Comparison of machine-learning and logistic regression models for prediction of 30-day unplanned readmission in electronic health records: A development and validation study.

It is expected but unknown whether machine-learning models can outperform regression models, such as a logistic regression (LR) model, especially when the number and types of predictor variables increase in electronic health records (EHRs). We aimed to compare the predictive performance of gradient-boosted decision tree (GBDT), random forest (RF), deep neural network (DNN), and LR with the least absolute shrinkage and selection operator (LR-LASSO) for unplanned readmission. We used EHRs of patients discharged alive from 38 hospitals in 2015-2017 for derivation and in 2018 for validation, including basic characteristics, diagnosis, surgery, procedure, and drug codes, and blood-test results. The outcome was 30-day unplanned readmission. We created six patterns of data tables having different numbers of binary variables (that ≥5% or ≥1% of patients or ≥10 patients had) with and without blood-test results. For each pattern of data tables, we used the derivation data to establish the machine-learning and LR models, and used the validation data to evaluate the performance of each model. The incidence of outcome was 6.8% (23,108/339,513 discharges) and 6.4% (7,507/118,074 discharges) in the derivation and validation datasets, respectively. For the first data table with the smallest number of variables (102 variables that ≥5% of patients had, without blood-test results), the c-statistic was highest for GBDT (0.740), followed by RF (0.734), LR-LASSO (0.720), and DNN (0.664). For the last data table with the largest number of variables (1543 variables that ≥10 patients had, including blood-test results), the c-statistic was highest for GBDT (0.764), followed by LR-LASSO (0.755), RF (0.751), and DNN (0.720), suggesting that the difference between GBDT and LR-LASSO was small and their 95% confidence intervals overlapped. In conclusion, GBDT generally outperformed LR-LASSO to predict unplanned readmission, but the difference of c-statistic became smaller as the number of variables was increased and blood-test results were used.

Timing of Direct Oral Anticoagulants Resumption Following Colorectal Endoscopic Submucosal Dissection: A Nationwide Study in Japan.

INTRODUCTION: With the increasing use of direct oral anticoagulants (DOACs), managing these agents around endoscopic submucosal dissection (ESD) is crucial. However, due to the need for a large number of cases, studies examining the timing of resumption are lacking, resulting in varied recommendations across international guidelines. We aimed to perform a comparative study about the resumption timing of DOACs after colorectal ESD using a nationwide database in Japan. METHODS: We conducted a retrospective cohort study on colorectal ESD using the Diagnosis Procedure Combination database from 2012 to 2023. Patients using anticoagulants other than DOACs were excluded, and only those who resumed DOACs within 3 days were included. From eligible patients, we divided them into early (the day after ESD) and delayed (2-3 days after ESD) resumption groups. We used inverse probability of treatment weighting to assess the delayed bleeding and thromboembolic events within 30 days. Delayed bleeding was defined as bleeding requiring endoscopic hemostasis or blood transfusion after ESD. RESULTS: Of 176,139 colorectal ESDs, 3,550 involved DOAC users, with 2,698 (76%) categorized as early resumption and 852 (24%) categorized as delayed resumption groups. After inverse probability of treatment weighting adjustment, the early resumption group did not significantly increase delayed bleeding compared with the delayed resumption group (odds ratio 1.05; 95% confidence interval 0.78-1.42; P = 0.73). However, it significantly reduced the risk of thromboembolic events (odds ratio 0.45; 95% confidence interval 0.25-0.82; P < 0.01). DISCUSSION: Resuming DOACs the day after colorectal ESD was associated with reduced thromboembolic events without significant increase in risk of delayed bleeding.

National trends in hospitalizations for gastrointestinal bleeding in Japan.

Gastrointestinal bleeding (GIB) is a significant public health concern, predominantly associated with high morbidity. However, there have been no reports investigating the trends of GIB in Japan using nationwide data. This study aims to identify current trends and issues in the management of GIB by assessing Japan's national data. We analyzed National Database sampling data from 2012 to 2019, evaluating annual hospitalization rates for major six types of GIB including hemorrhagic gastric ulcers, duodenal ulcers, esophageal variceal bleeding, colonic diverticular bleeding, ischemic colitis, and rectal ulcers. In this study, hospitalization rates per 100,000 indicated a marked decline in hemorrhagic gastric ulcers, approximately two-thirds from 41.5 to 27.9, whereas rates for colonic diverticular bleeding more than doubled, escalating from 15.1 to 34.0. Ischemic colitis rates increased 1.6 times, from 20.8 to 34.9. In 2017, the hospitalization rate per 100,000 for colonic diverticular bleeding and ischemic colitis surpassed those for hemorrhagic gastric ulcers (31.1, 31.3, and 31.0, respectively). No significant changes were observed for duodenal ulcers, esophageal variceal bleeding, or rectal ulcers. The findings of this study underscore a pivotal shift in hospitalization frequencies from upper GIB to lower GIB in 2017, indicating a potential shift in clinical focus and resource allocation.

Hospital and regional variations in intensive care unit admission for patients with invasive mechanical ventilation.

BACKGROUND: Patients who receive invasive mechanical ventilation (IMV) in the intensive care unit (ICU) have exhibited lower in-hospital mortality rates than those who are treated outside. However, the patient-, hospital-, and regional factors influencing the ICU admission of patients with IMV have not been quantitatively examined. METHODS: This retrospective cohort study used data from the nationwide Japanese inpatient administrative database and medical facility statistics. We included patients aged ≥ 15 years who underwent IMV between April 2018 and March 2019. The primary outcome was ICU admission on the day of IMV initiation. Multilevel logistic regression analyses incorporating patient-, hospital-, or regional-level variables were used to assess cluster effects by calculating the intraclass correlation coefficient (ICC), median odds ratio (MOR), and proportional change in variance (PCV). RESULTS: Among 83,346 eligible patients from 546 hospitals across 140 areas, 40.4% were treated in ICUs on their IMV start day. ICU admission rates varied widely between hospitals (median 0.7%, interquartile range 0-44.5%) and regions (median 28.7%, interquartile range 0.9-46.2%). Multilevel analyses revealed significant effects of hospital cluster (ICC 82.2% and MOR 41.4) and regional cluster (ICC 67.3% and MOR 12.0). Including patient-level variables did not change these ICCs and MORs, with a PCV of 2.3% and - 1.0%, respectively. Further adjustment for hospital- and regional-level variables decreased the ICC and MOR, with a PCV of 95.2% and 85.6%, respectively. Among the hospital- and regional-level variables, hospitals with ICU beds and regions with ICU beds had a statistically significant and strong association with ICU admission. CONCLUSIONS: Our results revealed that primarily hospital and regional factors, rather than patient-related ones, opposed ICU admissions for patients with IMV. This has important implications for healthcare policymakers planning interventions for optimal ICU resource allocation.

Clinical subtypes of older adults starting long-term care in Japan and their association with prognoses: a data-driven cluster analysis.

We aimed to identify the clinical subtypes in individuals starting long-term care in Japan and examined their association with prognoses. Using linked medical insurance claims data and survey data for care-need certification in a large city, we identified participants who started long-term care. Grouping them based on 22 diseases recorded in the past 6 months using fuzzy c-means clustering, we examined the longitudinal association between clusters and death or care-need level deterioration within 2 years. We analyzed 4,648 participants (median age 83 [interquartile range 78-88] years, female 60.4%) between October 2014 and March 2019 and categorized them into (i) musculoskeletal and sensory, (ii) cardiac, (iii) neurological, (iv) respiratory and cancer, (v) insulin-dependent diabetes, and (vi) unspecified subtypes. The results of clustering were replicated in another city. Compared with the musculoskeletal and sensory subtype, the adjusted hazard ratio (95% confidence interval) for death was 1.22 (1.05-1.42), 1.81 (1.54-2.13), and 1.21 (1.00-1.46) for the cardiac, respiratory and cancer, and insulin-dependent diabetes subtypes, respectively. The care-need levels more likely worsened in the cardiac, respiratory and cancer, and unspecified subtypes than in the musculoskeletal and sensory subtype. In conclusion, distinct clinical subtypes exist among individuals initiating long-term care.

Outcomes of Transcatheter Tricuspid Edge-to-Edge Repair in Patients With Right Ventricular Dysfunction.

BACKGROUND: We assessed the safety profile of tricuspid transcatheter edge-to-edge repair (TEER) in patients with right ventricular (RV) dysfunction. METHODS: We identified patients undergoing TEER to treat tricuspid regurgitation from June 2015 to October 2021 and assessed tricuspid annular plane systolic excursion (TAPSE) and RV fractional area change (RVFAC). RV dysfunction was defined as TAPSE <17 mm and RVFAC <35%. The primary end point was 30-day mortality after TEER. We also investigated the change in the RV function in the early phase and clinical outcomes at 2 years. RESULTS: The study participants (n=262) were at high surgical risk (EuroSCORE II, 6.2% [interquartile range, 4.0%-10.3%]). Among them, 44 patients met the criteria of RV dysfunction. Thirty-day mortality was 3.2% in patients with normal RV function and 2.3% in patients with RV dysfunction (P=0.99). Tricuspid regurgitation reduction to ≤2+ was consistently achieved irrespective of RV dysfunction (76.5% versus 70.5%; P=0.44). TAPSE and RVFAC declined after TEER in patients with normal RV function (TAPSE, 19.0±4.7 to 17.9±4.5 mm; P=0.001; RVFAC, 46.2%±8.1% to 40.3%±9.7%; P<0.001). In contrast, those parameters were unchanged or tended to increase in patients with RV dysfunction (TAPSE, 13.2±2.3 to 15.3±4.7 mm; P=0.011; RVFAC, 29.6%±4.1% to 31.6%±8.3%; P=0.14). Two years after TEER, compared with patients with normal RV function, patients with RV dysfunction had significantly higher mortality (27.0% versus 56.3%; P<0.001). CONCLUSIONS: TEER was safe and feasible to treat tricuspid regurgitation in patients with RV dysfunction. The decline in the RV function was observed in patients with normal RV function but not in patients with RV dysfunction.

Mortality After Alcohol Septal Ablation vs. Septal Myectomy in Patients With Obstructive Hypertrophic Cardiomyopathy.

Background: Alcohol septal ablation (ASA) and septal myectomy (SM) are 2 options for septal reduction therapy (SRT) to treat medication-resistant symptomatic obstructive hypertrophic cardiomyopathy (HCM). Because differences in mortality rates after these different SRT methods have not been extensively investigated in real-world settings, in this study compared the 1-year mortality rates after ASA and SM using population-based database. Methods and Results: Utilizing New York Statewide Planning and Research Cooperative System (SPARCS) data from 2005 to 2016, we performed a comparative effectiveness study of ASA vs. SM in patients with HCM. The outcome was all-cause death up to 360 days after SRT. We constructed a multivariable logistic regression model and performed sensitivity analysis with propensity score (PS)-matching and inverse probability of treatment weighting (IPTW) methods. We identified 755 patients with HCM who underwent SRT: 348 with ASA and 407 with SM. The multivariable analysis showed that all-cause deaths were significantly fewer in the ASA group at 360 days after SRT (adjusted odds ratio=0.34; 95% confidence interval [CI] 0.13-0.84; P=0.02). The PS-matching and IPTW methods also supported a lower mortality rate in the ASA group at 360 days post-SRT. Conclusions: In this population-based study of patients with HCM who underwent SRT in a real-world setting, the 1-year all-cause mortality rate was significantly lower in patients who underwent ASA compared with SM.

Development and validation of a scoring system for in-hospital mortality following band ligation in esophageal variceal bleeding.

OBJECTIVES: We aimed to develop and validate a simple scoring system to predict in-hospital mortality after endoscopic variceal ligation (EVL) for esophageal variceal bleeding. METHODS: Data from a 13-year study involving 46 Japanese institutions were split into development (initial 7 years) and validation (last 6 years) cohorts. The study subjects were patients hospitalized for esophageal variceal bleeding and treated with EVL. Variable selection was performed using least absolute shrinkage and selection operator regression, targeting in-hospital all-cause mortality as the outcome. We developed the Hospital Outcome Prediction following Endoscopic Variceal Ligation (HOPE-EVL) score from ß coefficients of multivariate logistic regression and assessed its discrimination and calibration. RESULTS: The study included 980 patients: 536 in the development cohort and 444 in the validation cohort. In-hospital mortality was 13.6% and 10.1% for the respective cohorts. The scoring system used five variables: systolic blood pressure (<80 mmHg: 2 points), Glasgow Coma Scale (≤12: 1 point), total bilirubin (≥5 mg/dL: 1 point), creatinine (≥1.5 mg/dL: 1 point), and albumin (<2.8 g/dL: 1 point). The risk groups (low: 0-1, middle: 2-3, high: ≥4) in the validation cohort corresponded to observed and predicted mortality probabilities of 2.0% and 2.5%, 19.0% and 22.9%, and 57.6% and 71.9%, respectively. In this cohort, the HOPE-EVL score demonstrated excellent discrimination ability (area under the curve [AUC] 0.890; 95% confidence interval [CI] 0.850-0.930) compared with the Model for End-stage Liver Disease score (AUC 0.853; 95% CI 0.794-0.912) and the Child-Pugh score (AUC 0.798; 95% CI 0.727-0.869). CONCLUSIONS: The HOPE-EVL score practically and effectively predicts in-hospital mortality. This score could facilitate the appropriate allocation of resources and effective communication with patients and their families.

Association of serum magnesium levels with renal prognosis in patients with chronic kidney disease.

BACKGROUND: Magnesium deficiency is associated with various health conditions, but its impact on the progression of chronic kidney disease (CKD) remains unclear. This study aimed to investigate the association between serum magnesium levels and prognosis of renal function in CKD patients. METHODS: This is an analysis of the Japan Chronic Kidney Disease Database Ex　(J-CKD-DB-Ex), which is a multicenter prospective cohort including CKD patients enrolled from January 1, 2014 to December 31, 2020. We included adult outpatients with CKD stage G3 and G4 at the time of initial magnesium measurement. Patients were classified by magnesium levels as low (<1.7 mg/dl), normal (1.7-2.6 mg/dl), or high (>2.6 mg/dl). The primary outcomes were the composite of an eGFR < 15 ml/min/1.73 m2 or a ≥30% reduction in eGFR from the initial measurement, which was defined as CKD progression. We applied the Kaplan-Meier analysis and Cox regression hazard model to examine the association between magnesium levels and CKD progression. RESULTS: The analysis included 9868 outpatients during the follow-up period. The low magnesium group was significantly more likely to reach CKD progression. Cox regression, adjusting for covariates and using the normal magnesium group as the reference, showed that the hazard ratio for the low magnesium group was 1.20 (1.08-1.34). High magnesium was not significantly associated with poor renal outcomes compared with normal magnesium. CONCLUSION: Based on large real-world data, this study demonstrated that low magnesium levels are associated with poorer renal outcomes.

Optimal cutoff values of drain amylase for predicting pancreatic fistula are different between open and laparoscopic distal pancreatectomy.

BACKGROUND: Drainage fluid amylase (DFA) is useful for predicting clinically relevant postoperative pancreatic fistula (CR-POPF) after distal pancreatectomy (DP). However, difference in optimal cutoff value of DFA for predicting CR-POPF between open DP (ODP) and laparoscopic DP (LDP) has not been investigated. This study aimed to identify the optimal cutoff values of DFA for predicting CR-POPF after ODP and LDP. METHODS: Data for 294 patients (ODP, n = 127; LDP, n = 167) undergoing DP at Kobe University Hospital between 2010 and 2021 were reviewed. Propensity score matching was performed to minimize treatment selection bias. Receiver operating characteristic (ROC) analysis was performed to determine the optimal cutoff values of DFA for predicting CR-POPF for ODP and LDP. Logistic regression analysis for CR-POPF was performed to investigate the diagnostic value of DFA on postoperative day (POD) three with identified cutoff value. RESULTS: In the matched cohort, CR-POPF rates were 24.7% and 7.9% after ODP and LDP, respectively. DFA on POD one was significantly lower after ODP than after LDP (2263 U/L vs 4243 U/L, p < 0.001), while the difference was not significant on POD three (543 U/L vs 1221 U/L, p = 0.171). ROC analysis revealed that the optimal cutoff value of DFA on POD one and three for predicting CR-POPF were different between ODP and LDP (ODP, 3697 U/L on POD one, 1114 U/L on POD three; LDP, 10564 U/L on POD one, 6020 U/L on POD three). Multivariate analysis showed that DFA on POD three with identified cutoff value was the independent predictor for CR-POPF both for ODP and LDP. CONCLUSIONS: DFA on POD three is an independent predictor for CR-POPF after both ODP and LDP. However, the optimal cutoff value for it is significantly higher after LDP than after ODP. Optimal threshold of DFA for drain removal may be different between ODP and LDP.

Effectiveness of antibiotic prophylaxis for acute esophageal variceal bleeding in patients with band ligation: A large observational study.

BACKGROUND: Esophageal variceal bleeding is a severe complication associated with liver cirrhosis and typically necessitates endoscopic hemostasis. The current standard treatment is endoscopic variceal ligation (EVL), and Western guidelines recommend antibiotic prophylaxis following hemostasis. However, given the improvements in prognosis for variceal bleeding due to advancements in the management of bleeding and treatments of liver cirrhosis and the global concerns regarding the emergence of multidrug-resistant bacteria, there is a need to reassess the use of routine antibiotic prophylaxis after hemostasis. AIM: To evaluate the effectiveness of antibiotic prophylaxis in patients treated for EVL. METHODS: We conducted a 13-year observational study using the Tokushukai medical database across 46 hospitals. Patients were divided into the prophylaxis group (received antibiotics on admission or the next day) and the non-prophylaxis group (did not receive antibiotics within one day of admission). The primary outcome was composed of 6-wk mortality, 4-wk rebleeding, and 4-wk spontaneous bacterial peritonitis (SBP). The secondary outcomes were each individual result and in-hospital mortality. A logistic regression with inverse probability of treatment weighting was used. A subgroup analysis was conducted based on the Child-Pugh classification to determine its influence on the primary outcome measures, while sensitivity analyses for antibiotic type and duration were also performed. RESULTS: Among 980 patients, 790 were included (prophylaxis: 232, non-prophylaxis: 558). Most patients were males under the age of 65 years with a median Child-Pugh score of 8. The composite primary outcomes occurred in 11.2% of patients in the prophylaxis group and 9.5% in the non-prophylaxis group. No significant differences in outcomes were observed between the groups (adjusted odds ratio, 1.11; 95% confidence interval, 0.61-1.99; P = 0.74). Individual outcomes such as 6-wk mortality, 4-wk rebleeding, 4-wk onset of SBP, and in-hospital mortality were not significantly different between the groups. The primary outcome did not differ between the Child-Pugh subgroups. Similar results were observed in the sensitivity analyses. CONCLUSION: No significant benefit to antibiotic prophylaxis for esophageal variceal bleeding treated with EVL was detected in this study. Global reassessment of routine antibiotic prophylaxis is imperative.

Identifying intense inflammatory subtype of esophageal squamous cell carcinoma using clustering approach.

OBJECTIVE: To establish a risk-stratification system for predicting the postoperative recurrence of esophageal squamous cell carcinoma, this study aimed to evaluate the prognostic value of clusters based on blood inflammation and coagulation markers and investigate their correlation with serum cytokines and genetic alteration. METHOD: This single-center, retrospective cohort study enrolled 491 patients with esophageal cancer who underwent subtotal esophagectomy between 2004 and 2012. For cluster exploration, nonhierarchical cluster analysis and k-means were applied using serum C-reactive protein, albumin, fibrinogen, and platelet-lymphocyte ratio as variables. Then, multivariate survival analysis was conducted to investigate the association of clusters with recurrence-free survival. To characterize the clusters, serum interleukin-6, interleukin-8, and genetic alteration in primary tumors, the PleSSision-Rapid panel, which can evaluate 160 representative driver genes, was used. RESULTS: Patients were classified into clusters 1, 2, and 3, which included 24 (5%), 161 (33%), and 306 (62%) patients, respectively. Compared with cluster 3, cluster 1 or 2 had significantly worse recurrence-free survival. Based on the multivariable analysis using cluster, pStage, and age as covariates, cluster was an independent prognostic factor for recurrence-free survival (hazard ratio, 1.55; 95% confidence interval, 1.08-2.21; P = 0.02). The percentage of serum interleukin-6 and interleukin-8 levels was the highest in cluster 1, followed by clusters 2 and 3. In 23 patients with available genomic profiles, no significant difference in representative genomic alterations was observed. CONCLUSIONS: Non-biased clustering using inflammation and coagulation markers identified the intense inflammatory subtype, which had an independent prognostic effect on recurrence-free survival.

Development of novel optical character recognition system to reduce recording time for vital signs and prescriptions: A simulation-based study.

Digital advancements can reduce the burden of recording clinical information. This intra-subject experimental study compared the time and error rates for recording vital signs and prescriptions between an optical character reader (OCR) and manual typing. This study was conducted at three community hospitals and two fire departments in Japan. Thirty-eight volunteers (15 paramedics, 10 nurses, and 13 physicians) participated in the study. We prepared six sample pictures: three ambulance monitors for vital signs (normal, abnormal, and shock) and three pharmacy notebooks that provided prescriptions (two, four, or six medications). The participants recorded the data for each picture using an OCR or by manually typing on a smartphone. The outcomes were recording time and error rate defined as the number of characters with omissions or misrecognitions/misspellings of the total number of characters. Data were analyzed using paired Wilcoxon signed-rank sum and McNemar's tests. The recording times for vital signs were similar between groups (normal state, 21 s [interquartile range (IQR), 17-26 s] for OCR vs. 23 s [IQR, 18-31 s] for manual typing). In contrast, prescription recording was faster with the OCR (e.g., six-medication list, 18 s [IQR, 14-21 s] for OCR vs. 144 s [IQR, 112-187 s] for manual typing). The OCR had fewer errors than manual typing for both vital signs and prescriptions (0/1056 [0%] vs. 14/1056 [1.32%]; p<0.001 and 30/4814 [0.62%] vs. 53/4814 [1.10%], respectively). In conclusion, the developed OCR reduced the recording time for prescriptions but not vital signs. The OCR showed lower error rates than manual typing for both vital signs and prescription data.

Prognostic value of initial serum sodium level in predicting disease severity in patients with COVID-19: A multicenter retrospective study.

INTRODUCTION: Early prediction of coronavirus disease (COVID-19) severity is crucial. Hyponatremia has been linked to poor outcomes in hospitalized COVID-19 patients, but its association with mild cases is unclear. This study aimed to investigate whether initial serum sodium level is a risk factor for COVID-19 severity in patients with mild-to-moderate disease. METHODS: A multicenter retrospective cohort study was conducted in 10 hospitals in Fukui City, Japan, from July 1, 2020, to October 31, 2021. The study included 1055 adult patients with asymptomatic, mild, or moderate COVID-19 confirmed by a positive RT-PCR test. The primary outcome was the need for oxygen therapy after hospitalization, and the secondary outcome was the composite of in-hospital death and critical care interventions. The association between initial serum sodium level (at the emergency department or on admission) and outcomes was examined, adjusting for age, sex, hypertension, and pneumonia presence. RESULTS: Of the 1267 patients diagnosed with COVID-19 during the study period, 1055 were eligible (median age: 45 years; 54 % male). Hyponatremia was observed in 5.2 % of patients upon admission. A lower initial serum sodium level was associated with an increased risk of the need for oxygen therapy after hospitalization (adjusted odds ratio [OR] per 1 mmol/L lower, 1.12 [95 % confidence interval {CI}, 1.05-1.19]) and the composite of critical care and in-hospital death (adjusted OR per 1 mmol/L lower, 1.09 [95 % CI, 0.99-1.20]). CONCLUSIONS: Among patients with mild COVID-19, lower initial serum sodium level was a risk factor for COVID-19 progression.

Refining accuracy of RV-PA coupling in patients undergoing transcatheter tricuspid valve treatment.

BACKGROUND: Assessing right ventricular (RV) function is paramount for risk stratification, which remains challenging in patients with tricuspid regurgitation (TR). We assessed RV-pulmonary artery (PA) coupling and its predictability of outcomes after transcatheter tricuspid valve repair (TTVR). METHODS: Study participants comprised patients undergoing transcatheter tricuspid valve repair to treat symptomatic TR from June 2015 to July 2021. We calculated an RV-PA coupling ratio using a formula, which is dividing tricuspid annular plane systolic excursion (TAPSE) by echocardiographically estimated (ePASP) or invasively measured PASP (iPASP) at baseline. The primary outcome was all-cause mortality or heart failure rehospitalization within one year. RESULTS: The study participants (n = 206) were at high surgical risk (EuroSCORE II: 7.4 ± 4.8%). The primary outcome occurred in 57 patients within one year. The c-statistics for the outcome were 0.565 (95% CI 0.488-0.643) for TAPSE/ePASP and 0.695 (95% CI 0.631-0.759) for TAPSE/iPASP. The correlation between the ePASP and iPASP was attenuated in patients with massive/torrential TR compared to those with severe TR (interaction p = 0.01). In the multivariable Cox proportional model, TAPSE/iPASP was inversely associated with the risk of the primary outcome (per 0.1-point increase: adjusted-HR 0.67, 95% CI 0.56-0.82, p < 0.001), independent of baseline demographics. According to the TAPSE/iPASP quartiles (i.e., ≤ 0.316; 0.317-0.407; 0.408-0.526; ≥ 0.527), the event-free survival was 43.4%, 48.3%, 77.9%, and 85.4% at one year after TTVR. CONCLUSION: RV-PA coupling predicts one-year mortality and heart failure rehospitalization after TTVR in patients with TR. The predictability is improved if invasively-measured PA pressure is included.

Corrigendum: Association of airway obstruction with first-pass success and intubation-related adverse events in the emergency department: multicenter prospective observational studies.

[This corrects the article DOI: 10.3389/fmed.2023.1199750.].

The Accuracy and Potential Racial and Ethnic Biases of GPT-4 in the Diagnosis and Triage of Health Conditions: Evaluation Study.

BACKGROUND: Whether GPT-4, the conversational artificial intelligence, can accurately diagnose and triage health conditions and whether it presents racial and ethnic biases in its decisions remain unclear. OBJECTIVE: We aim to assess the accuracy of GPT-4 in the diagnosis and triage of health conditions and whether its performance varies by patient race and ethnicity. METHODS: We compared the performance of GPT-4 and physicians, using 45 typical clinical vignettes, each with a correct diagnosis and triage level, in February and March 2023. For each of the 45 clinical vignettes, GPT-4 and 3 board-certified physicians provided the most likely primary diagnosis and triage level (emergency, nonemergency, or self-care). Independent reviewers evaluated the diagnoses as "correct" or "incorrect." Physician diagnosis was defined as the consensus of the 3 physicians. We evaluated whether the performance of GPT-4 varies by patient race and ethnicity, by adding the information on patient race and ethnicity to the clinical vignettes. RESULTS: The accuracy of diagnosis was comparable between GPT-4 and physicians (the percentage of correct diagnosis was 97.8% (44/45; 95% CI 88.2%-99.9%) for GPT-4 and 91.1% (41/45; 95% CI 78.8%-97.5%) for physicians; P=.38). GPT-4 provided appropriate reasoning for 97.8% (44/45) of the vignettes. The appropriateness of triage was comparable between GPT-4 and physicians (GPT-4: 30/45, 66.7%; 95% CI 51.0%-80.0%; physicians: 30/45, 66.7%; 95% CI 51.0%-80.0%; P=.99). The performance of GPT-4 in diagnosing health conditions did not vary among different races and ethnicities (Black, White, Asian, and Hispanic), with an accuracy of 100% (95% CI 78.2%-100%). P values, compared to the GPT-4 output without incorporating race and ethnicity information, were all .99. The accuracy of triage was not significantly different even if patients' race and ethnicity information was added. The accuracy of triage was 62.2% (95% CI 46.5%-76.2%; P=.50) for Black patients; 66.7% (95% CI 51.0%-80.0%; P=.99) for White patients; 66.7% (95% CI 51.0%-80.0%; P=.99) for Asian patients, and 62.2% (95% CI 46.5%-76.2%; P=.69) for Hispanic patients. P values were calculated by comparing the outputs with and without conditioning on race and ethnicity. CONCLUSIONS: GPT-4's ability to diagnose and triage typical clinical vignettes was comparable to that of board-certified physicians. The performance of GPT-4 did not vary by patient race and ethnicity. These findings should be informative for health systems looking to introduce conversational artificial intelligence to improve the efficiency of patient diagnosis and triage.

Challenges hindering emergency physicians; involvement in multicenter collaborative studies in Japan: A nationwide survey analysis.

Aim: Multicenter collaborative research accelerates patient recruitment and strengthens evidence. Nevertheless, the factors influencing emergency and critical care physicians' involvement in such research in Japan remain unclear. Methods: A nationwide web-based survey conducted in early 2023 targeted emergency physicians working a minimum of 3 days per week in Japan. The survey descriptively assessed their backgrounds, work and research environments, experiences, and perceived impediments and motivators for multicenter research. Results: Of the 387 respondents, 348 were included in the study, yielding a 5.1% response rate. Women comprised 11% of the participants; 33% worked in university hospitals, 65% served in both emergency departments and intensive care units, and 54% did shift work. Only 12% had designated research time during working hours, with a median of 1 hour per week (interquartile range 0-5 h), including time outside of work. While 73% had participated in multicenter research, 58% noted barriers to participation. The key obstacles were excessive data entry (72%), meeting time constraints (59%), ethical review at each facility (50%), and unique sample collection, such as bronchoalveolar lavage specimens or pathological tissues (51%). The major incentives were networking (70%), data sets reuse (65%), feedback on research results (63%), and recognition from academic societies (63%). Financial rewards were not highly prioritized (38%). Conclusions: While valuing clinical research, emergency physicians face barriers, especially data entry burden and limited research time. Networking and sharing research findings motivate them. These insights can guide strategies to enhance collaborative research in emergency and critical care in Japan.