CODEX Digest 11.05.26

Multiple sclerosis (MS) can be difficult to diagnose due to heterogeneous symptoms and lack of any single definitive MRI finding or blood test. In this news interview, a neurologist discussed actor Christina Applegate’s reported MS misdiagnosis and its emotional consequences. The segment underscores the impact of diagnostic uncertainty, need for reassessment when findings are atypical, and value of second opinions to support diagnostic excellence.

Measuring diagnostic safety requires identifying conditions for which missed or delayed emergency diagnoses may cause most harm. This retrospective cross-sectional study analyzes Medicare claims from 2022–2024 and used clinician review to identify 76 emergency diagnosis-sensitive conditions, representing nearly one in five nonelective hospitalizations. The classification offers a potential foundation for prioritizing diagnostic safety, evaluating variation, and targeting improvement efforts.

Understanding how clinicians actually use AI is essential before judging its effects on diagnostic excellence. This cross-sectional study analyzes 271 million queries from 952,008 verified clinicians using OpenEvidence over two and a half years. Treatment and management accounted for 33.1% of clinician queries, while diagnosis accounted for 13.9%. The findings document substantial adoption but did not establish accuracy, improved diagnostic decisions, or better patient outcomes.

Patient-facing conversational AI may change how diagnostic information is gathered during primary care encounters, but its safety and usefulness require real-world evaluation. This small prospective single center study examined the use of a conversational diagnostic AI system called AMIE during urgent ambulatory primary care appointments by experience rating and safety. This feasibilty study implies broader adoption will require further evaluation of the system impact on diagnostic accuracy, trust, communication, safety, and clinician workflow.

Identifying a condition does not guarantee that patients receive appropriate follow-up or treatment. This large population-based cohort study examined cardiovascular diagnoses and subsequent treatment among Medicare Advantage beneficiaries identified through in-home health risk assessments. The findings indicated that diagnoses made during these assessments were followed by lower treatment rates than diagnoses made during facility-based encounters. The results raise questions about continuity of care and whether additional diagnoses translate into meaningful clinical benefit.

Diagnostic delay can reflect broader inequities in healthcare access, recognition, and treatment. This conceptual human rights analysis examines endometriosis and the preventable suffering associated with delayed diagnosis. The authors connect diagnostic experiences with structural injustice and institutional responsibilities. They contend that institutional inaction, dismissal of patients’ experiences, and uneven accountability can contribute to prolonged diagnosis, offering a structural perspective on diagnostic equity and preventable suffering.

Psychiatric diagnosis requires contextual interpretation, differential diagnosis, and attention to uncertainty. This preprint evaluated multiple large language models for diagnostic accuracy and clinical reasoning in psychiatry. Its findings are intended to characterize model performance across psychiatric tasks, but the benchmark results alone did not establish clinical readiness or safety. Independent replication, clinician comparisons, bias assessment, and evaluation with representative real-world cases remain necessary.

Patients’ observations can expose diagnostic problems that fragmented clinical encounters overlook. This op-ed by a UCSF-affiliated patient advisor described experiences of delayed diagnosis, symptoms being attributed to existing conditions, and feeling dismissed within healthcare encounters. The account emphasizes the value of listening to patients, reviewing their full histories, and incorporating patient perspectives into institutional decision-making.

Diagnostic errors often arise not from unusual presentations but from premature acceptance of a plausible diagnosis. This retrospective study of 306 Japanese diagnostic-error cases assessed "typical" versus unusual or "chameleon" clinical presentations. Although 84.3% displayed features of the incorrect diagnosis, 64.0% also contained unusual or contradictory clues. The findings support diagnostic-safety strategies that emphasize disconfirming evidence and systematic reconsideration of initial diagnoses.

Diagnostic reasoning unfolds over time as clinicians gather information and revise hypotheses. This model evaluation examines frontier language models on consecutive diagnostic tasks, introducing a benchmark designed to examine performance across evolving clinical information rather than isolated questions. The work offers a more process-oriented approach to AI evaluation and suggests benchmark performance alone cannot improve patient outcomes or safety in clinical practice. Prospective comparisons with clinicians and real-world workflow testing are noted as important next steps

Diagnostic delays and misdiagnoses can reveal inequities that routine clinical records fail to capture. This international cross-sectional survey of 6,607 people with autoimmune diseases found that women report longer diagnostic journeys and more psychiatric and psychosomatic misattributions than men. The findings highlight potential sex-related inequities in diagnosis, although the predominantly White, UK-based sample limited generalizability.

Diagnostic reasoning can be influenced by patient characteristics even when clinical presentations are equivalent. In this French study of 357 fourth-year medical students, correct diagnostic hypotheses were more common with male standardized patients than female standardized patients. The results suggest that gender-related diagnostic bias could emerge during medical training. Further studies are needed to establish whether this pattern persists in routine clinical care and other diagnostic settings.