Skip to contents

Run a probabilistic sensitivity analysis (PSA)

Source: R/bim_psa.R
bim_run_psa.Rd

Performs a Monte Carlo PSA by repeatedly sampling uncertain parameters from their assumed statistical distributions and re-running the budget impact model for each draw. This produces a distribution of budget impact outcomes that reflects joint parameter uncertainty.

Distributional assumptions

  • Prevalence, diagnosed rate, treated rate, eligible rate – Beta distribution parameterised from the base-case value and a standard error.

  • Drug cost – LogNormal distribution parameterised from the base-case value and a coefficient of variation (CV).

Usage

bim_run_psa(
  model,
  n_sim = 1000L,
  prevalence_se = NULL,
  diagnosed_rate_se = NULL,
  treated_rate_se = NULL,
  eligible_rate_se = NULL,
  cost_cv = NULL,
  year = NULL,
  scenario = "base",
  seed = NULL
)

# S3 method for class 'bim_psa'
print(x, ...)

Arguments

model

A bim_model object (base case).

n_sim

integer(1). Number of Monte Carlo simulations. Default 1000L.

prevalence_se

numeric(1). Standard error for prevalence. If NULL (default), prevalence is held fixed.

diagnosed_rate_se

numeric(1) or NULL. SE for diagnosed rate.

treated_rate_se

numeric(1) or NULL. SE for treated rate.

eligible_rate_se

numeric(1) or NULL. SE for eligible rate.

cost_cv

numeric(1) or NULL. Coefficient of variation applied to all drug costs simultaneously. If NULL, costs are held fixed.

year

integer(1). Budget impact year to summarise. Defaults to the last year in the model.

scenario

character(1). Scenario to use. Default "base".

seed

integer(1) or NULL. Random seed for reproducibility.

Value

An object of class bim_psa: a list with elements:

simulations

data.frame with one row per simulation: sim, budget_impact, and the sampled parameter values.

summary

data.frame with mean, SD, median, and 95 \ interval of budget impact.

year

The year summarised.

scenario

The scenario used.

n_sim

Number of simulations run.

base_bi

Base-case budget impact for reference.

See also

bim_plot_psa(), bim_run_dsa()

Examples

pop <- bim_population(
  indication  = "Disease X", country = "custom",
  years = 1:5, prevalence = 0.003, n_total_pop = 42e6,
  diagnosed_rate = 0.60, treated_rate = 0.45, eligible_rate = 0.30
)
ms <- bim_market_share(
  population     = pop,
  treatments     = c("Drug C (SoC)", "Drug A (new)"),
  new_drug       = "Drug A (new)",
  shares_current = c("Drug C (SoC)" = 1.0, "Drug A (new)" = 0.0),
  shares_new     = c("Drug C (SoC)" = 0.8, "Drug A (new)" = 0.2)
)
costs <- bim_costs(
  treatments = c("Drug C (SoC)", "Drug A (new)"),
  drug_costs = c("Drug C (SoC)" = 500, "Drug A (new)" = 25000)
)
model <- bim_model(pop, ms, costs)
set.seed(1)
psa <- bim_run_psa(model, n_sim = 200L, prevalence_se = 0.0005,
                   eligible_rate_se = 0.05, cost_cv = 0.10)
print(psa)
#> htaBIM Probabilistic Sensitivity Analysis
#>   Year: 5 | Scenario: base | Simulations: 200 / 200 converged
#>   Base-case budget impact: GBP 50,004,500
#> 
#>   PSA summary (budget impact):
#>     Mean:   GBP 50,289,917
#>     Median: GBP 48,436,316
#>     SD:     GBP 12,182,224
#>     95% CrI: GBP 32,535,003  to  GBP 73,741,327