One-year mortality prediction for patients with sepsis: a nomogram integrating lactic dehydrogenase and clinical characteristics

Abstract

Abstract Background To explore the association between myocardial enzymes and one-year mortality, and establish a nomogram integrating myocardial enzymes and clinical characteristics to predict one-year mortality among sepsis patients. Methods Data of 1,983 sepsis patients were extracted from Medical Information Mart for Intensive Care III database in this retrospective cohort study. All participants were randomly split into the training set for the development of model and testing set for the internal validation at the ratio of 7:3. Univariate logistic regression was used to screen variables with statistical differences which were made for stepwise regression, obtaining the predictors associated with one-year mortality of sepsis patients. Adopted multivariate logistic regression to assess the relationship between myocardial enzymes and one-year mortality of sepsis patients. A nomogram was established in predicting the one-year survival status of sepsis patients, and the performance of developed model were compared with LDH alone, sequential organ failure assessment (SOFA), simplified acute physiology score II (SAPS II) by receiver operator characteristic, calibration, and decision curves analysis. Results The result found that LDH was associated with one-year mortality of sepsis patients [odds ratio = 1.28, 95% confidence interval (CI): 1.18–1.52]. Independent predictors, including age, gender, ethnicity, potassium, calcium, albumin, hemoglobin, alkaline phosphatase, vasopressor, Elixhauser score, respiratory failure, and LDH were identified and used to establish the nomogram (LDH-model) for predicting one-year mortality for sepsis patients. The predicted performance [area under curve (AUC) = 0.773, 95%CI: 0.748–0.798] of this developed nomogram in the training and testing sets (AUC = 0.750, 95%CI: 0.711–0.789), which was superior to that of LDH alone, SOFA score, SAPS II score. Additionally, calibration curve indicated that LDH-model may have a good agreement between the predictive and actual outcomes, while decision curve analysis demonstrated clinical utility of the LDH-model. Conclusion LDH level was related to the risk of one-year mortality in sepsis patients. A prediction model based on LDH and clinical features was developed to predict one-year mortality risk of sepsis patients, surpassing the predictive ability of LDH alone as well as conventional SAPS II and SOFA scoring systems.