CODEX Digest 03.19.26

Screening eligibility improves early cancer detection. This cohort study compares eight lung cancer screening strategies and examines algorithmic bias affecting Black/African American participants. The results show that most strategies identify fewer Black/African American patients as eligible compared to White patients. Race-aware methods may be needed to address bias and promote equity in lung cancer screening.

A international NEJM Catalyst Insights Council survey shows healthcare leaders are exploring diverse AI applications. US organizations are adopting AI faster than others. Drafting clinical notes is cited as AI's main use with participants expecting clinical decision support and diagnosis to have the greatest impact in the next three years. AI's influence on diagnostic imaging and pathology is also noted as emerging.

Timely breast cancer diagnosis is crucial for outcomes, and identifying process barriers helps to guide improvements. This UK retrospective cohort study reviews primary and secondary care data to assess racial and ethnic differences in diagnostic intervals. Results show Black patients experience longer diagnostic intervals than White patients regardless of breast lumps as a presenting symptom.

The impact of referral documentation on diagnostic accuracy is unclear. This single-center Japanese study examined whether referral letters influence diagnostic errors in patients referred to a general internal medicine clinic. Findings show that referral letters improve communication to support accurate diagnoses and help reduce errors.

Endometriosis patients commonly experience suboptimal diagnosis. A French cross-sectional study found that endometriosis patients typically face a ten-year diagnostic delay, with various clinical and sociodemographic factors linked to this prolonged timeframe. The research highlights patient groups at higher risk for delayed diagnosis.

Most clinical reasoning assessments focus on outcomes instead of the underlying cognitive reasoning processes, missing key factors that affect diagnostic accuracy. This pilot study offers initial psychometric support for using the Objective Structured Clinical Examination on Clinical Reasoning (OSCE-CR) to measure and give feedback on clinical reasoning steps linked to accurate diagnosis. The OSCE-CR helps students practice these essential reasoning behaviors, aiming to enhance skills and reduce diagnostic errors.

Understanding caregivers' views on diagnostic uncertainty helps medical providers manage and communicate more effectively. This qualitative study examines caregivers' perceptions during family-centered rounds in pediatric hospital medicine. The findings show that uncertainty is complex, but it can be reduced through symptom management and clear plans for follow-up.

Patient feedback on diagnostic experiences is essential to improving care. This survey found that one third of households reported diagnostic problems or mistakes, which led to significant emotional, physical, and financial effects. Using patient-centered language to communicate and understand patient-reported experiences is vital for enhancing diagnostic quality in learning health systems.

Motion artifacts lower MRI quality, but their contributing factors and effect on diagnostic accuracy are unclear. This Danish single-center retrospective study of suspected stroke patients found that motion artifacts in brain MRIs significantly decrease both human and AI detection of intracranial hemorrhages.

Overdiagnosis of disease results in harm to patients, systems and communities. The extent of thyroid cancer overdiagnosis in the US is significant (estimated at 72 to 94 %), even after considering increased true incidence. Reducing unnecessary thyroid ultrasounds—especially for nonpalpable nodules—could lower unnecessary diagnoses and treatment-related harms without raising mortality.

Emerging knowledge illustrates how patients and families use AI to support diagnostic excellence. This case study shares the caregiver and clinician perspective on using generative AI (e.g., ChatGPT, OpenAI) to understand test results, explore diagnoses, and communicate with care teams. The authors highlight lessons for patients and caregivers, demonstrating how AI can help navigate complex medical information.

Early detection of cardiovascular disease is a global health priority. This cluster-randomized trial assessed AI-enabled stethoscopes in UK primary care for heart condition detection. AI stethoscopes did not boost overall heart failure diagnoses over 12 months. The study design provides valuable real-world data on implementing AI healthcare innovations.