Internal and emergency medicine · 2023
A machine learning diagnostic model for Pneumocystis jirovecii pneumonia in patients with severe pneumonia.
Рандомизированное исследованиеСтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Li Xiaoqian, Xiong Xingyu, Liang Zongan, Tang Yongjiang
Аннотация
BACKGROUND: The diagnosis of Pneumocystis jirovecii pneumonia (PCP) in patients presenting with severe pneumonia is challenging and delays in treatment were associated with worse prognosis. This study aimed to develop a rapid, easily available, noninvasive machine learning diagnostic model for PCP among patients with severe pneumonia.
METHODS: A retrospective study was performed in West China Hospital among consecutive patients with severe pneumonia who had undergone bronchoalveolar lavage for etiological evaluation between October 2010 and April 2021. Factors associated with PCP were identified and four diagnostic models were established using machine learning algorithms including Logistic Regression, eXtreme Gradient Boosting, Random Forest (RF) and LightGBM. The performance of these models were evaluated by the area under the receiver operating characteristic curve (AUC).
RESULTS: Ultimately, 704 patients were enrolled and randomly divided into a training set (n = 564) and a testing set (n = 140). Four factors were ultimately selected to establish the model including neutrophil, globulin, β-D-glucan and ground glass opacity. The RF model exhibited the greatest diagnostic performance with an AUC of 0.907. The calibration curve and decision curve analysis also demonstrated its accuracy and applicability.
CONCLUSIONS: We constructed a PCP diagnostic model in patients with severe pneumonia using four easily available and noninvasive clinical indicators. With satisfying diagnostic performance and good clinical practicability, this model may help clinicians to make early diagnosis of PCP, reduce the delays of treatment and improve the prognosis among these patients.
METHODS: A retrospective study was performed in West China Hospital among consecutive patients with severe pneumonia who had undergone bronchoalveolar lavage for etiological evaluation between October 2010 and April 2021. Factors associated with PCP were identified and four diagnostic models were established using machine learning algorithms including Logistic Regression, eXtreme Gradient Boosting, Random Forest (RF) and LightGBM. The performance of these models were evaluated by the area under the receiver operating characteristic curve (AUC).
RESULTS: Ultimately, 704 patients were enrolled and randomly divided into a training set (n = 564) and a testing set (n = 140). Four factors were ultimately selected to establish the model including neutrophil, globulin, β-D-glucan and ground glass opacity. The RF model exhibited the greatest diagnostic performance with an AUC of 0.907. The calibration curve and decision curve analysis also demonstrated its accuracy and applicability.
CONCLUSIONS: We constructed a PCP diagnostic model in patients with severe pneumonia using four easily available and noninvasive clinical indicators. With satisfying diagnostic performance and good clinical practicability, this model may help clinicians to make early diagnosis of PCP, reduce the delays of treatment and improve the prognosis among these patients.