Clinical Decision Analysis (Decision Trees)

Description

Models a decision as a tree of choices, chance events with probabilities and outcomes with values, and computes expected values with sensitivity analysis.

Decision trees make uncertainty explicit: decision nodes, chance nodes with probabilities (from data or calibrated judgement) and terminal outcomes with values or utilities. Folding back the tree computes expected values; one- and two-way sensitivity analyses and threshold analysis show which parameters could change the decision; probabilistic sensitivity analysis via Monte Carlo propagates uncertainty. Clinical medicine uses this for test-or-treat decisions, which map neatly onto UX questions like whether to add a confirmation step (test) versus allow and offer undo (treat), or whether to show an AI warning given false-positive rates.

Attributes

Loss type
Unmigrated
Stage
Strategy, Design
Effort
Days

Steps

  1. Frame the decision, alternatives, time horizon and outcome metric.
  2. Draw the tree with decision and chance nodes, using realistic event sequences.
  3. Estimate probabilities from logs, research or calibrated experts, and values for outcomes.
  4. Fold back to compute expected values for each alternative.
  5. Run one-way, threshold and probabilistic sensitivity analyses.
  6. Present the decision, its robustness and the parameters worth researching further.

Sources

  • Weinstein & Fineberg, Clinical Decision Analysis (1980)
  • Hunink et al., Decision Making in Health and Medicine (2nd ed., 2014)
  • Raiffa, Decision Analysis: Introductory Lectures on Choices under Uncertainty (1968)

Also known as

  • Decision tree analysis
  • Expected value decision analysis
  • Decision modelling

Categories

  • Strategy & Collaboration
  • Game Theory & Decision Science