CODEX Digest - 2.12.26

1) Rethinking certainty: a retrospective study on diagnostic revisions in gynecologic pathology.

Cianfrini F, d’Amati A, Zamparese R, et al. Int J Gynecol Cancer. 2025;36(2):102826.

Pathology review is an important part of gynecologic oncology diagnostic excellence and care. This Italian study of second opinion referrals from one tertiary review center finds that an expert second opinion led to diagnostic changes in over half of gynecologic lesion cases. Routine specialist pathology reviews for complex or high-risk cases are recommended to improve accuracy, guide treatment, and lower medico-legal risks.

2) A scoping review of communication tools for risk-based breast cancer screening.

Hudson C, Isauiter J, Houssami N, et al. Patient Educ Couns. 2026;143:109425.

Clear screening communication can prompt timely breast cancer diagnosis. This review evaluates tools used to inform women about risk-based screening, outlines the information they provide, and notes a shortage of validated resources for effective risk-based screening communication.

3) From volume to value: leveraging artificial intelligence and deliberate practice to foster precision learning in radiology. (subscription required) 

Kelly BS, Duignan S, Booth CC, et al. Pediatr Radiol. Epub 2025 Nov 19.

AI's use in healthcare has raised the concern that clinicians will lose important clinical reasoning skills. This review explores how AI and the framework of "deliberate practice" interact in radiology training, offering recommendations for how AI can be used to integrate real-time diagnostic feedback into residency programs and enhance decision making, improve diagnostic safety, and limit deskilling and automation bias.

4) The impact of skin tone on performance of pulse oximeters used by NHS England COVID Oximetry @home scheme: measurement and diagnostic accuracy study.

Martin DS, Doidge JC, Gould D, et al. BMJ. 2026;392:e085535

Problems with pulse oximeters misdiagnosing patients with darker skin tones may have contributed to racial health disparities such as those during the COVID pandemic. This observational UK study finds that five different home pulse oximeters tested in an NHS initiative recorded overestimation of oxygen saturation and higher false negative rates in people with darker skin tones. This study has far reaching implications for the diagnostic accuracy of at-home devices and demonstrates the need to ensure devices are tested for participants of all skin tones.

5) How do we utilize information technology to reduce telephone calls in the clinical laboratory?

McCash SI, Korenstein D, Raoof N, et al. Am J Clin Pathol. 2025;164(6):854-860.

Workforce shortages, interruptions, and frequent information requests interfere with diagnostic test processing. This quality improvement project tested a web-based application to reduce phone communication, aiming to reduce the potential for errors and complaints during critical value follow-up. While critical lab turnaround time did not change, incoming call volume and experience ratings did improve.

6) Enhancing diagnostic precision for rare diseases using case-based reasoning.

Noll R, Berger A, Facchinello C, et al. J Am Med Inform Assoc. 2026;33(1):98-111.

Case-based reasoning (CBR) is a method that solves problems by using past cases. This German study compares three weighting methods using patient-similarity-based models to enhance rare disease diagnosis by enabling comparison of reference cases across three medical specialties. All three methods showed similar precision. Interestingly, less granular ICD codes led to higher precision. The paper highlights CBR's potential to improve diagnosis for rare disease diagnostics.

7) Unintended consequences of using ambient artificial intelligence scribes for billing.

Nong P, Neprash HT. JAMA Health Forum. 2026;7(1):e255771.

Ambient scribes can help lower clinician burnout and improve patient communication during visits, but their use could have broader unintended consequences. This commentary raises scenarios in which using ambient scribes for billing leads to higher healthcare spending through "up-billing," which can erode patient trust, qualify value-based care sites for additional funds, and (positively) identify patients due for reimbursable screening.

8) Early detection of high risk pregnancies using clinical and social data to improve health outcomes.

Patel SY, Akileswaran C, Basu S. NPJ Digit Public Health. 2026;1(1):3.

Maternal and infant mortality rates are significantly high in the United States. This machine learning study shows that using social determinants of health data can help predict adverse pregnancy outcomes. Modeling showed which components of social determinants were more influential in predicting and influencing outcomes.

​​​​​​

9) A cost consequence analysis of seven diagnostic strategies for ovarian cancer: a model-based economic evaluation.

Perry SJ, Griffin D, Williams EV, et al. BJOG. Epub 2025 Nov 5.

Ovarian cancer is frequently diagnosed late, leading to high mortality. This NHS study assessed seven screening strategies for pre- and post-menopausal women with symptoms in secondary care and found challenging tradeoffs between cost, diagnostic yield, and cancer mortality. The results indicate that screening decisions will require balancing cost and sensitivity to support effective diagnosis.

10) Quality indicators for screening and surveillance of colorectal cancer in adults: a review of performance measures by the American College of Physicians.

Qaseem A, Miller NL, Khetan RS, et al. Ann Intern Med. 2025;178(12):1779-1784.

Effective measurements can help physicians, payers, and policymakers select appropriate metrics to track diagnostic excellence. This article shares the analysis done by the American College of Physician's (ACP) performance review committee on five quality measures for colorectal cancer screening. Of the five measures, the committee endorsed "hospital admission after colonoscopy" on the basis of ensuring screening measurements are evidence based, actionable, and able to generate meaningful improvement metrics.

11) An autonomous agentic workflow for clinical detection of cognitive concerns using large language models.

Tian J, Fard P, Cagan C, et al. NPJ Digit Med. 2026;9(1):51.

Traditional screening tools and limited resources restrict early diagnosis of cognitive impairment. This study designed and evaluated an autonomous agentic workflow using large language models to identify signs of cognitive decline from clinical notes. However, sensitivity dropped when moving to an independent dataset, and the samples were predominantly White, limiting the generalizability of the study's findings across diverse groups and with different languages.

12) My nights eat my days: how my eating disorder overshadowed the autism no one saw. (subscription required)

Voltz W, Brown DA. Ann Intern Med. 2026;179(1):144-145. 

This patient story illustrates the diagnostic challenges patients can face when mental health disorders overlap with other conditions. Diagnostic overshadowing, premature closure, and patient-insensitive care are discussed and shown to acerbate a missed diagnosis over an extended period.