Analisi Stato Vegetativo
Danilo Lofaro
Tuesday, July 07, 2015
Data Processing
Read data
CRS_raw <- read.csv("./SAnnaCRS.csv", sep=";")
CRS_proc <- CRS_raw
dim(CRS_proc)## [1] 156 150str(CRS_proc[,1:28])## 'data.frame': 156 obs. of 28 variables:
## $ ID : Factor w/ 156 levels "P1","P10","P100",..: 1 69 80 91 102 113 124 135 146 2 ...
## $ Age : int 39 51 46 42 52 37 73 50 50 49 ...
## $ Etiology : Factor w/ 8 levels "Anossia","Infettivo",..: 5 7 7 4 7 4 7 1 7 4 ...
## $ Etiology_mod : Factor w/ 11 levels "Anossia_extraos",..: 8 10 10 7 10 7 10 1 10 7 ...
## $ DateOfEvent : Factor w/ 146 levels "01/01/2013","01/02/2010",..: 143 36 85 68 124 41 104 63 10 83 ...
## $ DateOfAdmission : Factor w/ 150 levels "01/02/2012","01/08/2011",..: 37 141 2 18 145 98 17 34 100 39 ...
## $ DipOrigin : Factor w/ 32 levels "GCA Pisa","NCH CS",..: 27 11 11 22 13 11 27 14 2 13 ...
## $ Craniectomy : Factor w/ 2 levels "NO","SI": 1 2 1 2 1 2 2 1 1 1 ...
## $ Cranioplasty : Factor w/ 5 levels "","NO","Progr NO",..: 1 2 1 2 1 2 2 1 1 1 ...
## $ ComplicationsNCH : Factor w/ 17 levels "","Cisti III Ventr",..: 15 8 1 15 1 1 7 1 1 1 ...
## $ Operations : Factor w/ 7 levels "","DVP","Embolizzazione",..: 4 1 1 2 1 1 1 1 1 1 ...
## $ OtherComplications: Factor w/ 58 levels "","Afasia","Aritmia Card",..: 1 2 1 30 1 1 1 1 1 1 ...
## $ Dysautonomia : Factor w/ 2 levels "","SI": 1 1 1 1 1 1 1 1 1 1 ...
## $ DRS_Adm : int 24 22 18 24 23 26 22 27 24 22 ...
## $ DRS_Dis : int 21 19 6 21 20 30 30 25 11 12 ...
## $ GOS : int 3 3 4 3 3 1 1 2 4 3 ...
## $ SNG_Adm : Factor w/ 2 levels "","SNG": 1 1 2 1 1 1 1 1 2 2 ...
## $ SNG_Dis : Factor w/ 3 levels "","OS","SNG": 1 1 2 1 1 1 1 1 2 1 ...
## $ PEG_Adm : Factor w/ 3 levels "","NO","SI": 3 2 1 3 2 2 3 2 1 1 ...
## $ PEG_Dis : Factor w/ 3 levels "","NO","SI": 3 3 1 3 3 2 3 3 1 2 ...
## $ TRACH_Adm : Factor w/ 4 levels "","NO","SI","SIS": 3 3 2 3 3 3 3 3 2 3 ...
## $ TRACH_Dis : Factor w/ 3 levels "","NO","SI": 3 2 2 2 3 3 3 3 2 2 ...
## $ DateOfDischarge : Factor w/ 144 levels "02/01/2013","02/02/2012",..: 67 16 12 53 63 69 23 136 37 144 ...
## $ ModeOfDischarge : Factor w/ 19 levels "AL LDS","AL LDS VIC RES",..: 15 3 6 17 17 5 5 17 16 8 ...
## $ LastConscStatus : Factor w/ 4 levels "EMERGED","EMERGED ",..: 3 1 1 3 3 4 4 4 1 1 ...
## $ Status : Factor w/ 2 levels "ALIVE","DECEASED": 1 1 1 1 1 1 2 1 1 1 ...
## $ LastTime : int 19 15 3 27 23 14 22 20 3 3 ...
## $ DateOfCRS1 : Factor w/ 152 levels "01/02/2012","01/07/2010",..: 43 149 10 39 151 112 22 66 117 54 ...Data transformation
Date transformation
CRS_proc$DateOfEvent <- as.Date.factor(CRS_proc$DateOfEvent,format="%d/%m/%Y")
CRS_proc$DateOfAdmission <- as.Date.factor(CRS_proc$DateOfAdmission,format="%d/%m/%Y")
CRS_proc$DateOfDischarge <- as.Date.factor(CRS_proc$DateOfDischarge,format="%d/%m/%Y")
CRS_proc$DateOfCRS1<- as.Date.factor(CRS_proc$DateOfCRS1,format="%d/%m/%Y")
# View(CRS_proc)Categorical Variables Transformation
- Etiology is stratified as: TCE, Vascular, Anoxia or others (infections, neoplastic)
require(plyr)
CRS_proc$Etiology <- as.character(CRS_proc$Etiology)
CRS_proc[grep("TCE",CRS_proc$Etiology),"Etiology"] <- "TCE"
CRS_proc[grep("Vasc",CRS_proc$Etiology),"Etiology"] <- "Vascular"
CRS_proc[grep("Anos",CRS_proc$Etiology),"Etiology"] <- "Anoxia"
CRS_proc[grep("(Inf|Neo)",CRS_proc$Etiology),"Etiology"] <- "Others"
CRS_proc$Etiology <- as.factor(CRS_proc$Etiology)- Origin Department is stratified as: ICU or NCH
CRS_proc$DipOrigin <- as.character(CRS_proc$DipOrigin)
CRS_proc[grep("^Rian",CRS_proc$DipOrigin),"DipOrigin"] <- "ICU"
CRS_proc[!(CRS_proc$DipOrigin=="ICU"),"DipOrigin"] <- "NCH"
CRS_proc$DipOrigin <- as.factor(CRS_proc$DipOrigin)- Change Craniectomy and Cranioplasty values as yes/no
CRS_proc$Craniectomy <- revalue(CRS_proc$Craniectomy,c("NO"="no","SI"="yes"))
CRS_proc$Cranioplasty <- revalue(CRS_proc$Cranioplasty,c("NO"="no","Progr NO"="no","Programm"="yes","SI"="yes"))
CRS_proc[CRS_proc$Cranioplasty=="","Cranioplasty"] <- "no"
CRS_proc$Cranioplasty <- factor(CRS_proc$Cranioplasty)- Change ComplicationsNCH empty values in no
CRS_proc$ComplicationsNCH <- as.factor(sapply(CRS_proc$ComplicationsNCH, function(x){
w <- as.character(x)
ifelse(w == "", w <- "no", w <- "yes")
return(w) } ))- Operations is stratified as: DVP, others or no
CRS_proc$Operations <- revalue(CRS_proc$Operations,c("Trasf X DVP"="DVP","Embolizzazione"="others","III cist-ventr-st"="others","Int NCH"="others", "NO DVP"="no"))
CRS_proc[CRS_proc$Operations=="","Operations"] <- "no"
CRS_proc$Operations <- factor(CRS_proc$Operations)- OtherComplications is stratified as: Blindness, CDI (Central Diabetes Insipidus), CVD (Cardio-Vascular Disease), Infection, Neuro (Neurological), Respiratory Failure, Tracheotoemy linked, others or none
CRS_proc$OtherComplications <- as.character(CRS_proc$OtherComplications)
CRS_proc[grep("[Ii]nf.*[Cc]er",CRS_proc$OtherComplications),"OtherComplications"] <- "Cerebral Infection"
CRS_proc[grep("[Ii]ns.*[Rr]es",CRS_proc$OtherComplications),"OtherComplications"] <- "Respiratory Failure"
CRS_proc[grep("[Ii]ns.*[Cc]ar",CRS_proc$OtherComplications),"OtherComplications"] <- "Heart Failure"
CRS_proc[grep("[Cc]ec",CRS_proc$OtherComplications),"OtherComplications"] <- "Blindness"
CRS_proc[grep("([Ii]nf|ite$)",CRS_proc$OtherComplications),"OtherComplications"] <- "Infection"
CRS_proc[grep("[Dd]ia.*[Ii]ns",CRS_proc$OtherComplications),"OtherComplications"] <- "CDI"
CRS_proc$OtherComplications <- as.character(CRS_proc$OtherComplications)
CRS_proc[grep("[Ii]nf.*[Cc]er",CRS_proc$OtherComplications),"OtherComplications"] <- "Cerebral Infection"
CRS_proc[grep("[Ii]ns.*[Rr]es",CRS_proc$OtherComplications),"OtherComplications"] <- "Respiratory Failure"
CRS_proc[grep("[Ii]ns.*[Cc]ar",CRS_proc$OtherComplications),"OtherComplications"] <- "Heart Failure"
CRS_proc[grep("[Cc]ec",CRS_proc$OtherComplications),"OtherComplications"] <- "Blindness"
CRS_proc[grep("([Ii]nf|ite$)",CRS_proc$OtherComplications),"OtherComplications"] <- "Infection"
CRS_proc[grep("[Dd]ia.*[Ii]ns",CRS_proc$OtherComplications),"OtherComplications"] <- "CDI"
CRS_proc[CRS_proc$OtherComplications=="","OtherComplications"] <- "none"
CRS_proc$OtherComplications <- as.factor(CRS_proc$OtherComplications)
CRS_proc$OtherComplications <- revalue(CRS_proc$OtherComplications,c("Heart Failure"="CVD","Aritmia Card"="CVD","Arresto C-R"="CVD","Ins Renale"="CVD"))
CRS_proc$OtherComplications <- revalue(CRS_proc$OtherComplications,c("Afasia"="Neuro","Spasticità"="Neuro","Paraplegia"="Neuro"))
CRS_proc$OtherComplications <- revalue(CRS_proc$OtherComplications,c("Fist TE"="Tracheotomy linked","Fistola TE"="Tracheotomy linked","Gran Trach"="Tracheotomy linked","Granulaz Trac"="Tracheotomy linked","Stenosi trac"="Tracheotomy linked"))
CRS_proc$OtherComplications <- as.character(CRS_proc$OtherComplications)
CRS_proc[!(CRS_proc$OtherComplications %in% c("Neuro","Blindness","CDI","CVD","Tracheotomy linked","Infection","none","Respiratory Failure")),"OtherComplications"] <- "others"
CRS_proc$OtherComplications <- as.factor(CRS_proc$OtherComplications)- Fix some other yes/no values
CRS_proc$Dysautonomia <- as.factor(sapply(CRS_proc$Dysautonomia, function(x){ w <- as.character(x)
ifelse(w == "", w <- "no", w <- "yes")
return(w) } ))
CRS_proc$SNG_Adm <- as.factor(sapply(CRS_proc$SNG_Adm, function(x){ w <- as.character(x)
ifelse(w == "", w <- "no", w <- "yes")
return(w) } ))
CRS_proc$SNG_Dis <- as.factor(sapply(CRS_proc$SNG_Dis, function(x){ w <- as.character(x)
ifelse(w == "SNG", w <- "yes", w <- "no")
return(w) } ))
CRS_proc$PEG_Adm <- as.factor(sapply(CRS_proc$PEG_Adm, function(x){ w <- as.character(x)
ifelse(w == "SI", w <- "yes", w <- "no")
return(w) } ))
CRS_proc$PEG_Dis <- as.factor(sapply(CRS_proc$PEG_Dis, function(x){ w <- as.character(x)
ifelse(w == "SI", w <- "yes", w <- "no")
return(w) } ))
CRS_proc$TRACH_Adm <- revalue(CRS_proc$TRACH_Adm,c("NO"="no","SI"="yes","SIS"="yes"))
CRS_proc$TRACH_Adm <- as.factor(sapply(CRS_proc$TRACH_Adm, function(x){ w <- as.character(x)
ifelse(w == "", w <- "no", w)
return(w) } ))
CRS_proc$TRACH_Dis <- revalue(CRS_proc$TRACH_Dis,c("NO"="no","SI"="yes"))
CRS_proc$TRACH_Dis <- as.factor(sapply(CRS_proc$PEG_Adm, function(x){ w <- as.character(x)
ifelse(w == "", w <- "no", w)
return(w) } ))- ModeOfDischarge is stratified as: Deceased, Home, LDS or NCH
CRS_proc$ModeOfDischarge <- as.character(CRS_proc$ModeOfDischarge)
CRS_proc[grep("(LDS|SUAP|HOSP|RIAB)",CRS_proc$ModeOfDismission),"ModeOfDismission"] <- "LDS"
CRS_proc[grep("(118|DECESSO)",CRS_proc$ModeOfDismission),"ModeOfDismission"] <- "Deceased"
CRS_proc[grep("DOMICILIO",CRS_proc$ModeOfDismission),"ModeOfDismission"] <- "Home"
CRS_proc[grep("NCH",CRS_proc$ModeOfDismission),"ModeOfDismission"] <- "NCH"
CRS_proc$ModeOfDismission <- as.factor(CRS_proc$ModeOfDismission)- Fix CRS values
x <- as.factor(sapply(as.list(CRS_proc[,29:89]),function(x) { w <- as.character(x)
w[grep("(EMERGED|SV|SMC|DECEASED)",w)] <- NA
ifelse(w == "", w[w==""] <- NA, w)
return(w) }))
CRS_proc[,29:89] <- x
rm(x)
CRS_proc[,29:89] <- lapply(CRS_proc[,29:89] , as.integer)
x <- sapply(as.list(CRS_proc[,90:150]),function(x) { w <- as.character(x)
ifelse(w == "", w[w==""] <- NA, w)
return(w) })
CRS_proc[,90:150] <- x
rm(x)
CRS_proc[CRS_proc$Consc_1=="EMERGED","Consc_1"] <- "SMC"
CRS_proc$Consc_1 <- factor(CRS_proc$Consc_1)
CRS_proc[,90:150] <- lapply(CRS_proc[,90:150] , factor)
CRS_proc <- CRS_proc[,1:150]Data Analysis
Classification Analysis
Analysis of prediction performance of Logistic Model vs. Classification Tree for the classification of Emerged/not Emerged patients
Define Complications as CardioRespiratory or others
CRS_proc$Deceased <- as.integer(CRS_proc$Status) - 1
CRS_proc[CRS_proc$Deceased==0 & CRS_proc$GOS==1,"Deceased"] <- 1
CRS_proc$Etiology <- relevel(CRS_proc$Etiology,"TCE")
CRS_proc$Consc_1 <- relevel(CRS_proc$Consc_1,"SV")
CRS_proc$Complications <- revalue(CRS_proc$OtherComplications,c("Blindness"="others","CDI"="others","Neuro"="others","Tracheotomy linked"="others","CVD"="CardioRespiratory","Respiratory Failure"="CardioRespiratory"))
CRS_proc$Complications <- factor(CRS_proc$Complications)
CRS_proc$Complications <- relevel(CRS_proc$Complications,"none")
names(CRS_proc)[29] <- "CRS_Adm"Create Emersion variable
CRS_proc$Emersion <- ifelse(CRS_proc$LastConscStatus=="EMERGED",1,0)
CRS_proc$Emersion_fac <- factor(CRS_proc$Emersion,levels=0:1,labels=c("not Emerged","Emerged"))Best Classification Tree 10 time 10-Fold-CrossValidation
set.seed(1)
class.control = trainControl(method = "repeatedCV", number = 10, repeats = 10, summaryFunction = twoClassSummary, classProbs = T, verboseIter = FALSE, savePredictions = TRUE)# 10-fold CV
tree.CRS = train(Emersion_fac ~ Etiology + Complications + + DRS_Adm + CRS_Adm + Age + Consc_1 + PEG_Adm, data=CRS_proc, method = "rpart", trControl = class.control, tuneLength = 10)Classification Tree Plot 
Best Logistic Regression Model 10 time 10-Fold-CrossValidation
log.CRS = train(Emersion_fac ~ Etiology + Complications + DRS_Adm + CRS_Adm + Age + Consc_1, data=CRS_proc, method = "glm", trControl = class.control, tuneLength = 30)ROC analysis: Logistic model performs much better than Trees
m3 = predict(tree.CRS, CRS_proc)
predTree.CRS <- tree.CRS$pred
names(predTree.CRS)[4] <- "not_Emerged"
probsTree.CRS <- summarise(group_by(predTree.CRS, rowIndex),Emerged=mean(Emerged),not_Emerged=mean(not_Emerged))[,2:3]
predLog.CRS <- log.CRS$pred
names(predLog.CRS)[3] <- "not_Emerged"
probsLog.CRS <- summarise(group_by(predLog.CRS, rowIndex),Emerged=mean(Emerged),not_Emerged=mean(not_Emerged))[,2:3]
rocTree.CRS <- roc(response = CRS_proc$Emersion_fac,predictor = probsTree.CRS$Emerged, levels = rev(levels(CRS_proc$Emersion_fac)))
rocTree.CRS##
## Call:
## roc.default(response = CRS_proc$Emersion_fac, predictor = probsTree.CRS$Emerged, levels = rev(levels(CRS_proc$Emersion_fac)))
##
## Data: probsTree.CRS$Emerged in 75 controls (CRS_proc$Emersion_fac Emerged) > 81 cases (CRS_proc$Emersion_fac not Emerged).
## Area under the curve: 0.6436rocLog.CRS <- roc(response = CRS_proc$Emersion_fac,predictor = probsLog.CRS$Emerged, levels = rev(levels(CRS_proc$Emersion_fac)))
rocLog.CRS##
## Call:
## roc.default(response = CRS_proc$Emersion_fac, predictor = probsLog.CRS$Emerged, levels = rev(levels(CRS_proc$Emersion_fac)))
##
## Data: probsLog.CRS$Emerged in 75 controls (CRS_proc$Emersion_fac Emerged) > 81 cases (CRS_proc$Emersion_fac not Emerged).
## Area under the curve: 0.7913
Survival Analysis
Analysis of mortality
Predictive analysis of factors associated with patient mortality in the first 70 weeks after admission
Kaplan-Meier curves by baseline Consciousness state (SV vs SMC): SMC patients show higher mortaity risk (!)
Kaplan-Meier curves by Complications: Cardio-Respiratory and Infection complication seem to be associated with higher mortality risk 
Kaplan-Meier curves by Etiology: No diffeences in mortality by Etiology
Cox Regression Models for Mortality vs. Etiology, Complications and baseline CRS adjusted by Age: Cardio-Respiratory and Infection complication, as well as Vascular Etiology and baseline CRS seem to be all positively associated with higher mortality risk
summary(coxph(Surv(LastTime,Deceased) ~ Etiology + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Deceased) ~ Etiology + Age, data = CRS_proc)
##
## n= 156, number of events= 18
##
## coef exp(coef) se(coef) z Pr(>|z|)
## EtiologyAnoxia -9.949e-01 3.698e-01 1.103e+00 -0.902 0.3669
## EtiologyOthers -1.604e+01 1.077e-07 5.671e+03 -0.003 0.9977
## EtiologyVascular 1.923e-01 1.212e+00 5.981e-01 0.322 0.7478
## Age 2.736e-02 1.028e+00 1.649e-02 1.659 0.0972 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## EtiologyAnoxia 3.698e-01 2.704e+00 0.0426 3.210
## EtiologyOthers 1.077e-07 9.285e+06 0.0000 Inf
## EtiologyVascular 1.212e+00 8.250e-01 0.3754 3.914
## Age 1.028e+00 9.730e-01 0.9950 1.062
##
## Concordance= 0.663 (se = 0.074 )
## Rsquare= 0.046 (max possible= 0.641 )
## Likelihood ratio test= 7.29 on 4 df, p=0.1214
## Wald test = 5.65 on 4 df, p=0.2266
## Score (logrank) test = 6.63 on 4 df, p=0.1568summary(coxph(Surv(LastTime,Deceased) ~ CRS_Adm + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Deceased) ~ CRS_Adm + Age, data = CRS_proc)
##
## n= 156, number of events= 18
##
## coef exp(coef) se(coef) z Pr(>|z|)
## CRS_Adm 0.15719 1.17022 0.07778 2.021 0.0433 *
## Age 0.02726 1.02763 0.01432 1.904 0.0569 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## CRS_Adm 1.170 0.8545 1.0047 1.363
## Age 1.028 0.9731 0.9992 1.057
##
## Concordance= 0.651 (se = 0.074 )
## Rsquare= 0.055 (max possible= 0.641 )
## Likelihood ratio test= 8.91 on 2 df, p=0.01164
## Wald test = 9.39 on 2 df, p=0.009163
## Score (logrank) test = 9.99 on 2 df, p=0.006778summary(coxph(Surv(LastTime,Deceased) ~ Complications + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Deceased) ~ Complications + Age,
## data = CRS_proc)
##
## n= 156, number of events= 18
##
## coef exp(coef) se(coef) z Pr(>|z|)
## Complicationsothers 0.40733 1.50280 1.01625 0.401 0.688553
## ComplicationsCardioRespiratory 2.74634 15.58544 0.78018 3.520 0.000431 ***
## ComplicationsInfection 2.04063 7.69545 0.88567 2.304 0.021220 *
## Age 0.01974 1.01994 0.01456 1.356 0.175137
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## Complicationsothers 1.503 0.66542 0.2051 11.014
## ComplicationsCardioRespiratory 15.585 0.06416 3.3777 71.914
## ComplicationsInfection 7.695 0.12995 1.3563 43.664
## Age 1.020 0.98045 0.9912 1.049
##
## Concordance= 0.831 (se = 0.074 )
## Rsquare= 0.159 (max possible= 0.641 )
## Likelihood ratio test= 27.08 on 4 df, p=1.915e-05
## Wald test = 23.65 on 4 df, p=9.375e-05
## Score (logrank) test = 38.27 on 4 df, p=9.841e-08summary(coxph(Surv(LastTime,Deceased) ~ Complications + Etiology + CRS_Adm + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Deceased) ~ Complications + Etiology +
## CRS_Adm + Age, data = CRS_proc)
##
## n= 156, number of events= 18
##
## coef exp(coef) se(coef) z Pr(>|z|)
## Complicationsothers 4.795e-01 1.615e+00 1.018e+00 0.471 0.6377
## ComplicationsCardioRespiratory 3.407e+00 3.017e+01 8.181e-01 4.164 3.12e-05 ***
## ComplicationsInfection 2.141e+00 8.504e+00 8.895e-01 2.406 0.0161 *
## EtiologyAnoxia -9.493e-01 3.870e-01 1.200e+00 -0.791 0.4288
## EtiologyOthers -1.514e+01 2.657e-07 7.297e+03 -0.002 0.9983
## EtiologyVascular 1.531e+00 4.621e+00 6.673e-01 2.294 0.0218 *
## CRS_Adm 1.983e-01 1.219e+00 9.078e-02 2.184 0.0289 *
## Age -4.785e-04 9.995e-01 1.659e-02 -0.029 0.9770
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## Complicationsothers 1.615e+00 6.191e-01 0.21950 11.887
## ComplicationsCardioRespiratory 3.017e+01 3.315e-02 6.07003 149.935
## ComplicationsInfection 8.504e+00 1.176e-01 1.48745 48.616
## EtiologyAnoxia 3.870e-01 2.584e+00 0.03685 4.065
## EtiologyOthers 2.657e-07 3.763e+06 0.00000 Inf
## EtiologyVascular 4.621e+00 2.164e-01 1.24947 17.089
## CRS_Adm 1.219e+00 8.201e-01 1.02057 1.457
## Age 9.995e-01 1.000e+00 0.96753 1.033
##
## Concordance= 0.877 (se = 0.074 )
## Rsquare= 0.227 (max possible= 0.641 )
## Likelihood ratio test= 40.12 on 8 df, p=3.049e-06
## Wald test = 32.27 on 8 df, p=8.328e-05
## Score (logrank) test = 49.28 on 8 df, p=5.609e-08Survival Tree model for mortality: Complications and DRS at admission are key factors for predicting mortality
Emersion Survival Analysis
Predictive analysis of factors associated with patient emersion (risk = probability of emersion)
Kaplan-Meier curves by baseline Consciousness state (SV vs SMC): SMC patients have higher rate of Emersion 
Kaplan-Meier curves by Complications: No clear differences in rate of Emersion by complications 
Kaplan-Meier curves by Etiology: Anoxia seems to be associated with lower rate of Emersion 
Cox Regression Models for Emersion vs. Etiology, Complications and baseline CRS (Category or Numeric) adjusted by Age: After adjustemt by Age, Cardio-Respiratory and Infection complication, as well as Anoxia Etiology are associated wit lower rate of Emersion, while baseline CRS with higher rate (!).
summary(coxph(Surv(LastTime,Emersion) ~ Etiology + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Emersion) ~ Etiology + Age, data = CRS_proc)
##
## n= 156, number of events= 75
##
## coef exp(coef) se(coef) z Pr(>|z|)
## EtiologyAnoxia -1.650237 0.192004 0.739360 -2.232 0.0256 *
## EtiologyOthers 0.330159 1.391189 0.742827 0.444 0.6567
## EtiologyVascular 0.265718 1.304367 0.294277 0.903 0.3666
## Age 0.006084 1.006102 0.007738 0.786 0.4318
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## EtiologyAnoxia 0.192 5.2082 0.04508 0.8178
## EtiologyOthers 1.391 0.7188 0.32441 5.9660
## EtiologyVascular 1.304 0.7667 0.73267 2.3221
## Age 1.006 0.9939 0.99096 1.0215
##
## Concordance= 0.62 (se = 0.038 )
## Rsquare= 0.094 (max possible= 0.986 )
## Likelihood ratio test= 15.38 on 4 df, p=0.003967
## Wald test = 9.72 on 4 df, p=0.04545
## Score (logrank) test = 12 on 4 df, p=0.01732summary(coxph(Surv(LastTime,Emersion) ~ CRS_Adm + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Emersion) ~ CRS_Adm + Age, data = CRS_proc)
##
## n= 156, number of events= 75
##
## coef exp(coef) se(coef) z Pr(>|z|)
## CRS_Adm 0.265976 1.304704 0.034378 7.737 1.02e-14 ***
## Age 0.008068 1.008100 0.006277 1.285 0.199
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## CRS_Adm 1.305 0.7665 1.2197 1.396
## Age 1.008 0.9920 0.9958 1.021
##
## Concordance= 0.782 (se = 0.038 )
## Rsquare= 0.319 (max possible= 0.986 )
## Likelihood ratio test= 59.83 on 2 df, p=1.019e-13
## Wald test = 62.94 on 2 df, p=2.154e-14
## Score (logrank) test = 72.42 on 2 df, p=2.22e-16summary(coxph(Surv(LastTime,Emersion) ~ Complications + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Emersion) ~ Complications + Age,
## data = CRS_proc)
##
## n= 156, number of events= 75
##
## coef exp(coef) se(coef) z Pr(>|z|)
## Complicationsothers -0.353661 0.702113 0.281935 -1.254 0.2097
## ComplicationsCardioRespiratory -0.826442 0.437603 0.436509 -1.893 0.0583 .
## ComplicationsInfection -1.424392 0.240655 0.605745 -2.351 0.0187 *
## Age 0.007131 1.007156 0.007300 0.977 0.3286
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## Complicationsothers 0.7021 1.4243 0.40404 1.2201
## ComplicationsCardioRespiratory 0.4376 2.2852 0.18600 1.0295
## ComplicationsInfection 0.2407 4.1553 0.07341 0.7889
## Age 1.0072 0.9929 0.99285 1.0217
##
## Concordance= 0.627 (se = 0.038 )
## Rsquare= 0.084 (max possible= 0.986 )
## Likelihood ratio test= 13.73 on 4 df, p=0.008193
## Wald test = 11.27 on 4 df, p=0.02374
## Score (logrank) test = 12.39 on 4 df, p=0.01469summary(coxph(Surv(LastTime,Emersion) ~ Complications + Etiology + CRS_Adm + Age, CRS_proc))## Call:
## coxph(formula = Surv(LastTime, Emersion) ~ Complications + Etiology +
## CRS_Adm + Age, data = CRS_proc)
##
## n= 156, number of events= 75
##
## coef exp(coef) se(coef) z Pr(>|z|)
## Complicationsothers -0.288234 0.749586 0.297057 -0.970 0.3319
## ComplicationsCardioRespiratory -0.976756 0.376531 0.454219 -2.150 0.0315 *
## ComplicationsInfection -1.268838 0.281158 0.610308 -2.079 0.0376 *
## EtiologyAnoxia -1.550739 0.212091 0.746912 -2.076 0.0379 *
## EtiologyOthers -0.436387 0.646368 0.781133 -0.559 0.5764
## EtiologyVascular 0.148335 1.159901 0.314418 0.472 0.6371
## CRS_Adm 0.267545 1.306753 0.036157 7.400 1.37e-13 ***
## Age 0.003492 1.003498 0.008330 0.419 0.6751
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## exp(coef) exp(-coef) lower .95 upper .95
## Complicationsothers 0.7496 1.3341 0.41876 1.3418
## ComplicationsCardioRespiratory 0.3765 2.6558 0.15459 0.9171
## ComplicationsInfection 0.2812 3.5567 0.08501 0.9299
## EtiologyAnoxia 0.2121 4.7150 0.04906 0.9168
## EtiologyOthers 0.6464 1.5471 0.13982 2.9880
## EtiologyVascular 1.1599 0.8621 0.62631 2.1481
## CRS_Adm 1.3068 0.7653 1.21735 1.4027
## Age 1.0035 0.9965 0.98725 1.0200
##
## Concordance= 0.812 (se = 0.038 )
## Rsquare= 0.409 (max possible= 0.986 )
## Likelihood ratio test= 82.15 on 8 df, p=1.799e-14
## Wald test = 74.91 on 8 df, p=5.135e-13
## Score (logrank) test = 92.23 on 8 df, p=1.11e-16Survival Tree model for emersion: DRS and Consciousness state at admission are key factors for predicting emersion rate
Conclusions
- Logistic model performs better than trees in classificating patients as Emerged or Not;
- SMC state at admission as well as cardio-respiratory and infection complication are associated with a higher mortality risk during the admission period;
- Admission DRS, more than CRS seems to be implicated in the prediction of both mortality and emersion rate.