Establishment of a nomogram prediction model for 28-day mortality of septic shock patients based on routine laboratory data mining
10.3760/cma.j.cn121430-20240202-00108
- VernacularTitle:基于常规检验数据挖掘构建脓毒性休克患者28 d内死亡列线图预测模型
- Author:
Qifen GUO
1
;
Tao DING
;
Ran ZENG
;
Min SHAO
Author Information
1. 安徽医科大学附属阜阳医院重症医学科，安徽阜阳 236000
- Keywords:
Data mining;
Septic shock;
Death;
Nomogram;
Prognosis
- From:
Chinese Critical Care Medicine
2024;36(11):1127-1132
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a nomogram prediction model for 28-day mortality in septic shock patients based on routine laboratory data mining and verify its predictive value.Methods:The clinical data of patients with septic shock admitted to Anhui Medical University Affiliated Fuyang Hospital from January 2018 to November 2023 were retrospectively analyzed. The patients were randomly divided into training set and validation set according to the ratio of 8∶2. The patient's gender, age, body mass index, underlying disease, smoking history, alcohol history, infection site, acute physiology and chronic health evaluation Ⅱ (APACHEⅡ), sequential organ failure assessment (SOFA), respiratory rate, heart rate, mean arterial pressure, blood lactate, procalcitonin, C-reactive protein, white blood cell count, platelet count, serum alanine aminotransferase, aspartate aminotransferase, urea nitrogen, serum creatinine, fibrinogen, D-dimer, albumin on the first day of admission to the intensive care unit (ICU), duration of mechanical ventilation, and length of ICU stay were collected. The patients were divided into survival and death groups based on their 28-day prognosis. The factors influencing 28-day mortality were analyzed, and routine laboratory data were used to develop a nomogram model for predicting the risk of 28-day mortality in septic shock patients. The model was validated and assessed using the Bootstrap method, calibration curve, and receiver operator characteristic curve (ROC curve).Results:Finally, 128 patients with septic shock were enrolled, and 32 (31.07%) death within 28-day of 103 patients in the training set, 8 (32.00%) death within 28-day of 25 patients in the validation set. Logistic regression analysis showed that APACHEⅡ score [odds ratio ( OR) = 5.254, 95% confidence interval (95% CI) was 2.161-12.769], SOFA score ( OR = 4.909, 95% CI was 2.020-11.930), blood lactate ( OR = 4.419, 95% CI was 1.818-10.741), procalcitonin ( OR = 4.358, 95% CI was 1.793-10.591) were significant factors influencing 28-day mortality in septic shock patients (all P < 0.01). Taking the above influencing factors as predictors, a nomogram model was established, with a total score of 89-374, corresponding to a mortality risk of 0.07-0.89. The results of nomogram model validation showed that the C-index was 0.801 (95% CI was 0.759-0.832), and the correction curve for predicting 28-day mortality in patients with septic shock was close to the ideal curve, Hosmer-Lemeshow test showed that χ 2 = 0.263, P = 0.512. The results of the ROC curve of the training set showed that the nomogram model had a sensitivity of 78.13% (95% CI was 59.57%-90.06%), a specificity of 80.28% (95% CI was 68.80%-88.43%) and area under the curve (AUC) of 0.854 (95% CI was 0.776-0.937) in predicting 28-day mortality in patients with septic shock. The results of the validation set ROC curve showed that the nomogram model had a sensitivity of 75.00% (95% CI was 35.58%-95.55%), a specificity of 88.23% (95% CI was 62.25%-97.94%) and AUC of 0.871 (95% CI was 0.793-0.946) in predicting 28-day mortality in patients with septic shock. Conclusion:A nomogram prediction model constructed based on routine laboratory data mining can effectively predict 28-day mortality in septic shock patients, and its prediction performance is good.
