Decision-analytic models to simulate health outcomes and costs in heart failure: a systematic review

Abstract

Chronic heart failure (CHF) is a critical public health issue with increasing effect on the healthcare budgets of developed countries. Various decision-analytic modelling approaches exist to estimate the cost effectiveness of health technologies for CHF. We sought to systematically identify these models and describe their structures. We performed a systematic literature review in MEDLINE/PreMEDLINE, EMBASE, EconLit and the Cost-Effectiveness Analysis Registry using a combination of search terms for CHF and decision-analytic models. The inclusion criterion required 'use of a mathematical model evaluating both costs and health consequences for CHF management strategies'. Studies that were only economic evaluations alongside a clinical trial or that were purely descriptive studies were excluded. We identified 34 modelling studies investigating different interventions including screening (n = 1), diagnostics (n = 1), pharmaceuticals (n = 15), devices (n = 13), disease management programmes (n = 3) and cardiac transplantation (n = 1) in CHF. The identified models primarily focused on middle-aged to elderly patients with stable but progressed heart failure with systolic left ventricular dysfunction. Modelling approaches varied substantially and included 27 Markov models, three discrete-event simulation models and four mathematical equation sets models; 19 studies reported QALYs. Three models were externally validated. In addition to a detailed description of study characteristics, the model structure and output, the manuscript also contains a synthesis and critical appraisal for each of the modelling approaches. Well designed decision models are available for the evaluation of different CHF health technologies. Most models depend on New York Heart Association (NYHA) classes or number of hospitalizations as proxy for disease severity and progression. As the diagnostics and biomarkers evolve, there is the hope for better intermediate endpoints for modelling disease progression as those that are currently in use all have limitations.