[Analysis of clinical treatment of acute respiratory distress syndrome assisted by artificial intelligence]
Zhiwei Yang,Yiwen Zu,Yuquan Luo,Quansheng Du
DOI: https://doi.org/10.3760/cma.j.cn121430-20231027-00916
Abstract:Objective: To evaluate the clinical practice of intensive care unit (ICU) physicians at Hebei General Hospital in identifying patients meeting the diagnostic criteria for acute respiratory distress syndrome (ARDS) and the current status of invasive mechanical ventilation management and adjunctive therapy in these patients, and to analyze the incidence and clinical outcomes of ARDS. Methods: A retrospective cohort study was conducted. The patients who were hospitalized in the ICU of Hebei General Hospital from April 10, 2017 to June 30, 2022 and met the Berlin definition diagnostic criteria for ARDS were enrolled as study subjects. Artificial intelligence (AI) technology was applied to search the basic information (age, gender, height, body weight, etc.), auxiliary examination, electronic medical record, non-drug doctor's advice, drug doctor's advice, critical report, scoring system, monitoring master table and other data of the above medical records in the electronic medical record system of the hospital. The first set of laboratory indicators sequentially retrieved from the system daily from 05:00 to 10:00 and vital signs and mechanical ventilation-related parameters recorded in the "critical care report" at 06:00 daily were extracted, and outcome indicators of the patients were collected. Results: After screening and analysis, a total of 255 patients who met the ARDS diagnostic criteria were finally enrolled. The overall incidence of ARDS in the ICU accounted for 3.4% (255/7 434) of the total number of ICU patients, of which mild, moderate and severe ARDS accounted for 22.4% (57/255), 49.0% (125/255), and 28.6% (73/255), respectively, while the recognition rates of clinical doctors were 71.9% (41/57), 58.4% (73/125) and 71.2% (52/73), respectively. During the ICU stay, 250 patients (98.0%) received only invasive mechanical ventilation, while 5 patients (2.0%) received both non-invasive and invasive mechanical ventilation. The tidal volume/ideal body weight of ARDS patients was 7.64 (6.49, 9.01) mL/kg, and the positive end-expiratory pressure (PEEP) was 8.0 (5.0, 10.0) cmH2O (1 cmH2O ≈ 0.098 kPa). In addition, during the diagnosis and detection of ARDS, only 7 patients were recorded the platform pressure and 6 patients were recorded the drive pressure. Regarding adjunctive therapies, 137 patients (53.7%) received deep sedation, 26 patients (10.2%) underwent lung recruitment, 55 patients (21.6%) received prone ventilation, 42 patients (16.5%) were treated with high-dose steroids, 19 patients (7.5%) were treated with neuromuscular blockade, and 8 patients (3.1%) were treated with extracorporeal membrane oxygenation (ECMO). Finally, 70 patients (27.5%) were discharged automatically, while 50 patients (19.6%) died in the ICU, of which the ICU mortality of mild, moderate, and severe ARDS patients were 15.8% (9/57), 22.4% (28/125), and 17.8% (13/73), respectively. After follow-up, it was found that all 70 patients discharged automatically died within 28 days after discharge, and the overall ICU mortality adjusted accordingly was 47.1% (120/255). Conclusions: The overall incidence of ARDS in ICU patients at Hebei General Hospital is relatively low, with a high recognition rate by clinical physicians. Despite the high level of compliance and implementation of lung protective ventilation strategies and auxiliary treatment measures, it is still necessary to further improve the level of standardization in the implementation of small tidal volume and respiratory mechanics monitoring. For the implementation of auxiliary measures such as prone ventilation, it is necessary to further improve the enthusiasm of medical staff. The mortality in ICU is relatively low in ARDS patients, while the rate of spontaneous discharge is relatively high.