CODEX Digest - 8.14.25

Operationalization of artificial intelligence applications in the intensive care unit: a systematic review. 

Berkhout WEM, van Wijngaarden JJ, Workum JD, et al. JAMA Netw Open. 2025;8(7):e2522866.  

Diagnosis in intensive care is one area that AI can improve. This review assesses the readiness of AI applications in a variety of critical care practices and found more than half of the included studies were at high risk of bias, only two percent of the studies were in clinical integration, and no AI models were found to be fully ready for integration. The overall findings indicate that while research on AI applicability is growing, more focus is needed on practical implementation of AI and its clinical impact.   

Understanding the AI Wave: Foundational Knowledge for Improving Diagnosis and Beyond.

Biro J, Salvador D. Agency for Healthcare Research and Quality; 2025.  

There is untapped potential in the application of AI in diagnostic processes, but the road to get there is yet unpaved. This issue brief introduces the form and function of AI, its promise, danger, and future and is a helpful overview for clinicians, system leaders, and patients who are new to the AI health field.  The report is part of a collection of AHRQ issue briefs developed on the topic of diagnostic safety.  

Leveraging diagnostic timeouts to foster interprofessional communication. (subscription required)    

Bowen J, Demeritt B, Ipsaro AJ, et al. Diagnosis (Berl). Epub 2025 Jul 8. 

Communication failures and cognitive biases degrade diagnostic excellence. This intervention engaged 90 bedside nurses ​​in one pediatric acute care unit in the use of a diagnostic timeout framework to reexamine conclusions for patients requiring additional attention. Teamwork, communication, and psychological safety were all positively affected using the framework, which triggered additional diagnostic work for half of the patients involved. 

Better understanding the clinical reasoning skills of 4th-year medical students through think aloud interviews: implications for theory and practice. 

Cuddy MM, Runyon C, Luciw-Dubas UA, et al. Adv Health Sci Educ Theory Pract. Epub 2025 Jun 25.  

Understanding student clinical reasoning steps can help design effective curriculum to enhance diagnostic skills. This qualitative study interviewed fourth year medical students while navigating patient charts and answering clinical questions. Thematic analysis revealed a new framework describing four skills: interpreting, framing, generating, and justifying. Educators can use this framework to design strategies for clinical reasoning skill development.  

Patient perspectives on artificial intelligence in health care: focus group study for diagnostic communication and tool implementation. 

Foresman G, Biro J, Tran A, et al. J Particip Med. 2025;17:e69564.  

Patient acceptance of AI as a diagnostic tool is emerging.​     ​This study illustrates the value of patient perspectives when developing AI healthcare processes. Across five AI scenarios, 17 patients were most comfortable with a digital scribe and least comfortable with a virtual human. Results show patients want human oversight, effective communication, tool development information availability. These findings can guide physicians, healthcare organizations, and policy makers to use AI ethically and put patients first. 

Evaluation of the accuracy and safety of machine translation of patient-specific discharge instructions: a comparative analysis.

Kong M, Fernandez A, Bains J, et al. BMJ Qual Saf. Epub 2025 Jul 9. 

Patients engaging in their care after hospitalization relies on an accurate understanding of what happened to them in the emergency department. However, language barriers can make that difficult. This study compares ChatGPT-4 and Google Translate to develop patient-centric discharge instructions for Spanish, Chinese, and Russian speaking patients. While translation quality was good for Spanish and Chinese, it was less for Russian, suggesting potential harm in some languages thus calling for human oversight in the outputs. Diagnostic relevant outcomes of the work include accuracy of primary diagnosis and return precautions. 

AI-based clinical decision support for primary care: a real-world study. (this is a preprint that has not gone through peer review) 

Korom R, Kiptinness S, Adan N, et al. arXiv. Epub 2025 Jul 22.

Active care rollouts of AI tool use in clinical environments that provide impact data are uncommon. This non-peer reviewed preprint quality improvement study examines the implementation of an AI clinical decision support tool in 15 primary care clinics in Kenya. Clinicians with access to AI Consult made 16% fewer diagnostic errors and 75% considered access to the tool to be substantial. The results provide insights into the use of real-world AI decision support and diagnostic error outcomes. 

Understanding ethnic inequalities in cancer diagnostic intervals: a cohort study of patients presenting suspected cancer symptoms to GPs in England.  

Martins T, Down L, Samuels A, et al. Br J Gen Pract. 2025;75(754):e333-e340.

The timeliness of cancer diagnosis is an element of excellence that is encumbered by socioeconomic forces. This cohort study of 70,971 cancer patients in English primary care found differences in how long it takes to get a diagnosis across racial and ethnic groups. Secondary care delays were seen for four common cancers and primary care delays for two. Although these delays were small, the inequities illustrate systemic weaknesses that affect Black and Asian-identified patients.  

Dental diagnostic errors and characteristics associated with claims in the United States, 1990-2020. 

Singhania R, Obadan-Udoh E. J Am Dent Assoc. 2025;156(7):563-570.  

Diagnostic excellence in dentistry is understudied. This cross-sectional analysis of dental paid malpractice claims from the National Practitioner Data Bank found 8.7% of 58,229 claims to be associated with diagnostic problems and missed diagnoses to be prevalent. The overall count of dental diagnostic errors was found to have been consistent since 2008. The results flag the need for improved policy to guide education and training in the specialty. 

Frequent missed opportunities for earlier diagnosis of advanced stage colorectal or lung cancer. (subscription required) 

Zimolzak AJ, Kapadia P, Upadhyay DK, et al. JAMA Intern Med. Epub 2025 Jul 21.

Effective measurement of diagnostic error is a work in progress. This cohort study uses data from cancer diagnosed over a five-year period at two integrated healthcare systems to develop a diagnostic quality measure related to advanced disease stage and missed diagnosis. Contributing factors to delayed advanced stage cancer care include missed screenings, problems in the patient-physician relationship, and testing issues. Cancer stage at diagnosis may serve as a valuable quality measure to design improvement mechanisms and reduce preventable diagnostic lapses.