ABSTRACT
Objective:To analyze the risk factors related to the prognosis of patients with sepsis in intensive care unit (ICU), construct a nomogram model, and verify its predictive efficacy.Methods:A retrospective cohort study was conducted using data from Medical Information Mart for Intensive Care-Ⅳ 0.4 [MIMIC-Ⅳ (version 2.0)]. The information of 6 500 patients with sepsis who meet the diagnostic criteria of Sepsis-3 were collected, including demography characteristics, complications, laboratory indicators within 24 hours after ICU admission, and final outcome. Using a simple random sampling method, the patients were divided into a training set and a validation set at a ratio of 7∶3. The restricted cubic spline (RCS) was used to explore whether there was a linear relationship between each variable and the prognosis, and the nonlinear variables were truncated into categorical variables. All variables were screened by LASSO regression and included in multivariate Cox regression analysis to analyze the death risk factors in ICU patients with sepsis, and construct a nomograph. The consistency index, calibration curve and receiver operator characteristic curve (ROC curve) were used to evaluate the prediction efficiency of nomogram model. The decision curve analysis (DCA) was used to validate the clinical value of the model and its impact on actual decision-making.Results:Among 6 500 patients with sepsis, 4 551 were in the training set and 1 949 were in the validation set. The 28-day, 90-day and 1-year mortality in the training set were 27.73% (1?262/4?551), 34.76% (1?582/4?551), and 42.98% (1?956/4?551), respectively, those in the validation set were 27.24% (531/1?949), 33.91% (661/1?949), and 42.23% (823/1?949), respectively. Both in training set and the validation set, compared with the final survival patients, the death patients were older, and had higher sequential organ failure assessment (SOFA) score and simplified acute physiology scoreⅡ (SAPSⅡ), more comorbidities, less urine output, and more use of vasoactive drugs, kidney replacement therapy, and mechanical ventilation. By RCS analysis, the variables with potential nonlinear correlation with the prognosis risk of septic patients were transformed into categorical variable. The variables screened by LASSO regression were enrolled in the multivariate Cox regression model. The results showed that age [hazard ratio ( HR) = 1.021, 95% confidence interval (95% CI) was 1.018-1.024], SOFA score ( HR = 1.020, 95% CI was 1.000-1.040), SAPSⅡ score > 44 ( HR = 1.480, 95% CI was 1.340-1.634), mean arterial pressure (MAP) ≤ 75 mmHg (1 mmHg ≈ 0.133 kPa; HR = 1.120, 95% CI was 1.026-1.222), respiratory rate (RR; HR = 1.044, 95% CI was 1.034-1.055), cerebrovascular disease ( HR = 1.620, 95% CI was 1.443-1.818), malignant tumor ( HR = 1.604, 95% CI was 1.447-1.778), severe liver disease ( HR = 1.330, 95% CI was 1.157-1.530), use of vasoactive drugs within 24 hours ( HR = 1.213, 95% CI was 1.101-1.336), arterial partial pressure of oxygen (PaO 2; HR = 0.999, 95% CI was 0.998-1.000), blood lactic acid (Lac; HR = 1.066, 95% CI was 1.053-1.079), blood urea nitrogen (BUN) > 8.9 mmol/L ( HR = 1.257, 95% CI was 1.144-1.381), total bilirubin (TBil; HR = 1.023, 95% CI was 1.015-1.031), and prothrombin time (PT) > 14.5 s ( HR = 1.232, 95% CI was 1.127-1.347) were associated with the death of ICU patients with sepsis (all P < 0.05). Based on the above factors, a nomogram model was constructed, and the model validation results showed that the consistency index was 0.730. The calibration curve showed a good consistency between the predicted results of the nomogram model and observed results in the training and validation sets. ROC curve analysis showed that the area under the ROC curve (AUC) predicted by the nomogram model in the training set and the validation set for 28-day, 90-day and 1-year death risk was 0.771 (95% CI was 0.756-0.786) and 0.761 (95% CI was 0.738-0.784), 0.777 (95% CI was 0.763-0.791) and 0.765 (95% CI was 0.744-0.787), 0.677 (95% CI was 0.648-0.707) and 0.685 (95% CI was 0.641-0.728), respectively. DCA analysis showed that the nomogram model had significant net benefits in predicting 28-day, 90-day, and 1-year death risk, verifying the clinical value of the model and its good impact on actual decision-making. Conclusions:The death risk factors related to ICU patients with sepsis include age, SOFA score, SAPSⅡ score > 44, MAP ≤ 75 mmHg, RR, cerebrovascular disease, malignant tumors, severe liver disease, use of vasoactive drugs within 24 hours, PaO 2, Lac, BUN, TBil, PT > 14.5 s. The nomogram model constructed based on this can predict the death risk of ICU patients with sepsis.
ABSTRACT
Objective:To evaluate the effect of invasive arterial blood pressure (IBP) monitoring on the prognosis of patients with sepsis.Methods:Patients with sepsis from the MIMIC-Ⅳ database were collected and divided into IBP and non-invasive blood pressure monitoring (NIBP) groups according to whether IBP monitoring was performed. Baseline variables that were considered clinically relevant or showed a univariate relationship with the outcome were entered into a multivariate logistic regression model as covariates.Propensity score matching(PSM) and inverse probability of treatment weighing(IPTW) were used to adjust confounders to ensure the robustness of findings.Subgroup analysis were conducted to evaluate the effect of differences in IBP onset and duration on outcome.Results:The 28-day mortality is lower in IBP group compared with NIBP group( OR=0.54, 95% CI 0.46-0.62, P<0.001), the conclusion maintain robust after PSM and IPTW.Then we conducted a series of logistic regression regarding to different initial IBP time(<24 h,24 h-48 h,>48 h) and the initial IBP time within 24 h showed the same results compared to primary outcoms( OR=0.42, 95% CI: 0.36-0.49, P<0.001). IBP duration varied (≤1day, ≤2days, ≤3days, ≤4days, >4days) all showed a statistically significant association with decreased 28-day mortality in the IBP group. Conclusions:IBP is associated with decreased 28-day mortality in patients with sepsis, and the optimal time of IBP is within 24 hours.