神田 直樹

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.

There have been few reports on the diagnostic performance of soluble interleukin-2 receptor (sIL-2R) for lymphoma. A cross-sectional study was conducted at a university hospital; all patients who were admitted to the Division of General Internal Medicine and underwent serum sIL-2R testing were included. Patients were divided into two groups based on the presence of fever (≥ 38.0 °C). Among 602 patients, 421 had fever and 76 were diagnosed with lymphoma (48 of the 76 were in the febrile group). In all patients, the area under the receiver operating characteristic curve (AUROC) of sIL-2R for the diagnosis of lymphoma was 0.81 [95% confidence interval (CI), 0.75-0.87]. The AUROC was significantly higher in the febrile group (0.88; 95% CI, 0.81-0.94) than in the afebrile group (0.75; 95% CI, 0.65-0.85). In the febrile group, the sensitivity and specificity were 81.2% and 82.3%, respectively, with an optimal cutoff value of 3,250 U/mL. In the afebrile group, they were 89.3% and 54.9%, respectively, with a cutoff value of 868 U/mL. Serum sIL-2R showed high performance as an adjunctive diagnostic marker for lymphoma, particularly among febrile patients. Different cutoff values should be used for patients with and without fever to maximize diagnostic performance.

INTRODUCTION: Days of antibiotic spectrum coverage (DASC), a novel metric for both antimicrobial volume and spectrum, was proposed to measure inpatient antimicrobial consumption in 2022. The DASC may better reflect efforts toward antimicrobial stewardship; however, no previous study has described the distribution of trends in DASC in hospitals or the association between the trend in DASC and days of therapy (DOT). This study assessed trends in antimicrobial consumption for inpatients at acute care hospitals in Japan using the DOT and DASC. METHODS: This retrospective observational study used the nationwide administrative inpatient claims database of Japanese acute care hospitals between 2014 and 2018. The prescriptions of all antibiotics were identified to calculate DOT/1000 patient-days, DASC/1000 patient-days, and DASC/DOT. We described the five-year trend in these metrics and examined the association between the trends in DOT and DASC. RESULTS: In total, 26,301,685 admissions from 634 hospitals were included. The DOT/1000 patient-days and DASC/1000 patient-days increased significantly by 6.1% and 5.6%, respectively. The DASC/DOT ratio did not change significantly (P = 0.35). Moreover, there was little correlation between DOT/1000 patient-days and DASC/DOT (R2 = 0.01). There was also little correlation between the five-year changes in DOT/1000 patient-days and those in DASC/DOT (R2 = 0.02). CONCLUSIONS: It may be difficult to assess trends in the antibiotic spectrum using DASC alone. However, the combination of DOT as a quantity indicator and DASC/DOT as a spectrum indicator may allow for a more appropriate evaluation of stewardship efforts.

BACKGROUND: Antimicrobial stewardship in dentistry and antibiotic prophylaxis for tooth extraction have been areas of concern in Japan, with limited research available. METHODS: This cross-sectional study utilized the regional health insurance claims database in Japan to examine antibiotic prescription trends by dentists, including antibiotic regimens, timing of prescription, and days of supply for prophylactic antibiotic use during tooth extraction. Antibiotic prophylaxis for patients with prosthetic heart valves was also investigated. FINDINGS: Antibiotic prescriptions by dentists decreased by 7% in 2019 compared to those in 2015, with third-generation cephalosporins still accounting for 48.5% in 2019. Amoxicillin prescription increased 3.9 times in 2019, although it only accounted for 8.4% of all antibiotic prescriptions. In 2019, amoxicillin was prescribed for 17.1% of all prophylactic antibiotics associated with tooth extraction, and 80% of prophylactic antibiotics were prescribed for 3 days or more, with 85% prescribed on the day of the procedure. However, only 60-70% of patients with prosthetic heart valves received antibiotic prophylaxis. INTERPRETATION: Despite the increasing trend after the implementation of the National Action Plan on antimicrobial resistance, the proportion of amoxicillin prescriptions in dentistry remains low. Antimicrobial stewardship issues related to long-term prescription and timing of administration of prophylactic antibiotics for tooth extraction should be addressed. Dentists must recognize the risks associated with high-risk patients with prosthetic heart valves who require antibiotic prophylaxis, and physicians providing valve replacement therapy should inform patients of the requirement for prophylaxis before invasive dental procedures.

Persistent inflammation, immunosuppression, and catabolism syndrome (PICS) is a serious condition after critical care. We examined the efficacy of antithrombin, which may attenuate coagulopathy with the control of inflammation, for PICS among patients with sepsis-induced disseminated intravascular coagulation (DIC). The present study used the inpatient claims database with laboratory findings to identify patients admitted to intensive care units and diagnosed with sepsis and DIC. A composite of the incidence of PICS on day 14 or 14-day mortality as the primary outcome was compared between the antithrombin and control groups using a propensity-score-matched analysis. Secondary outcomes were the incidence of PICS on day 28, 28-day mortality, and in-hospital mortality. A total of 324 well-balanced matched pairs were generated from 1622 patients. The primary outcome did not differ between the antithrombin and control groups (63.9% vs. 68.2%, respectively, p = 0.245). However, the incidences of 28-day and in-hospital mortality were significantly lower in the antithrombin group (16.0% vs. 23.5% and 24.4% vs. 35.8%, respectively). Similar results were obtained in a sensitivity analysis using overlap weighting. Antithrombin did not reduce the occurrence of PICS on day 14 in patients with sepsis-induced DIC; however, it was associated with a better mid-term (day 28) prognosis.