R/buildmarkov2.R
In StanWijn/cemtool: Step By Step Guide to Build a Markov Model
Defines functions
cemrun
Documented in
cemrun
#' Run the Markov model with input generated from \code{cemtool()}, \code{cemprob()},\code{cemtpm()}
#'
#'
#' The \code{cemtool()} package provides a step-by-step tool to guide users in building a default Markov model.
#' The tool guides the user though the steps of the development of a Markov model and will present the final result using graphs and tables.
#' The only prerequisite is that the user knows the structure of the model and the transition probabilities, no calculations or coding is required.
#'
#' @param HS Number of healthstates
#' @param HS1 String with name of healthstate 1
#' @param HS2 String with name of healthstate 2
#' @param HS3 String with name of healthstate 3
#' @param HS4 String with name of healthstate 4
#' @param HS5 String with name of healthstate 5
#' @param dead String with name of absorption / death state
#' @param n.t Number of cycles
#' @param control String with name of the usual care strategy
#' @param intervention String with name of the intervention strategy
#' @param d.rc Discount rate for costs
#' @param d.re Discount rate for effects
#' @param m.M Matrix showing the Markov trace of usual care, nrow = n.t + 1, ncol = HS
#' @param m.M_treatment Matrix showing the Markov trace of intervention strategy, nrow = n.t + 1, ncol = HS
#' @param m.P Matrix showing the transition probability matrix of the usual care, nrow = HS, ncol = HS
#' @param m.P_treatment Matrix showing the transition probability matrix of the intervention strategy, nrow = HS, ncol = HS
#' @param modelinput Matrix with 2 rows that include all the transition probabilities, costs and effects.
#'
#' @return The following variables will be created in the saved in the cemtool environment:
#' @return --- Full Markov trace (m.M and m.M_treatment)
#' @return --- Calculate the costs and effects for both strategies.
#' @return --- Results are saved (table_output) and shown in the console.
#' @return --- The model structure and markov trace are both plotted and saved (plot1 and plot2)
#'
#' @details
#' All input arguments can be generated and saved with the \code{cemtool()} function. (cemtool.env <- cemtool())
#' There are multiple software systems that can be used to build Markov models for cost-effectiveness analyses
#' like TreeAge, Excel or R. Although there are numerous advantages to use R over the others,
#' the biggest downside is the steep learning curve from R.
#' The \code{cemtool()} package aims to close this gap by introducing a step-by-step tool
#' to guide users in building a default Markov models.
#' The tool guides the user though the steps of the development of a Markov model
#' and will present the final result using graphs and tables.
#' The only prerequisite is that the user knows the structure of the model and the transition probabilities,
#' no calculations or coding is required.
#'
#'
#' @author S.R.W. Wijn MSc <stan.wijn@@radboudumc.nl>
#' @export
#' @examples
#' \dontrun{
#' cemtool() # Start from stratch
#' cemprob() # Start from the second phase (definding the parameters)
#' cemtpm() # Start from the third phase (modify the transition probability matrix)
#' cemrun() # Run the model with the current m.M markov trace and m.P transition probability matrix
#'
#' cemtool.env <- cemtool() # To save all input for further modification
#' }
cemrun <-function(HS = cemtool.env$HS, HS1 = cemtool.env$HS1, HS2 = cemtool.env$HS2, HS3 = cemtool.env$HS3,
HS4 = cemtool.env$HS4, HS5 = cemtool.env$HS5, dead = cemtool.env$dead, n.t = cemtool.env$n.t,
control = cemtool.env$control, intervention = cemtool.env$intervention,
d.rc = cemtool.env$d.rc, d.re = cemtool.env$d.rc,
m.M = cemtool.env$m.M, m.M_treatment = cemtool.env$m.M_treatment,
m.P = cemtool.env$m.P, m.P_treatment = cemtool.env$m.P_treatment, modelinput = cemtool.env$modelinput){
cemtool.env$Strategies <- c(control, intervention)
cemtool.env$v.dwc <- 1 / ((1 + d.rc) ^ (0:n.t))
cemtool.env$v.dwe <- 1 / ((1 + d.re) ^ (0:n.t))
if(HS==3){
cemtool.env$v.n <- c(HS1, HS2, dead)
} else if(HS==4){
cemtool.env$v.n <- c(HS1, HS2, HS3, dead)
} else if(HS==5){
cemtool.env$v.n <- c(HS1, HS2, HS3, HS4, dead)
} else if(HS==6){
cemtool.env$v.n <- c(HS1, HS2, HS3,
HS4, HS5, dead)}
input <- cemtool.env$modelinput
for (t in 1:cemtool.env$n.t){
######### using transition matrices ###########
# calculate the proportion of the cohort in each state at time t
cemtool.env$m.M[t + 1, ] <- t(cemtool.env$m.M[t, ]) %*% cemtool.env$m.P
cemtool.env$m.M_treatment[t + 1, ] <- t(cemtool.env$m.M_treatment[t, ]) %*% cemtool.env$m.P_treatment
}
matplot(0:cemtool.env$n.t, cemtool.env$m.M, type = 'l',
ylab = "Probability of state occupancy",
xlab = "Cycle",
main = "Markov Trace")
legend("topright", cemtool.env$v.n, col = 1:cemtool.env$HS,lty = 1:cemtool.env$HS, bty = "n") # add a legend to the graph
assign('plot2', recordPlot(), envir = cemtool.env)
cat("Inspect the Markov Trace (of the usual care strategy) in the plot window to the right and check if is as you expected.")
if(interactive()) readkey()
# calculate discounted costs and effects
inputtable <- input
if(cemtool.env$HS==3){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==4){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==5){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1, "c.4"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1, "u.4"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2, "c.4"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2, "u.4"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==6){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1, "c.4"], inputtable[1, "c.5"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1, "u.4"], inputtable[1, "u.5"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2, "c.4"], inputtable[2, "c.5"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2, "u.4"], inputtable[2, "u.5"], inputtable[2,"u.absorb"])
}
t.c <- cemtool.env$m.M %*% t.c
t.c_tr <- cemtool.env$m.M_treatment %*% t.c_tr
t.u <- cemtool.env$m.M %*% t.u
t.u_tr <- cemtool.env$m.M_treatment %*% t.u_tr
t.c <- t(t.c) %*% cemtool.env$v.dwc
t.c <- t.c + inputtable[1, "c.Tr"]
t.c_tr <- t(t.c_tr) %*% cemtool.env$v.dwc
t.c_tr <- t.c_tr + inputtable[2, "c.Tr"]
t.e <- t(t.u) %*% cemtool.env$v.dwe
t.e_tr <- t(t.u_tr) %*% cemtool.env$v.dwe
# calculate lifelong per patient discounted cost and QALYs.
costdiff <- t.c_tr - t.c # calculate the difference in discounted costs between the two strategies
names(costdiff) <- ""
effectdiff <- t.e_tr - t.e # calculate the difference in discounted effects between the two strategies
names(effectdiff) <- ""
ICER <- costdiff / effectdiff # calculate the ICER
names(ICER) <- "ICER"
results <- c(costdiff, effectdiff, ICER) # combine the results
# create full incremental cost-effectiveness analysis table
C <- round(c(t.c, t.c_tr), 2) # bind and round the total costs of the two strategies
E <- round(c(t.e, t.e_tr), 2) # bind and round the total effects of the two strategies
costdiff <- c("", as.character(round(costdiff, 2))) # round the delta of the costs (No Treatment is reference)
effectdiff <- c("", as.character(round(effectdiff, 2))) # round the delta of the effects (No Treatment is reference)
ICER <- c("", as.character(round(ICER, 2))) # round the ICER
table_output <- cbind(cemtool.env$Strategies, C, E, costdiff, effectdiff, ICER) # combine all data in a table
table_output <- as.data.frame(table_output) # create a data frame
names(table_output) <- c("", "Cost", "QALY", "Costs diff", "QALYs diff", "ICER")
cemtool.env$table_output <- table_output
cat("--------------------------------------------", "\n")
cat("Results:", "\n")
print(table_output)
cat("\n")
cat("--------------------------------------------", "\n")
if(interactive()) readkey()
cat("To rerun the analysis with alterations:
Type cemprob() in the console and press enter to redo step 2 (Markov model input)
Type cemtpm() in the console and press enter to redo step 3 (Transition probability matrix)
Type cemrun() in the console and press enter to run the model with the existing m.M and m.P tables (which you can alter by yourself)", " \n",
"To start over, type cemtool() in the console and press enter", "\n",
"To save the results: cemtool.env <- cemtool()", "\n",
"\n",
"\n",
"When you save the results, cemtool.env$table_output will show the result table.
cemtool.env$plot1 and cemtool.env$plot2 include the model structure and Markov trace.
cemtool.env$m.M and cemtool.env$m.M_treatment show the Markov trace
cemtool.env$m.P and cemtool.env$m.P_treatment show the transition probability matrix", "\n",
"\n",
"If you want to compare multiple models, save each model seperatly and then compare the table_output")
cat("--------------------------------------------", "\n")
cat("Finished", "\n")
return(cemtool.env)
}
StanWijn/cemtool documentation built on April 8, 2020, 1:42 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions cemrun
Documented in cemrun
#' Run the Markov model with input generated from \code{cemtool()}, \code{cemprob()},\code{cemtpm()}
#'
#'
#' The \code{cemtool()} package provides a step-by-step tool to guide users in building a default Markov model.
#' The tool guides the user though the steps of the development of a Markov model and will present the final result using graphs and tables.
#' The only prerequisite is that the user knows the structure of the model and the transition probabilities, no calculations or coding is required.
#'
#' @param HS Number of healthstates
#' @param HS1 String with name of healthstate 1
#' @param HS2 String with name of healthstate 2
#' @param HS3 String with name of healthstate 3
#' @param HS4 String with name of healthstate 4
#' @param HS5 String with name of healthstate 5
#' @param dead String with name of absorption / death state
#' @param n.t Number of cycles
#' @param control String with name of the usual care strategy
#' @param intervention String with name of the intervention strategy
#' @param d.rc Discount rate for costs
#' @param d.re Discount rate for effects
#' @param m.M Matrix showing the Markov trace of usual care, nrow = n.t + 1, ncol = HS
#' @param m.M_treatment Matrix showing the Markov trace of intervention strategy, nrow = n.t + 1, ncol = HS
#' @param m.P Matrix showing the transition probability matrix of the usual care, nrow = HS, ncol = HS
#' @param m.P_treatment Matrix showing the transition probability matrix of the intervention strategy, nrow = HS, ncol = HS
#' @param modelinput Matrix with 2 rows that include all the transition probabilities, costs and effects.
#'
#' @return The following variables will be created in the saved in the cemtool environment:
#' @return --- Full Markov trace (m.M and m.M_treatment)
#' @return --- Calculate the costs and effects for both strategies.
#' @return --- Results are saved (table_output) and shown in the console.
#' @return --- The model structure and markov trace are both plotted and saved (plot1 and plot2)
#'
#' @details
#' All input arguments can be generated and saved with the \code{cemtool()} function. (cemtool.env <- cemtool())
#' There are multiple software systems that can be used to build Markov models for cost-effectiveness analyses
#' like TreeAge, Excel or R. Although there are numerous advantages to use R over the others,
#' the biggest downside is the steep learning curve from R.
#' The \code{cemtool()} package aims to close this gap by introducing a step-by-step tool
#' to guide users in building a default Markov models.
#' The tool guides the user though the steps of the development of a Markov model
#' and will present the final result using graphs and tables.
#' The only prerequisite is that the user knows the structure of the model and the transition probabilities,
#' no calculations or coding is required.
#'
#'
#' @author S.R.W. Wijn MSc <stan.wijn@@radboudumc.nl>
#' @export
#' @examples
#' \dontrun{
#' cemtool() # Start from stratch
#' cemprob() # Start from the second phase (definding the parameters)
#' cemtpm() # Start from the third phase (modify the transition probability matrix)
#' cemrun() # Run the model with the current m.M markov trace and m.P transition probability matrix
#'
#' cemtool.env <- cemtool() # To save all input for further modification
#' }
cemrun <-function(HS = cemtool.env$HS, HS1 = cemtool.env$HS1, HS2 = cemtool.env$HS2, HS3 = cemtool.env$HS3,
HS4 = cemtool.env$HS4, HS5 = cemtool.env$HS5, dead = cemtool.env$dead, n.t = cemtool.env$n.t,
control = cemtool.env$control, intervention = cemtool.env$intervention,
d.rc = cemtool.env$d.rc, d.re = cemtool.env$d.rc,
m.M = cemtool.env$m.M, m.M_treatment = cemtool.env$m.M_treatment,
m.P = cemtool.env$m.P, m.P_treatment = cemtool.env$m.P_treatment, modelinput = cemtool.env$modelinput){
cemtool.env$Strategies <- c(control, intervention)
cemtool.env$v.dwc <- 1 / ((1 + d.rc) ^ (0:n.t))
cemtool.env$v.dwe <- 1 / ((1 + d.re) ^ (0:n.t))
if(HS==3){
cemtool.env$v.n <- c(HS1, HS2, dead)
} else if(HS==4){
cemtool.env$v.n <- c(HS1, HS2, HS3, dead)
} else if(HS==5){
cemtool.env$v.n <- c(HS1, HS2, HS3, HS4, dead)
} else if(HS==6){
cemtool.env$v.n <- c(HS1, HS2, HS3,
HS4, HS5, dead)}
input <- cemtool.env$modelinput
for (t in 1:cemtool.env$n.t){
######### using transition matrices ###########
# calculate the proportion of the cohort in each state at time t
cemtool.env$m.M[t + 1, ] <- t(cemtool.env$m.M[t, ]) %*% cemtool.env$m.P
cemtool.env$m.M_treatment[t + 1, ] <- t(cemtool.env$m.M_treatment[t, ]) %*% cemtool.env$m.P_treatment
}
matplot(0:cemtool.env$n.t, cemtool.env$m.M, type = 'l',
ylab = "Probability of state occupancy",
xlab = "Cycle",
main = "Markov Trace")
legend("topright", cemtool.env$v.n, col = 1:cemtool.env$HS,lty = 1:cemtool.env$HS, bty = "n") # add a legend to the graph
assign('plot2', recordPlot(), envir = cemtool.env)
cat("Inspect the Markov Trace (of the usual care strategy) in the plot window to the right and check if is as you expected.")
if(interactive()) readkey()
# calculate discounted costs and effects
inputtable <- input
if(cemtool.env$HS==3){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==4){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==5){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1, "c.4"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1, "u.4"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2, "c.4"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2, "u.4"], inputtable[2,"u.absorb"])
} else if (cemtool.env$HS==6){
t.c <- c(inputtable[1,"c.1"],inputtable[1,"c.2"], inputtable[1, "c.3"], inputtable[1, "c.4"], inputtable[1, "c.5"], inputtable[1,"c.absorb"])
t.u <- c(inputtable[1,"u.1"],inputtable[1,"u.2"], inputtable[1, "u.3"], inputtable[1, "u.4"], inputtable[1, "u.5"], inputtable[1,"u.absorb"])
t.c_tr <- c(inputtable[2,"c.1"],inputtable[2,"c.2"], inputtable[2, "c.3"], inputtable[2, "c.4"], inputtable[2, "c.5"], inputtable[2,"c.absorb"])
t.u_tr <- c(inputtable[2,"u.1"],inputtable[2,"u.2"], inputtable[2, "u.3"], inputtable[2, "u.4"], inputtable[2, "u.5"], inputtable[2,"u.absorb"])
}
t.c <- cemtool.env$m.M %*% t.c
t.c_tr <- cemtool.env$m.M_treatment %*% t.c_tr
t.u <- cemtool.env$m.M %*% t.u
t.u_tr <- cemtool.env$m.M_treatment %*% t.u_tr
t.c <- t(t.c) %*% cemtool.env$v.dwc
t.c <- t.c + inputtable[1, "c.Tr"]
t.c_tr <- t(t.c_tr) %*% cemtool.env$v.dwc
t.c_tr <- t.c_tr + inputtable[2, "c.Tr"]
t.e <- t(t.u) %*% cemtool.env$v.dwe
t.e_tr <- t(t.u_tr) %*% cemtool.env$v.dwe
# calculate lifelong per patient discounted cost and QALYs.
costdiff <- t.c_tr - t.c # calculate the difference in discounted costs between the two strategies
names(costdiff) <- ""
effectdiff <- t.e_tr - t.e # calculate the difference in discounted effects between the two strategies
names(effectdiff) <- ""
ICER <- costdiff / effectdiff # calculate the ICER
names(ICER) <- "ICER"
results <- c(costdiff, effectdiff, ICER) # combine the results
# create full incremental cost-effectiveness analysis table
C <- round(c(t.c, t.c_tr), 2) # bind and round the total costs of the two strategies
E <- round(c(t.e, t.e_tr), 2) # bind and round the total effects of the two strategies
costdiff <- c("", as.character(round(costdiff, 2))) # round the delta of the costs (No Treatment is reference)
effectdiff <- c("", as.character(round(effectdiff, 2))) # round the delta of the effects (No Treatment is reference)
ICER <- c("", as.character(round(ICER, 2))) # round the ICER
table_output <- cbind(cemtool.env$Strategies, C, E, costdiff, effectdiff, ICER) # combine all data in a table
table_output <- as.data.frame(table_output) # create a data frame
names(table_output) <- c("", "Cost", "QALY", "Costs diff", "QALYs diff", "ICER")
cemtool.env$table_output <- table_output
cat("--------------------------------------------", "\n")
cat("Results:", "\n")
print(table_output)
cat("\n")
cat("--------------------------------------------", "\n")
if(interactive()) readkey()
cat("To rerun the analysis with alterations:
Type cemprob() in the console and press enter to redo step 2 (Markov model input)
Type cemtpm() in the console and press enter to redo step 3 (Transition probability matrix)
Type cemrun() in the console and press enter to run the model with the existing m.M and m.P tables (which you can alter by yourself)", " \n",
"To start over, type cemtool() in the console and press enter", "\n",
"To save the results: cemtool.env <- cemtool()", "\n",
"\n",
"\n",
"When you save the results, cemtool.env$table_output will show the result table.
cemtool.env$plot1 and cemtool.env$plot2 include the model structure and Markov trace.
cemtool.env$m.M and cemtool.env$m.M_treatment show the Markov trace
cemtool.env$m.P and cemtool.env$m.P_treatment show the transition probability matrix", "\n",
"\n",
"If you want to compare multiple models, save each model seperatly and then compare the table_output")
cat("--------------------------------------------", "\n")
cat("Finished", "\n")
return(cemtool.env)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.