Today, there exists a high prevalence of obstructive sleep apnea (OSA) in the general population, a great proportion of which remains undiagnosed. The STOP-Bang questionnaire has been specifically developed to meet the need for a reliable, concise and easy-to-use screening tool. It consists of eight dichotomous (yes/no) items related to the clinical features of sleep apnea (Snoring, Tiredness, Observed apnea, high blood Pressure, BMI, age, neck circumference and male gender). The total score ranges from 0 to 8. Patients can be classified for their OSA risk based on their respective scores. The sensitivity of STOP-Bang score ≥3 to detect moderate-to-severe OSA (AHI > 15) and severe OSA (AHI > 30) is 93% and 100% respectively. The corresponding negative predictive values are 90% and 100%. As the STOP-Bang score increases from 0-2 to 7-8, the probability of moderate-to-severe OSA increases from 18% to 60%, and the probability of severe OSA rises from 4% to 38%. Patients with a STOP-Bang score of 0 to 2 can be classified as “low risk” for moderate-to-severe OSA, while those with a score of 5 to 8 can be classified as “high risk” for moderate-to-severe OSA. In patients whose STOP-Bang scores land in the mid-range (3 or 4), further criteria are required for classification. For example, a STOP score of ≥ 2 + BMI > 35 kg/m² would classify that patient as having a high risk for moderate-to-severe OSA. In this way, patients can be stratified for their OSA risk according to their STOP-Bang scores.

Objective: To compare the performance of machine learning models against the traditionally derived Combined Assessment of Risk Encountered in Surgery (CARES) model and the American Society of Anaesthesiologists-Physical Status (ASA-PS) in the prediction of 30-day postsurgical mortality and need for intensive care unit (ICU) stay >24 hours. Background: Prediction of surgical risk preoperatively is important for clinical shared decision-making and planning of health resources such as ICU beds. The current growth of electronic medical records coupled with machine learning presents an opportunity to improve the performance of established risk models. Methods: All patients aged 18 years and above who underwent noncardiac and nonneurological surgery at Singapore General Hospital (SGH) between 1 January 2012 and 31 October 2016 were included. Patient demographics, comorbidities, preoperative laboratory results, and surgery details were obtained from their electronic medical records. Seventy percent of the observations were randomly selected for training, leaving 30% for testing. Baseline models were CARES and ASA-PS. Candidate models were trained using random forest, adaptive boosting, gradient boosting, and support vector machine. Models were evaluated on area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC). Results: A total of 90,785 patients were included, of whom 539 (0.6%) died within 30 days and 1264 (1.4%) required ICU admission >24 hours postoperatively. Baseline models achieved high AUROCs despite poor sensitivities by predicting all negative in a predominantly negative dataset. Gradient boosting was the best performing model with AUPRCs of 0.23 and 0.38 for mortality and ICU admission outcomes respectively. Conclusions: Machine learning can be used to improve surgical risk prediction compared to traditional risk calculators. AUPRC should be used to evaluate model predictive performance instead of AUROC when the dataset is imbalanced.

Background: Cognitive biases in clinical decision-making significantly contribute to errors in diagnosis and suboptimal patient outcomes. Addressing these biases presents a formidable challenge in the medical field. Objective: This study aimed to explore the role of large language models (LLMs) in mitigating these biases through the use of the multi-agent framework. We simulate the clinical decision-making processes through multi-agent conversation and evaluate its efficacy in improving diagnostic accuracy compared with humans. Methods: A total of 16 published and unpublished case reports where cognitive biases have resulted in misdiagnoses were identified from the literature. In the multi-agent framework, we leveraged GPT-4 (OpenAI) to facilitate interactions among different simulated agents to replicate clinical team dynamics. Each agent was assigned a distinct role: (1) making the final diagnosis after considering the discussions, (2) acting as a devil’s advocate to correct confirmation and anchoring biases, (3) serving as a field expert in the required medical subspecialty, (4) facilitating discussions to mitigate premature closure bias, and (5) recording and summarizing findings. We tested varying combinations of these agents within the framework to determine which configuration yielded the highest rate of correct final diagnoses. Each scenario was repeated 5 times for consistency. The accuracy of the initial diagnoses and the final differential diagnoses were evaluated, and comparisons with human-generated answers were made using the Fisher exact test. Results: A total of 240 responses were evaluated (3 different multi-agent frameworks). The initial diagnosis had an accuracy of 0% (0/80). However, following multi-agent discussions, the accuracy for the top 2 differential diagnoses increased to 76% (61/80) for the best-performing multi-agent framework (Framework 4-C). This was significantly higher compared with the accuracy achieved by human evaluators (odds ratio 3.49; P=.002). Conclusions: The multi-agent framework demonstrated an ability to re-evaluate and correct misconceptions, even in scenarios with misleading initial investigations. In addition, the LLM-driven, multi-agent conversation framework shows promise in enhancing diagnostic accuracy in diagnostically challenging medical scenarios.

Objectives Studies in western healthcare settings suggest that preoperative anaemia is associated with poor outcomes after elective orthopaedic surgery. We investigated the prevalence of preoperative anaemia among patients with primary unilateral total knee arthroplasty (TKA) in Singapore and its association with length of hospital stay (LOS), perioperative blood transfusion and hospital readmission rates. Methods Retrospective cohort study performed in a tertiary academic medical centre in Singapore, involving patients who underwent primary unilateral TKA between January 2013 and June 2014. Demographics, comorbidities, preoperative haemoglobin (Hb) level, LOS and 30-day readmission data were collected. Anaemia was classified according to WHO definition. Prolonged LOS was defined as more than 6 days, which corresponds to >75th centile LOS of the data. Results We analysed 2394 patients. The prevalence of anaemia was 23.7%. 403 patients (16.8%) had mild anaemia and 164 patients (6.8%) had moderate to severe anaemia. Overall mean LOS was 5.4±4.8 days. Based on multivariate logistic regression, preoperative anaemia significantly increased LOS (mild anaemia, adjusted OR (aOR) 1.71, p<0.001; moderate/severe anaemia, aOR 2.29, p<0.001). Similar effects were seen when preoperative anaemia was defined by Hb level below 13 g/dL, regardless of gender. Transfusion proportionately increased prolonged LOS (1 unit: aOR 2.12, p=0.006; 2 or more units: aOR 6.71, p70 years were associated with prolonged LOS. Our 30-day related readmission rate was 1.7% (42) cases. Conclusion Anaemia is common among patients undergoing elective TKA in Singapore and is independently associated with prolonged LOS and increased perioperative blood transfusion. We suggest measures to correct anaemia prior to surgery, including the use of non-gender-based Hb cut-off for establishing diagnosis.

The Maslach Burnout Inventory for healthcare professionals (MBI-HSS) and its abbreviated version (aMBI), are the most common tools to detect burnout in clinicians. A wide range in burnout prevalence is reported in anesthesiology, so this study aimed to ascertain which of these two tools most accurately detected burnout in our anesthesiology residents. The MBI-HSS and aMBI were distributed amongst 86 residents across three hospitals, with a total of 58 residents completing the survey (67.4% response rate; 17 male and 41 female). Maslach-recommended cut-offs for the MBI-HSS and the aMBI with standard cut-offs were used to estimate burnout prevalence, and actual prevalence was established clinically by a thorough review of multiple data sources. Burnout proportions reported by the MBI-HSS and aMBI were found to be significantly different; 22.4% vs. 62.1% respectively (p < 0.0001). Compared to the actual prevalence of burnout in our cohort, the MBI-HSS detected burnout most accurately; area under receiver operating characteristic of 0.99 (95% confidence interval (CI): 0.92–1.0). Although there was a good correlation between the MBI-HSS and aMBI subscale scores, the positive predictive value of the aMBI was poor; 33.3% (95% CI:27.5–39.8%), therefore caution and clinical correlation are advised when using the aMBI tool because of the high rates of false-positives.