CODEX Digest 05.21.26

People with lived experience of mental ill-health (PWLE) encounter health care inequities, including stigma, discrimination, and diagnostic overshadowing. This interview study examines factors affecting the experience of PWLE across four European health systems. Three key themes emerged from the analysis highlighting systemic, personal, and diagnostic shadowing influences affecting PWLE care. Across all countries represented, patients reported not being taken seriously and facing stigma and discrimination in healthcare settings.

Reducing diagnostic error may require more standardized safety event reviews and better data sharing across health systems. This survey finds that diagnostic errors were commonly identified in safety reporting databases, but less consistently across other event reviews, with few systems tracking lessons across categories. The authors suggest more consistent review practices and shared databases could improve learning and reduce patient harm.

Ambient AI documentation tools create initial clinical note drafts from patient-provider conversations, which clinicians review and edit before adding to the EHR. Understanding clinicians' modifications is important for refining AI and improving efficiency. This study finds that changes were mainly made to enhance clinical accuracy, diagnostic assessment, specificity and history, specialty precision, and comply with medico-legal, billing, and documentation requirements.

AI in radiology has potential to improve diagnostic excellence, but limited cost-effectiveness data slows adoption. This review analyzes economic evaluations of radiology AI, highlighting inconsistent outcome measures and calling for standard metrics to enhance comparability, research, and responsible implementation in imaging services.

Patients’ views of clinician screening metrics can influence their willingness to be screened, follow up, or discuss results, affecting diagnosis. Survey results show half disagreement with metrics based on screening attendance, two-thirds opposed financial incentives, and most preferred physician scores tied to patient-centered discussions over care receipt.

Wearable technologies are increasingly used to detect atrial fibrillation (AF) outside clinical settings. This review finds that continuous screening with wearables nearly tripled AF diagnoses in asymptomatic, high-risk individuals, leading to more treatment but no clear improvement in outcomes. The findings suggest overdiagnosis and emphasize the need for careful integration of wearable technologies in primary care.

Cancer diagnosis disclosure is a process over time, and involving families in care is crucial. This commentary reviews a Chinese study with global relevance, revealing the complexities of nondisclosure as families and clinicians decide what to share or withhold. Individualized care, cultural humility, contextual awareness, and understanding of patient and family perspectives are vital. Interdisciplinary collaboration helps explore illness perceptions and preferences about cancer diagnosis communication.

Patient portals give patients access to diagnoses, test results, and clinical communication, but they often inherit EHR usability issues that hinder navigation and interpretation. This assessment of two portals found gaps in critical functions like diagnosis display and provider messaging—flaws that may reduce patient understanding and engagement. While most issues were minor, their accumulation in high-risk functions could threaten diagnostic safety.

Clinical reasoning is a useful benchmark for LLM task performance. This study shows that OpenAI's o1-preview model outperforms other LLMs in diagnostic and management reasoning, particularly generating differential diagnoses. The study included a real-world comparison of the LLM and human clinicians in emergency department patients. The LLMs performance in probabilistic reasoning and triage remains similar to previous models, suggesting strengths in complex critical thinking but no improvement in probability-based tasks. Based on these results, the authors advocate for prospective clinical trials to measure the impact of LLMs on diagnostic errors and other patient-centered outcomes.

Affected cognition is known to bias information and contribute to diagnostic error. This study examines how introducing distracting, believable information influences diagnostic accuracy in UAE medical students. When such information was provided early in the decision process, students made more diagnostic errors and processed information less thoroughly, indicating premature closure. The results find reduced processing time only when biased information appeared early in the decision making, highlighting its impact on error rates.

Risk-stratified cancer screening can reduce overdiagnosis in low-risk women, improve early detection in high-risk groups, and lower costs, though implementation is difficult. This trial shows that automated risk-based screening is feasible and acceptable with minimal impact on short-term psychological well-being, but these early results should be viewed cautiously since the main outcome will be measured after an additional 17 months of follow-up.

Patients often report better diagnostic experiences when imaging is performed, though the reasons remain unclear. This mixed-methods study uses the PRIME-ED (Patient-Reported Measure of Diagnostic Excellence in the Emergency Department) tool to assess patient perceptions of diagnosis and communication in one academic emergency department. Most patients valued thorough evaluations, while negative feedback centered on unresolved symptoms and a desire for additional testing.