Development Of AKI Electronic Alert System For Critically Ill Patients Based On MIMIC-? Database

Objective1.To study the independent risk factors affecting the outcomes in critically ill patients with acute kidney injury admitted to the intensive care unit by using MIMIC database,develop a predictive model,optimize data model further,and establish a prognostic scale.2.To develop AKI electronic alert system for critically ill patients based on MIMIC-? database.MethodsThis a retrospective and observational study.Using diagnostic codes to screen the discharge diagnosis of patients is acute renal failure in ICU during June 2001 to December 2012 from database(The definition is acute renal failure in database while acute kidney injury in this study).Inclusion criteria: the discharge diagnosis is acute renal failure;the first diagnosis during the same hospitalization;age > 18 years,no restriction on gender;with an ICU length of stay > 24 hours and have completeness for records of hospitalization.Exclusion criteria: the admission diagnosis is ARF;the cases with history of renal transplant and chronic renal failure;back to the ICU during the same hospitalization.Rejection criteria: with an ICU length of stay < 24 hours;age< 18 years;missing data.Using STATA software to obtain related data of the patients.First,the study carried out descriptive statistical analysis,and then patients were divided into two groups based on survival state at discharge.The risk factors for prognosis of patients with AKI were determined by using single factor and multivariate logistic regression analysis,and a predictive model was developed next.A scoring equation was developed according to the binary logistic regression results of the test group and verified in the verification group,also,a Quick Acute Kidney Injury Evaluation was established further.Last but not least,all above made an AKI electronic alert system for critically ill patients come true.Results1.According to the criteria,4554 patients were excluded in this study,862 patients died during hospitalization,with a mortality rate of 18.93%.Male patients accounted for 56.90%.The age groups of 41~60 years old and 61~80 years old were the highrisk group of critically ill patients with AKI.Among AKI patients included in this study,23.32% had sepsis or septic shock.2.The study validate that lymphoma,metastatic cancer,vancomycin,coagulopathy,cardiac arrest,sepsis or septic shock,mechanical ventilation,potassium,urea nitrogen,bilirubin and white blood cell were significantly associated with prognosis of AKI patients(P<0.05).3.The first 2/3 patients were selected from the whole group for modeling and the latter 1/3 patients for verification,which were regarded as test group and verification group.Binary Logistics regression analysis showed that coagulopathy,metastatic cancer,vancomycin,sepsis or septic shock,cardiac arrest,mechanical ventilation,elevated urea nitrogen,elevated bilirubin and abnormal potassium were independent risk factors.The AKI prediction model was 1/(1+exp-(-3.622+coagulopathy×0.374+ metastatic cancer ×0.963+ vancomycin ×0.401+ cardiac arrest ×0.906+ sepsis or septic shock ×0.861+ urea nitrogen ×0.006+ potassium ion ×0.137+ total bilirubin ×0.053+ mechanical ventilation ×0.738)).The AUROC was 0.733.4.Based on the odds ratios,a new prognostic scoring equation was derived,which was coagulopathy ×1+ metastatic cancer ×3+vancomycin ×1+ cardiac arrest ×2+ sepsis or septic shock ×2+ mechanical ventilation ×2+ urea nitrogen ×1+ potassium ion ×1+ bilirubin ×1.Using the chi-square trend to test,we can see statistically significant in mortality(P<0.05).The rank sum test was used to assess the goodness of fit between two groups.In view of the above results,an AKI electronic alert system for critically ill patients based on MIMIC-? database was developed.Conclusion1.The analysis results of data exported from MIMIC-? database suggest that the mortality of AKI patients in ICU is 18.93%.2.Coagulopathy,metastatic cancer,vancomycin,sepsis or septic shock,cardiac arrest,mechanical ventilation,elevated urea nitrogen,elevated bilirubin and abnormal potassium are independent risk factors for in-hospital death in patients with AKI.3.The development of Quick Acute Kidney Injury Evaluation and AKI electronic alert system for critically ill patients in this study may have strong practical significance for the prognosis risk assessment of acute kidney injury.