Performance of ChatGPT and Gemini Compared with Emergency Physicians in NSTEMI Cases: A Prospective Cross-sectional Study.

The European Society of Cardiology (ESC) guidelines provide detailed, evidence-based recommendations for managing cardiovascular diseases. However, their complexity and frequent updates can make them challenging to apply consistently in clinical settings. Artificial intelligence (AI), particularly large language models (LLMs), offers a novel solution by assisting in the interpretation and application of these guidelines more effectively. A narrative review was conducted to assess the role of large language models (LLMs) and related artificial intelligence (AI) systems in supporting the interpretation of ESC guidelines. From 102 records screened, seven studies met the inclusion criteria. Clinical Decision Support Systems (CDSSs) built on ESC guidelines demonstrated improvements in diagnostic accuracy and standardization. Comparative studies revealed that large language models (LLMs), including ChatGPT-4, showed high concordance with expert clinical decisions (up to 86% accuracy for acute coronary syndrome-related questions). Emerging tools, such as MedDoc-Bot, have highlighted the feasibility of direct ESC guideline interpretation by LLMs. LLMs show promise in enhancing clinician understanding and application of ESC guidelines. Although performance is encouraging, further validation and thoughtful integration into clinical practice are necessary to maximize their utility and safety.

e20011 Background: The therapeutic landscape of lung cancer is rapidly evolving, presenting oncologists with the challenge of staying updated amidst an overwhelming influx of data. Clinical decision support (CDS) tools, including artificial intelligence (AI) and large language models (LLMs), may help bridge this gap. Evaluating the accuracy of LLMs in complex, real-world oncology scenarios is crucial to understanding their potential. Methods: Twenty-five de-identified lung cancer cases from the fellows’ clinic at Karmanos Cancer Institute, Detroit, MI, were analyzed. Two LLMs, GPT-4 (OpenAI) and Claude Opus (Anthropic), were assessed using advanced prompting techniques like persona-based and chain-of-thought prompting. Five board-certified lung cancer oncologists from NCI-designated centers evaluated LLM-generated responses based on accuracy, treatment recommendation comprehensiveness, and supportive care planning, using a 1–5 scale. Novel insights, the presence of fabricated information, and harmful recommendations were flagged as binary outcomes. Oncologists were blinded to the LLM source and actual treatment decisions. Results: Table 1 presents patient characteristics. GPT-4 achieved an average accuracy score of 4.2 (95% CI, 3.9–4.4), with 3.7 for comprehensiveness of medical/surgical treatment recommendations and 3.7 for supportive care planning. Six responses (32%) were flagged as potentially harmful, and two (8%) contained inaccuracies. Sixteen GPT-4 responses (64%) were rated trustworthy as a CDS tool. Claude Opus had an average accuracy score of 3.6 (95% CI, 3.1–4.1), scoring 3.6 for treatment recommendation comprehensiveness and 3.5 for supportive care planning. Nine responses (36%) were flagged for potential harm, and five (20%) included inaccuracies. Eleven Claude responses (44%) were deemed trustworthy. Significant differences were observed in accuracy (p=0.04) and trustworthiness (p=0.03) between models using McNemar's test. Other factors showed no statistical significance. Conclusions: GPT-4 outperformed Claude Opus in accuracy and trustworthiness, but both models demonstrated limitations, including harmful recommendations and inaccuracies. These findings highlight the need for improved LLM refinement before routine use as CDS tools in lung cancer treatment. Patient demographics and clinical characteristics. Category subcategory Number Median Age (range)- Yr 65 (26-78) Female 7 Male 18 Histology Adenocarcinoma 10 Squamous Cell Carcinoma (SCC) 7 Small Cell Carcinoma 6 Poorly Differentiated 2 Total 25 Stage NSCLC Stage 3 7 NSCLC Stage 4 13 Small Cell limited stage 3 Small Cell Extensive Stage 2

Clinical Decision Support Tools

BackgroundA specific electrocardiographic (ECG) marker of ischemia would greatly improve the speed and accuracy of detecting and treating non-ST elevation myocardial infarction (NSTEMI). We hypothesize that ischemia induces ventricular repolarization dispersion (VRD), altering the T-wave before any ST segment deviation. We sought to evaluate the clinical utility of VRD to (1) detect NSTEMI cases in the emergency department (ED) and (2) identify NSTEMI cases at high risk for in-hospital major adverse cardiac events (MACEs).Methods and ResultsWe continuously recorded 12-lead Holter ECGs from chest pain patients upon their arrival to the ED. VRD was quantified using principal component analysis of the 12-lead ECG to compute a T-wave complexity ratio (ie, ratio of second to first eigenvectors of repolarization). Clinical outcomes were obtained from hospital records. The sample was composed mainly of older males (n=369; ages 63±12 years; 63% males), and 92 (25%) had NSTEMI and 26 (7%) had MACEs. Baseline T-wave complexity ratio modestly correlated with peak troponin levels (r=0.41; P<0.001) and was a good classifier of NSTEMI events (area under the curve=0.70). An increased T-wave complexity ratio on the presenting ECG was strongly associated with NSTEMI (odds ratio [OR]=3.8 [2.1 to 5.8]) and in-hospital MACE (OR=8.2 [3.1 to 21.5]).ConclusionsA simple measure of global VRD on the presenting 12-lead ECG correlates with ischemic myocardial injury and can discriminate NSTEMI cases very early during evaluation. Prospective studies should validate these findings and test whether VRD can guide therapy.

AIMTo assess the prevalence of depressed heart rate variability (HRV) after an acute myocardial infarction (MI), and to evaluate its prognostic significance in the present era of immediate reperfusion.METHODSTime-domain HRV (obtained from 24-h Holter recordings) was assessed in 326 patients (63.5 ± 12.1 years old; 80% males), two weeks after a complicated MI treated by early reperfusion: 208 ST-elevation myocardial infarction (STEMI) patients (in which reperfusion was successfully obtained within 6 h of symptoms in 94% of cases) and 118 non-ST-elevation myocardial infarction (NSTEMI) patients (percutaneous coronary intervention was performed within 24 h and successful in 73% of cases). Follow-up of the patients was performed via telephone interviews a median of 25 mo after the index event (95%CI of the mean 23.3-28.0). Primary end-point was occurrence of all-cause or cardiac death; secondary end-point was occurrence of major clinical events (MCE, defined as mortality or readmission for new MI, new revascularization, episodes of heart failure or stroke). Possible correlations between HRV parameters (mainly the standard deviation of all normal RR intervals, SDNN), clinical features (age, sex, type of MI, history of diabetes, left ventricle ejection fraction), angiographic characteristics (number of coronary arteries with critical stenoses, success and completeness of revascularization) and long-term outcomes were analysed.RESULTSMarkedly depressed HRV parameters were present in a relatively small percentage of patients: SDNN < 70 ms was found in 16% and SDNN < 50 ms in 4% of cases. No significant differences were present between STEMI and NSTEMI cases as regards to their distribution among quartiles of SDNN (χ2 =1.536, P = 0.674). Female sex and history of diabetes maintained a significant correlation with lower values of SDNN at multivariate Cox regression analysis (respectively: P = 0.008 and P = 0.008), while no correlation was found between depressed SDNN and history of previous MI (P = 0.999) or number of diseased coronary arteries (P = 0.428) or unsuccessful percutaneous coronary intervention (PCI) (P = 0.691). Patients with left ventricle ejection fraction (LVEF) < 40% presented more often SDNN values in the lowest quartile (P < 0.001). After > 2 years from infarction, a total of 10 patients (3.1%) were lost to follow-up. Overall incidence of MCE at follow-up was similar between STEMI and NSTEMI (P = 0.141), although all-cause and cardiac mortality were higher among NSTEMI cases (respectively: 14% vs 2%, P = 0.001; and 10% vs 1.5%, P = 0.001). The Kaplan-Meier survival curves for all-cause mortality and for cardiac deaths did not reveal significant differences between patients with SDNN in the lowest quartile and other quartiles of SDNN (respectively: P = 0.137 and P = 0.527). Also the MCE-free survival curves were similar between the group of patients with SDNN in the lowest quartile vs the patients of the other SDNN quartiles (P = 0.540), with no difference for STEMI (P = 0.180) or NSTEMI patients (P = 0.541). By the contrary, events-free survival was worse if patients presented with LVEF < 40% (P = 0.001).CONCLUSIONIn our group of patients with a recent complicated MI, abnormal autonomic parameters have been found with a prevalence that was similar for STEMI and NSTEMI cases, and substantially unchanged in comparison to what reported in the pre-primary-PCI era. Long-term outcomes did not correlate with level of depression of HRV parameters recorded in the subacute phase of the disease, both in STEMI and in NSTEMI patients. These results support lack of prognostic significance of traditional HRV parameters when immediate coronary reperfusion is utilised.

Distinguishing ST-elevation myocardial infarction (STEMI) and non-ST-elevation myocardial infarction (NSTEMI) is crucial in acute myocardial infarction (AMI) research due to their distinct characteristics. However, the accuracy of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes for STEMI and NSTEMI in Taiwan's National Health Insurance (NHI) database remains unvalidated. Therefore, we developed and validated case definition algorithms for STEMI and NSTEMI using ICD-10-CM and NHI billing codes. We obtained claims data and medical records of inpatient visits from 2016 to 2021 from the hospital's research-based database. Potential STEMI and NSTEMI cases were identified using diagnostic codes, keywords, and procedure codes associated with AMI. Chart reviews were then conducted to confirm the cases. The performance of the developed algorithms for STEMI and NSTEMI was assessed and subsequently externally validated. The algorithm that defined STEMI as any STEMI ICD code in the first three diagnosis fields had the highest performance, with a sensitivity of 93.6% (95% confidence interval [CI], 91.7-95.2%), a positive predictive value (PPV) of 89.4% (95% CI, 87.1-91.4%), and a kappa of 0.914 (95% CI, 0.900-0.928). The algorithm that used the NSTEMI ICD code listed in any diagnosis field performed best in identifying NSTEMI, with a sensitivity of 82.6% (95% CI, 80.7-84.4%), a PPV of 96.5% (95% CI, 95.4-97.4), and a kappa of 0.889 (95% CI, 0.878-0.901). The algorithm that included either STEMI or NSTEMI ICD codes listed in any diagnosis field showed excellent performance in defining AMI, with a sensitivity of 89.4% (95% CI, 88.2-90.6%), a PPV of 95.6% (95% CI, 94.7-96.4%), and a kappa of 0.923 (95% CI, 0.915-0.931). External validation confirmed these algorithms' efficacy. Our results provide valuable reference algorithms for identifying STEMI and NSTEMI cases in Taiwan's NHI database.

The aim of this study was to examine the application of artificial intelligence (AI) and generative AI (GAI) in dental education through a comprehensive bibliometric analysis, with the goal of identifying potential future research areas in this field. Data were obtained from the Web of Science and Scopus databases, with no date restrictions. The inclusion criteria encompassed research articles, reviews and proceedings papers focused on the use of AI in dental education. The Bibliometrix R package was used for data analysis. A total of 54 documents were analysed, revealing a significant increase in publications from 2021 to 2024. The Journal of Dental Education was identified as the most prolific source, with the United States leading in terms of contributions. The findings showed a growing interest in AI's potential to transform dental education, with significant contributions from a diverse set of global authors and institutions. Several key themes were identified, including 'large language models', 'chatbots' and 'clinical decision support systems'. The citation analysis highlighted influential papers, including those by Thurzo etal. (2023) and Yüzbaşıoğlu (2021). There is a growing interest in the use of AI and GAI in dental education. However, further research, particularly experimental studies, is essential to fully understand their impact on educational outcomes in dentistry. Key areas for exploration include the application of AI for personalised learning, the integration of chatbots as educational tools to support students, and the use of AI for training and assessing practical skills.

BackgroundAcute pulmonary embolism (APE) and non-ST elevation myocardial infarction (NSTEMI) are two of the most common cardiovascular emergencies with overlapping clinical presentations, which frequently produce diagnostic dilemmas. Both APE and NSTEMI share elevation in D-dimer and troponin I, rendering them less specific when utilized singly.ObjectiveThis study aimed to investigate whether the ratio of D-dimer to troponin I could serve as a discriminative biomarker for differentiating between APE and NSTEMI.MethodsAn observational cross-sectional analysis was performed in 40 patients with confirmed APE and NSTEMI cases. Demographic factors and laboratory variables (D-dimer, troponin I, and estimated D-dimer-to-troponin ratio) were recorded. Statistical analysis was conducted with IBM SPSS Statistics for Windows, Version 26.0 (Released 2019; IBM Corp., Armonk, New York, United States). Continuous data were reported as mean±SD, and the categorical data were given as frequencies and percentages. Comparisons between groups were made using t-tests and correlation with Pearson's coefficient, and p<0.05 was considered significant.ResultsPatients' mean age was 55.9±8.8 years for APE and 52.4±6.3 years for NSTEMI. APE patients showed much higher ratios of D-dimer to troponin (42.1±34.7) than NSTEMI patients (0.36±0.49; p<0.001). Sex-related differences were not statistically significant in both groups. Troponin I was negatively correlated with the ratio in APE (r=-0.590; p<0.001), while D-dimer was significantly correlated with the ratio in NSTEMI (r=0.798; p<0.001) in correlation analysis.ConclusionThe D-dimer-to-troponin ratio well differentiates between APE and NSTEMI, reflecting their unique pathophysiologic mechanisms. This ratio can perhaps offer clinicians an easy, inexpensive adjunct to enhance early diagnostic precision and direct proper treatment strategies.

A key problem in the emergency department is the prompt identification of patients with myocardial infarction. While this is not a diagnostic challenge in ST elevation myocardial infarction as the ECG is usually unmistakable, in non ST elevation myocardial infarction (NSTEMI) the diagnosis is more complex and based on information derived from detailed clinical assessment, ECG and biomarkers. The ideal biomarker should allow an accurate and immediate identification of patients with NSTEMI. We do not have this ideal biomarker yet. The best approximation is currently represented by high sensitivity cardiac troponin (hs-cTn). In particular, the hs-cTn 0h/1h-algorithm proposed in the European Society of Cardiology Guidelines on the management of NSTEMI might allow a rapid ‘rule in’ and ‘rule out’ of a large proportion of patients with suspected NSTEMI. A critical question is: ‘Should the 1h algorithm for rule in and rule out of acute myocardial infarction be used universally?’. This approach might substantially speed up the triage of patients with suspected NSTEMI. Yet, some concern has recently been raised on this universal application regardless of symptom duration, overall clinical risk and sex. In the following debate key opinion leaders give their answers to this important question with major clinical implications. Finally, it is worth mentioning that this is a rapidly moving field and I anticipate that a new stimulating debate will soon be needed. # Sometimes earlier may not be better {#article-title-2} The recent European Society of Cardiology (ESC) guidelines f or the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation (NSTE-ACS) added an avant-garde rapid one hour rule-in and rule-out algorithms using high sensitivity cardiac troponins (hscTn) as a clinical option.1 Although these algorithms may represent an advance in the early diagnosis of chest pain patients the guidelines, in our opinion, they overlook some gaps in the information that defined these approaches …

Evaluating artificial intelligence (AI) as a clinical decision support tool for AML patients

How Can <i>IJDS</i> Authors, Reviewers, and Editors Use (and Misuse) Generative AI?

Moving Beyond Tools and Building Bridges: Lessons Learned From a CKD Decision Support in Primary Care

The use of high-sensitivity troponin levels increases the sensitivity of the diagnosis of non-ST elevation myocardial infarction (NSTEMI). However, the inclusion of other factors in the differential diagnosis, apart from atherothrombosis causing myocardial injury, decreases the specificity of high-sensitivity troponin. In this study, we compared the efficacy of high-sensitivity troponin with serum oncostatin M in NSTEMI cases with elevated urea and creatinine. This study was performed with a prospective cross-sectional sample. Ninety participants with coronary angiography performed due to a preliminary diagnosis of NSTEMI were included. High-sensitivity troponin I, creatine kinase-MB, lactate dehydrogenase, serum transaminase and oncostatin M levels were quantitatively measured for the first 4-8 hours from the onset of symptoms. All participants had coronary angiography performed within the first 12 hours after attending the emergency service. Based on coronary angiography data, patients with significant coronary stenosis or occlusion detected during coronary angiography were defined as group A, and patients with no occlusion in the coronary artery and who did not require an additional interventional procedure were defined as group B. The SYNTAX 2 score was used to determine the severity of coronary artery disease. Patients in both groups A and B had similar age, sex distribution and comorbidities. Group A had higher serum urea, creatinine, oncostatin M and high-sensitivity troponin I values than group B. With 585 pg/ml as the cut-off value, serum oncostatin M had a sensitivity of 88.6% and specificity of 85% for the diagnosis of NSTEMI. Logistic regression multivariate analysis showed that serum oncostatin M and high-sensitivity troponin I values had diagnostic efficacy for NSTEMI. Serum oncostatin M was found to be more effective than high-sensitivity troponin I in patients with elevated urea and creatinine. Serum oncostatin M had similar sensitivity and specificity for NSTEMI diagnosis as high-sensitivity troponin I. Serum OSM can especially be considered as a complementary diagnostic biomarker for NSTEMI in patients with renal dysfunction.

Artificial intelligence (AI) augmentation of ECG assessment has significant potential to improve patient outcomes in acute coronary syndrome. We sought to evaluate the performance of a novel AI device (PMCardio) in assessing angiographic occlusion myocardial infarction (OMI) and predicting clinical outcomes. We used a 1-year retrospective cohort of angiographic data from patients presenting with ST-elevation myocardial infarction (STEMI) and non-ST-elevation myocardial infarction (NSTEMI). The device analyzed precatheterization ECGs to identify OMI, defined as a culprit vessel with thrombolysis In myocardial infarction (TIMI) 0-2 flow or TIMI 3 flow and peak cardiac troponin I > 10.0 ng/ml. A total of 217 patients were included: 72 STEMI (32%) and 145 NSTEMI (65%). Angiographic OMI was confirmed in 60 (83%) STEMI and 51 (35%) NSTEMI cases. The AI model achieved a sensitivity of 86.5%, specificity of 82.2%, and an area under the curve of 0.84. Traditional STEMI criteria had a sensitivity of 54.1% and a specificity of 88.7%. The AI model was 100% sensitive in detecting STEMI-OMI. The odds ratio for mortality in AI-detected OMI patients was 12.44 (1.56-98.98), unplanned readmissions 1.15 (0.53-2.51), and reduced ejection fraction at 1 year 0.24 (0.26-2.16). The AI model demonstrated higher sensitivity and similar specificity compared with traditional STEMI criteria, improving OMI detection while reducing false positives. These findings suggest potential benefits in triage accuracy and resource utilization, but further prospective validation is needed to determine its clinical impact.

Objective: To explore the feasibility and accuracy of 0-1 h high sensitivity cardiac troponin I (hs-cTnI) concentration and its changes in judging non-ST segment elevation myocardial infarction (NSTEMI), and to investigate the feasibility of a simplified process. Methods: Patients with acute chest pain and suspected NSTEMI who were admitted to the emergency department of Fuwai Hospital, the First Affiliated Hospital of Sun Yat-sen University and Nanjing First Hospital from January 2017 to September 2020 were selected. Hs-cTnI test was carried out for the selected patients at the time of visit (0 h) and 1 h after visit. According to the 0-1 h hs-cTnI diagnostic process and threshold standard recommended by European Society of Cardiology (ESC) guidelines in 2015, the laboratory adjudication was determined. Cardiologists who did not participate in the project design and did not know the results of hs-cTnI test performed the clinical judgment according to the routine diagnosis and treatment process of emergency department. Taking clinical judgment as the gold standard, the diagnostic efficacy of 0-1 h hs-cTnI concentration and its change recommended by the guidelines for judging NSTEMI in Chinese population was analyzed. The guide process was simplified. Under the condition of not considering the time of chest pain, the guideline threshold was used for test and judgement, and the diagnostic efficacy of the simplified process was evaluated. Results: A total of 1 534 patients were enrolled in the study, aged (62±12) years and 952 (62.1%) patients were male. Among them, 402 patients (26.2%) were clinically diagnosed as NSTEMI and 1 132 patients (73.8%) were diagnosed as non-NSTEMI. According to the diagnosis and determination process recommended by the guidelines, NSTEMI was excluded in 672 patients (42.8%), and 464 patients (30.2%) were diagnosed as NSTEMI. The consistency rate with clinical determination reached 92.4% (1 050/1 136), the sensitivity of excluding diagnosis was 99.5% (95%CI: 98.0%-99.9%), the negative predictive value was 99.7% (95%CI: 98.8%-99.9%), and the negative likelihood ratio was 0.008 (95%CI: 0.002-0.335). The diagnostic specificity was 92.6% (95%CI: 90.9%-94.0%), the positive predictive value was 81.9% (95%CI: 78.0%-85.2%), and the positive likelihood ratio was 12.739 (95%CI: 10.356-15.670). According to the simplified process, NSTEMI was excluded in 675 patients (44.0%), and 463 patients (30.2%) were diagnosed as NSTEMI. The consistency rate with clinical judgment was 92.4% (1 051/1 138), the sensitivity of exclusion diagnosis was 99.3% (95%CI: 97.6%-99.8%), the negative predictive value was 99.6% (95%CI: 98.6%-99.9%), and the negative likelihood ratio was 0.012 (95%CI: 0.004-0.389). The diagnostic specificity was 92.6% (95%CI: 90.9%-94.0%), the positive predictive value was 81.9% (95%CI: 78.0%-85.2%), and the positive likelihood ratio was 12.705 (95%CI: 10.328-15.630). There was no significant difference in diagnostic efficacy between the simplified process and the recommended process (all P>0.05). Conclusion: The diagnostic process for judging NSTEMI according to the 0-1 h hs-cTnI concentration and its change criteria recommended by the 2015 ESC guidelines is applicable in the Chinese population and remains highly accurate in judging NSTEMI without considering the duration of chest pain at the time of presentation.