神尾 直

OBJECTIVES: To operationalize and temporally validate an electronic medical record (EMR)-integrated machine learning system (Big data-driven Evaluation of Survival and Treatment in Acute Illness [BEST-AI]) that generates hourly predictions for multiple ICU outcomes, with emphasis on discrimination, calibration, and workflow integration. DESIGN: Single-center hybrid study with stepwise clinical deployment and forward-in-time temporal validation. SETTING: Thirty-bed tertiary mixed medical-surgical ICU in Japan. PATIENTS: All ICU admissions from August 2017 to March 2025. Exclusions: age younger than 16 years or ICU stay less than 4 hours. Development cohort (n = 11,176; from August 2017 to July 2024) and temporal validation cohort (n = 1,127; from August 2024 to March 2025). INTERVENTIONS: EMR-integrated deployment of BEST-AI providing hourly probabilistic predictions to clinicians within the EMR; no protocolized clinical interventions were mandated. MEASUREMENTS AND MAIN RESULTS: Six prediction tasks (in-hospital mortality, ICU mortality, ICU discharge ≤ 72 hr, intubation ≤ 72 hr, extubation ≤ 72 hr, tracheostomy at ICU discharge) were evaluated. In temporal validation, the area under the receiver operating characteristic curves ranged from 0.856 to 0.960, and the area under the precision-recall curves from 0.302 to 0.786. Decile-based calibration showed overall good agreement; hospital mortality was slightly overestimated at higher predicted probabilities, whereas ICU mortality remained well aligned. The intubation task had comparatively lower discrimination and greater deviation from perfect calibration, consistent with low event counts and heterogeneous timing. A 24-hour landmark sensitivity analysis (one prediction per patient at 24 hr after ICU admission) preserved discrimination and calibration relative to the main analysis, supporting robustness beyond repeated-measures evaluation. The system was successfully maintained with automated hourly updates and EMR-embedded patient- and unit-level visualizations, without prescriptive alerts. CONCLUSIONS: A continuously deployed, EMR-integrated ICU prediction system achieved strong temporal discrimination and generally good calibration. Embedding real-time predictions into routine workflow was feasible, and the system was maintained with automated hourly updates. Prospective multicenter studies are warranted to assess transportability and clinical impact.

BackgroundThe extent to which autopsies are conducted and their role in determining the cause of death has not been sufficiently examined. We aimed to evaluate the contribution of autopsy in determining the cause of death in cases of unexpected clinical deterioration or medical accidents, using data from the Japan Center for Quality Health Care database.MethodsWe analysed data from patients with unexpected clinical deterioration who required pathological autopsy between January 2010 and November 2024. Cases were categorised based on whether autopsies were performed, and the reasons for not performing autopsies were examined.ResultsIn total, 682 reports were identified. Pathological autopsies were performed in 341 patients, and the cause of death was determined in 64.2% of cases (219 patients). The cause of death remained undetermined in 25.2% of cases (86 patients) despite autopsy. Among the 341 patients in whom autopsies were not performed, the primary reason was the family refusal (81.5%, 278 patients).ConclusionsPathological autopsies are essential for determining the cause of death in cases of unexpected clinical deterioration. Family refusal was identified as the principal barrier to autopsy, whereas lack of autopsy proposal accounted for only a small proportion of cases.

Critical care patients require continuous monitoring of vital signs and test results, yet efficiently collecting and using this data poses challenges in the intensive care unit (ICU). Usability limitations in electronic health records (EHRs) within critical care settings can delay access to essential information, potentially jeopardising patient safety. To address these issues, we developed a bedside display system that provides ICU staff with real-time, accurate access to critical data. Our system extracts and reorganises key ICU data from the existing EHR, thus avoiding costly and time-consuming upgrades. By automatically updating information such as laboratory results, blood gas analysis, lactate levels, ratio of partial pressure of arterial oxygen to fractional inspired oxygen, fluid balance and body temperature in real-time, the display allows rapid access to essential information for managing critically ill patients without the need for personal computer-based EHR logins. Post-implementation surveys with physicians, nurses and clinical engineers showed predominantly positive responses, recognising improvements in workflow and care quality. Survey results also highlighted the need for customising the display format to meet the unique requirements of each professional role, thereby maximising the system's effectiveness in critical care. This bedside display system offers four key benefits. It enhances data reliability during multidisciplinary rounds, enables physicians with busy schedules to access critical information efficiently, helps nurses detect changes in patient status early and allows a complete transition from paper-based to digital data collection. This approach offers a fresh perspective and has the potential to encourage further research into optimal information presentation methods in critical care settings.

Hematologic malignancies are a global public health concern, with high mortality rates in patients requiring critical care. The role of chemotherapy during intensive care unit (ICU) admission in this context remains unclear. This study aimed to analyze trends in survival rates based on chemotherapy timing and examine patient characteristics, ICU treatments, and clinical outcomes in each group. Using the Japanese Diagnosis Procedure Combination inpatient database, data from 21,837 patients aged ≥ 18 years who were hospitalized for hematologic malignancies and admitted to ICUs between April 1, 2012, and March 31, 2022, were analyzed. Patients were categorized based on chemotherapy timing as follows: no chemotherapy (NC), chemotherapy before ICU admission (CB), chemotherapy during ICU admission (CD), and chemotherapy after ICU discharge (CA). Mortality trends were assessed, with in-hospital mortality as the primary outcome variable. The CB group had the highest mortality rate, which decreased over time (61.2% in 2012 to 46.2% in 2021). The CD group had stable mortality rates (24.2% in 2012 and 22.6% in 2021), with a notable proportion of patients (55.4%) discharged home. These findings highlight the need for further investigation into the factors influencing ICU outcomes in patients receiving chemotherapy.