Topic

Value of information

Value of information analysis estimates how much better decisions would be if more evidence were available, and so what further research is worth. The expected value of perfect information sets an upper limit, while partial and sample information measures show which parameters and which study designs offer the most.

Concepts in this topic

• Adaptive Value of InformationAdaptive value of information applies VOI analysis to staged research, re-valuing further data at each interim look to decide whether to continue.
• ENBS CalculationThe process of computing the expected net benefit of sampling, the expected value of sample information from a study minus its expected cost.
• Expected LossThe probability-weighted average cost of making an incorrect decision under uncertainty, calculated across all possible states and their probabilities.
• Expected Net Benefit of SamplingExpected net benefit of sampling is the expected population value of the decision improvement produced by a proposed sample, minus the financial and opportunity costs of obtaining that evidence.
• Expected Value of Partial Perfect InformationThe maximum value of completely resolving uncertainty in a specific subset of a model's parameters, while other parameters remain uncertain.
• Expected Value of Perfect InformationExpected value of perfect information is the expected increase in net benefit from eliminating all decision-relevant uncertainty before choosing among the available strategies.
• Expected Value of Sample InformationExpected value of sample information is the expected increase in the value of a decision when a specified study’s possible data can update uncertainty before a strategy is chosen.
• Gaussian ProcessA flexible statistical framework modelling an unknown function by treating any finite set of its values as jointly normally distributed.
• Linear PoolA method combining multiple experts' probability distributions by taking a weighted arithmetic average, producing a single pooled distribution.
• Logarithmic PoolA method combining multiple experts' probability distributions by taking a weighted geometric average rather than a simple arithmetic average.
• Opportunity LossThe difference in value between the outcome of the optimal decision and the outcome of the decision actually made under uncertainty.
• Optimal Sample SizeA study sample size selected to meet a stated design objective, such as adequate power or precision, or to maximise the expected value of information after accounting for research costs and decision delays.
• Population EVPIThe expected value of perfect information calculated for an entire affected patient population over time, rather than for a single patient.
• Response SurfaceA fitted mathematical approximation describing how a model's output varies with its inputs, built from limited runs and used to predict untested combinations.
• Sequential AnalysisA statistical approach evaluating data as it accumulates, allowing early stopping once a predefined level of evidence is reached, rather than a fixed sample size.
• Value of InformationValue of information (VOI) analysis estimates the expected benefit of reducing decision uncertainty, to judge whether more research is worth its cost.
• Value of Information AnalysisThe application of value of information methods to a health economic decision, typically used to judge whether further research is worth conducting.
• VOI AnalysisVOI analysis (value of information analysis) takes a model's PSA through EVPI, EVPPI, EVSI and ENBS in turn to decide if more research is worth its cost.