The Influence of Cost-Effectiveness and Other Factors on NICE Decisions

Abstract

Background: The National Institute for Health and Care Excellence (NICE) emphasises that cost-effectiveness is not the only consideration in health technology appraisal and is increasingly explicit about other factors considered relevant. Observing NICE decisions and the evidence considered in each appraisal allows us to “reveal” its implicit weights. Objectives: This study aims to investigate the influence of cost-effectiveness and other factors on NICE decisions and to investigate whether NICE’s decision-making has changed through time. Methods: We build on and extend the modelling approaches in Devlin and Parkin (2004) and Dakin et al (2006). Wemodel NICE’s decisions as binary choices: i.e. recommendations for or against use of a health care technology in a specific patient group. Independent variables comprised: the clinical and economic evidence regarding the technology; the characteristics of the patients, disease or treatment; and contextual factors affecting the conduct of health technology appraisal. Data on all NICE decisions published by December 2011 were obtained from HTAinSite.Results: Cost-effectiveness alone correctly predicted 82% of decisions; few other variables were significant and alternative model specifications led to very small variations in model performance. The odds of a positive NICE recommendation differed significantly between musculoskeletal disease, respiratory disease, cancer and other conditions. The accuracy with which the model predicted NICE recommendations was slightly improved by allowing for end of life criteria, uncertainty, publication date, clinical evidence, only treatment, paediatric population, patient group evidence, appraisal process, orphan status, innovation and use of probabilistic sensitivity analysis, although these variables were not statistically significant. Although there was a non-significant trend towards more recent decisions having a higher chance of a positive recommendation, there is currently no evidence that the threshold has changed over time. The model with highest prediction accuracy suggested that a technology costing £40,000/quality-adjusted life year (QALY) would have a 50% chance of NICE rejection (75% at £52,000/QALY; 25% at £27,000/QALY). Discussion: Past NICE decisions appear to have been based on a higher threshold than the £20,000-£30,000/QALY range that is explicitly stated. However, this finding may reflect consideration of other factors that drive a small number of NICE decisions or cannot be easily quantified.