The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

Decision Tree Herpes Virus Encephalopathy

This is an example that we use to illustrate decision trees in our DARTH teaching materials. The example involves the treatment of herpes simplex ecncephalopathy (HVE) by either treating everyone without biopsy (Treat), not treating or taking biopsy (DoNotTreat), and Taking biopsy and treat if biopsy posivite (Biopsy).

We will first define the twig model and the parameters and then we will show how one would code the model direclty in R and compare the outcomes.

twig

library(twig)

twig model definition

mytwig <- twig() + 
  decisions(names=c(DoNotTreat, Treat, Biopsy)) +  # treatment options
  event(name = DIE,  # first event
        options = c(yes, none), # either occurs or doesn't occur
        probs = c(pDie, leftover),  # occurs with prob pDie and doesn't occur with 1-pDie (leftover)
        transitions = c(Death, HVE_event)) + # if it occurs, transitions to Death, otherwise the HVE_event
  event(name = HVE_event,  # similarly, HVE_event occurs with f_HVE but if not it will be OVE
        options = c(yes, none), 
        probs = c(f_HVE, leftover), 
        transitions = c(get_HVE_comp, get_OVE_comp)) +
  event(name = get_HVE_comp, # evaluate whether HVE complications occured 
        options = c(yes, none), 
        probs = c(p_comp, leftover),
        transitions = c(HVE_comp, no_HVE_comp))  +
  event(name = get_OVE_comp, # evaluate whether other viral encephalitis (OVE) complications occured
        options = c(yes, none), 
        probs = c(p_comp, leftover),
        transitions = c(OVE_comp, no_OVE_comp)) + 
  payoffs(names = c(cost, utility)) # finally measure the cost and utilities

DecisionTwig

In DecisionTwig, this decision tree looks like this: DecisionTree

Model parameters

Define a data frame for model parameters as a list of scalar values. This can also ve a dataset for running a probabilistic analysis.

params <- list(
  # Probabilities,
  p_HVE          = 0.52   ,# prevalence of HVE
  p_HVE_comp     = 0.71   ,# complications with untreated HVE
  p_OVE_comp     = 0.01   ,# complications with untreated OVE
  p_HVE_comp_tx  = 0.36   ,# complications with treated HVE
  p_OVE_comp_tx  = 0.20   ,# complications with treated OVE
  p_biopsy_death = 0.005  ,# probability of death due to biopsy
  
  # Costs,
  c_VE           = 1200   ,# cost of viral encephalitis care without complications
  c_VE_comp      = 9000   ,# cost of viral encephalitis care with complications
  c_tx           = 9500   ,# cost of treatment
  c_biopsy       = 25000  ,# cost of brain biopsy
  
  # QALYs,
  q_VE           = 20     ,# remaining QALYs for those without VE-related complications
  q_VE_comp      = 19     ,# remaining QALYs for those with VE-related complications
  q_loss_biopsy  = 0.01   ,# one-time QALY loss due to brain biopsy
  q_death_biopsy = 0      # remaining QALYs for those who died during biopsy
)

Functions

Probability of death

We only look into death from biopsy in this decision tree.


pDie <- function(decision, p_biopsy_death){
  p_biopsy_death * (decision == "Biopsy")
}

Probability of complications

This depends on whether treatment received or not, HVE vs. OVE, and whether either HVE or OVE complications occured.

p_comp <- function(decision, HVE_event, p_HVE_comp, p_OVE_comp, 
                   p_HVE_comp_tx, p_OVE_comp_tx) {

    # complication of untreated HVE
    p_HVE_comp * (decision == "DoNotTreat" & HVE_event=="yes") + 

    # complication of untreated OVE
    p_OVE_comp * (decision %in% c("DoNotTreat", "Biopsy") & HVE_event=="none") + 

    # complications of treated HVE
    p_HVE_comp_tx * (decision %in% c("Treat", "Biopsy") & HVE_event=="yes") + 

    # complications of treated OVE
    p_OVE_comp_tx * (decision == "Treat" & HVE_event=="none")
}

Probability of HVE

All event probabilities must be function names. So, we can just create a simple wrapper around the p_HVE variable.

f_HVE <- function(p_HVE){
  p_HVE
}

Cost function

Cost is a function of the decision and the final model outcomes.

cost <- function(decision, outcome, c_biopsy, c_tx, c_VE_comp, c_VE){ 

  # cost of biopsy
  c_biopsy*(decision=="Biopsy") + 

    # cost of treatment if treated or biopsy was +ve for HVE
    c_tx*(decision=="Treat" | (decision=="Biopsy" & outcome %in% c("HVE_comp", "no_HVE_comp"))) + 

    # cost of complication if outcomes are in either HVE or OVE complications
    c_VE_comp*(outcome %in% c("HVE_comp", "OVE_comp")) + 

    # cost of viral encephalitis if complications didn't occur
    c_VE*(outcome %in% c("no_HVE_comp", "no_OVE_comp")) 
}

utility

Is a function of the decision and final outcome.

utility <- function(decision, outcome, q_loss_biopsy, q_VE_comp, q_VE){

  # apply utility discount for biopsy
  -q_loss_biopsy*(decision=="Biopsy") + 

  # apply utility values for complications
    q_VE_comp*(outcome %in% c("HVE_comp", "OVE_comp")) + 

    # apply utility values if complications didn't occur
    q_VE*(outcome %in% c("no_HVE_comp", "no_OVE_comp"))
}

Running the twig

We can run the model and check the results of the single parameter set under results$sim_ev

results <- run_twig(twig_obj = mytwig, params = params, parallel = FALSE, progress_bar = FALSE)
#> Checking Twig syntax ....
#> A states layer was not detected - Treatign the twig as a Decision Tree. Event transitions include the following event names:  HVE_event, get_HVE_comp, get_OVE_comp . These are valid. 
#> The following transitions are not event names and will be treated as final outcomes:  Death, HVE_comp, no_HVE_comp, OVE_comp, no_OVE_comp .
#> Twig syntax validation completed successfully.
#> Preprocessing started ...
#> Preprocessing completed. Starting simulation...
#> 
#> Total time: 0.00065 secs

results$sim_ev
#> , , sim = 1
#> 
#>             payoff
#> decision         cost  utility
#>   DoNotTreat  4117.20 19.62600
#>   Treat      12908.96 19.71680
#>   Biopsy     32599.41 19.69896

Incremental Cost-Effectiveness Ratio (ICER)

using the calculate_icers function adapted from the dampack package, we can retrieve ICER table by passing the payoffs summary table.

calculate_icers(results$mean_ev)
#>              decision     cost  utility inc_cost inc_utility     ICER status
#> DoNotTreat DoNotTreat  4117.20 19.62600       NA          NA       NA     ND
#> Treat           Treat 12908.96 19.71680  8791.76      0.0908 96825.55     ND
#> Biopsy         Biopsy 32599.41 19.69896       NA          NA       NA      D

Summary

This example illustrated the following features of twig * Decision tree * Multiple decisions * Multiple sequential events * probability dependency on prior events * transition payoffs * using sequential events to model complications * ICER

In addition, various intermediate objects and computations can be returned by enabling verbose.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.