rdecision

CRAN status Codecov test coverage

The goal of rdecision is to provide methods for assessing health care interventions using cohort models (decision trees and semi-Markov models) which can be constructed using only a few lines of R code. Mechanisms are provided for associating an uncertainty distribution with each source variable and for ensuring transparency of the mathematical relationships between variables. The package terminology follows Briggs et al “Decision Modelling for Health Economic Evaluation”.1

Installation

You can install the released version of rdecision from CRAN with:

install.packages("rdecision")

Examples

A decision tree with parameter uncertainty

Consider the fictitious and idealized decision problem of choosing between providing two forms of lifestyle advice, offered to people with vascular disease, which reduce the risk of needing an interventional procedure. The model has a time horizon of 1 year. The cost to a healthcare provider of the interventional procedure (e.g. inserting a stent) is 5000 GBP; the cost of providing the current form of lifestyle advice, an appointment with a dietician (“diet”), is 50 GBP and the cost of providing an alternative form, attendance at an exercise programme (“exercise”), is 500 GBP. If the advice programme is successful, there is no need for an interventional procedure. In a small trial of the “diet” programme, 12 out of 68 patients (17.6%) avoided having a procedure, and in a separate small trial of the “exercise” programme 18 out of 58 patients (31.0%) avoided the procedure. It is assumed that the baseline characteristics in the two trials were comparable, that the model is from the perspective of the healthcare provider and that the utility is the same for all patients.

A decision tree can be constructed to estimate the uncertainty of the cost difference between the two types of advice programme, due to the finite sample sizes of each trial. The proportions of each advice programme being successful (i.e. avoiding intervention) are represented by model variables with uncertainties which follow Beta distributions. Probabilities of the failure of the programmes are calculated using expression model variables to ensure that the total probability associated with each chance node is one.

library("rdecision")
# probabilities of programme success & failure
p.diet <- BetaModVar$new("P(diet)", "", alpha=12, beta=68-12)
p.exercise <- BetaModVar$new("P(exercise)", "", alpha=18, beta=58-18)
q.diet <- ExprModVar$new("1-P(diet)", "", rlang::quo(1-p.diet))
q.exercise <- ExprModVar$new("1-P(exercise)", "", rlang::quo(1-p.exercise))
# costs
c.diet <- 50
c.exercise <- ConstModVar$new("Cost of exercise programme", "GBP", 500)
c.stent <- 5000

The decision tree is constructed from nodes and edges as follows:

t.ds <- LeafNode$new("no stent")
t.df <- LeafNode$new("stent")
t.es <- LeafNode$new("no stent")
t.ef <- LeafNode$new("stent")
c.d <- ChanceNode$new("Outcome")
c.e <- ChanceNode$new("Outcome")
d <- DecisionNode$new("Programme")

e.d <- Action$new(d, c.d, cost=c.diet, label = "Diet")
e.e <- Action$new(d, c.e, cost=c.exercise, label = "Exercise")
e.ds <- Reaction$new(c.d, t.ds, p=p.diet, cost = 0, label = "success")
e.df <- Reaction$new(c.d, t.df, p=q.diet, cost=c.stent, label="failure")
e.es <- Reaction$new(c.e, t.es, p=p.exercise, cost=0, label="success")
e.ef <- Reaction$new(c.e, t.ef, p=q.exercise, cost=c.stent, label="failure")

DT <- DecisionTree$new(
  V = list(d, c.d, c.e, t.ds, t.df, t.es, t.ef),
  E = list(e.d, e.e, e.ds, e.df, e.es, e.ef)
)

The expected per-patient net cost of each option is obtained by evaluating the tree with expected values of all variables using DT$evaluate() and threshold values with DT$threshold(). Examination of the results of evaluation shows that the expected per-patient net cost of the diet advice programme is 4167.65 GBP and the per-patient net cost of the exercise programme is 3948.28 GBP, a point estimate saving of 219.37 GBP per patient if the exercise advice programme is adopted. By univariate threshold analysis, the exercise program will be cost saving when its cost of delivery is less than 719.73 GBP or when its success rate is greater than 26.6%.

The confidence interval of the cost saving is estimated by repeated evaluation of the tree, each time sampling from the uncertainty distribution of the two probabilities using, for example, DT$evaluate(setvars="random", N=1000) and inspecting the resulting data frame. From 1000 runs, the 95% confidence interval of the per patient cost saving is -507.18 GBP to 937.64 GBP, with 74.4% being cost saving, and it can be concluded that more evidence is required to be confident that the exercise programme is cost saving.

A three-state Markov model

Sonnenberg and Beck2 introduced an illustrative example of a semi-Markov process with three states: “Well”, “Disabled” and “Dead” and one transition between each state, each with a per-cycle probability. In rdecision such a model is constructed as follows. Note that transitions from a state to itself must be specified if allowed, otherwise the state would be a temporary state.

# create states
s.well <- MarkovState$new(name="Well", utility=1)
s.disabled <- MarkovState$new(name="Disabled",utility=0.7)
s.dead <- MarkovState$new(name="Dead",utility=0)
# create transitions, leaving rates undefined
E <- list(
  Transition$new(s.well, s.well),
  Transition$new(s.dead, s.dead),
  Transition$new(s.disabled, s.disabled),
  Transition$new(s.well, s.disabled),
  Transition$new(s.well, s.dead),
  Transition$new(s.disabled, s.dead)
)
# create the model
M <- SemiMarkovModel$new(V = list(s.well, s.disabled, s.dead), E)
# create transition probability matrix
snames <- c("Well","Disabled","Dead")
Pt <- matrix(
  data = c(0.6, 0.2, 0.2, 0, 0.6, 0.4, 0, 0, 1),
  nrow = 3, byrow = TRUE,
  dimnames = list(source=snames, target=snames)
)
# set the transition rates from per-cycle probabilities
M$set_probabilities(Pt)

With a starting population of 10,000, the model can be run for 25 years as follows. The output of the cycles function is the Markov trace, shown below, which replicates Table 2.2

# set the starting populations
M$reset(c(Well=10000, Disabled=0, Dead=0)) 
# cycle
MT <- M$cycles(25, hcc.pop=FALSE, hcc.cost=FALSE)
Years Well Disabled Dead Cumulative Utility
0 10000 0 0 0
1 6000 2000 2000 0.74
2 3600 2400 4000 1.268
3 2160 2160 5680 1.635
23 0 1 9999 2.375
24 0 0 10000 2.375
25 0 0 10000 2.375

Acknowledgements

In addition to using base R,3 redecision relies heavily on the R6 implementation of classes4 and the rlang package for error handling and non-standard evaluation used in expression model variables.5 Building the package vignettes and documentation relies on the testthat package,6 the devtools package7 and rmarkdown.10

Underpinning graph theory is based on terminology, definitions and algorithms from Gross et al,11 the Wikipedia glossary12 and links therein. Topological sorting of graphs is based on Kahn’s algorithm.13 Some of the terminology for decision trees was based on the work of Kaminski et al14 and an efficient tree drawing algorithm was based on the work of Walker.15 In semi-Markov models, representations are exported in the DOT language.16

Terminology for decision trees and Markov models in health economic evaluation was based on the book by Briggs et al1 and the output format and terminology follows ISPOR recommendations.18

Citations for examples used in vignettes are given in applicable vignette files.

References

1 Briggs A, Claxton K, Sculpher M. Decision modelling for health economic evaluation. Oxford, UK: Oxford University Press; 2006.

2 Sonnenberg FA, Beck JR. Markov Models in Medical Decision Making: A Practical Guide. Medical Decision Making 1993;13:322–38. https://doi.org/10.1177/0272989X9301300409.

3 R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2020.

4 Chang W. R6: Encapsulated classes with reference semantics. 2020.

5 Henry L, Wickham H. Rlang: Functions for base types and core r and ’tidyverse’ features. 2020.

6 Wickham H. Testthat: Get started with testing. The R Journal 2011;3:5–10.

7 Wickham H, Hester J, Chang W. Devtools: Tools to make developing r packages easier. 2020.

8 Xie Y, Allaire JJ, Grolemund G. R markdown: The definitive guide. Boca Raton, Florida: Chapman and Hall/CRC; 2018.

9 Allaire J, Xie Y, McPherson J, Luraschi J, Ushey K, Atkins A, et al. Rmarkdown: Dynamic documents for r. 2020.

10 Xie Y, Dervieux C, Riederer E. R markdown cookbook. Boca Raton, Florida: Chapman and Hall/CRC; 2020.

11 Gross JL, Yellen J, Zhang P. Handbook of Graph Theory. 2nd ed. Chapman and Hall/CRC.; 2013.

12 Wikipedia. Glossary of graph theory. Wikipedia 2021.

13 Kahn AB. Topological sorting of large networks. Communications of the ACM 1962;5:558–62. https://doi.org/10.1145/368996.369025.

14 Kamiński B, Jakubczyk M, Szufel P. A framework for sensitivity analysis of decision trees. Central European Journal of Operational Research 2018;26:135–59. https://doi.org/10.1007/s10100-017-0479-6.

15 Walker JQ. A node-positioning algorithm for general trees. Chapel Hill: University of North Carolina; 1989.

16 Gansner ER, Koutsofios E, North SC, Vo K-P. A technique for drawing directed graphs. IEEE Transactions on Software Engineering 1993;19:214–30. https://doi.org/10.1109/32.221135.

17 Briggs AH, Weinstein MC, Fenwick EAL, Karnon J, Sculpher MJ, Paltiel AD. Model Parameter Estimation and Uncertainty: A Report of the ISPOR-SMDM Modeling Good Research Practices Task Force-6. Value in Health 2012;15:835–42. https://doi.org/10.1016/j.jval.2012.04.014.

18 Siebert U, Alagoz O, Bayoumi AM, Jahn B, Owens DK, Cohen DJ, et al. State-Transition Modeling: A Report of the ISPOR-SMDM Modeling Good Research Practices Task Force-3. Value in Health 2012;15:812–20. https://doi.org/10.1016/j.jval.2012.06.014.