Abstract
Establishment and Validation of a Nomographic Model for Individualized Prediction of Invasive Pulmonary Fungal Infection after Chemotherapy for Hematologic Tumor
Department of Medical Laboratory, The First People’s Hospital of Chongqing Liang Jiang New Area, Chongqing 401121, China
Correspondence Address:
Shihai Zheng, Department of Medical Laboratory, The First People’s Hospital of Chongqing Liang Jiang New Area, Chongqing 401121, China, E-mail: liuling0710dym@163.com
To retrospectively analyze the risk factors of invasive pulmonary fungal infection after chemotherapy in patients with hematologic tumor and to establish and validate related nomographic models. 368 patients with hematologic tumor who were treated with chemotherapy were collected. The influencing factors of invasive pulmonary fungal infection after chemotherapy were analyzed by univariate and multivariate logistic regression. Based on the analyzed risk factors, the nomogram prediction model was established and verified internally. Of the 368 patients with hematologic tumor, 35.33 % (130/368) had fungal infection after chemotherapy and 64.67 % (238/368) had non-fungal infection. Univariate and multivariate logistic regression analysis showed that hypoproteinemia (Odds ratio=3.220, 95 % confidence interval: 1.886~5.496), broad-spectrum antibiotic use ≥2 kinds (Odds ratio=2.616, 95 % confidence interval: 1.514~4.518), broadspectrum antibiotic duration ≥7 d (Odds ratio=2.444, 95 % confidence interval: 1.455~4.103), duration of agranulocytosis ≥7 d (Odds ratio=3.112, 95 % confidence interval: 1.804~5.368), cytomegalovirus or Epstein-Barr virus infection (Odds ratio=6.170, 95 % confidence interval: 3.304~11.524) were risk factors for fungal infection after chemotherapy in patients with hematologic tumors. Based on the above five risk factors, the prediction model of nomogram was established and verified by Bootstrap method for 500 times. The calibration curve fitted well with the standard curve. The receiver operating characteristic curve area under the curve was 0.762, 95 % confidence interval: 0.7095-0.8146 and the Hosmer-Lemeshow test p>0.05, which indicated that the prediction model of nomogram had good discrimination and accuracy. Hypoproteinemia, broad-spectrum antibiotic use ≥2 kinds, broad-spectrum antibiotic duration ≥7 d, duration of agranulocytosis ≥7 d and cytomegalovirus or Epstein-Barr virus infection are the risk factors of concurrent fungal infection after chemotherapy in patients with hematologic tumor. After verification, the constructed nomogram has high discrimination and accuracy, which has dual guiding significance for clinical screening of high-risk population and making individualized prevention and treatment measures.