#Library loading

#library loading---------------------------------------------------
library(lme4)
library(ggplot2)
library(xlsx)
library(parallel)
library(sjmisc)
library(sjlabelled)
library(dplyr)
library(doParallel)
library(sjPlot)
#library(MuMIn)
library(boot)
library(table1)
library(partR2)
library(buildmer)
library(rptR)
library(furrr)
library(future)
library(glmmTMB)

#Options

CapNoLowFlow = FALSE
ExcludeDrWithFewOHCA = FALSE
ExcludeSex = FALSE

CategorizeVar = TRUE
quadratic = FALSE

computeRptR = FALSE
bootstrapSigma = FALSE

numberOfResidualSimulations = 10 #dharma
nBootPartR2 = 10 #nboot rptr and partR2
nBootPartR2Compare = 10
nbBootPrimary = 10

#Data importation & cleaning

Correction of doctor names and anonymisation has been done before the importation process

data <- read.xlsx("C:/Users/thoma/Desktop/reaclast.xlsx", 1)
nb <- nrow(data)

Delete corrupted/incomplete data

data <- filter(data, rythm >= 0)
print("Removing corrupt or unknown data about rythm")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, sexe < 3)
print("Removing corrupt or unknown data about sex")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, lowflow >= 0)
print("Removing corrupt or unknown data about low-flow")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, noflow >= 0)
print("Removing corrupt or unknown data about no-flow")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

Remove traumatic arrests

data <- filter(data, medical > 0)
print("Removing traumatic arrests")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

Options : Remove arrest linked to Dr with to few arrest (not in primary analysis but used in appendix)

if(ExcludeDrWithFewOHCA) {
  data <- data %>% group_by(dr) %>% filter(n() >= 40)
  
  data <- data %>% ungroup
  
  print("Removing arrests linked with Dr with less than 3 arrests")
  print(nrow(data))
  print(nb - nrow(data))
  nb <- nrow(data)
}

#Data preprocessing

Creation of binary score for the presence or absence for each observed medical history

data <- data %>% mutate("ATCD.Cardiovasculaire" = replace(ATCD.Cardiovasculaire, ATCD.Cardiovasculaire == -1, 0))
data <- data %>% mutate("ATCD.Respiratoire" = replace(ATCD.Respiratoire, ATCD.Respiratoire == -1, 0))
data <- data %>% mutate("ATCD.Diabete" = replace(ATCD.Diabete, ATCD.Diabete == -1, 0))
data <- data %>% mutate("ATCD.EOL" = replace(ATCD.EOL, ATCD.EOL == -1, 0))
data <- data %>% mutate("ATCD.Other" = replace(ATCD.Other, ATCD.Other == -1, 0))

Adjust sex to binary value

#Transforming men = 1 et women = 2 in men = 0 et women = 1
data <- data %>% mutate("sexe" = data$sexe-1)

Regroup asystole and pulseless electrical activity (two group remaining Asystole/PEA and other rythms)

data <- data %>% mutate("Asystolie/RSP" = replace(chocable, chocable == 3, 1))
data <- data %>% mutate("Asystolie/RSP" = data$`Asystolie/RSP`-1)

Calculate low-flow BEFORE SAMU-MICU arrival

#Creation of smurdelay useful for calculation just below
data <- data %>% mutate(delaysmur = difftime(data$smurHour, data$acHour, units ="mins"))
data <- data %>% mutate(delaysmur = ifelse(data$delaysmur > 720, difftime((as.Date(data$acHour)+1), data$smurHour, units ="mins"), data$delaysmur))
data <- data %>% mutate(delaysmur = ifelse(data$delaysmur < -720, difftime((as.Date(data$smurHour)+1), data$acHour, units ="mins"), data$delaysmur))

#Creation of lowflow before SMUR arrival
data <- data %>% mutate(low.flow = ifelse(pmin(data$delaysmur-data$noflow, data$lowflow) > 0, pmin(data$delaysmur-data$noflow, data$lowflow), data$lowflow))

Calculate the existence of a ROSC before SAMU-MICU arrival

data <- data %>% mutate(racs = ifelse((data$racsdelay < data$delaysmur) & (data$racsdelay != -1), 1, 0))

Creation of TOR column (termination of resuscitation)

data <- data %>% mutate(TOR = ifelse(!(data$reasmur == 1 | data$transport == 1), 1, 0))

Creation of data labels

label(data$age)   <- "Age"
label(data$sexe)   <- "Sex"
label(data$noflow)   <- "No-flow"
label(data$witness)   <- "Witness"
label(data$dr)   <- "Doctor"
label(data$low.flow)   <- "Low-flow"
label(data$racs)   <- "ROSC before SMUR"
label(data$TOR)   <- "Termination of ressusitation"
label(data$`ATCD.Cardiovasculaire`)   <- "Cardiac history"
label(data$`ATCD.Respiratoire`)   <- "Respiratory history"
label(data$`ATCD.EOL`)   <- "End of life / dependency"
label(data$`ATCD.Diabete`)   <- "Diabetes"
label(data$`ATCD.Other`)   <- "Other medical issue including oncologic"
label(data$`Asystolie/RSP`)   <- "Rythm"

print(paste0("Included cardiac arrests : ", nrow(data)))

Graph of continuous variables

hist(data$age, breaks = 50)
dataCapped <- data
dataCapped <- dataCapped %>% mutate("noflow" = replace(noflow, noflow >= 100, 100))
dataCapped <- dataCapped %>% mutate("low.flow" = replace(low.flow, low.flow >= 100, 100))
hist(dataCapped$noflow, breaks = 100)
hist(dataCapped$low.flow, breaks = 100)

Option : Capping low flow and no-flow

if(CapNoLowFlow) {
  dataCapped <- data
  dataCapped <- dataCapped %>% mutate("noflow" = replace(noflow, noflow >= 100, 100))
  dataCapped <- dataCapped %>% mutate("low.flow" = replace(low.flow, low.flow >= 100, 100))
  data <- dataCapped
}

Option : Categorize variables

dataSv <- data
if(CategorizeVar) {
  datacat <- data
  datacat$age <- cut(data$age, 
                   breaks=c(-Inf, 65, 75, 85, Inf), 
                   labels=c("< 65 years","65-75 years","75-85 years", "85 years and beyond"))
  datacat$noflow <- cut(data$noflow, 
                   breaks=c(-Inf, 1, 5, 10, 20, Inf), 
                   labels=c("< 1 minute", "1 to 5 minutes", "5 to 10 minutes","10 to 20 minutes", "20 minutes and beyond"))
  datacat$low.flow <- cut(data$low.flow, 
                   breaks=c(-Inf, 10, 20, 40, Inf), 
                   labels=c("< 10 minutes","10 to 20 minutes","20 to 40 minutes", "40 minutes and beyond"))
  data <- datacat
}

#Table1 : charasteristics of population Creation of table1 : Characteristics of population

if(CategorizeVar) {
  datacat <- dataSv
  datacat$age <- cut(dataSv$age, 
                   breaks=c(-Inf, 65, 75, 85, Inf), 
                   labels=c("Age ≤ 65","Age in [66, 75]","Age in [76, 85]", "Age > 85"))
  datacat$noflow <- cut(dataSv$noflow, 
                   breaks=c(-Inf, 1, 5, 10, 20, Inf), 
                   labels=c("No-flow ≤ 1", "No-flow in [2, 5]", "No-flow in [6, 10]","No-flow in [11, 20]", "No-flow > 20"))
  datacat$low.flow <- cut(dataSv$low.flow, 
                   breaks=c(-Inf, 10, 20, 40, Inf), 
                   labels=c("Low-flow ≤ 10","Low-flow in [11, 20]","Low-flow in [21, 40]", "Low-flow > 40"))
  dataT <- datacat
} else {
  dataT <- data
}

library(table1)
#dataT is a copy of data used for table display


render <- function(x, name, missing) {
  if (!is.numeric(x)) return(render.cat(x))
  else if (name == "age") {
    what <- switch(name,
        age = "Mean (SD)")
    parse.abbrev.render.code(c("", what))(x)
    
    # what <- switch(name,
    #     age = "Mean (SD)",
    #     low.flow  = "Median (IQR)",
    #     noflow  = "Median (IQR)")
    # parse.abbrev.render.code(c("", what))(x)
  } else {
    with(stats.apply.rounding(stats.default(x), digits=2), c("",
          "Median (IQR)"=sprintf("%s [%s-%s]", MEDIAN, Q1, Q3)))
  }
}


render.cat <- function(x) {
    c("", sapply(stats.default(x), function(y) with(y,
        sprintf("%d (%0.0f %%)", FREQ, PCT))))
}
  
dataT$TOR <- 
  factor(dataT$TOR, 
         levels=c(0,1),
         labels=c("Advanced life support", # Reference
                  "Termination of resuscitation"))



dataT$sexe <- 
  factor(dataT$sexe, levels=c(1,0),
         labels=c("Men", 
                  "Women"))

units(dataT$age) <- "Years"
units(dataT$noflow) <- "Minutes"
units(dataT$low.flow) <- "Minutes"

dataT$`Asystolie/RSP` <- 
  factor(dataT$`Asystolie/RSP`, levels=c(1,0),
         labels=c("Asystole/PEA", 
                  "VF/VT/SA"))

dataT$witness <- 
  factor(dataT$witness, levels=c(1),
         labels=c("Present"))

dataT$racs <- 
  factor(dataT$racs, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Cardiovasculaire` <- 
  factor(dataT$`ATCD.Cardiovasculaire`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.EOL` <- 
  factor(dataT$`ATCD.EOL`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Respiratoire` <- 
  factor(dataT$`ATCD.Respiratoire`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Diabete` <- 
  factor(dataT$`ATCD.Diabete`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Other` <- 
  factor(dataT$`ATCD.Other`, levels=c(1),
         labels=c("Present"))

label(dataT$age)   <- "Age"
label(dataT$sexe)   <- "Sex"
label(dataT$noflow)   <- "No-flow"
label(dataT$`witness`)   <- "Witness"
label(dataT$dr)   <- "Doctor"
label(dataT$low.flow)   <- "Low-flow"
label(dataT$racs)   <- "ROSC before SAMU-MICU arrival"
label(dataT$TOR)   <- "Termination of resuscitation"
label(dataT$`ATCD.Cardiovasculaire`)   <- "Cardiac history"
label(dataT$`ATCD.Respiratoire`)   <- "Respiratory history"
label(dataT$`ATCD.EOL`)   <- "End of life / dependancy"
label(dataT$`ATCD.Diabete`)   <- "Diabetes history"
label(dataT$`ATCD.Other`)   <- "Other medical issue including oncologic history"
label(dataT$`Asystolie/RSP`)   <- "Rhythm"

table <- table1( ~ age + sexe + noflow + low.flow + `Asystolie/RSP` + witness + racs + `ATCD.Cardiovasculaire` + `ATCD.Respiratoire` + `ATCD.EOL` + `ATCD.Diabete` + `ATCD.Other` | TOR, data=dataT,  topclass="Rtable1-zebra", overall="Total", render=render)

#Mean and median of OHCA by doctor

median(aggregate(data$age, by=list(data$dr), FUN=length)$x)
mean(aggregate(data$age, by=list(data$dr), FUN=length)$x)

t.test(aggregate(data$age, by=list(data$dr), FUN=length)$x)
quantile(aggregate(data$age, by=list(data$dr), FUN=length)$x)

#Creating formulae

formula = `TOR` ~ `racs` + `noflow` + `Asystolie/RSP` + `low.flow` + `sexe` + `witness` + `age` + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other` + (1|`dr`)

if(quadratic) {
  formulaNonLinear = `TOR` ~ `racs` + `noflow` + I(`noflow`^2) + `Asystolie/RSP` + `low.flow` + I(`low.flow`^2) + `sexe` + `witness` + `age` + I(`age`^2) + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other` + (1|`dr`)
}

formulaGLM = `TOR` ~ `racs` + `noflow` + `Asystolie/RSP` + `low.flow` + `sexe` + `witness` + `age` + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other`

if(ExcludeSex) {
  formula = remove.terms(formula, "sexe")
  formulaGLM = remove.terms(formulaGLM, "sexe")
  if(quadratic) {
    formulaNonLinear = remove.terms(formulaNonLinear, "sexe")
  }
}

#Implementation of GLMM and GLM models (FitGLMM and ReducedGLM)

FitGLMM : - Binomial family since outcome is binary - Dr effect (random effect) implemented without random slope, only intersect.

library(lme4)
label(data$TOR)   <- "TOR"
FitGLMM <- glmer(formula, family=binomial, data=data)

if(quadratic)
  FitGLMMnonLinear <- glmer(formulaNonLinear, family=binomial, data=data)

ReducedGLM - Same as fit GLMM without Dr effect (no random effect)

label(data$TOR)   <- "TOR (without random effect)"
ReducedGLM <- glm(formulaGLM, family=binomial, data=data)
label(data$TOR)   <- "TOR"

#Fitting concurrent models

We fit concurrent GLMM models with some predictors missing to assess the impact.

label(data$TOR)   <- "TOR (without medical history / end of life)"
formulaWithoutHistoryEOL = remove.terms(formula, "`ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL`+ `ATCD.Respiratoire` + `ATCD.Other`")
glmmWithoutHistoryEOL <- glmer(formulaWithoutHistoryEOL, family=binomial, data=data)
label(data$TOR)   <- "TOR (without medical history)"
formulaWithoutHistory = remove.terms(formula, "ATCD.EOL")
glmmWithoutHistory <- glmer(formulaWithoutHistory, family=binomial, data=data)
label(data$TOR)   <- "TOR (without no-flow)"
formulaWithoutNF = remove.terms(formula, "noflow")
glmmWithoutNF <- glmer(formulaWithoutNF, family=binomial, data=data)
label(data$TOR)   <- "TOR (without low-flow)"
formulaWithoutLF = remove.terms(formula, "low.flow")
glmmWithoutLF <- glmer(formulaWithoutLF, family=binomial, data=data)
label(data$TOR)   <- "TOR (without rythm and ROSC)"
formulaWithoutRythm = remove.terms(formula, "`racs` + `Asystolie/RSP`")
glmmWithoutRythm <- glmer(formulaWithoutRythm, family=binomial, data=data)
label(data$TOR)   <- "TOR (without sex)"
formulaWithoutSex = remove.terms(formula, "sexe")
glmmWithoutSex <- glmer(formulaWithoutSex, family=binomial, data=data)
label(data$TOR)   <- "TOR (without age)"
formulaWithoutAge = remove.terms(formula, "age")
glmmWithoutAge <- glmer(formulaWithoutAge, family=binomial, data=data)
label(data$TOR)   <- "TOR (without witness)"
formulaWithoutWitness = remove.terms(formula, "witness")
glmmWithoutWitness <- glmer(formulaWithoutWitness, family=binomial, data=data)

#Checking GLMM model asumptions

s <- summary(FitGLMM, correlation = TRUE)
s
summary(ReducedGLM, correlation = TRUE)
confint(FitGLMM, method = "Wald")

Using Dharma which create visualy interpretable residuals by simulating many response and and using the cumulative density for the observed data point to create residuals (more information in package vignette). (Refit = T pour un bootstrap paramétrique)

From this graph, we check : normality of residus, hemegeneity of residus for the whole package and also for each individual predictors

library(DHARMa)
#simFit <- simulateResiduals(fittedModel = FitGLMM, n = numberOfResidualSimulations, refit = T)
simFit <- simulateResiduals(fittedModel = FitGLMM, n = numberOfResidualSimulations)
plot(simFit)

simReduced <- simulateResiduals(fittedModel = ReducedGLM, n = numberOfResidualSimulations)
plot(simReduced)

plotResiduals(simulationOutput = FitGLMM, form=data$low.flow)
plotResiduals(simulationOutput = FitGLMM, form=data$noflow)
plotResiduals(simulationOutput = FitGLMM, form=data$racs)
plotResiduals(simulationOutput = FitGLMM, form=data$`Asystolie/RSP`)
plotResiduals(simulationOutput = FitGLMM, form=data$`sexe`)
plotResiduals(simulationOutput = FitGLMM, form=data$`witness`)
plotResiduals(simulationOutput = FitGLMM, form=data$`age`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Cardiovasculaire`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Diabete`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.EOL`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Respiratoire`)
plotResiduals(simulationOutput = FitGLMM)
plotResiduals(simulationOutput = ReducedGLM)
#Quantile normalization of scaled residuals to visualize them in a 
residualsNorm <- residuals(simFit, quantileFunction = qnorm, outlierValues = c(0,1))

h <- hist(residualsNorm, breaks = 40, density = 10,
          col = "black", xlab = "Scaled residues (after quantile normalisation)", ylab = "Number of observation", main = "Distribution of residues") 
xfit <- seq(min(residualsNorm), max(residualsNorm), length = 40) 
yfit <- dnorm(xfit, mean = mean(residualsNorm), sd = sd(residualsNorm)) 
yfit <- yfit * diff(h$mids[1:2]) * length(residualsNorm) 

lines(xfit, yfit, col = "red", lwd = 2)
plot(h)
plot_model(FitGLMM, type="diag", sort.est=TRUE,
           vline.color="#A9A9A9", dot.size=1.5,
           show.values=T, value.offset=.2)
save_plot("qqplot.jpg", dpi = 500, width = 20, height = 15)

Here we look at the variance-covariance table which informs us on linear correlations

print(s, correlation = TRUE)
summary(ReducedGLM, correlation = TRUE)
if(computeRptR) {
  rptRadjust100 <- rptR::rpt(formula, "dr", data = data, datatype = "Binary", nboot = nBootPartR2, npermut = nBootPartR2, parallel = TRUE, ncores = 6, adjusted = FALSE)
  rptRadjust100
  
  sprintf("Part of explained variance of Dr effect for full model : %s %%", trunc(rptRadjust100[["R"]][["dr"]][2]*10^4)/10^2)
}
if(computeRptR) {
  rptRadjustWithoutLF <- rptR::rpt(formulaWithoutLF, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without low flow) : %s %%", trunc(rptRadjustWithoutLF[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutNF <- rptR::rpt(formulaWithoutNF, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without no flow) : %s %%", trunc(rptRadjustWithoutNF[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutHistoryEOL <- rptR::rpt(formulaWithoutHistoryEOL, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without history) : %s %%", trunc(rptRadjustWithoutHistoryEOL[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutEOL <- rptR::rpt(formulaWithoutHistory, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without end of life) : %s %%", trunc(rptRadjustWithoutEOL[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutRythm <- rptR::rpt(formulaWithoutRythm, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without rythm) : %s %%", trunc(rptRadjustWithoutRythm[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutSex <- rptR::rpt(formulaWithoutSex, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without sex) : %s %%", trunc(rptRadjustWithoutSex[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutAge <- rptR::rpt(formulaWithoutAge, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without age) : %s %%", trunc(rptRadjustWithoutAge[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutWitness <- rptR::rpt(formulaWithoutWitness, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without witness) : %s %%", trunc(rptRadjustWithoutWitness[["R"]][["dr"]][2]*10^4)/10^2)
}

#Checking FitGLMM model performance

Comparing all models via AIC, pseudo-R2 to check for : - random effect independance to fixed effects choice - Pertinence of chosen fixed variables

anova(FitGLMM, glmmWithoutHistoryEOL, glmmWithoutHistory, glmmWithoutNF, glmmWithoutLF, glmmWithoutRythm, glmmWithoutSex, glmmWithoutAge, glmmWithoutWitness, ReducedGLM)

if(computeRptR) {
  R2cFull <- partR2(FitGLMM, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (full) : %s %%", trunc(R2cFull[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutLF <- partR2(glmmWithoutLF, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without low flow) : %s %%", trunc(R2cWithoutLF[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutNF <- partR2(glmmWithoutNF, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without no flow) : %s %%", trunc(R2cWithoutNF[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutEOL <- partR2(glmmWithoutHistory, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without end of life) : %s %%", trunc(R2cWithoutEOL[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutHistoryEOL <- partR2(glmmWithoutHistoryEOL, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without history and end of life) : %s %%", trunc(R2cWithoutHistoryEOL[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutRythm <- partR2(glmmWithoutRythm, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without rythm) : %s %%", trunc(R2cWithoutRythm[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutSex <- partR2(glmmWithoutSex, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without sex) : %s %%", trunc(R2cWithoutSex[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutAge <- partR2(glmmWithoutAge, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without age) : %s %%", trunc(R2cWithoutAge[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutWitness <- partR2(glmmWithoutWitness, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without witness) : %s %%", trunc(R2cWithoutWitness[["R2"]][["estimate"]][1]*10^4)/10^2)
}

compareTab = tab_model(FitGLMM, glmmWithoutHistoryEOL, glmmWithoutHistory, glmmWithoutNF, glmmWithoutLF, glmmWithoutRythm, glmmWithoutSex, glmmWithoutAge, glmmWithoutWitness, ReducedGLM, p.style = "stars", pred.labels = c("Intercept","ROSC","No-flow", "Asystole/PEA", "Low-flow", "Sex", "Witness", "Age", "Cardiac history", "Diabetes", "End of life / dependancy", "Respiratory history", "Other medical issue including oncologic history"))

#RESULTS #Primary endpoint

#Evaluating the significance of doctor effect through comparison of ReducedGLM and FitGLMM through wilk’s test (Maximum likelihood ratio test)

anova(FitGLMM,ReducedGLM, test="LRT")

#Computing sd of random effect and it’s p-value with parametric bootstrap

Computing primary objective and it’s confidence interval. Parametric bootstrap between FitGLMM et ReducedGLM.

if(bootstrapSigma){
  library(doParallel)
  # Bootstrap iterations
  nsamples <- nbBootPrimary
  
  # Multithreading
  ncores=6
  cl = makeCluster(ncores)
  registerDoParallel(cl)
  
  # Vector of random effect variance of FitGLMM on data simulated through ReducedGLM
  estimated_var<- c() 
  
  # Bootstrap loop
  estimated_var = foreach(i=1:nsamples) %dopar% { 
    library(lme4)
    
    # Creation of simulated data from experimental data by replacing
    sim_data <- data[sample(1:nrow(data), nrow(data), replace=TRUE), ]
    
    # Simulation of response (TOR) for simulated data using ReducedGLM and adding a noise similar to the one found in experimental data
    sim_reasmur <- simulate(ReducedGLM, nsim = nrow(data), newdata=sim_data)
    
    # Integration of the responses to simulated data frame 
    sim_data["TOR"] <- sim_reasmur
    # We apply a model similar in every way to FitGLMM called BootGlmm and we fit it on simdata
    BootGlmm <- glmer(formula, family=binomial, data=sim_data)
  
    # We add sd of random effect to estimated_var
    var <- as.data.frame(VarCorr(BootGlmm))["sdcor"][1]
    var
  }
  
  # Threads closing
  stopCluster(cl)
  
  estimated_var_list <- estimated_var
  estimated_var <- as.numeric(unlist(estimated_var))
  
  # Statistic test between sd of random effect of experimental data and simulated data
  testEffectDr <- VarCorr(FitGLMM) > quantile(estimated_var,.95) 
  testEffectDr["dr"]
  
  # Extraction of quantile of sd
  quantile(estimated_var, probs = c(0.05, 0.95))
  
  # Calculation of p-value of sd of random effect
  pvalueSdDevDrEffect<- mean(VarCorr(FitGLMM) < estimated_var) 
  pvalueSdDevDrEffect[1]
  
  qplot(estimated_var, geom="histogram")
}

#Secondary endpoint

#OR of fixed effects (standardized for 1 increase of SD value)

#OR_LF = exp(coef(summary(FitGLMM))["low.flow", "Estimate"]*sqrt(var(data$low.flow)))
#exp(confint(FitGLMM))
plot_model(FitGLMM, sort.est = TRUE, show.values = TRUE, value.offset = .3, type = "std")

#OR of random effect #point estimate for the odds ratio can be obtained for a doctor one SD above the mean, relative to a doctor at the mean, by exponentiating the value of sigma

#exp(sd(ranef(FitGLMM)$dr[1, ]))

#OR sans interval de confiance

racsOR <- exp(coef(summary(FitGLMM))["racs", "Estimate"])#*sqrt(var(dataB$racs)))
sprintf("OR Racs : %s", racsOR)
  
AsystolieOR <- exp(coef(summary(FitGLMM))["`Asystolie/RSP`", "Estimate"])
sprintf("OR asystolie : %s", AsystolieOR)

witnessOR <- exp(coef(summary(FitGLMM))["witness", "Estimate"])
sprintf("OR witness : %s", witnessOR)

ATCDcardioOR <- exp(coef(summary(FitGLMM))["ATCD.Cardiovasculaire", "Estimate"])
sprintf("OR atcd cardio : %s", ATCDcardioOR)

ATCDdiaOR <- exp(coef(summary(FitGLMM))["ATCD.Diabete", "Estimate"])
sprintf("OR atcd dia : %s", ATCDdiaOR)

ATCD_EOL_OR <- exp(coef(summary(FitGLMM))["ATCD.EOL", "Estimate"])
sprintf("OR EOL : %s", ATCD_EOL_OR)

ATCDrespOR <- exp(coef(summary(FitGLMM))["ATCD.Respiratoire", "Estimate"])
sprintf("OR atcd resp : %s", ATCDrespOR)

ATCDotherOR <- exp(coef(summary(FitGLMM))["ATCD.Other", "Estimate"])
sprintf("OR atcd other : %s", ATCDotherOR)

sexeOR <- 0
  if(ExcludeSex == FALSE)
    sexeOR <- exp(coef(summary(FitGLMM))["sexe", "Estimate"])

sprintf("OR sex : %s", sexeOR)

if(CategorizeVar) {
  LF_OR10 <- exp(coef(summary(FitGLMM))["low.flow10 to 20 minutes", "Estimate"])
  LF_OR20 <- exp(coef(summary(FitGLMM))["low.flow20 to 40 minutes", "Estimate"])
  LF_OR40 <- exp(coef(summary(FitGLMM))["low.flow40 minutes and beyond", "Estimate"])
  print(sprintf("OR lf < 20 : %s", LF_OR10))
  print(sprintf("OR lf < 40 : %s", LF_OR20))
  print(sprintf("OR lf > 40 : %s", LF_OR40))
  
  noflowOR1 <- exp(coef(summary(FitGLMM))["noflow1 to 5 minutes", "Estimate"])
  noflowOR5 <- exp(coef(summary(FitGLMM))["noflow5 to 10 minutes", "Estimate"])
  noflowOR10 <- exp(coef(summary(FitGLMM))["noflow10 to 20 minutes", "Estimate"])
  noflowOR20 <- exp(coef(summary(FitGLMM))["noflow20 minutes and beyond", "Estimate"])
  print(sprintf("OR nf > 1 : %s", noflowOR1))
  print(sprintf("OR nf > 5 : %s", noflowOR5))
  print(sprintf("OR nf > 10 : %s", noflowOR10))
  print(sprintf("OR nf > 20 : %s", noflowOR20))
  
  ageOR65 <- exp(coef(summary(FitGLMM))["age65-75 years", "Estimate"])
  ageOR75 <- exp(coef(summary(FitGLMM))["age75-85 years", "Estimate"])
  ageOR85 <- exp(coef(summary(FitGLMM))["age85 years and beyond", "Estimate"])
  print(sprintf("OR age > 65 : %s", ageOR65))
  print(sprintf("OR age > 75 : %s", ageOR75))
  print(sprintf("OR age > 85 : %s", ageOR85))
} else {
  LF_OR <- exp(coef(summary(FitGLMM))["low.flow", "Estimate"]*sqrt(var(dataB$low.flow)))
  ageOR <- exp(coef(summary(FitGLMM))["age", "Estimate"]*sqrt(var(dataB$age)))
  noflowOR <- exp(coef(summary(FitGLMM))["noflow", "Estimate"]*sqrt(var(dataB$noflow)))
  print(sprintf("OR lf : %s", LF_OR))
  print(sprintf("OR nf : %s", noflowOR))
  print(sprintf("OR age : %s", ageOR))
}

#bootstrapped confidence interval

nbBootPrimary = 1000

library(doParallel)
# Bootstrap iterations
nsamples <- nbBootPrimary

# Multithreading
ncores=6
cl = makeCluster(ncores)
registerDoParallel(cl)

# Define the bootstrapping function
estimated_var <- function(dataB, index) {
  sim_data <- dataB[sample(1:nrow(dataB), nrow(dataB), replace=TRUE), ]
  if(quadratic)
    BootGLMM <- glmer(formulaNonLinear, family=binomial, data=sim_data)
  else
    BootGLMM <- glmer(formula, family=binomial, data=sim_data)
  
  #OR of mixed effect (dr)
  Mixed <- exp(sd(ranef(BootGLMM)$dr[,]))
  
  #OR of fixed effects
  racsOR <- exp(coef(summary(BootGLMM))["racs", "Estimate"])#*sqrt(var(dataB$racs)))
  AsystolieOR <- exp(coef(summary(BootGLMM))["`Asystolie/RSP`", "Estimate"])#*sqrt(var(dataB$`Asystolie/RSP`)))
  witnessOR <- exp(coef(summary(BootGLMM))["witness", "Estimate"])#*sqrt(var(dataB$witness)))
  ATCDcardioOR <- exp(coef(summary(BootGLMM))["ATCD.Cardiovasculaire", "Estimate"])#*sqrt(var(dataB$ATCD.Cardiovasculaire)))
  ATCDdiaOR <- exp(coef(summary(BootGLMM))["ATCD.Diabete", "Estimate"])#*sqrt(var(dataB$ATCD.Diabete)))
  ATCD_EOL_OR <- exp(coef(summary(BootGLMM))["ATCD.EOL", "Estimate"])#*sqrt(var(dataB$ATCD.EOL)))
  ATCDrespOR <- exp(coef(summary(BootGLMM))["ATCD.Respiratoire", "Estimate"])#*sqrt(var(dataB$ATCD.Respiratoire)))
  ATCDotherOR <- exp(coef(summary(BootGLMM))["ATCD.Other", "Estimate"])#*sqrt(var(dataB$ATCD.Other)))
  
  
  
  sexeOR <- 0
  if(ExcludeSex == FALSE)
    sexeOR <- exp(coef(summary(BootGLMM))["sexe", "Estimate"])#*sqrt(var(dataB$sexe)))
  
  if(CategorizeVar) {
    LF_OR10 <- exp(coef(summary(BootGLMM))["low.flow10 to 20 minutes", "Estimate"])
    LF_OR20 <- exp(coef(summary(BootGLMM))["low.flow20 to 40 minutes", "Estimate"])
    LF_OR40 <- exp(coef(summary(BootGLMM))["low.flow40 minutes and beyond", "Estimate"])
    
    noflowOR1 <- exp(coef(summary(BootGLMM))["noflow1 to 5 minutes", "Estimate"])
    noflowOR5 <- exp(coef(summary(BootGLMM))["noflow5 to 10 minutes", "Estimate"])
    noflowOR10 <- exp(coef(summary(BootGLMM))["noflow10 to 20 minutes", "Estimate"])
    noflowOR20 <- exp(coef(summary(BootGLMM))["noflow20 minutes and beyond", "Estimate"])
    
    ageOR65 <- exp(coef(summary(BootGLMM))["age65-75 years", "Estimate"])
    ageOR75 <- exp(coef(summary(BootGLMM))["age75-85 years", "Estimate"])
    ageOR85 <- exp(coef(summary(BootGLMM))["age85 years and beyond", "Estimate"])
  } else {
    
    LF_OR <- exp(coef(summary(BootGLMM))["low.flow", "Estimate"]*sqrt(var(dataB$low.flow)))
    ageOR <- exp(coef(summary(BootGLMM))["age", "Estimate"]*sqrt(var(dataB$age)))
    noflowOR <- exp(coef(summary(BootGLMM))["noflow", "Estimate"]*sqrt(var(dataB$noflow)))
    
    noflowORnoSD <- exp(coef(summary(BootGLMM))["noflow", "Estimate"])
    LF_ORnoSD <- exp(coef(summary(BootGLMM))["low.flow", "Estimate"])
    ageORnoSD <- exp(coef(summary(BootGLMM))["age", "Estimate"])
  }
  
if(CategorizeVar) {
      Fixed <- data.frame("racs" = racsOR, "noflow1 to 5 minutes" = noflowOR1, "noflow5 to 10 minutes" = noflowOR5, "noflow10 to 20 minutes" = noflowOR10, "noflow20 minutes and beyond" = noflowOR20, "Asystolie/RSP" = AsystolieOR, "low.flow10 to 20 minutes" = LF_OR10, "low.flow20 to 40 minutes" = LF_OR20, "low.flow40 minutes and beyond" = LF_OR40, "sexe"= sexeOR, "witness" = witnessOR, "age65-75 years"= ageOR65, "age75-85 years"= ageOR75, "age85 years and beyond"= ageOR85, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
      
      NoSD <- Fixed
      
} else {
  #named vector of fixed effects OR
    Fixed <- data.frame("racs" = racsOR, "noflow" = noflowOR, "Asystolie/RSP" = AsystolieOR, "low.flow" = LF_OR, "sexe"= sexeOR, "witness" = witnessOR, "age"= ageOR, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
    
    NoSD <- data.frame("racs" = racsOR, "noflow" = noflowORnoSD, "Asystolie/RSP" = AsystolieOR, "low.flow" = LF_ORnoSD, "sexe"= sexeOR, "witness" = witnessOR, "age"= ageORnoSD, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
}

  
  
  
  #Fixed <- exp(coef(summary(BootGLMM))[, "Estimate"]*sqrt(var(dataB$low.flow)))
  output<-list(Mixed,Fixed, NoSD)
  return(output)
}

# Bootstrapped estimates
#results <- boot(data=sim_data, statistic=estimated_var, R=nbBootPrimary, progress="text") #nbBootPrimary

# Initialize a vector to store the bootstrapped estimates
resultsMixed <- numeric(nbBootPrimary)

if(CategorizeVar) {
  resultsFixed <- data.frame("racs" = numeric(), "noflow1 to 5 minutes" = numeric(), "noflow5 to 10 minutes" = numeric(), "noflow10 to 20 minutes" = numeric(), "noflow20 minutes and beyond" = numeric(), "Asystolie/RSP" = numeric(), "low.flow10 to 20 minutes" = numeric(), "low.flow20 to 40 minutes" = numeric(), "low.flow40 minutes and beyond" = numeric(), "sexe"= numeric(), "witness" = numeric(), "age65-75 years"= numeric(), "age75-85 years"= numeric(), "age85 years and beyond"= numeric(), "ATCD.Cardiovasculaire"= numeric(), "ATCD.Diabete" = numeric(), "ATCD.EOL" = numeric(), "ATCD.Respiratoire" = numeric(), "ATCD.Other"= numeric())
  
  resultsNoSD <- resultsFixed
} else {
  resultsFixed <- data.frame("racs"= numeric(), "noflow"= numeric(), "Asystolie/RSP"= numeric(), "low.flow"= numeric(), "sexe"= numeric(), "witness"= numeric(), "age"= numeric(), "ATCD.Cardiovasculaire"= numeric(), "ATCD.Diabete"= numeric(), "ATCD.EOL"= numeric(), "ATCD.Respiratoire"= numeric(), "ATCD.Other"= numeric())
  resultsNoSD <- resultsFixed
}
#colnames(resultsFixed) = c("Intercept", "racs", "noflow", "Asystolie/RSP", "low.flow", "sexe", "witness", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")

  if(ExcludeSex) {
    resultsFixed <- select(resultsFixed, -sexe)
    resultsNoSD <- select(resultsNoSD, -sexe)
  }

resultsB<- c() 

# Bootstrapped estimates with progress counter
resultsB = foreach(i=1:nbBootPrimary) %dopar% { 
#for (i in 1:nbBootPrimary) {
  library(lme4)
  index <- sample(1:nrow(data), replace=TRUE)
  results <- estimated_var(data, index)
  
  
  # Display the progress counter
  # if (i %% 25 == 0) {
  #   cat("Iteration", i, "of ", nbBootPrimary, "\n time : ", Sys.time())
  # }
  
  return(results)
}

# resultsMixed[i] <- results[1]
# resultsFixed <- rbind(resultsFixed, data.frame(as.list(results[2][[1]])))
# resultsNoSD <- rbind(resultsNoSD, data.frame(as.list(results[3][[1]])))

# Threads closing
  stopCluster(cl)

resultsMixed <- c()
for (i in 1:length(resultsB)) {
  resultsMixed[i] = resultsB[[i]][1]
  resultsFixed <- rbind(resultsFixed, data.frame(as.list(resultsB[[i]][2][[1]])))
  resultsNoSD <- rbind(resultsNoSD, data.frame(as.list(resultsB[[i]][3][[1]])))
}


  if(ExcludeSex) {
    resultsFixed <- select(resultsFixed, -sexe)
    resultsNoSD <- select(resultsNoSD, -sexe)
  }
resultsMixed <- as.numeric(unlist(resultsMixed))

mean(resultsMixed)
median(resultsMixed)
quantile(resultsMixed, probs = c(0.025, 0.975))
hist(resultsMixed, main = "Frequency Plot of Bootstrapped Means", xlab = "Bootstrapped Means", ylab = "Frequency", breaks = 100)
statsBootstrap <- data.frame(Predictor = "Dr Effect", OR = mean(resultsMixed), Lower = quantile(resultsMixed, probs = c(0.025, 0.975))["2.5%"][[1]], Upper = quantile(resultsMixed, probs = c(0.025, 0.975))["97.5%"][[1]])

for (i in colnames(resultsFixed)){
  #print(resultsFixed[[i]])
  #as.numeric(unlist(resultsMixed))
  print(i)
  cat("\n")

  print(mean(resultsFixed[[i]]))
  print(median(resultsFixed[[i]]))
  print(quantile(resultsFixed[[i]], probs = c(0.025, 0.975)))
  cat("\n\n")
  
  statsBootstrap <- statsBootstrap %>% 
   add_row(Predictor = i, OR = mean(resultsFixed[[i]]), Lower = quantile(resultsFixed[[i]], probs = c(0.025, 0.975))["2.5%"][[1]], Upper = quantile(resultsFixed[[i]], probs = c(0.025, 0.975))["97.5%"][[1]])
}

print("NON NORMALIZED")

for (i in colnames(resultsNoSD)){
  #print(resultsNoSD[[i]])
  #as.numeric(unlist(resultsMixed))
  print(i)
  cat("\n")

  print(mean(resultsNoSD[[i]]))
  print(median(resultsNoSD[[i]]))
  print(quantile(resultsNoSD[[i]], probs = c(0.025, 0.975)))
  cat("\n\n")
  
}

#Prepare data for plot

library(ggplot2)

stats <- statsBootstrap[order(statsBootstrap$OR, decreasing=TRUE),]
stats <- stats %>% mutate(names=NA) %>% add_row(Predictor = "Dr Effect beneath", OR = 1/stats$OR[stats$Predictor == "Dr Effect"], Lower =1/stats$Lower[stats$Predictor == "Dr Effect"], Upper = 1/stats$Upper[stats$Predictor == "Dr Effect"])

if(CategorizeVar) {
  stats$names[stats$Predictor == "age65.75.years"] <- "Age in [66, 75] years" 
  stats$names[stats$Predictor == "age75.85.years"] <- "Age in [76, 85] years" 
  stats$names[stats$Predictor == "age85.years.and.beyond"] <- "Age > 85 years"
  
  stats$names[stats$Predictor == "noflow1.to.5.minutes"] <- "No-flow in [2, 5] min" 
  stats$names[stats$Predictor == "noflow5.to.10.minutes"] <- "No-flow in [6, 10] min" 
  stats$names[stats$Predictor == "noflow10.to.20.minutes"] <- "No-flow in [11, 20] min" 
  stats$names[stats$Predictor == "noflow20.minutes.and.beyond"] <- "No-flow > 20 min" 
  
  stats$names[stats$Predictor == "low.flow10.to.20.minutes"] <- "Low-flow in [11, 20] min" 
  stats$names[stats$Predictor == "low.flow20.to.40.minutes"] <- "Low-flow in [21, 40] min" 
  stats$names[stats$Predictor == "low.flow40.minutes.and.beyond"] <- "Low-flow > 40 min" 
} else {
  stats$names[stats$Predictor == "age"] <- "Age" 
  stats$names[stats$Predictor == "low.flow"] <- "Low-Flow" 
  stats$names[stats$Predictor == "noflow"] <- "No-Flow" 
}

stats$names[stats$Predictor == "Dr Effect beneath"] <- "Doctor effect, one SD beneath mean"
stats$names[stats$Predictor == "Dr Effect"] <- "Doctor effect, one SD above mean"
stats$names[stats$Predictor == "ATCD.EOL"] <- "Dependency for activities of daily living" 
stats$names[stats$Predictor == "Asystolie.RSP"] <- "Non-shockable inital rhythm" 
stats$names[stats$Predictor == "ATCD.Cardiovasculaire"] <- "Cardiovascular disease" 
stats$names[stats$Predictor == "ATCD.Diabete"] <- "Diabetes" 
stats$names[stats$Predictor == "ATCD.Respiratoire"] <- "Respiratory disease" 
stats$names[stats$Predictor == "racs"] <- "ROSC" 
stats$names[stats$Predictor == "witness"] <- "Witness" 
stats$names[stats$Predictor == "ATCD.Other"] <- "Oncologic or other relevant disease" 

if(!ExcludeSex) {
  stats$names[stats$Predictor == "sexe"] <- "Male gender" 
}



plotF <- stats %>%
  arrange(OR) %>%    # First sort by val. This sort the dataframe but NOT the factor levels
  mutate_if(is.numeric, round, digits = 2) %>%
  mutate(names=factor(names, levels=names)) %>%   # This trick update the factor levels
  ggplot( aes(y = names, x = OR, xmin = Lower, xmax = Upper, label=OR, size = 40)) +
    scale_x_log10() +
    geom_vline(xintercept = 1, color = "red") +

    geom_text(hjust=0.5, vjust=-1, size = 3) +
    
    geom_errorbar(width = 0.3, size = 0.5, color = "darkgrey") +
      geom_point( size=2, color="black") + 
    ylab("Factor") +
    xlab("Odds Ratio") +
    #ggtitle("Odds Ratios of TOR factors in OHCA") +
    theme_minimal() + 
  theme(axis.text.y = element_text(lineheight = 20)) +
  theme(plot.margin = margin(0.5,0.5,0.5,0.5, "cm")) 
  #theme(axis.x=element_text(margin = margin(t = 20))

plotF

ggsave(filename = "resultplot.jpg", plotF, dpi = 500, width = 9, height = 8, device = "jpg")
stats %>%
  arrange(OR) %>%    # First sort by val. This sort the dataframe but NOT the factor levels
  mutate_if(is.numeric, round, digits = 2) %>%
  mutate(names=factor(names, levels=names)) %>%   # This trick update the factor levels
  ggsave(filename = "resultplot", 
         ggplot(aes(y = names, x = OR, xmin = Lower, xmax = Upper, label=OR)) +
    scale_x_log10() +
    geom_vline(xintercept = 1, color = "red") +

    geom_text(hjust=0.5, vjust=-1, size = 3) +
    
    geom_errorbar(width = 0.3, size = 0.5, color = "darkgrey") +
      geom_point( size=2, color="black") + 
    ylab("Factor") +
    xlab("Odds Ratio") +
    #ggtitle("Odds Ratios of TOR factors in OHCA") +
    theme_minimal(), dpi = 500, device = "png") #+
  #theme(axis.x=element_text(margin = margin(t = 20))
#library(htmltools)
#library(htmlTable)
#stats %>% htmlTable

#Calculate pseudo-R2 of FitGLMM

if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  R2_confint_marginal <- partR2(FitGLMM, partvars = c("noflow"),
  R2_type = "marginal", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
  R2_confint_conditional <- partR2(FitGLMM, partvars = c("noflow"),
  R2_type = "conditional", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
  
  R2_confint_marginal
  R2_confint_conditional
}
if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  partvars = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")
  if(ExcludeSex) {
    partvars = partvars[partvars != "sexe"];
  }
  partR2resultn100 <- partR2(FitGLMM, partvars = partvars,
  R2_type = "conditional", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
}
if(computeRptR) {
  library(ggplot2)
  library(partR2)
  partR2test <- partR2resultn100
  #partR2test$R2 <- partR2test$R2 %>% add_row(term = "Marginal R2", estimate = 0.563, CI_lower = 0.531, CI_upper = 0.581, ndf = 13)
  #partR2test$R2 <- partR2test$R2 %>% add_row(term = "Conditional R2", estimate = 0.06, CI_lower = 0.022, CI_upper = 0.08, ndf = 13)
  partR2test$R2 <- partR2test$R2 %>% add_row(term = "Doctor effect", estimate = rptRadjust100[["R"]][["dr"]][2], CI_lower = rptRadjust100[["CI_emp"]][["CI_link"]][["2.5%"]], CI_upper = rptRadjust100[["CI_emp"]][["CI_link"]][["97.5%"]], ndf = 13)
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "racs", "ROSC"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "noflow", "No-flow"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "witness", "Witness"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "`Asystolie/RSP`", "Asystole"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "low.flow", "Low-flow"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "sexe", "Sex"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "age", "Age"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Cardiovasculaire", "Cardiovascular disease"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Diabete", "Diabetes"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.EOL", "Poor autonomy / end of life"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Respiratoire", "Respiratory disease"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Other", "Oncologic and other relevant disease"))
  partR2test$R2 <- arrange(partR2test$R2, desc(estimate), .by_group = FALSE)
  
  p1test <- forestplot(partR2test, type = "R2", text_size = 10)
  p1test + geom_text(aes(label = sprintf("%.2f [%.2f-%.2f]", partR2test$R2$estimate, partR2test$R2$CI_lower, partR2test$R2$CI_upper), hjust = -0.01, vjust = 0.5), nudge_x = (partR2test$R2$CI_upper - partR2test$R2$estimate)+0.005, nudge_y = 0.05, size = 3) + scale_x_continuous(expand = expansion(mult = 0.2))
  p1test
  
  partR2test
}
if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  partR2resultn5 <- partR2(FitGLMM, partvars = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other"),
  R2_type = "conditional", max_level = 1, nboot = 5, CI = 0.95, parallel = TRUE, data=data)
}
nbBootPrimary = 100

library(doParallel)
# Bootstrap iterations
nsamples <- nbBootPrimary

# Multithreading
ncores=12
cl = makeCluster(ncores)
registerDoParallel(cl)

factorToRemove = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")



# Define the bootstrapping function
estimated_varF <- function(dataB, index, formulaT) {
  sim_data <- dataB[sample(1:nrow(dataB), nrow(dataB), replace=TRUE), ]
  
  
  BootGLMM <- glmer(formulaT, family=binomial, data=sim_data)
  
  #OR of mixed effect (dr)
  Mixed <- exp(sd(ranef(BootGLMM)$dr[,]))
  
  return(Mixed)
}

# Bootstrapped estimates
#results <- boot(data=sim_data, statistic=estimated_var, R=nbBootPrimary, progress="text") #nbBootPrimary

# Initialize a vector to store the bootstrapped estimates



for (factor in factorToRemove) {
    resultsBF <- c() 

    # Bootstrapped estimates with progress counter
    resultsBF = foreach(i=1:nbBootPrimary) %dopar% { 
    #for (i in 1:nbBootPrimary) {
      library(lme4)
      library(buildmer)
      index <- sample(1:nrow(data), replace=TRUE)
      formulaT = remove.terms(formula, factor)
      resultsF <- estimated_varF(data, index, formulaT)
    
      return(resultsF)
    }
    
    resultsMixedF <- numeric(nbBootPrimary)
    
    resultsMixedF <- as.numeric(unlist(resultsBF))

    sprintf("Result %s", factor)
    print(mean(resultsMixedF))
    print(median(resultsMixedF))
    print(quantile(resultsMixedF, probs = c(0.025, 0.975)))
}

# Threads closing
  stopCluster(cl)
---
title: "Quantifying physician’s bias to terminate resuscitation. The TERMINATOR Study"
output:
  html_notebook: default
  word_document: default
  pdf_document: default
---

#Library loading

```{r}
#library loading---------------------------------------------------
library(lme4)
library(ggplot2)
library(xlsx)
library(parallel)
library(sjmisc)
library(sjlabelled)
library(dplyr)
library(doParallel)
library(sjPlot)
#library(MuMIn)
library(boot)
library(table1)
library(partR2)
library(buildmer)
library(rptR)
library(furrr)
library(future)
library(glmmTMB)
```

#Options

```{r}
CapNoLowFlow = FALSE
ExcludeDrWithFewOHCA = FALSE
ExcludeSex = FALSE

CategorizeVar = TRUE
quadratic = FALSE

computeRptR = FALSE
bootstrapSigma = FALSE

numberOfResidualSimulations = 10 #dharma
nBootPartR2 = 10 #nboot rptr and partR2
nBootPartR2Compare = 10
nbBootPrimary = 10
```


#Data importation & cleaning

Correction of doctor names and anonymisation has been done before the importation process

```{r warning=FALSE}
data <- read.xlsx("C:/Users/thoma/Desktop/reaclast.xlsx", 1)
nb <- nrow(data)
```

Delete corrupted/incomplete data

```{r}
data <- filter(data, rythm >= 0)
print("Removing corrupt or unknown data about rythm")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, sexe < 3)
print("Removing corrupt or unknown data about sex")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, lowflow >= 0)
print("Removing corrupt or unknown data about low-flow")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)

data <- filter(data, noflow >= 0)
print("Removing corrupt or unknown data about no-flow")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)
```

Remove traumatic arrests

```{r}
data <- filter(data, medical > 0)
print("Removing traumatic arrests")
print(nrow(data))
print(nb - nrow(data))
nb <- nrow(data)
```

Options : 
Remove arrest linked to Dr with to few arrest (not in primary analysis but used in appendix)

```{r Appendix : Doctors with > 3 OHCA only}
if(ExcludeDrWithFewOHCA) {
  data <- data %>% group_by(dr) %>% filter(n() >= 40)
  
  data <- data %>% ungroup
  
  print("Removing arrests linked with Dr with less than 3 arrests")
  print(nrow(data))
  print(nb - nrow(data))
  nb <- nrow(data)
}
```

#Data preprocessing

Creation of binary score for the presence or absence for each observed medical history

```{r}
data <- data %>% mutate("ATCD.Cardiovasculaire" = replace(ATCD.Cardiovasculaire, ATCD.Cardiovasculaire == -1, 0))
data <- data %>% mutate("ATCD.Respiratoire" = replace(ATCD.Respiratoire, ATCD.Respiratoire == -1, 0))
data <- data %>% mutate("ATCD.Diabete" = replace(ATCD.Diabete, ATCD.Diabete == -1, 0))
data <- data %>% mutate("ATCD.EOL" = replace(ATCD.EOL, ATCD.EOL == -1, 0))
data <- data %>% mutate("ATCD.Other" = replace(ATCD.Other, ATCD.Other == -1, 0))
```

Adjust sex to binary value

```{r}
#Transforming men = 1 et women = 2 in men = 0 et women = 1
data <- data %>% mutate("sexe" = data$sexe-1)
```

Regroup asystole and pulseless electrical activity (two group remaining Asystole/PEA and other rythms)

```{r}
data <- data %>% mutate("Asystolie/RSP" = replace(chocable, chocable == 3, 1))
data <- data %>% mutate("Asystolie/RSP" = data$`Asystolie/RSP`-1)
```

Calculate low-flow BEFORE SAMU-MICU arrival

```{r}
#Creation of smurdelay useful for calculation just below
data <- data %>% mutate(delaysmur = difftime(data$smurHour, data$acHour, units ="mins"))
data <- data %>% mutate(delaysmur = ifelse(data$delaysmur > 720, difftime((as.Date(data$acHour)+1), data$smurHour, units ="mins"), data$delaysmur))
data <- data %>% mutate(delaysmur = ifelse(data$delaysmur < -720, difftime((as.Date(data$smurHour)+1), data$acHour, units ="mins"), data$delaysmur))

#Creation of lowflow before SMUR arrival
data <- data %>% mutate(low.flow = ifelse(pmin(data$delaysmur-data$noflow, data$lowflow) > 0, pmin(data$delaysmur-data$noflow, data$lowflow), data$lowflow))
```

Calculate the existence of a ROSC before SAMU-MICU arrival

```{r}
data <- data %>% mutate(racs = ifelse((data$racsdelay < data$delaysmur) & (data$racsdelay != -1), 1, 0))
```

Creation of TOR column (termination of resuscitation)

```{r}
data <- data %>% mutate(TOR = ifelse(!(data$reasmur == 1 | data$transport == 1), 1, 0))
```

Creation of data labels

```{r message=FALSE, warning=FALSE}
label(data$age)   <- "Age"
label(data$sexe)   <- "Sex"
label(data$noflow)   <- "No-flow"
label(data$witness)   <- "Witness"
label(data$dr)   <- "Doctor"
label(data$low.flow)   <- "Low-flow"
label(data$racs)   <- "ROSC before SMUR"
label(data$TOR)   <- "Termination of ressusitation"
label(data$`ATCD.Cardiovasculaire`)   <- "Cardiac history"
label(data$`ATCD.Respiratoire`)   <- "Respiratory history"
label(data$`ATCD.EOL`)   <- "End of life / dependency"
label(data$`ATCD.Diabete`)   <- "Diabetes"
label(data$`ATCD.Other`)   <- "Other medical issue including oncologic"
label(data$`Asystolie/RSP`)   <- "Rythm"

print(paste0("Included cardiac arrests : ", nrow(data)))
```
Graph of continuous variables
```{r}
hist(data$age, breaks = 50)
dataCapped <- data
dataCapped <- dataCapped %>% mutate("noflow" = replace(noflow, noflow >= 100, 100))
dataCapped <- dataCapped %>% mutate("low.flow" = replace(low.flow, low.flow >= 100, 100))
hist(dataCapped$noflow, breaks = 100)
hist(dataCapped$low.flow, breaks = 100)
```



Option :
Capping low flow and no-flow
```{r Appendix : capping low flow and no flow}
if(CapNoLowFlow) {
  dataCapped <- data
  dataCapped <- dataCapped %>% mutate("noflow" = replace(noflow, noflow >= 100, 100))
  dataCapped <- dataCapped %>% mutate("low.flow" = replace(low.flow, low.flow >= 100, 100))
  data <- dataCapped
}
```

Option :
Categorize variables
```{r}
dataSv <- data
if(CategorizeVar) {
  datacat <- data
  datacat$age <- cut(data$age, 
                   breaks=c(-Inf, 65, 75, 85, Inf), 
                   labels=c("< 65 years","65-75 years","75-85 years", "85 years and beyond"))
  datacat$noflow <- cut(data$noflow, 
                   breaks=c(-Inf, 1, 5, 10, 20, Inf), 
                   labels=c("< 1 minute", "1 to 5 minutes", "5 to 10 minutes","10 to 20 minutes", "20 minutes and beyond"))
  datacat$low.flow <- cut(data$low.flow, 
                   breaks=c(-Inf, 10, 20, 40, Inf), 
                   labels=c("< 10 minutes","10 to 20 minutes","20 to 40 minutes", "40 minutes and beyond"))
  data <- datacat
}

```

#Table1 : charasteristics of population
Creation of table1 : Characteristics of population

```{r Table1, message=FALSE, warning=FALSE}
if(CategorizeVar) {
  datacat <- dataSv
  datacat$age <- cut(dataSv$age, 
                   breaks=c(-Inf, 65, 75, 85, Inf), 
                   labels=c("Age ≤ 65","Age in [66, 75]","Age in [76, 85]", "Age > 85"))
  datacat$noflow <- cut(dataSv$noflow, 
                   breaks=c(-Inf, 1, 5, 10, 20, Inf), 
                   labels=c("No-flow ≤ 1", "No-flow in [2, 5]", "No-flow in [6, 10]","No-flow in [11, 20]", "No-flow > 20"))
  datacat$low.flow <- cut(dataSv$low.flow, 
                   breaks=c(-Inf, 10, 20, 40, Inf), 
                   labels=c("Low-flow ≤ 10","Low-flow in [11, 20]","Low-flow in [21, 40]", "Low-flow > 40"))
  dataT <- datacat
} else {
  dataT <- data
}

library(table1)
#dataT is a copy of data used for table display


render <- function(x, name, missing) {
  if (!is.numeric(x)) return(render.cat(x))
  else if (name == "age") {
    what <- switch(name,
        age = "Mean (SD)")
    parse.abbrev.render.code(c("", what))(x)
    
    # what <- switch(name,
    #     age = "Mean (SD)",
    #     low.flow  = "Median (IQR)",
    #     noflow  = "Median (IQR)")
    # parse.abbrev.render.code(c("", what))(x)
  } else {
    with(stats.apply.rounding(stats.default(x), digits=2), c("",
          "Median (IQR)"=sprintf("%s [%s-%s]", MEDIAN, Q1, Q3)))
  }
}


render.cat <- function(x) {
    c("", sapply(stats.default(x), function(y) with(y,
        sprintf("%d (%0.0f %%)", FREQ, PCT))))
}
  
dataT$TOR <- 
  factor(dataT$TOR, 
         levels=c(0,1),
         labels=c("Advanced life support", # Reference
                  "Termination of resuscitation"))



dataT$sexe <- 
  factor(dataT$sexe, levels=c(1,0),
         labels=c("Men", 
                  "Women"))

units(dataT$age) <- "Years"
units(dataT$noflow) <- "Minutes"
units(dataT$low.flow) <- "Minutes"

dataT$`Asystolie/RSP` <- 
  factor(dataT$`Asystolie/RSP`, levels=c(1,0),
         labels=c("Asystole/PEA", 
                  "VF/VT/SA"))

dataT$witness <- 
  factor(dataT$witness, levels=c(1),
         labels=c("Present"))

dataT$racs <- 
  factor(dataT$racs, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Cardiovasculaire` <- 
  factor(dataT$`ATCD.Cardiovasculaire`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.EOL` <- 
  factor(dataT$`ATCD.EOL`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Respiratoire` <- 
  factor(dataT$`ATCD.Respiratoire`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Diabete` <- 
  factor(dataT$`ATCD.Diabete`, levels=c(1),
         labels=c("Present"))

dataT$`ATCD.Other` <- 
  factor(dataT$`ATCD.Other`, levels=c(1),
         labels=c("Present"))

label(dataT$age)   <- "Age"
label(dataT$sexe)   <- "Sex"
label(dataT$noflow)   <- "No-flow"
label(dataT$`witness`)   <- "Witness"
label(dataT$dr)   <- "Doctor"
label(dataT$low.flow)   <- "Low-flow"
label(dataT$racs)   <- "ROSC before SAMU-MICU arrival"
label(dataT$TOR)   <- "Termination of resuscitation"
label(dataT$`ATCD.Cardiovasculaire`)   <- "Cardiac history"
label(dataT$`ATCD.Respiratoire`)   <- "Respiratory history"
label(dataT$`ATCD.EOL`)   <- "End of life / dependancy"
label(dataT$`ATCD.Diabete`)   <- "Diabetes history"
label(dataT$`ATCD.Other`)   <- "Other medical issue including oncologic history"
label(dataT$`Asystolie/RSP`)   <- "Rhythm"

table <- table1( ~ age + sexe + noflow + low.flow + `Asystolie/RSP` + witness + racs + `ATCD.Cardiovasculaire` + `ATCD.Respiratoire` + `ATCD.EOL` + `ATCD.Diabete` + `ATCD.Other` | TOR, data=dataT,  topclass="Rtable1-zebra", overall="Total", render=render)

```




#Mean and median of OHCA by doctor

```{r}
median(aggregate(data$age, by=list(data$dr), FUN=length)$x)
mean(aggregate(data$age, by=list(data$dr), FUN=length)$x)

t.test(aggregate(data$age, by=list(data$dr), FUN=length)$x)
quantile(aggregate(data$age, by=list(data$dr), FUN=length)$x)
```

#Creating formulae

```{r}
formula = `TOR` ~ `racs` + `noflow` + `Asystolie/RSP` + `low.flow` + `sexe` + `witness` + `age` + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other` + (1|`dr`)

if(quadratic) {
  formulaNonLinear = `TOR` ~ `racs` + `noflow` + I(`noflow`^2) + `Asystolie/RSP` + `low.flow` + I(`low.flow`^2) + `sexe` + `witness` + `age` + I(`age`^2) + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other` + (1|`dr`)
}

formulaGLM = `TOR` ~ `racs` + `noflow` + `Asystolie/RSP` + `low.flow` + `sexe` + `witness` + `age` + `ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL` + `ATCD.Respiratoire` + `ATCD.Other`

if(ExcludeSex) {
  formula = remove.terms(formula, "sexe")
  formulaGLM = remove.terms(formulaGLM, "sexe")
  if(quadratic) {
    formulaNonLinear = remove.terms(formulaNonLinear, "sexe")
  }
}
```


#Implementation of GLMM and GLM models (FitGLMM and ReducedGLM)

FitGLMM :
- Binomial family since outcome is binary
- Dr effect (random effect) implemented without random slope, only intersect.

```{r warning=FALSE}
library(lme4)
label(data$TOR)   <- "TOR"
FitGLMM <- glmer(formula, family=binomial, data=data)

if(quadratic)
  FitGLMMnonLinear <- glmer(formulaNonLinear, family=binomial, data=data)
```

ReducedGLM
- Same as fit GLMM without Dr effect (no random effect)

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without random effect)"
ReducedGLM <- glm(formulaGLM, family=binomial, data=data)
label(data$TOR)   <- "TOR"
```


#Fitting concurrent models

We fit concurrent GLMM models with some predictors missing to assess the impact.

- Without medical history / end of life

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without medical history / end of life)"
formulaWithoutHistoryEOL = remove.terms(formula, "`ATCD.Cardiovasculaire` + `ATCD.Diabete`+ `ATCD.EOL`+ `ATCD.Respiratoire` + `ATCD.Other`")
glmmWithoutHistoryEOL <- glmer(formulaWithoutHistoryEOL, family=binomial, data=data)
```

- Without medical history

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without medical history)"
formulaWithoutHistory = remove.terms(formula, "ATCD.EOL")
glmmWithoutHistory <- glmer(formulaWithoutHistory, family=binomial, data=data)
```

- Without no-flow

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without no-flow)"
formulaWithoutNF = remove.terms(formula, "noflow")
glmmWithoutNF <- glmer(formulaWithoutNF, family=binomial, data=data)
```

- Without low-flow

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without low-flow)"
formulaWithoutLF = remove.terms(formula, "low.flow")
glmmWithoutLF <- glmer(formulaWithoutLF, family=binomial, data=data)
```

- Without rythm and ROSC

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without rythm and ROSC)"
formulaWithoutRythm = remove.terms(formula, "`racs` + `Asystolie/RSP`")
glmmWithoutRythm <- glmer(formulaWithoutRythm, family=binomial, data=data)
```

- Without sex

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without sex)"
formulaWithoutSex = remove.terms(formula, "sexe")
glmmWithoutSex <- glmer(formulaWithoutSex, family=binomial, data=data)
```

- Without age

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without age)"
formulaWithoutAge = remove.terms(formula, "age")
glmmWithoutAge <- glmer(formulaWithoutAge, family=binomial, data=data)
```

- Without witness

```{r warning=FALSE}
label(data$TOR)   <- "TOR (without witness)"
formulaWithoutWitness = remove.terms(formula, "witness")
glmmWithoutWitness <- glmer(formulaWithoutWitness, family=binomial, data=data)
```

#Checking GLMM model asumptions

- Basic information of models

```{r}
s <- summary(FitGLMM, correlation = TRUE)
s
summary(ReducedGLM, correlation = TRUE)
```

```{r}
confint(FitGLMM, method = "Wald")
```


- Checking residuals

Using Dharma which create visualy interpretable residuals by simulating many response and and using the cumulative density for the observed data point to create residuals (more information in package vignette).
(Refit = T pour un bootstrap paramétrique)

From this graph, we check : normality of residus, hemegeneity of residus for the whole package and also for each individual predictors

```{r}
library(DHARMa)
#simFit <- simulateResiduals(fittedModel = FitGLMM, n = numberOfResidualSimulations, refit = T)
simFit <- simulateResiduals(fittedModel = FitGLMM, n = numberOfResidualSimulations)
plot(simFit)

simReduced <- simulateResiduals(fittedModel = ReducedGLM, n = numberOfResidualSimulations)
plot(simReduced)

plotResiduals(simulationOutput = FitGLMM, form=data$low.flow)
plotResiduals(simulationOutput = FitGLMM, form=data$noflow)
plotResiduals(simulationOutput = FitGLMM, form=data$racs)
plotResiduals(simulationOutput = FitGLMM, form=data$`Asystolie/RSP`)
plotResiduals(simulationOutput = FitGLMM, form=data$`sexe`)
plotResiduals(simulationOutput = FitGLMM, form=data$`witness`)
plotResiduals(simulationOutput = FitGLMM, form=data$`age`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Cardiovasculaire`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Diabete`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.EOL`)
plotResiduals(simulationOutput = FitGLMM, form=data$`ATCD.Respiratoire`)
plotResiduals(simulationOutput = FitGLMM)
plotResiduals(simulationOutput = ReducedGLM)
```


- Normality of residuals
To test for normality of residuals the simplest is to transform the data given by DHARMa and plot it with an histogram

```{r}
#Quantile normalization of scaled residuals to visualize them in a 
residualsNorm <- residuals(simFit, quantileFunction = qnorm, outlierValues = c(0,1))

h <- hist(residualsNorm, breaks = 40, density = 10,
          col = "black", xlab = "Scaled residues (after quantile normalisation)", ylab = "Number of observation", main = "Distribution of residues") 
xfit <- seq(min(residualsNorm), max(residualsNorm), length = 40) 
yfit <- dnorm(xfit, mean = mean(residualsNorm), sd = sd(residualsNorm)) 
yfit <- yfit * diff(h$mids[1:2]) * length(residualsNorm) 

lines(xfit, yfit, col = "red", lwd = 2)
plot(h)
```

- Checking random effect (doctor effect) normal distribution

```{r warning=FALSE}
plot_model(FitGLMM, type="diag", sort.est=TRUE,
           vline.color="#A9A9A9", dot.size=1.5,
           show.values=T, value.offset=.2)
```
```{r}
save_plot("qqplot.jpg", dpi = 500, width = 20, height = 15)
```

- Checking fixed variable independance

Here we look at the variance-covariance table which informs us on linear correlations

```{r}
print(s, correlation = TRUE)
summary(ReducedGLM, correlation = TRUE)
```

- Checking of random effect independence to fixed predictors is done below by checking the independance of repeatability from variables

```{r warning=FALSE}
if(computeRptR) {
  rptRadjust100 <- rptR::rpt(formula, "dr", data = data, datatype = "Binary", nboot = nBootPartR2, npermut = nBootPartR2, parallel = TRUE, ncores = 6, adjusted = FALSE)
  rptRadjust100
  
  sprintf("Part of explained variance of Dr effect for full model : %s %%", trunc(rptRadjust100[["R"]][["dr"]][2]*10^4)/10^2)
}
```

```{r warning=FALSE}
if(computeRptR) {
  rptRadjustWithoutLF <- rptR::rpt(formulaWithoutLF, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without low flow) : %s %%", trunc(rptRadjustWithoutLF[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutNF <- rptR::rpt(formulaWithoutNF, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without no flow) : %s %%", trunc(rptRadjustWithoutNF[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutHistoryEOL <- rptR::rpt(formulaWithoutHistoryEOL, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without history) : %s %%", trunc(rptRadjustWithoutHistoryEOL[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutEOL <- rptR::rpt(formulaWithoutHistory, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without end of life) : %s %%", trunc(rptRadjustWithoutEOL[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutRythm <- rptR::rpt(formulaWithoutRythm, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without rythm) : %s %%", trunc(rptRadjustWithoutRythm[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutSex <- rptR::rpt(formulaWithoutSex, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without sex) : %s %%", trunc(rptRadjustWithoutSex[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutAge <- rptR::rpt(formulaWithoutAge, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without age) : %s %%", trunc(rptRadjustWithoutAge[["R"]][["dr"]][2]*10^4)/10^2)
  
  rptRadjustWithoutWitness <- rptR::rpt(formulaWithoutWitness, "dr", data = data, datatype = "Binary", nboot = nBootPartR2Compare, npermut = nBootPartR2Compare, parallel = TRUE, ncores = 6, adjusted = FALSE)
  sprintf("Part of explained variance of Dr effect (without witness) : %s %%", trunc(rptRadjustWithoutWitness[["R"]][["dr"]][2]*10^4)/10^2)
}
```


#Checking FitGLMM model performance

Comparing all models via AIC, pseudo-R2 to check for :
- random effect independance to fixed effects choice
- Pertinence of chosen fixed variables

```{r warning=FALSE}
anova(FitGLMM, glmmWithoutHistoryEOL, glmmWithoutHistory, glmmWithoutNF, glmmWithoutLF, glmmWithoutRythm, glmmWithoutSex, glmmWithoutAge, glmmWithoutWitness, ReducedGLM)

if(computeRptR) {
  R2cFull <- partR2(FitGLMM, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (full) : %s %%", trunc(R2cFull[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutLF <- partR2(glmmWithoutLF, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without low flow) : %s %%", trunc(R2cWithoutLF[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutNF <- partR2(glmmWithoutNF, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without no flow) : %s %%", trunc(R2cWithoutNF[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutEOL <- partR2(glmmWithoutHistory, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without end of life) : %s %%", trunc(R2cWithoutEOL[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutHistoryEOL <- partR2(glmmWithoutHistoryEOL, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without history and end of life) : %s %%", trunc(R2cWithoutHistoryEOL[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutRythm <- partR2(glmmWithoutRythm, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without rythm) : %s %%", trunc(R2cWithoutRythm[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutSex <- partR2(glmmWithoutSex, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without sex) : %s %%", trunc(R2cWithoutSex[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutAge <- partR2(glmmWithoutAge, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without age) : %s %%", trunc(R2cWithoutAge[["R2"]][["estimate"]][1]*10^4)/10^2)
  
  R2cWithoutWitness <- partR2(glmmWithoutWitness, R2_type = "conditional", max_level = 1, nboot = NULL, CI = 0.95, parallel = TRUE, data=data)
  sprintf("Part of explained variance of model (without witness) : %s %%", trunc(R2cWithoutWitness[["R2"]][["estimate"]][1]*10^4)/10^2)
}

compareTab = tab_model(FitGLMM, glmmWithoutHistoryEOL, glmmWithoutHistory, glmmWithoutNF, glmmWithoutLF, glmmWithoutRythm, glmmWithoutSex, glmmWithoutAge, glmmWithoutWitness, ReducedGLM, p.style = "stars", pred.labels = c("Intercept","ROSC","No-flow", "Asystole/PEA", "Low-flow", "Sex", "Witness", "Age", "Cardiac history", "Diabetes", "End of life / dependancy", "Respiratory history", "Other medical issue including oncologic history"))
```

#RESULTS
#Primary endpoint

#Evaluating the significance of doctor effect through comparison of ReducedGLM and FitGLMM through wilk's test (Maximum likelihood ratio test)

```{r}
anova(FitGLMM,ReducedGLM, test="LRT")
```


#Computing sd of random effect and it's p-value with parametric bootstrap

Computing primary objective and it's confidence interval.
Parametric bootstrap between FitGLMM et ReducedGLM.

```{r warning=FALSE}
if(bootstrapSigma){
  library(doParallel)
  # Bootstrap iterations
  nsamples <- nbBootPrimary
  
  # Multithreading
  ncores=6
  cl = makeCluster(ncores)
  registerDoParallel(cl)
  
  # Vector of random effect variance of FitGLMM on data simulated through ReducedGLM
  estimated_var<- c() 
  
  # Bootstrap loop
  estimated_var = foreach(i=1:nsamples) %dopar% { 
    library(lme4)
    
    # Creation of simulated data from experimental data by replacing
    sim_data <- data[sample(1:nrow(data), nrow(data), replace=TRUE), ]
    
    # Simulation of response (TOR) for simulated data using ReducedGLM and adding a noise similar to the one found in experimental data
    sim_reasmur <- simulate(ReducedGLM, nsim = nrow(data), newdata=sim_data)
    
    # Integration of the responses to simulated data frame 
    sim_data["TOR"] <- sim_reasmur
    # We apply a model similar in every way to FitGLMM called BootGlmm and we fit it on simdata
    BootGlmm <- glmer(formula, family=binomial, data=sim_data)
  
    # We add sd of random effect to estimated_var
    var <- as.data.frame(VarCorr(BootGlmm))["sdcor"][1]
    var
  }
  
  # Threads closing
  stopCluster(cl)
  
  estimated_var_list <- estimated_var
  estimated_var <- as.numeric(unlist(estimated_var))
  
  # Statistic test between sd of random effect of experimental data and simulated data
  testEffectDr <- VarCorr(FitGLMM) > quantile(estimated_var,.95) 
  testEffectDr["dr"]
  
  # Extraction of quantile of sd
  quantile(estimated_var, probs = c(0.05, 0.95))
  
  # Calculation of p-value of sd of random effect
  pvalueSdDevDrEffect<- mean(VarCorr(FitGLMM) < estimated_var) 
  pvalueSdDevDrEffect[1]
  
  qplot(estimated_var, geom="histogram")
}
```

#Secondary endpoint

#OR of fixed effects (standardized for 1 increase of SD value)

```{r}
#OR_LF = exp(coef(summary(FitGLMM))["low.flow", "Estimate"]*sqrt(var(data$low.flow)))
```

```{r}
#exp(confint(FitGLMM))

```


```{r}
plot_model(FitGLMM, sort.est = TRUE, show.values = TRUE, value.offset = .3, type = "std")
```


#OR of random effect
#point estimate for the odds ratio can be obtained for a doctor one SD above the mean, relative to a doctor at the mean, by exponentiating the value of sigma

```{r}
#exp(sd(ranef(FitGLMM)$dr[1, ]))
```
#OR sans interval de confiance
```{r}
racsOR <- exp(coef(summary(FitGLMM))["racs", "Estimate"])#*sqrt(var(dataB$racs)))
sprintf("OR Racs : %s", racsOR)
  
AsystolieOR <- exp(coef(summary(FitGLMM))["`Asystolie/RSP`", "Estimate"])
sprintf("OR asystolie : %s", AsystolieOR)

witnessOR <- exp(coef(summary(FitGLMM))["witness", "Estimate"])
sprintf("OR witness : %s", witnessOR)

ATCDcardioOR <- exp(coef(summary(FitGLMM))["ATCD.Cardiovasculaire", "Estimate"])
sprintf("OR atcd cardio : %s", ATCDcardioOR)

ATCDdiaOR <- exp(coef(summary(FitGLMM))["ATCD.Diabete", "Estimate"])
sprintf("OR atcd dia : %s", ATCDdiaOR)

ATCD_EOL_OR <- exp(coef(summary(FitGLMM))["ATCD.EOL", "Estimate"])
sprintf("OR EOL : %s", ATCD_EOL_OR)

ATCDrespOR <- exp(coef(summary(FitGLMM))["ATCD.Respiratoire", "Estimate"])
sprintf("OR atcd resp : %s", ATCDrespOR)

ATCDotherOR <- exp(coef(summary(FitGLMM))["ATCD.Other", "Estimate"])
sprintf("OR atcd other : %s", ATCDotherOR)

sexeOR <- 0
  if(ExcludeSex == FALSE)
    sexeOR <- exp(coef(summary(FitGLMM))["sexe", "Estimate"])

sprintf("OR sex : %s", sexeOR)

if(CategorizeVar) {
  LF_OR10 <- exp(coef(summary(FitGLMM))["low.flow10 to 20 minutes", "Estimate"])
  LF_OR20 <- exp(coef(summary(FitGLMM))["low.flow20 to 40 minutes", "Estimate"])
  LF_OR40 <- exp(coef(summary(FitGLMM))["low.flow40 minutes and beyond", "Estimate"])
  print(sprintf("OR lf < 20 : %s", LF_OR10))
  print(sprintf("OR lf < 40 : %s", LF_OR20))
  print(sprintf("OR lf > 40 : %s", LF_OR40))
  
  noflowOR1 <- exp(coef(summary(FitGLMM))["noflow1 to 5 minutes", "Estimate"])
  noflowOR5 <- exp(coef(summary(FitGLMM))["noflow5 to 10 minutes", "Estimate"])
  noflowOR10 <- exp(coef(summary(FitGLMM))["noflow10 to 20 minutes", "Estimate"])
  noflowOR20 <- exp(coef(summary(FitGLMM))["noflow20 minutes and beyond", "Estimate"])
  print(sprintf("OR nf > 1 : %s", noflowOR1))
  print(sprintf("OR nf > 5 : %s", noflowOR5))
  print(sprintf("OR nf > 10 : %s", noflowOR10))
  print(sprintf("OR nf > 20 : %s", noflowOR20))
  
  ageOR65 <- exp(coef(summary(FitGLMM))["age65-75 years", "Estimate"])
  ageOR75 <- exp(coef(summary(FitGLMM))["age75-85 years", "Estimate"])
  ageOR85 <- exp(coef(summary(FitGLMM))["age85 years and beyond", "Estimate"])
  print(sprintf("OR age > 65 : %s", ageOR65))
  print(sprintf("OR age > 75 : %s", ageOR75))
  print(sprintf("OR age > 85 : %s", ageOR85))
} else {
  LF_OR <- exp(coef(summary(FitGLMM))["low.flow", "Estimate"]*sqrt(var(dataB$low.flow)))
  ageOR <- exp(coef(summary(FitGLMM))["age", "Estimate"]*sqrt(var(dataB$age)))
  noflowOR <- exp(coef(summary(FitGLMM))["noflow", "Estimate"]*sqrt(var(dataB$noflow)))
  print(sprintf("OR lf : %s", LF_OR))
  print(sprintf("OR nf : %s", noflowOR))
  print(sprintf("OR age : %s", ageOR))
}



```



#bootstrapped confidence interval

```{r warning=FALSE}
nbBootPrimary = 1000

library(doParallel)
# Bootstrap iterations
nsamples <- nbBootPrimary

# Multithreading
ncores=6
cl = makeCluster(ncores)
registerDoParallel(cl)

# Define the bootstrapping function
estimated_var <- function(dataB, index) {
  sim_data <- dataB[sample(1:nrow(dataB), nrow(dataB), replace=TRUE), ]
  if(quadratic)
    BootGLMM <- glmer(formulaNonLinear, family=binomial, data=sim_data)
  else
    BootGLMM <- glmer(formula, family=binomial, data=sim_data)
  
  #OR of mixed effect (dr)
  Mixed <- exp(sd(ranef(BootGLMM)$dr[,]))
  
  #OR of fixed effects
  racsOR <- exp(coef(summary(BootGLMM))["racs", "Estimate"])#*sqrt(var(dataB$racs)))
  AsystolieOR <- exp(coef(summary(BootGLMM))["`Asystolie/RSP`", "Estimate"])#*sqrt(var(dataB$`Asystolie/RSP`)))
  witnessOR <- exp(coef(summary(BootGLMM))["witness", "Estimate"])#*sqrt(var(dataB$witness)))
  ATCDcardioOR <- exp(coef(summary(BootGLMM))["ATCD.Cardiovasculaire", "Estimate"])#*sqrt(var(dataB$ATCD.Cardiovasculaire)))
  ATCDdiaOR <- exp(coef(summary(BootGLMM))["ATCD.Diabete", "Estimate"])#*sqrt(var(dataB$ATCD.Diabete)))
  ATCD_EOL_OR <- exp(coef(summary(BootGLMM))["ATCD.EOL", "Estimate"])#*sqrt(var(dataB$ATCD.EOL)))
  ATCDrespOR <- exp(coef(summary(BootGLMM))["ATCD.Respiratoire", "Estimate"])#*sqrt(var(dataB$ATCD.Respiratoire)))
  ATCDotherOR <- exp(coef(summary(BootGLMM))["ATCD.Other", "Estimate"])#*sqrt(var(dataB$ATCD.Other)))
  
  
  
  sexeOR <- 0
  if(ExcludeSex == FALSE)
    sexeOR <- exp(coef(summary(BootGLMM))["sexe", "Estimate"])#*sqrt(var(dataB$sexe)))
  
  if(CategorizeVar) {
    LF_OR10 <- exp(coef(summary(BootGLMM))["low.flow10 to 20 minutes", "Estimate"])
    LF_OR20 <- exp(coef(summary(BootGLMM))["low.flow20 to 40 minutes", "Estimate"])
    LF_OR40 <- exp(coef(summary(BootGLMM))["low.flow40 minutes and beyond", "Estimate"])
    
    noflowOR1 <- exp(coef(summary(BootGLMM))["noflow1 to 5 minutes", "Estimate"])
    noflowOR5 <- exp(coef(summary(BootGLMM))["noflow5 to 10 minutes", "Estimate"])
    noflowOR10 <- exp(coef(summary(BootGLMM))["noflow10 to 20 minutes", "Estimate"])
    noflowOR20 <- exp(coef(summary(BootGLMM))["noflow20 minutes and beyond", "Estimate"])
    
    ageOR65 <- exp(coef(summary(BootGLMM))["age65-75 years", "Estimate"])
    ageOR75 <- exp(coef(summary(BootGLMM))["age75-85 years", "Estimate"])
    ageOR85 <- exp(coef(summary(BootGLMM))["age85 years and beyond", "Estimate"])
  } else {
    
    LF_OR <- exp(coef(summary(BootGLMM))["low.flow", "Estimate"]*sqrt(var(dataB$low.flow)))
    ageOR <- exp(coef(summary(BootGLMM))["age", "Estimate"]*sqrt(var(dataB$age)))
    noflowOR <- exp(coef(summary(BootGLMM))["noflow", "Estimate"]*sqrt(var(dataB$noflow)))
    
    noflowORnoSD <- exp(coef(summary(BootGLMM))["noflow", "Estimate"])
    LF_ORnoSD <- exp(coef(summary(BootGLMM))["low.flow", "Estimate"])
    ageORnoSD <- exp(coef(summary(BootGLMM))["age", "Estimate"])
  }
  
if(CategorizeVar) {
      Fixed <- data.frame("racs" = racsOR, "noflow1 to 5 minutes" = noflowOR1, "noflow5 to 10 minutes" = noflowOR5, "noflow10 to 20 minutes" = noflowOR10, "noflow20 minutes and beyond" = noflowOR20, "Asystolie/RSP" = AsystolieOR, "low.flow10 to 20 minutes" = LF_OR10, "low.flow20 to 40 minutes" = LF_OR20, "low.flow40 minutes and beyond" = LF_OR40, "sexe"= sexeOR, "witness" = witnessOR, "age65-75 years"= ageOR65, "age75-85 years"= ageOR75, "age85 years and beyond"= ageOR85, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
      
      NoSD <- Fixed
      
} else {
  #named vector of fixed effects OR
    Fixed <- data.frame("racs" = racsOR, "noflow" = noflowOR, "Asystolie/RSP" = AsystolieOR, "low.flow" = LF_OR, "sexe"= sexeOR, "witness" = witnessOR, "age"= ageOR, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
    
    NoSD <- data.frame("racs" = racsOR, "noflow" = noflowORnoSD, "Asystolie/RSP" = AsystolieOR, "low.flow" = LF_ORnoSD, "sexe"= sexeOR, "witness" = witnessOR, "age"= ageORnoSD, "ATCD.Cardiovasculaire"= ATCDcardioOR, "ATCD.Diabete" = ATCDdiaOR, "ATCD.EOL" = ATCD_EOL_OR, "ATCD.Respiratoire" = ATCDrespOR, "ATCD.Other"= ATCDotherOR)
}

  
  
  
  #Fixed <- exp(coef(summary(BootGLMM))[, "Estimate"]*sqrt(var(dataB$low.flow)))
  output<-list(Mixed,Fixed, NoSD)
  return(output)
}

# Bootstrapped estimates
#results <- boot(data=sim_data, statistic=estimated_var, R=nbBootPrimary, progress="text") #nbBootPrimary

# Initialize a vector to store the bootstrapped estimates
resultsMixed <- numeric(nbBootPrimary)

if(CategorizeVar) {
  resultsFixed <- data.frame("racs" = numeric(), "noflow1 to 5 minutes" = numeric(), "noflow5 to 10 minutes" = numeric(), "noflow10 to 20 minutes" = numeric(), "noflow20 minutes and beyond" = numeric(), "Asystolie/RSP" = numeric(), "low.flow10 to 20 minutes" = numeric(), "low.flow20 to 40 minutes" = numeric(), "low.flow40 minutes and beyond" = numeric(), "sexe"= numeric(), "witness" = numeric(), "age65-75 years"= numeric(), "age75-85 years"= numeric(), "age85 years and beyond"= numeric(), "ATCD.Cardiovasculaire"= numeric(), "ATCD.Diabete" = numeric(), "ATCD.EOL" = numeric(), "ATCD.Respiratoire" = numeric(), "ATCD.Other"= numeric())
  
  resultsNoSD <- resultsFixed
} else {
  resultsFixed <- data.frame("racs"= numeric(), "noflow"= numeric(), "Asystolie/RSP"= numeric(), "low.flow"= numeric(), "sexe"= numeric(), "witness"= numeric(), "age"= numeric(), "ATCD.Cardiovasculaire"= numeric(), "ATCD.Diabete"= numeric(), "ATCD.EOL"= numeric(), "ATCD.Respiratoire"= numeric(), "ATCD.Other"= numeric())
  resultsNoSD <- resultsFixed
}
#colnames(resultsFixed) = c("Intercept", "racs", "noflow", "Asystolie/RSP", "low.flow", "sexe", "witness", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")

  if(ExcludeSex) {
    resultsFixed <- select(resultsFixed, -sexe)
    resultsNoSD <- select(resultsNoSD, -sexe)
  }

resultsB<- c() 

# Bootstrapped estimates with progress counter
resultsB = foreach(i=1:nbBootPrimary) %dopar% { 
#for (i in 1:nbBootPrimary) {
  library(lme4)
  index <- sample(1:nrow(data), replace=TRUE)
  results <- estimated_var(data, index)
  
  
  # Display the progress counter
  # if (i %% 25 == 0) {
  #   cat("Iteration", i, "of ", nbBootPrimary, "\n time : ", Sys.time())
  # }
  
  return(results)
}

# resultsMixed[i] <- results[1]
# resultsFixed <- rbind(resultsFixed, data.frame(as.list(results[2][[1]])))
# resultsNoSD <- rbind(resultsNoSD, data.frame(as.list(results[3][[1]])))

# Threads closing
  stopCluster(cl)

```


#
```{r}
resultsMixed <- c()
for (i in 1:length(resultsB)) {
  resultsMixed[i] = resultsB[[i]][1]
  resultsFixed <- rbind(resultsFixed, data.frame(as.list(resultsB[[i]][2][[1]])))
  resultsNoSD <- rbind(resultsNoSD, data.frame(as.list(resultsB[[i]][3][[1]])))
}


  if(ExcludeSex) {
    resultsFixed <- select(resultsFixed, -sexe)
    resultsNoSD <- select(resultsNoSD, -sexe)
  }

```



```{r}
resultsMixed <- as.numeric(unlist(resultsMixed))

mean(resultsMixed)
median(resultsMixed)
quantile(resultsMixed, probs = c(0.025, 0.975))
```
```{r}
hist(resultsMixed, main = "Frequency Plot of Bootstrapped Means", xlab = "Bootstrapped Means", ylab = "Frequency", breaks = 100)
```

```{r}
statsBootstrap <- data.frame(Predictor = "Dr Effect", OR = mean(resultsMixed), Lower = quantile(resultsMixed, probs = c(0.025, 0.975))["2.5%"][[1]], Upper = quantile(resultsMixed, probs = c(0.025, 0.975))["97.5%"][[1]])

for (i in colnames(resultsFixed)){
  #print(resultsFixed[[i]])
  #as.numeric(unlist(resultsMixed))
  print(i)
  cat("\n")

  print(mean(resultsFixed[[i]]))
  print(median(resultsFixed[[i]]))
  print(quantile(resultsFixed[[i]], probs = c(0.025, 0.975)))
  cat("\n\n")
  
  statsBootstrap <- statsBootstrap %>% 
   add_row(Predictor = i, OR = mean(resultsFixed[[i]]), Lower = quantile(resultsFixed[[i]], probs = c(0.025, 0.975))["2.5%"][[1]], Upper = quantile(resultsFixed[[i]], probs = c(0.025, 0.975))["97.5%"][[1]])
}

print("NON NORMALIZED")

for (i in colnames(resultsNoSD)){
  #print(resultsNoSD[[i]])
  #as.numeric(unlist(resultsMixed))
  print(i)
  cat("\n")

  print(mean(resultsNoSD[[i]]))
  print(median(resultsNoSD[[i]]))
  print(quantile(resultsNoSD[[i]], probs = c(0.025, 0.975)))
  cat("\n\n")
  
}

```
#Prepare data for plot
```{r}

```




```{r}
library(ggplot2)

stats <- statsBootstrap[order(statsBootstrap$OR, decreasing=TRUE),]
stats <- stats %>% mutate(names=NA) %>% add_row(Predictor = "Dr Effect beneath", OR = 1/stats$OR[stats$Predictor == "Dr Effect"], Lower =1/stats$Lower[stats$Predictor == "Dr Effect"], Upper = 1/stats$Upper[stats$Predictor == "Dr Effect"])

if(CategorizeVar) {
  stats$names[stats$Predictor == "age65.75.years"] <- "Age in [66, 75] years" 
  stats$names[stats$Predictor == "age75.85.years"] <- "Age in [76, 85] years" 
  stats$names[stats$Predictor == "age85.years.and.beyond"] <- "Age > 85 years"
  
  stats$names[stats$Predictor == "noflow1.to.5.minutes"] <- "No-flow in [2, 5] min" 
  stats$names[stats$Predictor == "noflow5.to.10.minutes"] <- "No-flow in [6, 10] min" 
  stats$names[stats$Predictor == "noflow10.to.20.minutes"] <- "No-flow in [11, 20] min" 
  stats$names[stats$Predictor == "noflow20.minutes.and.beyond"] <- "No-flow > 20 min" 
  
  stats$names[stats$Predictor == "low.flow10.to.20.minutes"] <- "Low-flow in [11, 20] min" 
  stats$names[stats$Predictor == "low.flow20.to.40.minutes"] <- "Low-flow in [21, 40] min" 
  stats$names[stats$Predictor == "low.flow40.minutes.and.beyond"] <- "Low-flow > 40 min" 
} else {
  stats$names[stats$Predictor == "age"] <- "Age" 
  stats$names[stats$Predictor == "low.flow"] <- "Low-Flow" 
  stats$names[stats$Predictor == "noflow"] <- "No-Flow" 
}

stats$names[stats$Predictor == "Dr Effect beneath"] <- "Doctor effect, one SD beneath mean"
stats$names[stats$Predictor == "Dr Effect"] <- "Doctor effect, one SD above mean"
stats$names[stats$Predictor == "ATCD.EOL"] <- "Dependency for activities of daily living" 
stats$names[stats$Predictor == "Asystolie.RSP"] <- "Non-shockable inital rhythm" 
stats$names[stats$Predictor == "ATCD.Cardiovasculaire"] <- "Cardiovascular disease" 
stats$names[stats$Predictor == "ATCD.Diabete"] <- "Diabetes" 
stats$names[stats$Predictor == "ATCD.Respiratoire"] <- "Respiratory disease" 
stats$names[stats$Predictor == "racs"] <- "ROSC" 
stats$names[stats$Predictor == "witness"] <- "Witness" 
stats$names[stats$Predictor == "ATCD.Other"] <- "Oncologic or other relevant disease" 

if(!ExcludeSex) {
  stats$names[stats$Predictor == "sexe"] <- "Male gender" 
}



plotF <- stats %>%
  arrange(OR) %>%    # First sort by val. This sort the dataframe but NOT the factor levels
  mutate_if(is.numeric, round, digits = 2) %>%
  mutate(names=factor(names, levels=names)) %>%   # This trick update the factor levels
  ggplot( aes(y = names, x = OR, xmin = Lower, xmax = Upper, label=OR, size = 40)) +
    scale_x_log10() +
    geom_vline(xintercept = 1, color = "red") +

    geom_text(hjust=0.5, vjust=-1, size = 3) +
    
    geom_errorbar(width = 0.3, size = 0.5, color = "darkgrey") +
      geom_point( size=2, color="black") + 
    ylab("Factor") +
    xlab("Odds Ratio") +
    #ggtitle("Odds Ratios of TOR factors in OHCA") +
    theme_minimal() + 
  theme(axis.text.y = element_text(lineheight = 20)) +
  theme(plot.margin = margin(0.5,0.5,0.5,0.5, "cm")) 
  #theme(axis.x=element_text(margin = margin(t = 20))

plotF

ggsave(filename = "resultplot.jpg", plotF, dpi = 500, width = 9, height = 8, device = "jpg")
```
```{r}
stats %>%
  arrange(OR) %>%    # First sort by val. This sort the dataframe but NOT the factor levels
  mutate_if(is.numeric, round, digits = 2) %>%
  mutate(names=factor(names, levels=names)) %>%   # This trick update the factor levels
  ggsave(filename = "resultplot", 
         ggplot(aes(y = names, x = OR, xmin = Lower, xmax = Upper, label=OR)) +
    scale_x_log10() +
    geom_vline(xintercept = 1, color = "red") +

    geom_text(hjust=0.5, vjust=-1, size = 3) +
    
    geom_errorbar(width = 0.3, size = 0.5, color = "darkgrey") +
      geom_point( size=2, color="black") + 
    ylab("Factor") +
    xlab("Odds Ratio") +
    #ggtitle("Odds Ratios of TOR factors in OHCA") +
    theme_minimal(), dpi = 500, device = "png") #+
  #theme(axis.x=element_text(margin = margin(t = 20))
```

```{r}
#library(htmltools)
#library(htmlTable)
#stats %>% htmlTable

```

```{r}


```

#Calculate pseudo-R2 of FitGLMM

- Calculation of marginal and conditional pseudo-R2 of full model using partR2
R2_confint_conditional - R2_confint_marginal = rptR unadjusted repeatability

```{r warning=FALSE}
if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  R2_confint_marginal <- partR2(FitGLMM, partvars = c("noflow"),
  R2_type = "marginal", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
  R2_confint_conditional <- partR2(FitGLMM, partvars = c("noflow"),
  R2_type = "conditional", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
  
  R2_confint_marginal
  R2_confint_conditional
}
```

- Pseudo part-R² of individual predictors

```{r warning=FALSE}
if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  partvars = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")
  if(ExcludeSex) {
    partvars = partvars[partvars != "sexe"];
  }
  partR2resultn100 <- partR2(FitGLMM, partvars = partvars,
  R2_type = "conditional", max_level = 1, nboot = nBootPartR2, CI = 0.95, parallel = TRUE, data=data)
}
```

- Display result in a nice forest plot
Since ICC and part-R2 are both measure of proportion of varicv fgbfance of the model, they are both included in the results

```{r}
if(computeRptR) {
  library(ggplot2)
  library(partR2)
  partR2test <- partR2resultn100
  #partR2test$R2 <- partR2test$R2 %>% add_row(term = "Marginal R2", estimate = 0.563, CI_lower = 0.531, CI_upper = 0.581, ndf = 13)
  #partR2test$R2 <- partR2test$R2 %>% add_row(term = "Conditional R2", estimate = 0.06, CI_lower = 0.022, CI_upper = 0.08, ndf = 13)
  partR2test$R2 <- partR2test$R2 %>% add_row(term = "Doctor effect", estimate = rptRadjust100[["R"]][["dr"]][2], CI_lower = rptRadjust100[["CI_emp"]][["CI_link"]][["2.5%"]], CI_upper = rptRadjust100[["CI_emp"]][["CI_link"]][["97.5%"]], ndf = 13)
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "racs", "ROSC"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "noflow", "No-flow"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "witness", "Witness"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "`Asystolie/RSP`", "Asystole"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "low.flow", "Low-flow"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "sexe", "Sex"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "age", "Age"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Cardiovasculaire", "Cardiovascular disease"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Diabete", "Diabetes"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.EOL", "Poor autonomy / end of life"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Respiratoire", "Respiratory disease"))
  partR2test$R2 <- partR2test$R2 %>% mutate(term = replace(term, term == "ATCD.Other", "Oncologic and other relevant disease"))
  partR2test$R2 <- arrange(partR2test$R2, desc(estimate), .by_group = FALSE)
  
  p1test <- forestplot(partR2test, type = "R2", text_size = 10)
  p1test + geom_text(aes(label = sprintf("%.2f [%.2f-%.2f]", partR2test$R2$estimate, partR2test$R2$CI_lower, partR2test$R2$CI_upper), hjust = -0.01, vjust = 0.5), nudge_x = (partR2test$R2$CI_upper - partR2test$R2$estimate)+0.005, nudge_y = 0.05, size = 3) + scale_x_continuous(expand = expansion(mult = 0.2))
  p1test
  
  partR2test
}
```


```{r warning=FALSE}
if(computeRptR) {
  plan(multisession, workers = parallel::detectCores())
  partR2resultn5 <- partR2(FitGLMM, partvars = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other"),
  R2_type = "conditional", max_level = 1, nboot = 5, CI = 0.95, parallel = TRUE, data=data)
}
```


```{r warning=FALSE}
nbBootPrimary = 100

library(doParallel)
# Bootstrap iterations
nsamples <- nbBootPrimary

# Multithreading
ncores=12
cl = makeCluster(ncores)
registerDoParallel(cl)

factorToRemove = c("racs", "noflow", "witness", "`Asystolie/RSP`", "low.flow", "sexe", "age", "ATCD.Cardiovasculaire", "ATCD.Diabete", "ATCD.EOL", "ATCD.Respiratoire", "ATCD.Other")



# Define the bootstrapping function
estimated_varF <- function(dataB, index, formulaT) {
  sim_data <- dataB[sample(1:nrow(dataB), nrow(dataB), replace=TRUE), ]
  
  
  BootGLMM <- glmer(formulaT, family=binomial, data=sim_data)
  
  #OR of mixed effect (dr)
  Mixed <- exp(sd(ranef(BootGLMM)$dr[,]))
  
  return(Mixed)
}

# Bootstrapped estimates
#results <- boot(data=sim_data, statistic=estimated_var, R=nbBootPrimary, progress="text") #nbBootPrimary

# Initialize a vector to store the bootstrapped estimates



for (factor in factorToRemove) {
    resultsBF <- c() 

    # Bootstrapped estimates with progress counter
    resultsBF = foreach(i=1:nbBootPrimary) %dopar% { 
    #for (i in 1:nbBootPrimary) {
      library(lme4)
      library(buildmer)
      index <- sample(1:nrow(data), replace=TRUE)
      formulaT = remove.terms(formula, factor)
      resultsF <- estimated_varF(data, index, formulaT)
    
      return(resultsF)
    }
    
    resultsMixedF <- numeric(nbBootPrimary)
    
    resultsMixedF <- as.numeric(unlist(resultsBF))

    sprintf("Result %s", factor)
    print(mean(resultsMixedF))
    print(median(resultsMixedF))
    print(quantile(resultsMixedF, probs = c(0.025, 0.975)))
}

# Threads closing
  stopCluster(cl)

```
