Modelling future outcomes to estimate the cost-effectiveness of interventions related to childhood/adolescent overweight/obesity

Background Childhood overweight and obesity is a major public health priority. Overweight or obese children are likely to become overweight or obese adults, are likely to be at increased risk of a wide range of chronic diseases (eg, type 2 diabetes, cardiovascular disease, and some cancers), and can have broader effects on lifestyle, health, and wellbeing compared with their peers. A clear need exists for cost-effective interventions or programmes that can help with the problems associated with childhood obesity. However, the conduct of cost-effectiveness analyses (and health technology assessment more broadly) in childhood obesity presents challenges. Translation of measures of effectiveness, for example body-mass index, into the potential effects of interventions on future health outcomes, such as prevention of diabetes, is one such challenge. The current methods available to model future outcomes in this way are limited. We aimed to develop a decision-analytic model to estimate the effect of interventions aimed at childhood overweight and obesity, for use in cost-effectiveness analyses in a public health context. Methods We developed a decision-analytic modelling framework using a simple model structure (in the first instance) consisting of three stages: prediction of adult weight status (healthy, overweight, or obese) from childhood weight status; prediction of obesity-related disease or events in adult years (type 2 diabetes, coronary heart disease, stroke, and colorectal cancer); and estimation of the effects of interventions for childhood obesity on prevention of these future events. The model was developed with longitudinal data for the 1958 UK birth cohort in which participants were followed up between ages 11 and 33 years, applying UK90 growth reference standards. Incidence rates for obesity-related health events were calculated on the basis of a 2009 systematic review and meta-analysis. The model was applied to a case study with data from an exploratory trial of a novel, drama-based, school-located obesity intervention, the Healthy Lifestyles Programme (HeLP), to present an example of the potential use of the model in a public health setting. HeLP was delivered to children aged 9–10 years during three school terms, and aimed to deliver a general healthy lifestyle message, seeking to change behaviours in the family, at school, and individually. Three key behaviours were emphasised: decrease in consumption of sweetened fizzy drinks, increase in proportion of healthy snacks to unhealthy snacks, and a reduction in television viewing and other screen-based activities. The empirical or mechanistic element of the model, prediction of adult weight status, was a major challenge, and sensitivity analyses explore the effect of the use of different sources of longitudinal data for prediction of adult weight status. Findings In a base-case scenario with a typical control cohort of 1000 children aged 11 years, the model predicted adult weight status at age 33 years, and thereafter an expected minimum of 274 health events (mainly coronary heart disease and type 2 diabetes) in this cohort between age 33 and 63 years (30 year time horizon). The model was used to estimate the expected effect of the HeLP intervention (a change in the distribution of weight status at age 11 years) on the expected number of health events over time, compared with standard practice and applying a third party payer (National Health Service or Personal Social Services) perspective. Outputs from the model can be used to estimate cost-effectiveness based on various input measures and assumptions. The sensitivity of judgments on cost effectiveness to different structural and data inputs were considered. Interpretation Findings suggest a reduction in future obesity-related health effects even when interventions are used in a broad public health context, and when average population effects are small. The model developed for use in cost-effectiveness analyses is a simple first stage, and the parsimonious nature of the model is subject to obvious limitations, but its potential usefulness in a public health decision making context is promising