Skip to content
Scintria
ToolsGrant writerFunding
Start free
ScintriaStart free
Simulation Lab

Cost-effectiveness model

Is a new intervention worth its cost? A Markov model follows a cohort through being well, becoming sick and dying, and compares costs and quality-adjusted life years (QALYs) with and without the intervention, then shows how sure that answer is.

Below 1 means the intervention helps.

$
$
$

Often set from a country's cost-effectiveness threshold.

More options
%
%
%
years
%
%

The spread used in the sensitivity analysis, as a share of each input.

Cost per QALY gained

$588

Cost-effective at your threshold

Chance it is cost-effective

66%

At $1,500 per QALY, across 600 draws of the inputs.

Extra cost per person

$240

QALYs gained per person

0.409

Net monetary benefit

$373

Life-years, with / without

16.52 / 16.07

Cost-effectiveness plane

Each dot is one draw of the inputs. Dots below the dashed line are cost-effective at your threshold.

-1.0k01.0k-0.5000.501ThresholdQALYs gained per personExtra cost per person

Acceptability curve

0%20%40%60%80%100%01.0k2.0k3.0k4.0kYouWillingness to pay per QALY ($)Probability cost-effective

For your protocol

A three-state Markov cohort model (well, sick, dead) with yearly cycles over 20 years, discounting costs and outcomes at 3.0% a year, compared current care with an intervention costing $60 per person per year that changes the yearly risk of becoming sick by a relative risk of 0.75. Per person, the intervention costs $240 more and gains 0.409 QALYs (incremental cost-effectiveness ratio $588 per QALY). In a probabilistic sensitivity analysis of 600 draws, the probability that it is cost-effective at a willingness to pay of $1,500 per QALY is 66%.

How this simulation works

A cohort starts well. Each year some die of other causes, some become sick (less often with the intervention, by the relative risk), and some sick people die. Costs and quality of life accrue each year for the people in each state, discounted, with states counted at the start of each yearly cycle (no half-cycle correction). The probabilistic sensitivity analysis redraws every input (beta distributions for probabilities and utilities, gamma for costs, lognormal for the relative risk) with the spread you choose, and reruns the model each time. A teaching and early-planning model: a real evaluation needs inputs from the literature, more health states and often a longer horizon.