Pharmacoeconomic Models in Disease Management: A Guide for the Novice or the Perplexed

Clinical trials and meta-analyses of trials are models of clinical reality. A pharmacoeconomic model is a logical, quantitative blend of therapeutic and/or disease management strategies, evidence-based clinical outcomes, patient survival data and/or quality-of-life (utility) data, epidemiological data and costs. Pharmacoeconomic models can link evidence-based medicine to the local environment. They require locally appropriate resource consumption and cost information, so that the economic outcomes (e.g. cost and cost-effectiveness of therapy) are current and locally relevant. Decision analytical models represent a sequence of chance events and decisions over time and are appropriate for acute episodes of illness, whereas Markov models represent recurring health states and are useful in describing chronic illness. Epidemiological models combine clinical trial data with observational data, and can be used for predicting the efficiency of risk management strategies such as vaccination and antihypertensive therapy. User-friendly commercial modelling software is available. For maximum credibility, pharmacoeconomic models should build on validated disease management protocols and/or landmark clinical trials or meta-analyses of trials. They should also adhere to published standards for economic analysis, including the use of locally relevant comparators, discounting to present value, extensive sensitivity analysis, and appropriate health utility values. Models should be presented fully with the logical structure plus all decision probabilities and/or state transition probabilities plus unit costs and resource consumption. A standard reporting format for publication in peer-reviewed journals has been suggested. Models can be timely, adaptable, relatively inexpensive, and often the only way to obtain appropriate information on the clinical, economic and humanistic outcomes of disease management protocols. However, without due care they can be obscure and open to bias and misunderstanding. Both the analyst and the user must avoid mistaking obscurity for profundity.