CODEX Digest - 3.5.26

1) Exploratory Cancer-Related Diagnostic Excellence Measures.

Agency for Healthcare Research and Quality. February 11, 2026.

Developing diagnostic excellence measures is vital for tracking progress and identifying process weaknesses. This guidance draft outlines cancer-related measures from the AHRQ Quality Indicators Program, focusing on timely follow-up after screening, late-stage diagnoses, and new diagnoses following acute presentations to guide system-level safety improvements.

2) Core Elements of Hospital Diagnostic Excellence (DxEx).

Centers for Disease Control and Prevention. February 4, 2026.

Diagnostic excellence is a significant patient safety objective. This updated framework supports efforts to reduce missed, delayed, and incorrect diagnoses. It outlines a structured six-element strategy defining leadership and clinician roles to improve diagnostic reasoning, minimize unnecessary tests, and foster a culture of learning from errors. It also adds specific tools and assessments for hospitals to track progress toward excellence.

3) Artificial intelligence in depression diagnostics: a systematic review of methodologies and clinical applications.

Ghorbankhani M, Safara M. Artif Intell Med. 2026;172:103320.

AI is gaining attention in mental health diagnosis. This review examines current AI methods, data types, and performance metrics. The authors outline both the potential as well as the methodological, ethical, and practical challenges of applying AI to mental health and discuss options to advance AI as a safe tool for depression detection.

4) Colorectal cancer screening among people with intellectual disabilities.

Horsbøl TA, Michelsen SI, Sørensen TT, et al. JAMA Netw Open. 2026;9(1):e2557013.

Colorectal cancer mortality is higher in people with intellectual disabilities, possibly due to delayed diagnosis. This Danish cohort study finds that they are less likely to participate in screening and face more challenges with sample collection and colonoscopy than people without intellectual challenges. The results highlight the need for tailored strategies to enhance equitable access to colorectal cancer screening.

5) Overdiagnosis in screening for gestational diabetes: a scoping review.

Karentius S, Brodersen JB, Bjørch MF, et al. Midwifery. 2026;153:104646.

Overdiagnosis is a growing problem in healthcare for patients and the system at large. This review explores the potential for overdiagnosis in gestational diabetes mellitus (GDM) screening for healthy pregnancies. The analysis shows that intensified GDM screening increases incidence without clear benefits, suggesting universal screening could be a contributor to unnecessary diagnoses. The review also identifies the lack of an established mechanism to define or measure overdiagnosis in GDM screening.

6) Exposing the fragility of LLM reasoning through bias-inducing prompts: evidence from BiasMedQA.

Kim SH, Ziegelmayer S, Busch F, et al. BMJ Digit Health. 2026;2(1):e000189.

LLMs are seen as promising tools to support clinical decision-making but can be vulnerable to implicit bias. Recent LLM capabilities mimicking human reasoning could potentially address this problem. This model study demonstrates that LLM system design prompts do not consistently reduce cognitive bias in human clinical decision-making. The findings highlight the potential risk of clinicians and researchers over-relying on LLMs that appear to mimic human reasoning.

7) Exploring diagnostic challenges and performance feedback in older adult emergency general surgery.

Liu JK, Peters XD, Remer SL, et al. World J Surg. Epub 2026 Feb 1.

Older adults frequently present as emergency surgical cases often because of missed diagnoses. This interview study summarizes clinicians' views on diagnostic challenges in geriatric emergency surgery, improvement strategies for feedback, and barriers and enablers to using tools that address limitations, including multidisciplinary collaboration, increased family involvement, and heightened attention to patients' social needs.

8) Incidence, contributing factors, and predictors of diagnostic errors in medical inpatients: a retrospective cohort study.

Marx CE, Hofmann E, Perrig M, et al. J Hosp Med. Epub 2026 Feb 8

Diagnostic error is a major safety risk in hospitals. This Swiss study of general medical inpatients—a group rarely examined because more science is available on high-risk subgroups—found errors were common and often harmful. They frequently resulted from not considering the correct diagnosis, not ordering appropriate tests, or missing key physical exam findings.

​​​​​​

9) Unveiling dental diagnostic dilemmas: a national survey of US dentists.

Obadan-Udoh E, Howard R. BMC Oral Health. 2025;26(1):172.

Diagnostic errors in dentistry are not well understood. This study finds 40% of dentists regularly observe errors made by others, while only 12.4% admitted making mistakes themselves. Younger dentists and those seeing over 61 patients weekly admitted reporting more personal errors. The authors recommend developing strategies to reduce these incidents.

10) Large language model diagnostic assistance for physicians in a lower-middle-income country: a randomized controlled trial.

Qazi IA, Ali A, Khawaja AU, et al. Nat Health. 2026;1(2):198-205.

LLMs may support clinicians with diagnostic reasoning but can generate inaccurate information, prompting the need for AI-literacy training. This real-world Pakistani study shows that AI-trained physicians using LLMs enhanced their diagnostic skills without slowing reviews, indicating LLMs may help close diagnostic gaps in resource-limited areas.

11) Thinking critically about AI documentation quality in primary care.

Schiff GD, Khazen M. BMJ Qual Saf. Epub 2026 Feb 9.

AI-driven clinical documentation could improve patient–clinician interactions and workflow, with important implications for diagnostic safety. However, because clinical note generation is central to diagnosis, reflection and communication, this commentary urges caution. Drawing on past overconfidence in health IT innovations, the authors highlight risks such as diminished patient trust, missing or distorted content, and loss of clinician reflection opportunities if AI-generated documentation is implemented at scale.

12) Does clinical exposure to different skin tones during training improve diagnostic ability?

Shammoon Y, Coulson A, Trigg B, et al. Med Educ. Epub 2025 Dec 12.

Evidence shows that medical students are less accurate at diagnosing conditions in skin of color compared to in white skin, regardless of their clinical exposure. This British study suggests that exposure alone does not address this gap and concludes that targeted educational strategies are needed to improve diagnostic equity and accuracy across different skin tones.