The Comparison Of Using Various Scoring System In General Intensive Care Unit

Objective:In our study,we aim to describe and evaluate not validated and calibrated the performance of original predictive score system:APACHEâ…¡,APACHEâ…£,SAPSâ…¡,MPMâ…¡0,MPMâ…¡24 by retrospective study.5 scoring systems were used in the general ICU of the first affiliated hospital of kunming medical college. By combined with at present the medical field about using critical scoring system xisted problem, clinical application and development tendency.we will try to explore the issue about the critical scoring system's applied and future development in China.Methods and material:We use a series of critical ill scoring software package and an individual APACHEâ…£predictive scoring software package.which are made by Jiangxuecheng who is a medicine professor in the NO.97 hospital of PLA,XuZhou.JangSu province.We selected following 5 predictive scoring system:APACHEâ…¡, APACHEâ…£,SAPSâ…¡,MPMâ…¡0,MPMâ…¡24.Study population come from the critical ill patients in the general ICU of the first affiliated hospital of kunming medical college from Jan,2009 to Dec,2010.Excluded criterion:age<18year, the length of stay(LOS)<24hours and >365 days.and automatic discharge patients.Besides.cardial surgical patiens and transplant postoprated patients are all not include. Collecting the worst information and phsiologic variable of patients within 24 hours after admitted ICU,we use above illustrated 5 critical ill predictive scoring tool to analyse them.Then,SPSS 17.0 statistic caculate and analyse these outcome.Calibration of the prognostic models was assessed by using Hosmer-Lemeshow chi square test and by SMR and its 95% confidence interval. We used the ROC area under the curve to estimate Model discrimination.Result:The study included 197 patients.Observed hospital mortality was 56.9% in this study.Selected 5 scoring systems that they predicted patients PHM was following: APACHEâ…¡52.8%,APACHEâ…£45.7%,SAPSâ…¡43.1%,MPMâ…¡0 34.5%,MPMâ…¡24 34%.SMR of every risk predictive model was computed:APACHEâ…¡1.07,APACHEâ…£1.21,SAPSâ…¡1.17,MPMâ…¡0 1.31, MPMâ…¡24 1.25,which are all more than 1 and indicate 5scoring models to underestimate the actual PHM.Each of model have good diagnostic capability. Discrimination was generally good for all models[area under the receiver operating characteristic curve ranged from 0.762(MPMâ…¡0) to 0.871 (APACHEâ…¡)]. APACHEâ…¡among of these models presented the best discrimination and APACHEâ…¡'s upper limit of confidence interval is 0.925(>0.9).Next were APACHEâ…£(AUC=0.829) and SAPSâ…¡(AUC=0.828), their AUC are almost not difference.The last were MPMâ…¡24,MPMâ…¡0.SMR's confidence showed:SAPSâ…¡,MPMâ…¡24T and MPMâ…¡0 three models'performance were unsatisfactory(95% CI>1),however.two other models(APACHEâ…¡,APACHEâ…£) appeared to common performance(95% CI included 1).Hosmer-Lemeshow chi square goodness-of-fit test displaied poor calibration for APACHEâ…£.SAPSâ…¡. MPMâ…¡24,MPMâ…¡0.Only APACHEâ…¡presented good calibration.Conclusions:1. New gerneral assessment model:APACHE IV not showed superior ideal performance than older model.2. Over 25 years.APACHEâ…¡still presented better performance than others and not showed "model fade".3. MPMâ…¡0,MPMâ…¡24 that were two subsystem of the 3th generation MPMâ…¡.were almost not difference about the discrimination of the patient survival or not. Although their performance were inferior to APACHEâ…¡.and had a little bit inferior to APACHEâ…£and SAPSâ…¡.their discrimination still showed moderate.Otherwise.they also showed poor calibration.As a result of MPMâ…¡'s cost and manual collected data were significantly less than others.the MPMâ…¡model offers a viable alternative without a substantial loss in accuracv.