In n8thangreen/VANTDET: Cost-Effectiveness Decision Tree

options(width = 1000)

An R package for a cost-effectiveness analysis using decision tree models of novel active TB diagnostics.

The tests investigated are:

transcriptomics
proteomic SELDI
proteomic flowassay
molecular
IGRA

The pathways are:

rule-in
rule-out
dual-test

Installing `VANTDET`

To install the development version from github:

# library(devtools)
install_github("n8thangreen/VANTDET")

Example: Rule-out tests

library(VANTDET)
library(BCEA)
library(ggplot2)
library(purrr)
library(dplyr)
library(formattable)
library(IDEAdectree)
# library(ceplot)

Load the package data.

data("TBdata_clinical_cleaned", package = "IDEAdectree")

data("drug", package = "VANTDET")
data("QALYloss",  package = "VANTDET")
data("costs", package = "VANTDET")
data("time_res", package = "VANTDET")
data("performance", package = "VANTDET")

Initialise variables.

dat <- list()

QUANT <- 0.5  #0.65
yearindays <- 365.25
WTP <- c(20000, 30000)/yearindays

Run model

We pass the structure of the tree to dectree() as a function argument.

terminal_cost_ruleout <- function(cost) {

  c(cost$visit1 + cost$std.TB + cost$newtest,
    cost$visit1 + cost$std.TB + cost$newtest + cost$visit2,
    cost$visit1 + cost$std.nonTB + cost$newtest,
    cost$visit1 + cost$newtest)
}

terminal_health_ruleout <- function(health) {

  c(health$std.TB + health$newtest,
    health$std.TB + health$newtest + health$followup,
    health$std.nonTB + health$newtest,
    health$newtest)
}

The run the model and rearrange the output so we can plot.

dat <- list()

dat[['transcript']] <-
  dectree(
    data = data,
    name.newtest = "transcript",
    costDistns = costs,
    performance = list(performance$transcript_ruleout),
    time_res = list(time_res$transcript),
    drug = drug,
    QALYloss = QALYloss,
    terminal_cost = terminal_cost_ruleout,
    terminal_health = terminal_health_ruleout)

dat[['IGRA']] <-
  dectree(
    data = data,
    name.newtest = "IGRA",
    costDistns = costs,
    performance = list(performance$IGRA),
    time_res = list(time_res$IGRA),
    drug = drug,
    QALYloss = QALYloss,
    terminal_cost = terminal_cost_ruleout,
    terminal_health = terminal_health_ruleout)


e_df <- do.call(cbind, purrr::map(dat, 'e'))
c_df <- do.call(cbind, purrr::map(dat, 'c'))

evens <- seq(from = 2, to = 2*length(dat), 2)
odds <- evens - 1

QALYgain <- as.matrix(data.frame(0, e_df[ ,odds] - e_df[ ,evens]))
cost_incur <- as.matrix(data.frame(0, c_df[ ,evens] - c_df[ ,odds]))

res_bcea <- bcea(e = -QALYgain,
                 c = -cost_incur,
                 interventions = c("status-quo", names(dat)))

Output plots

ceplot::my_contour2(
  res_bcea,
  graph = "ggplot2", CONTOUR_PC = '5%') +
  coord_cartesian(xlim = c(-0.002, 0.002)) +
  theme(legend.position = "none")