CODEX Digest - 1.22.26

1) Incentivizing co-occurring disorder diagnoses through blended payments.

Baslock D, Manuel JI, Stanhope V. Soc Sci Med. 2026;389:118849.

Diagnosing co-occurring mental health and substance use disorders is difficult. This study examines how blended insurance payment strategies affect diagnosis rates in a community mental health system, finding that such models increased the likelihood of diagnosing co-occurring disorders.

2) Are Smartwatches in healthcare a panacea or a Pandora's box?

Boyle T. The Diagnostic Detective. January 9, 2026.

Consumer-facing health monitoring tools can inform patient involvement in timely diagnosis. This Substack article summarizes a New England Journal case report of a smartwatch finding heart block, and opines on the pitfalls of considering wearables as a diagnostic tool versus a screening tool.

3) Socio-demographic inequalities in diagnostic delays of breast cancer: a multistage time-to-diagnosis analysis.

Burciu OM, Gramada T, Gramada-Stefurac S, et al. Curr Oncol. 2025;32(12):674.

Limited access to healthcare delays breast cancer diagnosis. This Romanian study finds that living in rural areas and social disparities slow cancer detection in a regional screening program. Improving screening processes can reduce barriers, promoting early diagnosis and equitable care.

4) Collaborative artificial intelligence for the diagnosis and management of acute ischemic stroke.

Fan Z, Chen Q, Lu W, et al. Ann Med. 2026;58(1):2594356.

Diagnostic or predictive AI advocates consider health IT privacy and data fragmentation a major barrier to improving accuracy. This review advocates for collaborative AI data sharing across institutions, to include organizational learning and secure data exchange, to improve acute ischemic stroke diagnosis.

5) Pediatric emergency department provider experience using the Revised Safer Dx for self-directed feedback on diagnostic performance. (subscription required)   

Grubenhoff JA, Dillon M, Geanacopoulos AT, et al. J Contin Educ Health Prof. Epub 2025 Dec 17. 

Diagnostic feedback is an underused mechanism for decision-making improvement. This study explores the use of the Revised Safer Dx tool to enable clinician self-assessment of possible diagnostic misses. The participants found the tool resulted in clinicians altering their diagnostic approach among those who found an error.

6) Scoping review of patient and family engagement interventions in diagnosis: a paradox of too much, yet so little.

Hill MA, Haskell H, Dainty KN, et al. BMJ Qual Saf. Epub 2025 Dec 23.

Patient and family engagement in diagnosis is core to excellence, yet the needle has still not moved to rigorously prove its benefit. This review uses the NAM sociotechnical process to assess current engagement strategies, finding limited evidence for the interventions, particularly those regarding equity, patient co-design, and effectiveness.

7) Suboptimal diagnostic decisions in hospitalized patients with fever: a prospective record review with physician interviews.

Hooftman J, Sikkens JJ, van Wingerden N, et al. Jt Comm J Qual Patient Saf. Epub 2025 Nov 14.

Suboptimal diagnostic decisions are when a case deviates from diagnostic excellence during the diagnostic journey. This study reviewed 53 case records of fever and found over half had suboptimal diagnostic decisions. Suboptimal decisions were associated either with diagnostic error or discrepancy, and were human decisions during assessment and testing.

8) Impact of AI recommendation correctness on diagnostic accuracy in clinical decision-making.

Kücking F, Busch DA, Przysucha M, et al. Int J Med Inform. 2025;207:106223.

Overreliance on technology can bias healthcare decisions. This German study investigated healthcare provider characteristics, such as age and work experience, and correct vs incorrect AI recommendations on providers' diagnostic accuracy in a simulated task. Results show participants had lower diagnostic accuracy when receiving incorrect AI recommendations, showing evidence of automation bias, and age did not have a significant effect.

​​​​​​

9) How large language models can affect clinical reasoning: a randomized clinical trial. (This is a preprint that has not yet gone through peer review.)

Levels M, et al. medRxiv. Epub 2025 Dec 18.

Evidence on the ability of AI to enhance decision making is emerging, but AI research has not been globally representative. This pre-print analysis evaluates the impact of LLMs on physicians’ clinical cognition in primary care scenarios in Indonesia, Kenya, and the Netherlands. Access to LLMs enhanced information gathering and management reasoning across all countries.

10) Reported patient safety incidents in radiology - understanding the relation between risk areas and their underlying factors.

Wallin A, Lundén M, Ringdal M, et al. Radiography (Lond). 2025;31(6):103149.

Patient safety reporting systems help identify problems and guide improvements. This study analyzes 923 radiology safety reports revealing risks like misdiagnosis, radiation exposure, and delayed examinations. The results highlight the value of a reporting framework to identify latent risks in radiologic practice and enhance interdisciplinary collaboration.

11) Economic evaluations of early detection strategies for pancreatic cancer: a systematic review.

Wittram R, Kreis L, König H-H, Brettschneider C. Eur J Health Econ. 2025;26(9):1655-1670.

Economic screening strategies are a valuable public health approach supporting timely cancer diagnosis. This international review finds that screening for pancreatic cancer may be cost-effective for some high-risk patients, but more consistent evidence is needed to fully understand the costs of early detection methods.

12) AI as a Healthcare Ally - How Americans Are Navigating the System with ChatGPT. 

OpenAI. OpenAI; 2026.

AI is increasingly used to answer healthcare questions. This OpenAI industry white paper offers a little glimpse under the hood on some of its data on how Americans use ChatGPT. Three in five adults in the US say they’ve used AI tools for their health or healthcare, and many use ChatGPT to navigate insurance or seek help outside of regular office hours.