Scholarlink

Background <p>Non-human immunodeficiency virus (HIV) immunocompromised patients with Pneumocystis <i>jirovecii</i> pneumonia (PJP) face rapid progression and high mortality, necessitating a predictive model to identify patients at risk of adverse clinical outcomes for timely interventions and improved stratification.</p> Methods <p>Patients admitted between January 2011 and June 2024 at Beijing Chao-Yang Hospital were retrospectively analyzed. Collected data included patients’ demographics, smoking status, comorbidities, immunosuppressive diseases, blood laboratory tests, in-hospital treatment, and adverse clinical outcomes. Predictor selection was performed using the least absolute shrinkage and selection operator (LASSO) and logistic regression, with selected features incorporated into a nomogram. Internal validation was conducted using a 500-bootstrap resampling method to ensure model robustness. Model performance was assessed via the area under the receiver operating curve (AUC), calibration plots, decision curve analysis (DCA), and clinical impact curve (CIC).</p> Results <p>Among the 431 patients, 243 (56.4%) experienced adverse clinical outcomes. LASSO regression screened 21 variables, selecting 9 predictors with non-zero coefficients through 10-fold cross-validation at lambda.1se = 0.0453 (log(lambda.1se) = -3.092). Multivariate logistic regression identified 7 independent risk factors for adverse clinical outcomes: smoking status, cytomegalovirus infection, diabetes, neutrophil-lymphocyte ratio (NLR), lactate dehydrogenase (LDH), PaO<sub>2</sub>/FiO<sub>2</sub> (PFR), and lymphocyte subset. These factors were incorporated into a nomogram, achieving an AUC of 0.89 (95% CI: 0.86–0.92), with the Hosmer–Lemeshow test (<i>p</i> = 0.134) and calibration curves showing strong agreement between predicted and observed outcomes. Internal validation via 500-bootstrap resampling yielded a bias-corrected AUC of 0.83 (95% CI: 0.80–0.86). DCA demonstrated strong clinical decision-making utility, while the CIC confirmed its practical reliability.</p> Conclusions <p>Regression analysis identified smoking status, CMV infection, diabetes, NLR, LDH, PFR, and lymphocyte subset as independent risk factors for adverse clinical outcomes in non-HIV PJP patients. The predictive model constructed from these factors exhibited robust accuracy and reliability.</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 nomogram-based clinical prediction model for adverse clinical outcomes in non-HIV Pneumocystis jirovecii pneumonia patients

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月17日

查看全文
  • Dong Wang,
  • Lujia Guan,
  • Qian Yin,
  • Xiaoxia Hou,
  • Xi Zhan,
  • Zhaohui Tong

摘要

Background

Non-human immunodeficiency virus (HIV) immunocompromised patients with Pneumocystis jirovecii pneumonia (PJP) face rapid progression and high mortality, necessitating a predictive model to identify patients at risk of adverse clinical outcomes for timely interventions and improved stratification.

Methods

Patients admitted between January 2011 and June 2024 at Beijing Chao-Yang Hospital were retrospectively analyzed. Collected data included patients’ demographics, smoking status, comorbidities, immunosuppressive diseases, blood laboratory tests, in-hospital treatment, and adverse clinical outcomes. Predictor selection was performed using the least absolute shrinkage and selection operator (LASSO) and logistic regression, with selected features incorporated into a nomogram. Internal validation was conducted using a 500-bootstrap resampling method to ensure model robustness. Model performance was assessed via the area under the receiver operating curve (AUC), calibration plots, decision curve analysis (DCA), and clinical impact curve (CIC).

Results

Among the 431 patients, 243 (56.4%) experienced adverse clinical outcomes. LASSO regression screened 21 variables, selecting 9 predictors with non-zero coefficients through 10-fold cross-validation at lambda.1se = 0.0453 (log(lambda.1se) = -3.092). Multivariate logistic regression identified 7 independent risk factors for adverse clinical outcomes: smoking status, cytomegalovirus infection, diabetes, neutrophil-lymphocyte ratio (NLR), lactate dehydrogenase (LDH), PaO2/FiO2 (PFR), and lymphocyte subset. These factors were incorporated into a nomogram, achieving an AUC of 0.89 (95% CI: 0.86–0.92), with the Hosmer–Lemeshow test (p = 0.134) and calibration curves showing strong agreement between predicted and observed outcomes. Internal validation via 500-bootstrap resampling yielded a bias-corrected AUC of 0.83 (95% CI: 0.80–0.86). DCA demonstrated strong clinical decision-making utility, while the CIC confirmed its practical reliability.

Conclusions

Regression analysis identified smoking status, CMV infection, diabetes, NLR, LDH, PFR, and lymphocyte subset as independent risk factors for adverse clinical outcomes in non-HIV PJP patients. The predictive model constructed from these factors exhibited robust accuracy and reliability.