Ask Mallard
Analysis method

Decision tree model

4 min read

A decision tree model combines choices, chance events and their consequences to estimate expected costs or outcomes over a defined pathway.

Fit

When to use it, and when not to

Use it when

  • Your decision has a finite pathway and a clearly defined endpoint.
  • You can estimate branch probabilities and consequences for each strategy.

Not when

  • Events recur over long follow-up. Consider a Markov model. Read: Markov model
  • You want to predict an individual outcome from features. A predictive classification tree answers that different question.

Before you start

What you need

Decide these first
  1. Strategies, population and perspectiveState whose costs count and the time horizon.
  2. Mutually exclusive, complete chance branchesProbabilities leaving each chance node must sum to one.
  3. Path-specific costs and outcomesGive sources, units, dates and uncertainty. Avoid counting the same cost twice.
It rests on these
  1. The pathways capture the important consequences.How to check: Check the structure with people who understand the decision and test plausible alternatives.
  2. Each branch probability fits its preceding history.How to check: Use conditional probabilities; do not multiply unrelated marginal risks.
  3. Inputs and assumptions have credible ranges.How to check: Vary parameters and structure, and preserve dependence between related inputs.

Plain language

How it works

Start with a choice among strategies. Chance nodes then branch into events. Terminal paths carry the costs and outcomes that follow the full path.

Reading the method

Multiply conditional probabilities along each path. Weight terminal consequences by those path probabilities and add them within each strategy. Compare strategies using the same perspective and horizon.

Scope and evidence

Economic evaluation is not one of the five research areas this hierarchy covers. That describes the hierarchy, not the quality of a model. The cited applications show finite-path economic models. Model reporting should distinguish assumed inputs from observed data.

The method in outline
A choice leads to chance events and consequences. Branches must cover the possibilities without overlap. The full path determines the consequence.A choice leads to chance events and consequencesChoose a strategyChance event under that strategyPath A consequencePath B consequenceWeight and sum path consequencesBranches must cover the possibilities without overlap. The full path determines the consequence.
Conceptual diagram. It shows the structure of the idea, not data from any study.

Worked example

A worked example

Invented for illustration
Setting
An analyst compares two short diagnostic pathways from a payer perspective.
Question
What is the expected cost of the proposed pathway?
Approach
Its mutually exclusive branches have probabilities 0.8 and 0.2, with total path costs of $100 and $600. The comparator costs $300.
Result
The proposed pathway costs $200 on average versus $300, a saving of $100 per person.
What it means
These are assumed modeled costs. Costs alone do not establish which pathway has better health outcomes.

In practice

Recent studies PubMed search · not vetted

Searching PubMed…

Watch out

Common mistakes

The one to remember

A decision tree carries the assumptions and evidence supplied to it. Expected costs or outcomes are modeled results, not observed treatment effects; repeated events can make a tree unwieldy.

  • Confusing the decision model with a predictive tree.Instead: Define strategies and consequences rather than fitting a classifier.
  • Choosing the cheapest strategy without outcomes.Instead: Compare relevant benefits, harms and uncertainty as well as costs.

Inside Mallard

How Mallard uses this

Plans do not link to this method yet. Mallard does not build or run economic decision models.

No Mallard code example points to this page.
Atlas node analysis.decision_tree_model, the id a plan cue records.

Related Learn pages

Related designs and methods

Evidence

Sources & further reading

  1. Cost-effectiveness study of therapeutic approaches for mucosal leishmaniasis (opens in a new tab) · Cad Saude Publica · doi 10.1590/0102-311XEN132523
  2. Cost-Effectiveness Analysis of Early versus Late Debridement of Superficial Triangular Fibrocartilage Complex Tears (opens in a new tab) · J Hand Microsurg · doi 10.1055/s-0042-1757179
  3. Modeling good research practices--overview: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force--1 (opens in a new tab) · Value Health · doi 10.1016/j.jval.2012.06.012
Prepared with AI assistance. PubMed records were checked by PMID, title, authors, year and DOI on 2026-11-06. The leishmaniasis and wrist-injury studies are primary decision-tree applications with explicit perspectives and sensitivity analyses. Caro and colleagues support transparent structure, uncertainty and model reporting. No independent professional approval is claimed.
Last source review: 2026-11-06
AI-generated educational content
Mallard uses reputable methodological and reporting sources, but this material may contain errors or omit important context. Verify important methodological decisions with the cited sources and, when appropriate, a qualified research methodologist or statistician.