Multi Criteria Decision Analysis (MCDA)

Multi criteria decision analysis (MCDA) is a structured, transparent approach to decision making that evaluates alternatives against multiple criteria simultaneously. It systematically incorporates both quantitative evidence and value judgments, assigning weights to criteria that reflect their relative importance and scoring alternatives on each criterion to produce an overall ranking.

Many decisions in health care, policy, and research involve tradeoffs among competing objectives that cannot be reduced to a single metric. A new treatment may offer survival benefits but at high cost and with significant side effects. Traditional cost effectiveness analysis captures the tradeoff between cost and health, but it may not reflect other considerations that decision makers value, such as equity, innovation, unmet need, or patient preferences. MCDA provides a framework for making these tradeoffs explicit and transparent. It is increasingly used by some HTA bodies (for example, the Lombardy region in Italy has used an MCDA based on the EVIDEM framework in its HTA process), pharmaceutical companies, and hospital formulary committees as a complement to or alternative to standard cost effectiveness analysis.

The MCDA process typically follows several steps. First, the decision problem is defined, including the alternatives being compared and the criteria against which they will be evaluated. Common criteria in health care MCDA include clinical effectiveness, safety, cost effectiveness, burden of disease severity, quality of evidence, innovativeness, equity, and patient reported outcomes.

Second, a performance matrix is constructed, showing how each alternative performs on each criterion using available evidence, usually on natural scales (such as survival in months or cost in dollars). The performance on each criterion is then converted into a score on a common scale (such as 0 to 100) that reflects stakeholders' preferences, for example with a value function.

Third, criteria are weighted to reflect their relative importance. Weighting methods include direct rating or point allocation (for example, stakeholders distribute 100 points across the criteria), swing weighting (stakeholders assess the importance of moving from worst to best on each criterion), and discrete choice experiments (stakeholders choose between hypothetical alternatives, and weights are inferred from their choices). Different stakeholder groups (patients, clinicians, payers, manufacturers) may assign different weights, and exploring this variation is a key feature of MCDA.

Fourth, scores and weights are combined using an aggregation rule. The simplest approach is a linear additive model, where the overall score for each alternative is the weighted sum of its scores on each criterion. Other approaches handle what the additive model cannot: multiplicative models allow interactions between criteria, and outranking methods allow veto thresholds, so that a very poor result on one criterion cannot be offset by good results on others. Sensitivity analysis explores how changes in weights or scores affect the ranking of alternatives.

MCDA is appropriate when decisions involve multiple objectives that cannot be captured by a single metric, when stakeholder values and preferences are important inputs to the decision, when transparency in the decision making process is desired, and when traditional economic evaluation methods (such as cost effectiveness analysis alone) are insufficient. It is increasingly used in HTA (particularly when QALY based analysis is controversial or when additional criteria are relevant), in formulary decisions, in health priority setting, and in research funding allocation.

A common error is selecting criteria that overlap conceptually (double counting), such as including both "clinical effectiveness" and "efficacy in subgroups" as separate criteria. Another mistake is using weights derived from one stakeholder group (such as manufacturers) without testing the sensitivity of results to alternative weightings. Some MCDA applications fail to explicitly define the measurement scale for each criterion, leading to ambiguous scores. The aggregation model chosen (linear additive, multiplicative) embeds assumptions about the tradeoff structure that should be justified rather than adopted by default. Finally, using MCDA as a black box that produces a ranking without explaining the reasoning behind the scores, weights, and aggregation model undermines the transparency that MCDA is designed to provide.

In a hypothetical example, a hospital formulary committee evaluates three biologic agents for moderate to severe psoriasis using MCDA with six criteria: efficacy (PASI 75 response rate), safety (serious adverse event rate), cost per patient per year, quality of evidence (number and quality of supporting RCTs), convenience of administration (subcutaneous injection or intravenous infusion, and dosing interval), and patient reported outcomes (such as itch and quality of life scores). Clinicians and patient representatives assign weights through swing weighting, and each drug is scored on each criterion using published trial data. The linear additive model produces overall scores of 72, 68, and 61 for the three agents. Sensitivity analysis shows that the ranking is stable across all plausible weight distributions except when cost is weighted above 40%, at which point the second agent overtakes the first.