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}.

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

`install.packages("rdecision")`

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
<- BetaModVar$new("P(diet)", "", alpha = 12L, beta = 68L - 12L)
p.diet <- BetaModVar$new("P(exercise)", "", alpha = 18L, beta = 58L - 18L)
p.exercise <- ExprModVar$new("1-P(diet)", "", rlang::quo(1.0 - p.diet))
q.diet <- ExprModVar$new("1-P(exercise)", "", rlang::quo(1.0 - p.exercise))
q.exercise # costs
<- 50.0
c.diet <- ConstModVar$new("Cost of exercise programme", "GBP", 500.0)
c.exercise <- 5000.0 c.stent
```

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

```
<- LeafNode$new("no stent")
t.ds <- LeafNode$new("stent")
t.df <- LeafNode$new("no stent")
t.es <- LeafNode$new("stent")
t.ef <- ChanceNode$new("Outcome")
c.d <- ChanceNode$new("Outcome")
c.e <- DecisionNode$new("Programme")
d
<- Action$new(d, c.d, cost=c.diet, label = "Diet")
e.d <- Action$new(d, c.e, cost=c.exercise, label = "Exercise")
e.e <- Reaction$new(c.d, t.ds, p = p.diet, cost = 0.0, label = "success")
e.ds <- Reaction$new(c.d, t.df, p = q.diet, cost = c.stent, label="failure")
e.df <- Reaction$new(c.e, t.es, p = p.exercise, cost = 0.0, label="success")
e.es <- Reaction$new(
e.ef p = q.exercise, cost = c.stent, label = "failure"
c.e, t.ef,
)
<- DecisionTree$new(
DT 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 -563.42 GBP to 1032.17 GBP, with 70.9% being
cost saving, and it can be concluded that more evidence is required to
be confident that the exercise programme is cost saving.

Sonnenberg and Beck^{2} 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
<- MarkovState$new(name = "Well", utility = 1.0)
s.well <- MarkovState$new(name = "Disabled", utility = 0.7)
s.disabled <- MarkovState$new(name = "Dead", utility = 0.0)
s.dead # create transitions, leaving rates undefined
<- list(
E $new(s.well, s.well),
Transition$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)
Transition
)# create the model
<- SemiMarkovModel$new(V = list(s.well, s.disabled, s.dead), E)
M # create transition probability matrix
<- c("Well","Disabled","Dead")
snames <- matrix(
Pt data = c(0.6, 0.2, 0.2, 0.0, 0.6, 0.4, 0.0, 0.0, 1.0),
nrow = 3L, byrow = TRUE,
dimnames = list(source=snames, target=snames)
)# set the transition rates from per-cycle probabilities
$set_probabilities(Pt) M
```

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
$reset(c(Well = 10000L, Disabled = 0L, Dead = 0L))
M# cycle
<- M$cycles(25L, hcc.pop = FALSE, hcc.cost = FALSE) MT
```

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 |

In addition to using base R^{3}, `redecision`

relies heavily on the `R6`

implementation of
classes^{4} 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`

package^{7} and
`rmarkdown`

^{10}.

Underpinning graph theory is based on terminology, definitions and
algorithms from Gross *et al*^{11}, the Wikipedia
glossary^{12} 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
al*^{14} 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 al*^{1}
and the output format and terminology follows ISPOR
recommendations^{18}.

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

1.

Briggs, A., Claxton, K. & Sculpher, M.
*Decision modelling for health economic evaluation*. (Oxford
University Press, 2006).

2.

Sonnenberg, F. A. & Beck, J. R. Markov Models in
Medical Decision Making: A Practical Guide. *Medical Decision
Making* **13,** 322–338 (1993).

3.

R
Core Team. *R: A language and environment for statistical
computing*. (R Foundation for Statistical Computing, 2020). at
<https://www.R-project.org/>

4.

Chang, W. *R6: Encapsulated classes with
reference semantics*. (2020). at <https://CRAN.R-project.org/package=R6>

5.

Henry, L. & Wickham, H. *Rlang:
Functions for base types and core r and ’tidyverse’ features*.
(2020). at <https://CRAN.R-project.org/package=rlang>

6.

7.

Wickham, H., Hester, J. & Chang, W.
*Devtools: Tools to make developing r packages easier*. (2020).
at <https://CRAN.R-project.org/package=devtools>

8.

Xie,
Y., Allaire, J. J. & Grolemund, G. *R markdown: The definitive
guide*. (Chapman and Hall/CRC, 2018). at <https://bookdown.org/yihui/rmarkdown>

9.

Allaire, J., Xie, Y., McPherson, J., Luraschi,
J., Ushey, K., Atkins, A., Wickham, H., Cheng, J., Chang, W. &
Iannone, R. *Rmarkdown: Dynamic documents for r*. (2020). at
<https://github.com/rstudio/rmarkdown>

10.

Xie, Y., Dervieux, C. & Riederer, E. *R
markdown cookbook*. (Chapman and Hall/CRC, 2020). at <https://bookdown.org/yihui/rmarkdown-cookbook>

11.

Gross, J. L., Yellen, J. & Zhang, P.
*Handbook of Graph Theory*. (Chapman and Hall/CRC., 2013). at
<https://doi.org/10.1201/b16132>

12.

Wikipedia. Glossary of graph theory.
*Wikipedia* (2021). at <https://en.wikipedia.org/wiki/Glossary_of_graph_theory>

13.

14.

Kamiński, B., Jakubczyk, M. & Szufel, P. A framework for
sensitivity analysis of decision trees. *Central European Journal
of Operational Research* **26,** 135–159 (2018).

15.

Walker, J. Q. *A node-positioning algorithm
for general trees*. (University of North Carolina, 1989). at <http://www.cs.unc.edu/techreports/89-034.pdf>

16.

Gansner, E. R., Koutsofios, E., North, S. C.
& Vo, K.-P. A technique
for drawing directed graphs. *IEEE Transactions on Software
Engineering* **19,** 214–230 (1993).

17.

Briggs, A. H., Weinstein, M. C., Fenwick, E. A.
L., Karnon, J., Sculpher, M. J. & Paltiel, A. D. Model Parameter
Estimation and Uncertainty: A Report of the ISPOR-SMDM Modeling Good
Research Practices Task Force-6. *Value in Health*
**15,** 835–842 (2012).

18.

Siebert, U., Alagoz, O., Bayoumi, A. M., Jahn,
B., Owens, D. K., Cohen, D. J. & Kuntz, K. M. State-Transition
Modeling: A Report of the ISPOR-SMDM Modeling Good Research Practices
Task Force-3. *Value in Health* **15,** 812–820
(2012).