Economic evaluation in health encompasses the set of analytical methods used to compare the costs and consequences of alternative health interventions. The three primary types — cost-effectiveness analysis (CEA), cost-utility analysis (CUA), and cost-benefit analysis (CBA) — share a common analytical framework but differ in how outcomes are measured. All economic evaluations require defining the perspective (whose costs and benefits count?), the comparator (what is the alternative?), the time horizon (how far into the future?), and the discount rate (how are future costs and benefits valued relative to present ones?). Decision-analytic models — decision trees for short-term outcomes and Markov models for chronic diseases — project costs and outcomes beyond the trial period using the best available evidence. Sensitivity analysis tests whether conclusions are robust to uncertainty in key parameters.
Economic evaluation provides the analytical backbone for health resource allocation decisions. While the three main types (CEA, CUA, CBA) have different outcome measures, they share a common analytical framework that involves several key methodological choices.
The perspective determines whose costs and benefits are counted. A healthcare system perspective includes only direct medical costs (drugs, hospitalizations, physician visits). A societal perspective adds patient costs (transportation, lost wages), caregiver costs, and productivity losses. The same intervention can look cost-effective from a healthcare perspective but not from a societal perspective (or vice versa) depending on how non-medical costs distribute. Most HTA bodies specify a required perspective; the US Second Panel on Cost-Effectiveness recommends reporting both healthcare and societal perspectives.
The comparator is what the new intervention replaces. An intervention must be compared to the relevant alternative — typically the current standard of care, not placebo. A drug that beats placebo may not beat an existing generic; evaluating it against placebo would overstate its added value. The choice of comparator profoundly affects the ICER: if the comparator is ineffective, the new intervention looks excellent; if the comparator is already good, incremental gains are small and the ICER rises.
Decision-analytic models extend the analysis beyond the evidence directly observed in trials. A decision tree maps short-term decision points and their probabilistic outcomes (e.g., surgery succeeds or fails, with different downstream costs and health states). A Markov model adds time: patients transition between health states (e.g., well → mild disease → severe disease → death) at each time cycle (typically annual), accumulating costs and QALYs in each state. Markov models are standard for chronic diseases where long-term outcomes matter. Parameters come from clinical trials, observational studies, administrative databases, and expert opinion, each with uncertainty.
Sensitivity analysis is the critical quality control step. One-way analysis varies each parameter individually to identify which drives the result. Threshold analysis finds the value at which the conclusion changes. Probabilistic sensitivity analysis (PSA) simultaneously varies all parameters according to their probability distributions across thousands of Monte Carlo simulations, producing a distribution of ICERs that reflects the full uncertainty in the model. The results are displayed as cost-effectiveness scatter plots (cloud of simulated ICERs) and cost-effectiveness acceptability curves (probability of cost-effectiveness at each WTP threshold). PSA transforms the analysis from a point estimate ("the ICER is $50,000/QALY") into a probabilistic statement ("there is a 75% probability that the intervention is cost-effective at a $50,000 threshold").