Concept Architecture
Concept
Theoretically, Stochastic Programming is a mathematical optimisation framework for decision-making under uncertainty in which one or more model parameters are represented as random variables with known or estimated probability distributions. It is founded on probability theory, optimisation theory and decision analysis, enabling optimal decisions to be identified before and after uncertain events occur. In health economics, stochastic programming is used to optimise healthcare resource allocation, capacity planning and intervention strategies when future demand, costs or epidemiological conditions are uncertain.
Mathematically, stochastic programming formulates optimisation problems by incorporating uncertain parameters into the objective function and/or constraints. The optimisation seeks a decision that maximises expected benefit or minimises expected cost across a probability distribution of future scenarios while satisfying feasibility requirements. Two-stage and multi-stage stochastic programming models are the most widely recognised formulations, with recourse decisions allowing adaptation after uncertainty is realised.
In practice, stochastic programming is implemented by defining uncertain parameters, generating representative scenarios with associated probabilities, formulating the optimisation model, and solving it using specialised optimisation algorithms. In health economics it is applied to hospital capacity planning, workforce scheduling, pharmaceutical supply chain management, pandemic preparedness, vaccine allocation and healthcare budget planning under uncertainty.
Purpose
Used to optimise healthcare decisions under uncertainty by explicitly incorporating probabilistic future scenarios into resource allocation and planning models.
Mathematical Formulae
Primary Formula
Two-stage stochastic programming:
min? {c?x + ???[Q(x, ?)]}
where
Q(x, ?) = min? {q(?)?y : W(?)y = h(?) ? T(?)x}
where:
- x = first-stage decision variables
- y = second-stage (recourse) decision variables
- ? = random parameters
- c = first-stage cost vector
- q(?) = second-stage cost vector
- ?? = expectation operator
Supporting Formulae
Scenario approximation:
min? [c?x + ????? p?Q(x, ??)]
where:
- p? = probability of scenario s
- ?p? = 1
Related Mathematical Methods
- Scenario analysis
- Linear programming
- Integer programming
- Robust optimisation
- Dynamic programming
- Monte Carlo simulation
- Numerical optimisation
Example
A health authority must determine the number of intensive care beds to prepare before influenza season.
Three demand scenarios are estimated:
| Scenario | Probability | Additional Cost (� million) |
|---|---|---|
| Low demand | 0.20 | 2.0 |
| Medium demand | 0.50 | 4.5 |
| High demand | 0.30 | 8.0 |
The expected second-stage cost is
0.20(2.0) + 0.50(4.5) + 0.30(8.0) = �5.05 million.
The optimisation identifies the initial bed capacity that minimises total expected cost while allowing recourse decisions once actual demand is observed.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| SUMPRODUCT | =SUMPRODUCT(B2:B4,C2:C4) | Calculates expected costs or benefits across scenarios. |
| Solver Add-in | Minimise expected total cost subject to constraints | Optimises first-stage healthcare decisions. |
| RAND | =RAND() | Generates random values for scenario testing or simulation. |
| IF | =IF(B2>Threshold,Action1,Action2) | Implements scenario-dependent recourse decisions in simplified models. |
VBA (Optional)
A VBA procedure can automate stochastic programming by generating scenarios, solving successive optimisation problems, and summarising expected costs and optimal first-stage decisions.
Sources
- Birge JR, Louveaux F. Introduction to Stochastic Programming. Springer.
- Shapiro A, Dentcheva D, Ruszczy?ski A. Lectures on Stochastic Programming: Modeling and Theory. SIAM.
- Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. Oxford University Press.
- Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. Oxford University Press.
- NICE. Health Technology Evaluation Manual.
- ISPOR Good Practice Reports on decision-analytic modelling.
Related Concepts (5)
Frequently Asked Questions (6)
What is stochastic programming?
An optimisation framework incorporating parameter uncertainty directly into the process, seeking a solution that performs well across possible parameter values.
Source: Birge & Louveaux 1997
How does stochastic programming use probabilities of future states?
Stochastic programming builds the probabilities of different future conditions directly into the optimisation, choosing the decision that performs best on average across those weighted possibilities rather than under a single assumed scenario. It can also allow later decisions to be adjusted once uncertainty resolves, so an initial choice is made knowing that corrective action will follow. This use of probability distinguishes it from approaches that guard only against the worst case. Birge and Louveaux (2011) set out the framework.
Source: Birge & Louveaux 2011
How does stochastic programming handle uncertainty?
Stochastic programming handles uncertainty by representing the uncertain parameters through a set of scenarios or probability distributions and optimising a measure of performance, commonly the expected value, across them. It often distinguishes decisions made before the uncertainty is resolved from recourse decisions made after, allowing the plan to adapt as information arrives. By building the uncertainty and the possibility of later adjustment into the optimisation, it finds a solution that performs well on average, or by another criterion, over the range of possible outcomes.
Source: Birge & Louveaux 1997
What is recourse in stochastic programming?
Recourse in stochastic programming refers to decisions that can be made after some of the uncertainty has been resolved, adjusting to the outcome that occurs. A two-stage stochastic program has first-stage decisions made before the uncertainty is known and second-stage, or recourse, decisions made after, which correct or adapt the plan given the realised outcome. Modelling recourse captures the value of being able to respond to information, so the first-stage decision anticipates the adjustments possible later, giving a more realistic and flexible plan under uncertainty.
Source: Birge & Louveaux 1997
How does stochastic programming differ from robust optimisation?
Stochastic programming and robust optimisation both address parameter uncertainty but differ in approach. Stochastic programming represents uncertainty by probability distributions or scenarios and typically optimises expected performance, using the likelihood of outcomes and often allowing recourse. Robust optimisation defines an uncertainty set of possible values and seeks a solution good in the worst case within it, without requiring probabilities. Stochastic programming uses probabilistic information and averages over it, while robust optimisation guards against the worst case, so they suit different attitudes to uncertainty.
Source: Birge & Louveaux 1997
How does stochastic programming apply to health decisions?
Stochastic programming applies to health planning and resource decisions made under uncertainty, such as allocating capacity, planning services, or managing supply when demand, costs, or effects are uncertain, and where later adjustments are possible. By optimising over scenarios and allowing recourse, it produces plans that perform well across possible futures and can adapt as uncertainty resolves. This suits decisions where commitments must be made before outcomes are known but can be adjusted later, offering a structured way to plan under uncertainty in health systems.
Source: Birge & Louveaux 1997
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 3 Oct 2025
Content version: 1.0.0
Canonical Identity
- Term code
- HE-EM-DM-097
Stable URI · Machine-readable · Resolvable · CC BY 4.0