CODEX Digest - 9.25.25

Role of overdiagnosis in the rising incidence of endometrial cancer: a population-based ecological study

Barmayehvar B, Behzadnia A, Nicholson B, et al. BMJ Evid Based Med. Epub 2025 Jul 24.  

Examining diagnostic trends helps show whether rising case numbers reflect a real increase in illness or are the result of changes in testing and diagnosis methods. This Norwegian study analyzed endometrial cancer (EC) trends across decades and found a substantial rise in incidence while mortality remained relatively stable. The increase was seen across both early- and late-stage diagnoses, especially in older age groups. Although transvaginal ultrasound (TVUS) usage increased over time, it showed only a weak correlation with early-stage EC detection. The findings suggest that diagnostic patterns and stage migration may contribute to the observed trends, but TVUS alone does not appear to fully explain them.  

Implementation of diagnostic stewardship to improve urinary tract infection antibiotic use across three medical centers. (subscription required) 

Claeys KC, Brown CH, Pineles L, et al. Clin Infect Dis. Epub 2025 Jul 26.

Overtreatment of urinary tract infections (UTIs) remains common and can lead to bacterial resistance and unnecessary side effects. This study tested diagnostic stewardship strategies across three veteran hospitals and found that conditional reflex urine culturing, wherein ordered urine cultures were only performed if urine white blood cells were at a certain count, led to a nearly 40% drop in UTI antibiotic use—with no increase in adverse outcomes like bacteremia. Other interventions at the ordering or reporting level weren’t as effective.  

Human factors in diagnostic radiology: practical challenges and cognitive biases. (subscription required) 

Cowen JE, Vigneswaran G, Bekker J, et al. Eur J Radiol. 2025;190:112248.

Radiologists are facing increasing workloads and time pressure. This narrative review investigates how human factors—like fatigue, distractions, and interruptions—alongside common cognitive biases, undermine diagnostic accuracy in radiology. The authors categorize these challenges into two domains (practical environmental issues and cognitive biases) and synthesize evidence-based strategies such as environmental optimization, structured reporting, and cognitive debiasing to mitigate them.  

Appendicitis: common and commonly missed - the story of Alice Tapper. 

Dajer AJ, Olson APJ. Diagnosis (Berl). Epub 2025 Aug 1.   

Appendicitis is commonly misdiagnosed. This case study shares the story of a 12-year-old girl, Alice Tapper, whose appendicitis was initially misdiagnosed as gastroenteritis, leading to a delayed diagnosis and subsequent perforation. The authors use this case to highlight the challenges in diagnosing appendicitis, particularly in pediatric patients, and the potential consequences of misdiagnosis.  

**Alice Tapper will be a featured speaker at DEX25.

Use of ChatGPT for Urinary Symptom Management Among People With Spinal Cord Injury or Disease: Qualitative Study.

Hose BZ, Rounds AK, Nandwani I, et al. JMIR Rehabil Assist Technol. Epub 2025 May 29.  

AI tools like ChatGPT are increasingly supporting patient-driven understanding of their diagnosis and symptom management. This two-phase study engaged 30 patients with a spinal cord injury or disease in scenario-based testing and follow-up interviews to assess their experience using ChatGPT for urinary-symptom related queries. Participants trusted and agreed with ChatGPT's advice and found it comparable to professional medical guidance, but concerns were raised about the lack of personalized health data integration and source transparency. However, the findings are promising for centering patients with lived experience regarding a condition that has difficult to interpret symptoms.  

Cancer screening as a conversation, not a checklist: enhancing professional cancer education through narrative medicine. (subscription required)   

Jerjes W, Williams S. J Cancer Educ. Epub 2025 Aug 6. 

Cancer screening guidelines are intended for early detection but can result in too heavy a focus on checklists and logistics rather than talking to patients about their fears, misconceptions, and personal concerns. This reflective piece argues for incorporating narrative medicine, including reflective listening and eliciting patient stories, to deepen provider-patient communication and thereby improving patient trust and screening adherence. The authors suggest structural changes in medical education and clinical workflows to better support more meaningful conversations between patients and clinicians.  

CARE-AD: a multi-agent large language model framework for Alzheimer’s disease prediction using longitudinal clinical notes. 

Li R, Wang X, Berlowitz D, Mez J, et al. NPJ Digital Med. 2025;8(1):541.   

LLMs offer a promising approach to improving early Alzheimer’s diagnosis by analyzing unstructured clinical notes and detecting risk factors before the condition is formally coded. This retrospective study introduces CARE-AD, a multi-agent LLM-based framework that mimics multidisciplinary clinical reasoning to predict Alzheimer’s disease using longitudinal clinical notes. It outperformed single-LLM approaches including for long-term prediction up to 10 years before diagnosis. However, the model’s generalizability is limited by its reliance on a predominantly male population and selection bias toward patients with extensive documentation.  

Clinical Reasoning: Coaching the Struggling Medical Learner.

Parsons A, Warburton K. Association for Hospital Medical Education; 2025. 

This peer-reviewed book features multiple contributors from the clinical reasoning education and research community and is intended to guide educators in helping medical learners achieve competence in clinical reasoning. The book includes case examples, coaching tools, real‑world scenarios, and strategies for scaffolded feedback in both classroom and clinical settings. The authors emphasize a holistic, growth‑oriented approach: helping learners take ownership of their reasoning process rather than simply correcting errors. 

Inpatient and outpatient radiology report access after the 21st century cures act.

Pollock JR, Tariq A, Schmitz JJ, et al. JAMA Netw Open. 2025;8(8):e2528683.

This retrospective, cohort study examined how quickly patients accessed their radiology reports before and after the implementation of the 21st Century Cures Act, which requires immediate release of medical records to patients. Using data from over 1.2 million imaging exams, the study found a 70% reduction in the median time to patient access. Outpatients accessed results fastest, with a median drop from 4.9 to 1.1 hours post-implementation. Patients reported positive benefits, while clinicians were concerned about increases in call volumes and notification of cancer before discussion with a clinician. These findings highlight how policy changes can shift patient engagement with diagnostic information. 

Harnessing AI for improved diagnosis and management of pediatric sepsis: current advances, challenges, and future directions. (subscription required) 

Siolos P, Pasha S, Triantafyllou M, et al. Pediatr Emerg Care. 2025;41(7):576-585. 

AI may help transform pediatric sepsis diagnosis and management. This continuing medical education article explores how AI and machine learning are being applied to pediatric sepsis—covering tools such as early screening systems, risk stratification algorithms, personalized treatment models, and wearable or virtual assistant technologies integrated into clinical workflows. The authors assess current implementations and trials while also discussing practical considerations like bias mitigation and seamless integration into healthcare systems.  

Artificial intelligence in prediction of postpartum hemorrhage: a primer and review.

Wakefield BM, Zapf MA, Ende HB. Int J Obstet Anesth. 2025;63:104694.  

This article provides a review on how AI is currently being used to predict postpartum hemorrhage (PPH) and a practical primer on interpreting these AI tools.  The authors note the importance of standardized risk assessment in improving PPH recognition and prevention, while also calling for stronger validation and implementation science to bring these AI tools into routine obstetric care. 

Leveraging large language models to advance certification, physician learning, and diagnostic excellence.

Wang T, Price DW, Bazemore AW. J Am Board Fam Med. Epub 2025 Aug 25.    

Longitudinal assessment platforms support diagnostic education by providing continuous assessments to ensure physicians continue to enhance their medical knowledge and diagnostic skills. This article introduces a project that evaluates the use of LLMs in the American Board of Family Medicine’s Continuous Knowledge Self-Assessment platform to generate new questions, enhance knowledge transfer, and provide personalized feedback to physicians.