The Data

A state health department is considering a new diabetes prevention program. The evidence suggests the program works, but with substantial uncertainty. The effect could be large, modest, or even zero. (Data are simulated for illustration.)

Estimated Program Effect: Distribution of Possible Outcomes
Probability distribution of true effect
Mean estimate

Next: What happens if we implement the program now? What happens if we wait? The answer depends on which scenario turns out to be true.

Decision Under Uncertainty

Every decision carries risk. Implementing a program that doesn't work wastes resources. Waiting to implement a program that works delays health benefits. The expected value calculation weighs these trade-offs.

Go

Implement Now

Act on current evidence
If effective (85%) Gain 7,500 QALYs, cost $250M. Net benefit: +$125M at $50K/QALY threshold
If ineffective (15%) Gain 0 QALYs, cost $250M. Net benefit: -$250M
Expected Net Benefit
+$68.8M
Wait

Wait for Better Evidence

Delay decision by 2 years
If effective (85%) Eventually implement, but 2 years of benefits lost. Net: +$87M (delayed)
If ineffective (15%) Avoid wasting $250M. Net benefit: $0
Expected Net Benefit
+$74.0M

Expected Value of a Decision

The expected value is the probability-weighted average of all possible outcomes. It accounts for both:

  • Upside risk: The chance of gains if the decision is right
  • Downside risk: The chance of losses if the decision is wrong

In this case, both decisions have positive expected value. But which is higher depends on how uncertain we are and how costly waiting is.

Next: If we could learn the truth before deciding, how much would that information be worth? This is the Expected Value of Perfect Information.

EVPI Explained

The Expected Value of Perfect Information (EVPI) tells us the maximum we should spend on research. It's the difference between deciding with perfect knowledge and deciding under uncertainty.

Decision Tree: With and Without Perfect Information

Next: How do the key parameters affect the value of information? Explore interactively to build intuition.

Interactive Value of Information

Adjust the parameters below to see how uncertainty, costs, and population size affect the optimal decision and the value of additional research.

Probability Program is Effective 85%
Effect Size if Effective (QALYs per person) 0.15
Cost per Person $5,000
Target Population 50,000
+$68.8M
EV: Implement Now
+$74.0M
EV: Wait 2 Years
$37.5M
Max Research Value
Optimal Decision
Implement Now

Next: What's the key insight for health economists? "Wait for better evidence" is itself a decision that must be evaluated.

Key Insight

Waiting for more evidence has real costs. Every day of delay means foregone benefits for people who would have been helped. Economics treats the decision to wait as itself a choice with consequences.

Decision diagram showing two paths: Implement Now (with uncertain outcomes) and Wait for Evidence (with delayed benefits and research costs). Both paths lead to Net Health Benefit, with Value of Information connecting uncertainty to the wait path.

The decision to wait is itself a choice with expected costs and benefits. VOI analysis makes these trade-offs explicit.

Questions Economists Ask About Evidence and Action

Expected value: What's the probability-weighted average of all possible outcomes?
Opportunity cost of waiting: How many people are harmed by delayed implementation if the program works?
Value of information: How much would we pay to know the truth before deciding?
Decision threshold: At what level of uncertainty should we act versus wait?
Research prioritization: Which uncertainties matter most for the decision?

Value of Information Analysis

Value of Information (VOI) quantifies how much additional research is worth. It provides:

  • EVPI (Expected Value of Perfect Information): Upper bound on what any research could be worth
  • EVPPI (Partial Perfect Information): Value of resolving uncertainty about a specific parameter
  • EVSI (Sample Information): Value of a specific study with a specific sample size

VOI transforms vague calls for "more research" into precise questions about which research, how much, and whether it's worth the investment.

Key Takeaway

"Wait for better evidence" is not a neutral position. It's a decision with real costs: delayed benefits for patients, continued spending on possibly inferior alternatives, and the expense of the research itself. Value of Information analysis makes these trade-offs explicit and quantifiable. The goal isn't to eliminate uncertainty before acting. The goal is to act wisely given the uncertainty we face. This is what economists mean by "decision-making under uncertainty."