CODEX Digest - 9.11.25

Prevalence and accuracy of nursing diagnoses in patients with malignant bronchial and lung cancer: a retrospective observational study

Cesare M, Magliozzi E, D’Agostino F, et al. Eur J Oncol Nurs. 2025;77:102931.   

Nurses play distinct diagnostic roles that impact cancer care. This observational study at one university hospital in Rome examined nursing diagnosis in patients with malignant bronchial and lung cancer. Infection risk, fall risk, and acute pain were the most common nursing diagnoses for this patient population. Analysis using a standardized score found nursing documentation accuracy to be high, but case formulations to be less accurate. The authors call for nursing diagnostic reasoning improvement to enhance case documentation.

Identifying opportunities to reduce imaging overuse in hospitalized children.

Desai S, Treasure J, Richardson T, et al. J Hosp Med. 2025;20(6):573-580.  

Excellent diagnosis is cost effective, safe, and reliable. This cross-sectional study examined pediatric admissions across 50 hospitals. The goal was to identify common conditions that generate high volumes of imaging, costs, and hospital-level variation. They found high variability for scoliosis imaging and pre-septal cellulitis. These findings help flag instances where guidance and performance efforts can target organizational work and improve imaging practice value. 

Racial and ethnic disparities in pediatric emergency department patients with missed opportunities for diagnostic excellence. (subscription required) 

Eyal K, Leonard J, Dominguez F, et al. Diagnosis (Berl). 2025;12(3):396-401.  

Thorough emergency department (ED) evaluation enables timely identification of disease. This single center, retrospective review of pediatric ED or urgent care cases sought to document disparate diagnostic processes across multiethnic patient populations. Missed opportunities for diagnostic errors were not more prevalent among non-White versus White patients. However, White patients had significantly more testing and workup, highlighting a possible mechanism of diagnostic disparities.   

Population health colorectal cancer screening strategies in adults aged 45 to 49 years: a randomized clinical trial. (subscription required) 

Galoosian A, Dai H, Croymans D, et al. JAMA. Epub 2025 Aug 4. 

Patient uptake of standard screening, now starting at age 45, is an important contributor to timely colon cancer diagnosis. This large-scale, single center study sought to improve colonoscopy participation of 45-49 age adults using one of four outreach tactics. Researchers found that their usual care arm (mailing an actual test kit to patients) resulted in higher participation than inviting participants via other “active choice” test strategies including explicitly asking patients to accept or decline a screening offer.  

Incidental finding of coronary and non-coronary artery calcium: what do clinicians need to know? (subscription required) 

Haudenchild C, Parsa S, Rodriguez F. Curr Atheroscler Rep. 2025;27(1):71. 

Incidental findings serve as a potential avenue to initiate the diagnostic process. This review examines how analyzing and communicating incidental findings of coronary artery and other vascular calcification can contribute to timely diagnosis and management of cardiovascular disease. The authors explore the evidence on the potential of a variety of AI-centered solutions to automate, optimize, and scale the sharing of incidental findings to improve care.  

Clinical Reasoning in the Health Professions.

Higgs J, Jensen GM, Loftus S, et al, eds. 5th ed. Elsevier; 2025. 

Clinical reasoning involves the application and reinterpretation of data and information throughout the diagnostic process to determine a patient’s condition. This textbook provides an entryway into factors affecting clinical reasoning for health professions students and educators. It discusses the future of the study of clinical reasoning and educational strategies to embed the skills needed in learners and assessment methods.  

Diagnostic accuracy differences in detecting wound maceration between humans and artificial intelligence: the role of human expertise revisited.

Kücking F, Hübner UH, Busch D. J Am Med Inform Assoc. 2025;32(9):1425-1433.  

Expertise may be a delineator in understanding the accuracy differences between AI-powered diagnostic imaging analysis and clinicians. This study examined the role that expertise has among 481 clinicians looking at 30 chronic wound assessments compared to convolutional neural network (CNN)-assisted wound image review. Results show clinician accuracy was significantly lower than those of the CNN in the groups without formal qualifications, low diagnostic self-confidence, no focus on wound care, and short work experience. These results raise questions over what defines human expertise and who most benefits from AI-assisted diagnostics.  

Simulated emergency room sign-outs: what they tell us about diagnostic accuracy and opportunities for improvement.

Matin R, King BR, Mehta AM, et al. J Patient Saf. Epub 2025 Aug 18.  

Sign-out documentation is central to effective communication as care teams change during a patient journey. This vignette study examined the quality of ED sign-outs information and their effect on diagnostic accuracy. Results show the extensiveness of the information in the sign-out had no impact on clinicians' diagnostic accuracy or confidence. Instead, stage of care was the primary driver of uncertainty. Patients early in their care journeys should be recognized as more vulnerable to diagnostic inaccuracy. 

Sustained impact: long-term application of diagnostic uncertainty communication training. 

McCarthy DM, Hernandez J, Papanagnou D, et al. AEM Edu Train. 2025;9(4):e70086.  

Challenges involved in communicating diagnostic uncertainty can affect the doctor/patient relationship. This study reports on the robustness of uncertainty communication skills training over time. The authors surveyed participants in an emergency medicine resident simulation learning program focused on a checklist with best practices on communicating diagnostic uncertainty to patients at discharge. They found that even four years after the training, learners continued to apply concepts from the training in their daily practice. 

Machine learning techniques for stroke prediction: a systematic review of algorithms, datasets, and regional gaps.

Soladoye AA, Aderinto N, Popoola MR, et al. Int J Med Inform. 2025;203:106041.  

Stroke is a common and serious condition that can be challenging to predict, diagnose, and prevent. This systematic review from researchers in Nigeria and Ireland examined machine learning (ML) stroke prediction methods and their usefulness in the direct care environment. The analysis found that several ML tools demonstrated promising accuracy in predicting stroke risk in retrospective studies; however, prospective validation in diverse clinical settings and among high-risk populations, including in sub-Saharan Africa, is still lacking. 

From margin to center: transforming the landscape of diagnosis through community engagement. (subscription required) 

Srivarathan A, Giardina TD. J Patient Saf Risk Manag. Epub 2025 Aug 12. 

The patient is the ultimate initiator of a diagnostic course and the community around them can affect action. This commentary discusses social and relational elements that construct the diagnostic process. The authors highlight the role of communities—both tangible and online—that initiate patient participation in diagnosis and enable improvements. They encourage healthcare workers to embrace the community outside the healthcare setting as a partner to achieve patient-centered diagnostic excellence.   

Generative medical event models improve with scale. (This is a preprint that has not gone through peer review). 

Waxler S, Blazek P, White D, et al. arXiv. Epub 2025 Aug 16.

Training AI on large sets of medical data offers opportunities for early diagnosis. This preprint describes CoMET, a set of transformer models trained on 115 billion medical events from 300 million patients in the Epic system and is designed to predict future medical events. Results show CoMET could outperform specialized models without extra modification, showing the potential of large generative models to improve clinical decisions and healthcare operations. Amongst the capabilities discussed, analysis of predictive differential diagnosis demonstrated weeks earlier identification of disease 25% of the time.