CODEX Digest - 6.4.26

1) Patient perspectives on trust in artificial intelligence-powered tools in prostate cancer diagnostics.

Berger SA, Håland E, Solbjør M. Qual Health Res. 2026;36(2-3):276-288.

AI holds potential to optimize prostate cancer diagnostics, yet knowledge on how patients feel about these technologies is limited. This Norwegian qualitative study interviewed 18 men with low-grade prostate cancer, who ranged in educational background, employment, and age, about trust in AI-powered prostate cancer diagnostics. Findings reveal three themes: foundational trust in the healthcare system built from prior experience, interpersonal trust with professionals as the gateway to AI acceptance, and recognition of AI's diagnostic potential, tempered by concerns over accountability and AI's limited intuition.

2) Blood culture contamination and subsequent unnecessary health care utilization in pediatrics.

Childress K, Hartmann M, Kwan ML, et al. Hosp Pediatr. 2026;16(4):331-338.

Contaminated blood cultures drive billions in annual US healthcare costs, but the pediatric burden is understudied. The results of this retrospective cohort study across outpatient, inpatient, and hospital settings find that contamination was linked to more emergency department visits, admissions, and longer stays, increasing costs and patient impact particularly in the outpatient setting.

3) Viewpoint on the consequences and mitigation of cognitive bias in the radiological interpretation of breast cancer imaging using artificial intelligence.

Conti L, Capetti B, Battaglia O, et al. JMIR Med Inform. 2026;14:e78955.

AI is increasingly used in breast imaging, but successful adoption depends on radiologists’ trust and willingness to integrate it into practice. This commentary examines cognitive and system sources of diagnostic error and AI’s potential to support—and sometimes reshape—interpretation. The authors argue for context-sensitive implementation paired with ongoing education and bias awareness to improve accuracy while preserving radiologists’ critical judgment.

4) Sex differences in inappropriate imaging requests: insights from the Medical Imaging Decision And Support (MIDAS) study.

Dijk SW, Wollny C, Kroencke T, et al. Eur Radiol. 2026;36(4):3292-3299.

Inappropriate imaging can cause unnecessary radiation, delayed diagnoses, and higher costs, yet sex-based disparities are understudied. In this German, multicenter, cluster-randomized trial, women were more likely than men to receive inappropriate imaging requests, suggesting disparities in diagnostic decision-making.

5) Efficiency pitfalls of explainable AI in clinical diagnostic and treatment human-AI workflows.

Hunsicker T, Schulz A, Leist RA, et al. Hum Factors. Epub 2026 Apr 21.

The transparency of AI decision content is heralded as important but its integration into clinical workflow may harbor unintended consequences. This mixed methods study looked at AI-supported diabetic retinopathy diagnosis, both with and without visual explanations (highlighting lesions). The visual explanations increased diagnostic time and decreased efficiency, showing how benefits of AI may actually decrease productivity when needed "human in loop" and transparency are incorporated.

6) Cultivating competence in error management: The development and impact of a tailored quality improvement and patient safety curriculum in pathology training.

Khorsandi N, Tanaka K, Nedelcu E, et al. Acad Pathol. 2026;13(2):100245.

A gap in pathology training regarding quality improvement and patient safety affects diagnostic safety. This study evaluates a structured curriculum aimed at enhancing pathology trainee competence and confidence. Its success in shifting attitudes and skills supports wider adoption in residency programs. Integrating such education nationally could promote a culture of safety and open error management in pathology.

7) Gender bias in the clinical reasoning steps of medical students: a critical examination.

Le Boudec J, Potarca G, Félix S, et al. BMC Med Educ. Epub April 17, 2026.

Gender prejudice in medical education may affect student patient assessment and decision making. This study examines gender bias in male and female medical students' clinical reasoning for male and female standardized patients. Female patients were asked less about alcohol, work occupation, received fewer heart exams, and had increased prescription of lab tests. While the study didn't include non-binary or gender non conforming individuals, it is a provocative signal to how sexism pervades diagnostic care.

8) Digital health safety surveillance: improving patient support throughout the diagnostic trajectory.

Moran B, Weon JL, Hinshelwood M, et al. Jt Comm J Qual Patient Saf. Epub 2026 Apr 21.

Missed and delayed diagnoses are fundamental detractors from patient safety that too commonly affects minority populations. This 2025 Eisenberg Award-winning program describes Parkland Health's Safety Net Surveillance system, addressing diagnostic error inequities among underserved populations. By centralizing digital oversight, the program improved identification and management of missed or delayed diagnoses, offering a replicable blueprint for healthcare organizations to systematically reduce diagnostic errors and care disparities.

9) The association between sepsis and diagnostic errors: a secondary analysis of the Utility of Predictive Systems for Diagnostic Error study.

Prasad PA, Hubbard C, Lee T, et al. J Hosp Med. Epub 2026 May 14.

Diagnostic error occurrence in hospitalized sepsis patients has yet to be determined. This retrospective cohort study at 29 academic medical centers found that sepsis patients who died in the hospital or were admitted to the ICU had diagnostic error rates similar to those without sepsis. Using Safer Dx and DEER taxonomy, the authors found sepsis was less strongly associated with diagnostic error when history-taking or consultation gaps were present, suggesting these errors reflect broader care problems rather than sepsis-specific factors.

**UCSF CODEX’s director Sumant Ranji, MD is an author for this publication. 

10) Portrayals of depression on TikTok: content analysis of diagnostic accuracy, creator type, and stylistic features.

Rainer E, van der Wal A, Beyens I. JMIR Infodemiology. 2026;6:e85323.

Depression content on TikTok is often diagnostically incomplete across creator types, and engagement of young patients is apt to be driven more by style than accuracy. This quantitative content analysis findings illustrate concerns about the gradual broadening of depression examples and content on the social media platform that may support premature self-diagnosis. Diagnostic accuracy received an average score of 1 on a 0-4 scale.

11) The effect of feedback on the diagnostic process of physicians at the emergency department: a systematic review.

Schols LA, Zwaan L, Woltman AM, et al. Eur J Emerg Med. Epub 2026 Apr 17.

Systematic feedback may improve diagnostic performance in emergency departments (ED), though evidence is limited. This review looks at feedback characteristics provided to ED physicians and their impact on performance and confidence. Despite few low-quality studies, the findings suggest that structured feedback could positively influence ED physicians’ diagnostic abilities, highlighting its potential as a valuable enhancement tool.

12) Patterns of diagnostic errors using hierarchical cluster analysis: a single-center, cross-sectional study in general internal medicine.

Suzuki Y, Nishizawa T, Ozawa H, et al. Diagnosis (Berl). Epub 2026 Apr 27.

To effectively reduce the potential for error, it is crucial to examine diagnostic process failures contextually rather than by single contributing factors. This Japanese cross-sectional study examines diagnostic errors linked to unexpected readmissions in general internal medicine, emphasizing the need to analyze recurring failure patterns rather than isolated factors. Two key weaknesses emerged: diagnostic prioritization and urgency appraisal and information acquisition and synthesis. Despite a small sample, identified patterns suggest targetable improvement areas applicable to similar outpatient settings.