| Title: | Discrete Event Simulation for Cost-Effectiveness Modelling |
|---|---|
| Description: | A package designed to perform discrete event simulation for cost-effectiveness modelling. |
| Authors: | Valerie Aponte Ribero [aut, cre], Javier Sanchez Alvarez [aut], F. Hoffmann La Roche Ltd [cph] |
| Maintainer: | Valerie Aponte Ribero <[email protected]> |
| License: | Apache License (>= 2) |
| Version: | 0.1.2 |
| Built: | 2024-11-08 05:15:02 UTC |
| Source: | https://github.com/Roche/descem |
Defining costs for events and intervention
add_cost(.data = NULL, cost, evt, trt, cycle_l = NULL, cycle_starttime = 0)add_cost(.data = NULL, cost, evt, trt, cycle_l = NULL, cycle_starttime = 0)
.data |
Existing cost data |
cost |
Value or expression to calculate the cost estimate |
evt |
Vector of events for which this cost is applicable |
trt |
Vector of interventions for which this cost is applicable |
cycle_l |
Cycle length; only needed if costs are calculated per cycle |
cycle_starttime |
Cycle when costs start being accrued; only needed if costs are calculated per cycle |
Costs can be defined by writing expressions and objects in the cost argument whose execution will be delayed until the model runs.
This function accepts the use of pipes (%>%) to define multiple costs.
A list of costs
add_cost(evt = c("start","idfs","ttot"),trt = "int",cost = cost.int*fl.int + cost.idfs)add_cost(evt = c("start","idfs","ttot"),trt = "int",cost = cost.int*fl.int + cost.idfs)
Defining parameters that may be used in model calculations
add_item(.data = NULL, ...)add_item(.data = NULL, ...)
.data |
Existing data |
... |
Items to define for the simulation |
The functions to add/modify events/inputs use lists. Whenever several inputs/events are added or modified, it's recommended to group them within one function, as it reduces the computation cost.
So rather than use two add_item with a list of one element, it's better to group them into a single add_item with a list of two elements.
A list of items
add_item(fl.idfs = 0) add_item(util_idfs = if(psa_bool){rnorm(1,0.8,0.2)} else{0.8}, util.mbc = 0.6, cost_idfs = 2500)add_item(fl.idfs = 0) add_item(util_idfs = if(psa_bool){rnorm(1,0.8,0.2)} else{0.8}, util.mbc = 0.6, cost_idfs = 2500)
Define the modifications to other events, costs, utilities, or other items affected by the occurrence of the event
add_reactevt(.data = NULL, name_evt, input)add_reactevt(.data = NULL, name_evt, input)
.data |
Existing data for event reactions |
name_evt |
Name of the event for which reactions are defined. |
input |
Expressions that define what happens at the event, using functions as defined in the Details section |
There are a series of objects that can be used in this context to help define the event reactions.
The following functions may be used to define event reactions within this add_reactevt() function:
modify_item() | Adds & Modifies items/flags/variables for future events
new_event() | Adds events to the vector of events for that patient
modify_event() | Modifies existing events by changing their time
Apart from the items defined with add_item(), we can also use standard variables that are always defined within the simulation:
curtime | Current event time (numeric)
prevtime | Time of the previous event (numeric)
cur_evtlist | Named vector of events that is yet to happen for that patient (named numeric vector)
evt | Current event being processed (character)
i | Patient being iterated (character)
simulation | Simulation being iterated (numeric)
The model will run until curtime is set to Inf, so the event that terminates the model should modify curtime and set it to Inf.
add_reactevt(name_evt = "start",input = {}) add_reactevt(name_evt = "idfs",input = {modify_item(list("fl.idfs"= 0))})add_reactevt(name_evt = "start",input = {}) add_reactevt(name_evt = "idfs",input = {modify_item(list("fl.idfs"= 0))})
Define events and the initial event time
add_tte(.data = NULL, trt, evts, other_inp = NULL, input)add_tte(.data = NULL, trt, evts, other_inp = NULL, input)
.data |
Existing data for initial event times |
trt |
The intervention for which the events and initial event times are defined |
evts |
A vector of the names of the events |
other_inp |
A vector of other input variables that should be saved during the simulation |
input |
The definition of initial event times for the events listed in the evts argument |
Events need to be separately defined for each intervention.
For each event that is defined in this list, the user needs to add a reaction to the event using the add_reactevt() function which will determine what calculations will happen at an event.
A list of initial events and event times
add_tte(trt="int",evts = c("start","ttot","idfs","os"), input={ start <- 0 idfs <- draw_tte(1,'lnorm',coef1=2, coef2=0.5) ttot <- min(draw_tte(1,'lnorm',coef1=1, coef2=4),idfs) os <- draw_tte(1,'lnorm',coef1=0.8, coef2=0.2) })add_tte(trt="int",evts = c("start","ttot","idfs","os"), input={ start <- 0 idfs <- draw_tte(1,'lnorm',coef1=2, coef2=0.5) ttot <- min(draw_tte(1,'lnorm',coef1=1, coef2=4),idfs) os <- draw_tte(1,'lnorm',coef1=0.8, coef2=0.2) })
Defining utilities for events and interventions
add_util(.data = NULL, util, evt, trt, cycle_l = NULL, cycle_starttime = 0)add_util(.data = NULL, util, evt, trt, cycle_l = NULL, cycle_starttime = 0)
.data |
Existing utility data |
util |
Value or expression to calculate the utility estimate |
evt |
Events for which this utility is applicable |
trt |
Interventions for which this utility is applicable |
cycle_l |
Cycle length; only needed if utilities are calculated per cycle |
cycle_starttime |
Cycle when utilities start being accrued; only needed if utilities are calculated per cycle |
Utilities can be defined by writing expressions and objects in the cost argument whose execution will be delayed until the model runs.
This function accepts the use of pipes (%>%) to define multiple utilities.
A list of utilities
add_util(evt = c("start","idfs","ttot"), trt = c("int", "noint"), util = util.idfs.ontx * fl.idfs.ontx + util.idfs.offtx * (1-fl.idfs.ontx))add_util(evt = c("start","idfs","ttot"), trt = c("int", "noint"), util = util.idfs.ontx * fl.idfs.ontx + util.idfs.offtx * (1-fl.idfs.ontx))
Calculate the cost-effectiveness acceptability curve (CEAC) for a DES model with a PSA result
ceac_des(wtp, results, interventions = NULL)ceac_des(wtp, results, interventions = NULL)
wtp |
Vector of length >=1 with the willingness to pay |
results |
The list object returned by |
interventions |
A character vector with the names of the interventions to be used for the analysis |
A data frame with the CEAC results
## Not run: ceac_des(seq(from=10000,to=500000,by=10000),results) ## End(Not run)## Not run: ceac_des(seq(from=10000,to=500000,by=10000),results) ## End(Not run)
Draw from a beta distribution based on mean and se
draw_beta(value, se, seed = NULL)draw_beta(value, se, seed = NULL)
value |
A vector of the mean values |
se |
A vector of the standard errors of the means |
seed |
An integer which will be used to set the seed for this draw. |
A single estimate from the beta distribution based on given parameters
draw_beta(value=0.8,se=0.2)draw_beta(value=0.8,se=0.2)
Draw from a gamma distribution based on mean and se
draw_gamma(value, se, seed = NULL)draw_gamma(value, se, seed = NULL)
value |
A vector of the mean values |
se |
A vector of the standard errors of the means |
seed |
An integer which will be used to set the seed for this draw. |
A single estimate from the gamma distribution based on given parameters
draw_gamma(value=0.8,se=0.2)draw_gamma(value=0.8,se=0.2)
Draw from a restricted Gompertz distribution
draw_resgompertz( n, shape, rate, lower_bound = 0, upper_bound = Inf, seed = NULL )draw_resgompertz( n, shape, rate, lower_bound = 0, upper_bound = Inf, seed = NULL )
n |
The number of observations to be drawn |
shape |
The shape parameter of the Gompertz distribution, defined as in the coef() output on a flexsurvreg object |
rate |
The rate parameter of the Gompertz distribution, defined as in the coef() output on a flexsurvreg object |
lower_bound |
The lower bound of the restricted distribution |
upper_bound |
The upper bound of the restricted distribution |
seed |
An integer which will be used to set the seed for this draw. |
Estimate(s) from the restricted Gompertz distribution based on given parameters
draw_resgompertz(1,shape=0.05,rate=0.01,lower_bound = 50)draw_resgompertz(1,shape=0.05,rate=0.01,lower_bound = 50)
Draw a time to event from a list of parametric survival functions
draw_tte( n_chosen = 1, dist = "exp", coef1 = 1, coef2 = NULL, coef3 = NULL, hr = 1, seed = NULL )draw_tte( n_chosen = 1, dist = "exp", coef1 = 1, coef2 = NULL, coef3 = NULL, hr = 1, seed = NULL )
n_chosen |
The number of observations to be drawn |
dist |
The distribution; takes values 'lnorm','weibullPH','weibull','llogis','gompertz','gengamma','gamma','exp' |
coef1 |
First coefficient of the distribution, defined as in the coef() output on a flexsurvreg object |
coef2 |
Second coefficient of the distribution, defined as in the coef() output on a flexsurvreg object |
coef3 |
Third coefficient of the distribution, defined as in the coef() output on a flexsurvreg object |
hr |
A hazard ratio |
seed |
An integer which will be used to set the seed for this draw. |
A vector of time to event estimates from the given parameters
draw_tte(n_chosen=1,dist='exp',coef1=1,hr=1)draw_tte(n_chosen=1,dist='exp',coef1=1,hr=1)
Calculate the Expected Value of Perfect Information (EVPI) for a DES model with a PSA result
evpi_des(wtp, results, interventions = NULL)evpi_des(wtp, results, interventions = NULL)
wtp |
Vector of length >=1 with the willingness to pay |
results |
The list object returned by |
interventions |
A character vector with the names of the interventions to be used for the analysis |
A data frame with the EVPI results
## Not run: evpi_des(seq(from=10000,to=500000,by=10000),results) ## End(Not run)## Not run: evpi_des(seq(from=10000,to=500000,by=10000),results) ## End(Not run)
Extract PSA results from a treatment
extract_psa_result(x, element, trt)extract_psa_result(x, element, trt)
x |
The output_psa data frame from the list object returned by |
element |
Variable for which PSA results are being extracted (single string) |
trt |
Intervention for which PSA results are being extracted (single string) |
A dataframe with PSA results from the specified intervention
## Not run: extract_psa_result(results$output_psa,"costs","int") ## End(Not run)## Not run: extract_psa_result(results$output_psa,"costs","int") ## End(Not run)
Modify the time of existing events
modify_event(evt, env_ch = NULL)modify_event(evt, env_ch = NULL)
evt |
A list of events and their times |
env_ch |
Environment in which to save list (should not be defined by user) |
The functions to add/modify events/inputs use lists. Whenever several inputs/events are added or modified, it's recommended to group them within one function, as it reduces the computation cost.
So rather than use two modify_event with a list of one element, it's better to group them into a single modify_event with a list of two elements.
## Not run: modify_event(list("os"=40, "ttot"=curtime+0.0001)) ## End(Not run)## Not run: modify_event(list("os"=40, "ttot"=curtime+0.0001)) ## End(Not run)
Modify the value of existing items
modify_item(list_item, env_ch = NULL)modify_item(list_item, env_ch = NULL)
list_item |
A list of items and their values or expressions |
env_ch |
Environment in which to save list (should not be defined by user) |
The functions to add/modify events/inputs use lists. Whenever several inputs/events are added or modified, it's recommended to group them within one function, as it reduces the computation cost.
So rather than use two modify_item with a list of one element, it's better to group them into a single modify_item with a list of two elements.
## Not run: modify_item(list(cost.idfs = 500, cost.tx = cost.tx + 4000)) ## End(Not run)## Not run: modify_item(list(cost.idfs = 500, cost.tx = cost.tx + 4000)) ## End(Not run)
Generate new events to be added to existing vector of events
new_event(evt, env_ch = NULL)new_event(evt, env_ch = NULL)
evt |
Event name and event time |
env_ch |
Environment in which to save list (should not be defined by user) |
The functions to add/modify events/inputs use lists. Whenever several inputs/events are added or modified, it's recommended to group them within one function, as it reduces the computation cost.
So rather than use two new_event with a list of one element, it's better to group them into a single new_event with a list of two elements.
## Not run: new_event(list("ae"=5)) new_event(list("ae"=5,"nat.death" = 100)) ## End(Not run)## Not run: new_event(list("ae"=5)) new_event(list("ae"=5,"nat.death" = 100)) ## End(Not run)
Run the simulation
RunSim( trt_list = c("int", "noint"), common_all_inputs = NULL, common_pt_inputs = NULL, unique_pt_inputs = NULL, init_event_list = NULL, evt_react_list = evt_react_list, util_ongoing_list = NULL, util_instant_list = NULL, util_cycle_list = NULL, cost_ongoing_list = NULL, cost_instant_list = NULL, cost_cycle_list = NULL, npats = 500, n_sim = 1, psa_bool = NULL, ncores = 1, drc = 0.035, drq = 0.035, input_out = NULL, ipd = TRUE, debug = FALSE )RunSim( trt_list = c("int", "noint"), common_all_inputs = NULL, common_pt_inputs = NULL, unique_pt_inputs = NULL, init_event_list = NULL, evt_react_list = evt_react_list, util_ongoing_list = NULL, util_instant_list = NULL, util_cycle_list = NULL, cost_ongoing_list = NULL, cost_instant_list = NULL, cost_cycle_list = NULL, npats = 500, n_sim = 1, psa_bool = NULL, ncores = 1, drc = 0.035, drq = 0.035, input_out = NULL, ipd = TRUE, debug = FALSE )
trt_list |
A vector of the names of the interventions evaluated in the simulation |
common_all_inputs |
A list of inputs common across patients that do not change within a simulation |
common_pt_inputs |
A list of inputs that change across patients but are not affected by the intervention |
unique_pt_inputs |
A list of inputs that change across each intervention |
init_event_list |
A list of initial events and event times. If no initial events are given, a "Start" event at time 0 is created automatically |
evt_react_list |
A list of event reactions |
util_ongoing_list |
A list of utilities that are accrued at an ongoing basis |
util_instant_list |
A list of utilities that are accrued instantaneously at an event |
util_cycle_list |
A list of utilities that are accrued in cycles |
cost_ongoing_list |
A list of costs that are accrued at an ongoing basis |
cost_instant_list |
A list of costs that are accrued instantaneously at an event |
cost_cycle_list |
A list of costs that are accrued in cycles |
npats |
The number of patients to be simulated |
n_sim |
The number of simulations to run per patient |
psa_bool |
A boolean to determine if PSA should be conducted. If n_sim > 1 and psa_bool = FALSE, the differences between simulations will be due to sampling |
ncores |
The number of cores to use for parallel computing |
drc |
The discount rate for costs |
drq |
The discount rate for LYs/QALYs |
input_out |
A vector of variables to be returned in the output data frame |
ipd |
A boolean to determine if individual patient data should be returned. If set to false, only the main aggregated outputs will be returned (slightly speeds up code) |
debug |
A boolean to determine if non-parallel RunEngine function should be used, which facilitates debugging. Setting this option to true will ignore the value of ncores |
A list of data frames with the simulation results
## Not run: RunSim(trt_list=c("int","noint"), common_all_inputs = common_all_inputs, common_pt_inputs = common_pt_inputs, unique_pt_inputs = unique_pt_inputs, init_event_list = init_event_list, evt_react_list = evt_react_list, util_ongoing_list = util_ongoing_list, util_instant_list = util_instant_list, cost_ongoing_list = cost_ongoing_list, cost_instant_list = cost_instant_list, npats = 500, n_sim = 1, psa_bool = FALSE, ncores = 1, drc = 0.035, drq = 0.035, ipd = TRUE) ## End(Not run)## Not run: RunSim(trt_list=c("int","noint"), common_all_inputs = common_all_inputs, common_pt_inputs = common_pt_inputs, unique_pt_inputs = unique_pt_inputs, init_event_list = init_event_list, evt_react_list = evt_react_list, util_ongoing_list = util_ongoing_list, util_instant_list = util_instant_list, cost_ongoing_list = cost_ongoing_list, cost_instant_list = cost_instant_list, npats = 500, n_sim = 1, psa_bool = FALSE, ncores = 1, drc = 0.035, drq = 0.035, ipd = TRUE) ## End(Not run)
Deterministic results for a specific treatment
summary_results_det(out = final_output, trt = NULL)summary_results_det(out = final_output, trt = NULL)
out |
The final_output data frame from the list object returned by |
trt |
The reference treatment for calculation of incremental outcomes |
A dataframe with absolute costs, LYs, QALYs, and ICER and ICUR for each intervention
## Not run: summary_results_det(results$final_output,trt="int") ## End(Not run)## Not run: summary_results_det(results$final_output,trt="int") ## End(Not run)
Summary of PSA outputs for a treatment
summary_results_psa(out = output_psa, trt = NULL)summary_results_psa(out = output_psa, trt = NULL)
out |
The output_psa data frame from the list object returned by |
trt |
The reference treatment for calculation of incremental outcomes |
A data frame with mean and 95% CI of absolute costs, LYs, QALYs, ICER and ICUR for each intervention from the PSA samples
## Not run: summary_results_psa(results$output_psa, trt="int") ## End(Not run)## Not run: summary_results_psa(results$output_psa, trt="int") ## End(Not run)