Scholarlink

<p>Co-infection with <i>Mycoplasma pneumoniae</i> (MP) represents a clinically significant complication that often leads to prolonged hospital stay, increased mortality risk, and a higher demand for mechanical ventilation. This study constructed a risk prediction model for early identification of SARS-CoV-2 and MP co-infections. We retrospectively analyzed SARS-CoV-2 patients admitted between December 2022 and February 2023. Patients were stratified into co-infection and mono-infection groups based on MP antibody results. LASSO regression screened 55 variables, followed by multicollinearity checks. Restricted cubic splines (RCS) analyzed nonlinear relationships between continuous variables and infection risk. Conventional logistic and integrated RCS-logistic models were constructed and compared. LASSO identified seven predictors: age, globulin, anion gap, blood urea nitrogen (BUN), uric acid, prothrombin time (PT), and thrombin time (TT). Multivariate analysis showed globulin, anion gap, uric acid, and TT were independent risk factors, whereas BUN was protective. RCS revealed significant nonlinear associations between globulin, PT, and TT levels. The RCS-logistic model outperformed the conventional linear model, with a higher AUC of 0.827, better calibration (Brier score = 0.169), and greater net clinical benefit on decision curve analysis. This model enables early risk assessment and optimizes treatment, offering a methodological reference for predicting co-infections with emerging respiratory pathogens.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A retrospective clinical risk prediction model for co‑infection with Mycoplasma pneumoniae in patients with COVID‑19 based on restricted cubic splines

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月19日

查看全文
  • Kailong Ye,
  • Yanling Su,
  • Xiaoqing Hu,
  • Xiu Chen,
  • Bin Song,
  • Qinghua Zhang,
  • Hui Lin,
  • Linmiao Zeng,
  • Yiqun Dai,
  • Jianhong Xiao

摘要

Co-infection with Mycoplasma pneumoniae (MP) represents a clinically significant complication that often leads to prolonged hospital stay, increased mortality risk, and a higher demand for mechanical ventilation. This study constructed a risk prediction model for early identification of SARS-CoV-2 and MP co-infections. We retrospectively analyzed SARS-CoV-2 patients admitted between December 2022 and February 2023. Patients were stratified into co-infection and mono-infection groups based on MP antibody results. LASSO regression screened 55 variables, followed by multicollinearity checks. Restricted cubic splines (RCS) analyzed nonlinear relationships between continuous variables and infection risk. Conventional logistic and integrated RCS-logistic models were constructed and compared. LASSO identified seven predictors: age, globulin, anion gap, blood urea nitrogen (BUN), uric acid, prothrombin time (PT), and thrombin time (TT). Multivariate analysis showed globulin, anion gap, uric acid, and TT were independent risk factors, whereas BUN was protective. RCS revealed significant nonlinear associations between globulin, PT, and TT levels. The RCS-logistic model outperformed the conventional linear model, with a higher AUC of 0.827, better calibration (Brier score = 0.169), and greater net clinical benefit on decision curve analysis. This model enables early risk assessment and optimizes treatment, offering a methodological reference for predicting co-infections with emerging respiratory pathogens.