imaging (EUSOBI)-Parts 1 & 2. (subscription required) 

Thomassin-Naggara I, Athanasiou A, Kilburn-Toppin F, et al.  Eur Radiol. 2025;35(5):2387-2411. 

Poor mammography techniques and interpretation can lead to missed breast cancer diagnoses. This pair of organizational guidelines provides recommendations to enhance excellence in diagnostic breast imaging. Part 1 reviews practice mistakes that contribute to missed breast cancer misdiagnosis and part 2 discusses the reasons for those errors. Suggestions for improvement include training all providers, enhancing collaboration, and focusing on team communication to enhance diagnosis.