Omran Bakoush

Background: Most clinical reasoning evaluations in large language models (LLMs) score only the final answer, usually multiple-choice accuracy, which explains little about model reasoning. Two developments stress this gap: reasoning-optimized models are now common, and several expose an explicit extended thinking control. Whether that deliberation improves the reasoning process remains untested with a validated instrument. Objective: This protocol introduces and aims to validate the Rapid Evaluation Assessment of Clinical Reasoning Tool (REACT)–AI, a Behaviorally Anchored Rating Scale (BARS) with 13 subdomains for process-oriented assessment of AI clinical reasoning, that is, the quality of the externalized reasoning rather than its faithfulness to the model’s internal computation. It also builds a conflict-of-interest–controlled LLM-as-judge pipeline and tests whether extended thinking improves reasoning quality. Methods: This 2-phase, prospective, comparative study has a within-model thinking-mode factor. Six flagship models are run in standard and extended thinking or reasoning modes, giving 12 conditions: 4 providers are toggled within the same model, and 2 pair a standard model with a dedicated reasoning model. In phase 1, 5 standardized urgent care vignettes are answered under all 12 conditions, 3 runs per condition (180 AI outputs), and by an independent expert clinician panel (n=5) and senior and junior medical students. Every output is scored by dual-blind human raters and, in parallel, by LLM judges, with no model judging its own family. Acceptance criteria for deploying the judge at scale are a weighted κ of at least 0.60 and an intraclass correlation coefficient above 0.75. In phase 2, the validated pipeline scores 3600 AI outputs. A self-correction turn and a paired Gulf English condition run alongside and are reported separately, with the AI metacognition assessment rubric and disinformation generation rate as secondary instruments. Results: As of June 2026, ethics approval was obtained (ERSC_2025_6124), and the study is registered on the Open Science Framework under embargo. Responses to the 5 phase 1 vignettes were collected from senior and junior medical students between January and June 2026; 173 scripts were received and remain sealed, unopened, and unscored. No AI outputs have been generated, and the expert panel remains unrecruited. Following the June 1, 2026, model-version lock, AI generation and scoring will begin in September 2026, with phase 1 calibration through December 2026 and phase 2 from January to April 2027; results are expected in winter 2027. The study will report the validated instrument, judge calibration against human experts, self-preference bias by model family, and the effect of extended thinking on reflection and metacognition, including null findings. Conclusions: REACT-AI and its judge pipeline are built to be reusable. They close the gap between accuracy benchmarks and genuine reasoning assessment and provide a template for studying reasoning mode effects. Trial Registration: OSF Registries osf.io/jep5t; https://osf.io/jep5t International Registered Report Identifier (IRRID): DERR1-10.2196/103220

BACKGROUND Most clinical reasoning evaluations in large language models (LLMs) score only the final answer, usually multiple-choice accuracy, which explains little about model reasoning. Two developments stress this gap: reasoning-optimized models are now common, and several expose an explicit extended thinking control. Whether that deliberation improves the reasoning process remains untested with a validated instrument. OBJECTIVE This protocol introduces and aims to validate the Rapid Evaluation Assessment of Clinical Reasoning Tool (REACT)–AI, a Behaviorally Anchored Rating Scale (BARS) with 13 subdomains for process-oriented assessment of AI clinical reasoning, that is, the quality of the externalized reasoning rather than its faithfulness to the model’s internal computation. It also builds a conflict-of-interest–controlled LLM-as-judge pipeline and tests whether extended thinking improves reasoning quality. METHODS This 2-phase, prospective, comparative study has a within-model thinking-mode factor. Six flagship models are run in standard and extended thinking or reasoning modes, giving 12 conditions: 4 providers are toggled within the same model, and 2 pair a standard model with a dedicated reasoning model. In phase 1, 5 standardized urgent care vignettes are answered under all 12 conditions, 3 runs per condition (180 AI outputs), and by an independent expert clinician panel (n=5) and senior and junior medical students. Every output is scored by dual-blind human raters and, in parallel, by LLM judges, with no model judging its own family. Acceptance criteria for deploying the judge at scale are a weighted κ of at least 0.60 and an intraclass correlation coefficient above 0.75. In phase 2, the validated pipeline scores 3600 AI outputs. A self-correction turn and a paired Gulf English condition run alongside and are reported separately, with the AI metacognition assessment rubric and disinformation generation rate as secondary instruments. RESULTS As of June 2026, ethics approval was obtained (ERSC_2025_6124), and the study is registered on the Open Science Framework under embargo. Responses to the 5 phase 1 vignettes were collected from senior and junior medical students between January and June 2026; 173 scripts were received and remain sealed, unopened, and unscored. No AI outputs have been generated, and the expert panel remains unrecruited. Following the June 1, 2026, model-version lock, AI generation and scoring will begin in September 2026, with phase 1 calibration through December 2026 and phase 2 from January to April 2027; results are expected in winter 2027. The study will report the validated instrument, judge calibration against human experts, self-preference bias by model family, and the effect of extended thinking on reflection and metacognition, including null findings. CONCLUSIONS REACT-AI and its judge pipeline are built to be reusable. They close the gap between accuracy benchmarks and genuine reasoning assessment and provide a template for studying reasoning mode effects. CLINICALTRIAL OSF Registries osf.io/jep5t; https://osf.io/jep5t INTERNATIONAL REGISTERED REPORT DERR1-10.2196/103220

Introduction: The coronavirus disease 2019 (COVID-19) pandemic resulted in significant global mortality and morbidity, with emerging mutant strains continuing to potentially precipitate severe respiratory illness. Two clinical assessment tools, namely, the COVID-19 Risk of Complications Score (CRS), based on 13 comorbidities, and the ALKA (age, lactate dehydrogenase, kidney function, and albumin) score have been developed to predict disease severity among patients who are symptomatic at presentation. This study aimed to compare the performance of these two risk-scoring systems in predicting hospital admission, critical illness, and mortality.Methods: This retrospective study included 368 patients diagnosed with COVID-19 at SEHA hospitals in Al Ain over a six-month period. The CRS and ALKA scores were calculated to predict hospital admission, critical illness, and mortality. Predictive ability was assessed using receiver operating characteristic (ROC) curve analysis. Odds ratios (ORs) were calculated to assess the risk of hospital admission, critical illness, and mortality.Results: The mean age of the patients was 51 ± 19.42 years, and 145 (39.4%) of them were male. Among the patients, 162 required inpatient care, 13 required invasive ventilation, and the mortality rate was 4.9% (eight patients). ROC analysis revealed that ALKA outperformed CRS in predicting hospital admission (ALKA area under the curve [AUC] 0.79 vs. CRS AUC 0.71), critical illness (ALKA AUC 0.76 vs. CRS AUC 0.67), and mortality (ALKA AUC 0.96 vs. CRS AUC 0.82). The OR for ALKA outperformed CRS in predicting hospital admission (ALKA 3.12 vs. CRS 1.12), critical illness (ALKA 2.9 vs. CRS 2.01), and mortality (ALKA 6.25 vs. CRS 1.1).Conclusion: Our study demonstrated that ALKA score outperforms CRS in predicting hospital admission, critical illness, and mortality among patients with symptomatic COVID-19 at initial presentation. Further external validation of both tools is required to assess their effectiveness in different healthcare settings.

Introduction: Organ transplantation is a vital treatment for end-stage organ failure. However, the shortage of available organs remains a significant challenge worldwide. This study aimed to explore university students’ willingness to donate organs, including their views on brain death, in the United Arab Emirates (UAE).Methods: The study is cross-sectional survey on the respondents’ knowledge of, attitudes toward, and perceived barriers to and facilitators of organ donation. A total of 521 students completed the survey. Chi-square tests and logistic regression models were used to identify the factors associated with their willingness to donate their organs after death.Results: Most of the respondents (69%) were willing to donate their organs after death, and 79% were willing to donate their organs to a loved one during their lifetime. However, only 42.8% accepted brain death as equivalent to death. The most reported reasons behind the respondents’ willingness to donate organs were the belief that it is something everyone should do (adjusted odds ratio [aOR]: 4.68) and a responsibility to help loved ones (aOR: 2.63). Meanwhile, the significant barriers to organ donation included a preference for whole-body burial (aOR: 0.079), religious objections (aOR: 0.195), and family objections (aOR: 0.326).Discussion: University students in the UAE show a positive attitude toward organ donation. However, significant barriers, including family and religious objections, remain to be addressed. Increasing public awareness about brain death and establishing mechanisms for securing family consent in advance are crucial steps for the successful implementation of a deceased organ donation program in the UAE.

Medical school attrition is a complex problem, and improvement can only be achieved through the integrated involvement of all stakeholders. Mental health disorders are perceived to be prevalent among medical students in UAE. This study aimed to explore educators’ perspectives on the high rate of early medical school attrition in UAE. The findings suggest that the early medical school attrition rate is correlated with students’ inadequate academic readiness and their inability to cope with the psychological demands of studying medicine in English. To reduce the attrition rate, the educators proposed preparatory courses to enhance students’ self-learning skills and academic readiness for university studies, along with provision of targeted academic support throughout their medical studies. Psychological well-being and academic preparedness for studying medicine in English is a prime target for a program to reduce dropout from medical schools.

Background There is increasing concern about the quality, integrity, and accessibility to research published in the developing world. This study explores the editorial practices and editors’ perspectives to gain insight into the standard of scholarly publishing in Libya. Methods Between 21st January and 12th February, 2022, the editors-in-chief (EC) of Libyan academic journals were invited to complete a questionnaire on editorial practices, degree of satisfaction with submitted and published manuscripts, review processes, and journal performance, as well as challenges facing the journals. Journal websites were examined for quality, and indexation coverage and citations were assessed. We examined the number of citations in Google Scholar for all 2019 articles published in each journal. Descriptive statistics were used to quantitatively summarize the data and thematic analysis was used for the narrative text. Results 48 EC completed the questionnaire. The EC was affiliated with the institution that owns the journal in 92% of cases. Most EC (83%) were satisfied with the peer-review quality, 69% believed that most of their published papers add new ideas or findings, and 96% were satisfied with their journal’s performance. However, despite the high degree of satisfaction, only one journal was indexed in Web of Science or Scopus and only 17% of the journals were indexed in Google Scholar. A qualitative assessment of journal websites revealed shortcomings in publishing practices in a large proportion of the journals. Conclusions The discordance between the satisfaction of the journal editors and the journal quality indicators points to a break in the quality system of Libyan academic publishing. Similar expedient publishing practices might exist in other countries as well. A comprehensive action plan led by academic institutions to enforce high standards for scholarly publishing is needed to advance research and high-quality scholarly publications in developing countries.

Background There is increasing concern about the quality, integrity, and accessibility to research published in the developing world. This study explores the editorial practices and editors’ perspectives to gain insight into the standard of scholarly publishing in Libya. Methods Between 21st January and 12th February, 2022, the editors-in-chief (EC) of Libyan academic journals were invited to complete a questionnaire on editorial practices, degree of satisfaction with submitted and published manuscripts, review processes, and journal performance, as well as challenges facing the journals. Journal websites were examined for quality, and indexation coverage and citations were assessed. We examined the number of citations in Google Scholar for all 2019 articles published in each journal. Descriptive statistics were used to quantitatively summarize the data and thematic analysis was used for the narrative text. Results 48 EC completed the questionnaire. The EC was affiliated with the institution that owns the journal in 92% of cases. Most EC (83%) were satisfied with the peer-review quality, 69% believed that most of their published papers add new ideas or findings, and 96% were satisfied with their journal’s performance. However, despite the high degree of satisfaction, only one journal was indexed in Web of Science or Scopus and only 17% of the journals were indexed in Google Scholar. A qualitative assessment of journal websites revealed shortcomings in publishing practices in a large proportion of the journals. Conclusions The discordance between the satisfaction of the journal editors and the journal quality indicators points to a break in the quality system of Libyan academic publishing. Similar expedient publishing practices might exist in other countries as well. A comprehensive action plan led by academic institutions to enforce high standards for scholarly publishing is needed to advance research and high-quality scholarly publications in developing countries.

Background: Scholarly literature from the developing world is low and the scientific impact in some countries, including Libya, is much lower compared with their peers. This study explores the editorial practices and editors’ perspectives to gain insight into the standard of scholarly publishing in Libya. Methods: Between 21st January and 12th February, 2022, the editors-in-chief (EC) of Libyan academic journals were invited to complete a questionnaire on editorial practices, degree of satisfaction with submitted and published manuscripts, review processes, and journal performance, as well as challenges facing the journals. Journal websites were examined for quality, and indexation coverage and citations were assessed. We examined the number of citations in Google Scholar for all 2019 articles published in each journal. Descriptive statistics were used to quantitatively summarize the data and thematic analysis was used for the narrative text. Results: 48 ECs completed the questionnaire. The EC was affiliated with the institution that owns the journal in 92% of cases. Most EC (83%) were satisfied with the peer-review quality, 69% believed that most of their published papers add new ideas or findings, and 96% were satisfied with their journal performance. However, despite the high degree of satisfaction, only one journal was indexed in Web of Science or Scopus and only 17% of the journals were indexed in Google Scholar. A qualitative assessment of journal websites revealed shortcomings in publishing practices in a large proportion of the journals. Conclusions: The discordance between the satisfaction of the journal editors and the journal quality indicators points to a break in the quality system of Libyan academic publishing. Similar expedient publishing practices might exist in other countries as well. A comprehensive action plan led by academic institutions to enforce high standards for scholarly publishing is needed to advance research and high-quality scholarly publications in developing countries.

BACKGROUND: Early detection of iron overload in transfusion-dependent thalassemia (TDT) patients is critical to prevent complications and improve survival. OBJECTIVES: Evaluate the utility of serum ferritin (SF) in the prediction of hepatic and myocardial iron overload (HIO and MIO) compared to T2*-MRI. DESIGN: Retrospective SETTINGS: Governmental hospitals. PATIENTS AND METHODS: Patients with TDT who had T2*-MRI examinations between January 2016 to October 2019 were included. The predictive value of SF for detection of HIO and MIO was assessed by measuring area under the curve (AUC). A sample size of 123 cases was calculated to detect a correlation of 0.25 with 90% power and a two-sided type I error of 0.05. MAIN OUTCOME MEASURES: The correlation between SF and estimated hepatic iron concentration. SAMPLE SIZE: 137 TDT patients who required regular blood transfusions. RESULTS: The predictive value of SF was excellent for detection of HIO (AUC=0.83-0.87) but fair for detection of MIO (AUC=0.67). The two independent predictors of MIO were age and SF. The log of (age × SF) enhanced the SF predictive value for MIO (AUC=0.78). SF values of 700 and 1250 mg/L effectively excluded mild and moderate HIO with a sensitivity of 97.8% and 94.2%, respectively (LR−=0.1). While SF values of 1640 and 2150 mg/L accurately diagnosed mild and moderate HIO with a specificity of 95.55% and 96.4%, respectively (LR+>10). A log of (age × SF) cut-off value of 4.15 effectively excluded MIO (LR−=0.1), while a value of 4.65 moderately confirmed MIO (LR+=3.2). CONCLUSIONS: SF is an excellent predictor of hepatic IO in TDT. Age adjustment enhanced its myocardial IO predictive accuracy. Likelihood ratio-based SF cut-off values may help clinicians in risk stratification and treatment decision-making. LIMITATIONS: The laboratory data were gathered retrospectively and although the risk of selection bias for T2*-MRI examination is thought to be low, it cannot be ignored. CONFLICT OF INTEREST: None.

Background The impact of clinical proficiency on individual student scores on the National Board of Medical Examiners (NBME) Subject Examinations remains uncertain. We hypothesised that increasing the length of time spent in a clinical environment would augment students’ performance. Methods Performance on the NBME Subject Examination in Internal Medicine (NBME-IM) of three student cohorts was observed longitudinally. Scores at the end of two unique internal medicine clerkships held at the third and fourth years were compared. The score differences between the two administrations were compared using paired t-tests, and the effect size was measured using Cohen’s d. Moreover, linear regression was used to assess the correlation between the NBME-IM score gains and performance on a pre-clinical Comprehensive Basic Science Examination (CBSE). A two-tailed p-value d of 0.47. However, a larger effect size with a Cohen’s d value of 0.96 was observed among poorly scoring students. Performance on the CBSE was a significant predictor of score gain on the NBME-IM (R 0.51, R² 0.26, p-value < 0.001). Conclusions Despite the increased length of clinical exposure, modest improvement in students’ performance on repeated NBME-IM examination was observed. Medical educators need to reconsider how the NBME-IM is used in clerkship assessments.