CODEX Digest 07.09.26

The clinical problem list is central to inpatient and outpatient diagnostic documentation. Despite widespread EHR use, audits still find problem lists incomplete, inaccurate, and poorly written. These shortcomings can undermine patient safety, continuity of care, and medicolegal accountability. This commentary shares the history of the clinical "SOAP note" and problem list and offers practical guidance for trainees and supervisors on turning raw clinical information into prioritized, actionable problem statements that make diagnostic reasoning visible.

Unclear definitions can hinder progress in diagnostic safety. Using an interdisciplinary Delphi panel with diverse clinical and demographic backgrounds, this study defines diagnostic excellence and missed opportunities for emergency care, recognizing that provisional diagnoses are appropriate when uncertainty is communicated, the working diagnosis broadly fits the final condition, and patients receive next-step guidance.

Timely and accurate diagnosis of Alzheimer's disease can be limited by rural location, clinician shortages, and socioeconomic disadvantages. This commentary conceptualizes these regions as “dementia diagnostic deserts,” where barriers can lead to delayed recognition of true cognitive decline or over testing of those without impairment. The authors propose policy solutions that address workforce, financial, and healthcare system challenges together to improve equitable access to diagnosis and follow-up care.

Diagnostic uncertainty is common and challenging for both patients and clinicians. This rapid scoping review found that communication, clinical reasoning, and collaboration help health professionals manage it in practice. Educational and workplace settings can support this by normalizing uncertainty, encouraging collaboration, and adapting strategies to specific care contexts.

Radiology report errors are uncommon, making automated detection prone to false alarms. This retrospective study evaluates a multi-step LLM approach designed to review radiology reports more efficiently. The findings illustrate that the method can reduce false alarms while maintaining the ability to identify potential errors, suggesting that AI tools may have a role in supporting quality assurance processes, rather than replacing radiologist review.

Pathologist errors occur when diagnostic findings are missed during review of physical test specimens. This guideline update reviews the role of second opinions in detecting and preventing errors and calls for defined review procedures to improve safety. The authors examine emerging evidence on digital pathology and AI tools for error detection and case triage.

Online access to test results may improve diagnostic communication, but cancer patients’ experiences remain underexplored. In this bronchoscopy patient quality improvement survey, two-thirds used their portal, yet only 9% of users were satisfied with explanations and 47% felt more worried. About half preferred a phone call plus educational video, suggesting biopsy results should be released in ways that match patient preferences and include clinician support. Patients surveyed were predominantly White.

Prostate cancer screening disparities by race and sociodemographic factors persist across the US. This single-center retrospective cohort study found that a safety-net program tailored specifically for underserved populations reduced these disparities. The findings suggest that targeted support can improve screening equity and may offer a successful approach for uptake in other settings.

Premature closure—reaching a conclusion before enough information is available—is a known source of diagnostic error and may also affect LLM output. This preprint study tested five advanced LLMs on structured and open-ended medical questions, including challenging physician-designed cases. The models often answered despite unresolved uncertainty. Although safety-oriented prompts reduced these errors, important failures remained, highlighting the need to assess not only what medical AI systems know, but also whether they can recognize when to withhold judgment.

Surgical site infections are sentinel events that can be identified through labor-intensive chart review. In this retrospective study of 1,157 colorectal surgery cases, an LLM reviewed narrative electronic health record notes and showed reasonable accuracy in identifying overall infections - except incisional infections -- while ruling out many cases without infection. The findings suggest a possible role for AI-assisted surveillance to support, rather than replace, human chart review.

Primary care doctors play a key role in cancer prevention and early detection by promoting awareness, identifying high-risk patients, and recommending screening. This review finds that physicians across countries and healthcare systems generally follow colorectal, breast, and cervical cancer screening guidelines, but adherence is shaped by multiple facilitating and limiting factors. The review also raises the need for more research in low and middle income countries.

Wrong-side imaging orders are preventable but can go undetected due to reporting gaps. This six-site retrospective cohort study validated a decently accurate EHR-based method to flag these errors, which were more common in outpatient settings and patients with unknown race/ethnicity. The scalable approach supports surveillance and targeted interventions to improve diagnostic safety and reduce patient harm.