[Value of sepsis single-disease manage system in predicting mortality in patients with sepsis]

Objective: To observe the effect of sepsis single-disease manage system on the improvement of sepsis treatment and the value in predicting mortality in patients with sepsis. Methods: A retrospective study was conducted. Patients with sepsis admitted to the Department of Surgical Intensive Care Unit of Sun Yat-Sen University First Affiliated Hospital from September 22, 2013 to May 5, 2015 were enrolled in this study. Sepsis single-disease manage system (Rui Xin clinical data manage system, China data, China) was used to monitor 25 clinical quality parameters, consisting of timeliness, normalization and outcome parameters. Based on whether these quality parameters could be completed or not, the clinical practice was evaluated by the system. The unachieved quality parameter was defined as suspicious parameters, and these suspicious parameters were used to predict mortality of patients with receiver operating characteristic curve (ROC). Results: A total of 1 220 patients with sepsis were enrolled, included 805 males and 415 females. The mean age was (59±17) years, and acute physiology and chronic health evaluation (APACHE Ⅱ) scores was 19±8. The area under ROC curve of total suspicious numbers for predicting 28-day mortality was 0.70; when the suspicious parameters number was more than 6, the sensitivity was 68.0% and the specificity was 61.0% for predicting 28-day mortality. In addition, the area under ROC curve of outcome suspicious number for predicting 28-day mortality was 0.89; when the suspicious outcome parameters numbers was more than 1, the sensitivity was 88.0% and the specificity was 78.0% for predicting 28-day mortality. Moreover, the area under ROC curve of total suspicious number for predicting 90-day mortality was 0.73; when the total suspicious parameters number was more than 7, the sensitivity was 60.0% and the specificity was 74.0% for predicting 90-day mortality. Finally, the area under ROC curve of outcome suspicious numbers for predicting 90-day mortality was 0.92; when suspicious outcome parameters numbers was more than 1, the sensitivity was 88.0% and the specificity was 81.0% for predicting 90-day mortality. Conclusion: The single center study suggests that this sepsis single-disease manage system could be used to monitor the completion of clinical practice for intensivist in managing sepsis, and the number of quality parameters failed to complete could be used to predict the mortality of the patients.

Predictive model of death for hospitalized patients with esophageal varices bleeding due to liver cirrhosis based on routine laboratory results

AIM: To analyze the relationship between routine laboratory results and death risk in patients with cirrhotic esophageal variceal bleeding (CEVB), and to create a simple and timely predictive model of death for the patients.

METHODS: The medical documents of CEVB patients were reviewed retrospectively and the data were collected. Univariate and multivariate Logistic regression analyses were performed, in which the discharged results (survival or death) as dependent variables and the results of liver function, kidney function, serum electrolytes and blood cell analysis as independent variables were included. The multivariate regression equation was as the model for the prediction of patient outcome and its predictive performance was evaluated.

RESULTS: In univariate regression, the significant positive variables for death outcome were DBIL, AKP, K, WBC, PLT, and the negative variables were TP, AP, A/G, Na, Cl, Ca. The variables entered the multivariate regression included ALT, TBIL, DBIL, GP, A/G, Cr, Na, Cl, Ca, WBC, Hb, PLT. The sensitivity, specificity and accuracy of the regression equation for predicting death of CEVB patients were 97.1%, 95.1% and 95.8%.

CONCLUSION: The liver function, kidney function, serum electrolytes and blood cell count are generally independent factors for death risk of CEVB patients, especially DBIL, A/G and Ca. The created regression model shows good prediction performance.