Method: Model structure:- a generally applicable UK (18+) population based individual patient-level simulation model with annual time cycles including survival state (alive/dead) over time. Health Survey for England (HSE) 2008 data (N = 14,925) informs modelled trajectories over time for diet (average daily fruit and vegetable portions), alcohol (average weekly units), smoking (yes/no) and physical activity (average number of minutes per week), by assuming that the person maintains the same percentile rank as at baseline for each behaviour unless intervention occurs. Annual probability of dying uses UK life tables, adjusted for the behaviours using a UK study by Kvaavik 2010, which published an integrated analysis of hazard ratios for four behaviours (N=4,886 followed for 20 years). Each simulated individual has an associated EQ-5D health-related quality of life conditional on the behaviours and age/gender using a regression based on HSE data. To model an intervention requires evidence on changes in the four behaviours, plus assumptions for time lags between behaviour change and achieving full effect in risk reductions, which we elicited from epidemiological / modelling experts. Case Study:- RCT of an online health behaviour intervention targeted at young people starting Sheffield University (“U@Uni”) (N = 1,445). Intervention costs were from a staff time survey within the RCT. For outcomes, 1-month follow-up effect of U@Uni on behaviours is assumed sustained for the first year. For extrapolation, formal elicitation of maximum duration of the treatment effect was undertaken with two psychological experts.
Result: U@Uni provides an estimated +0.0002 QALYs at £102 per person over lifetime. The incremental cost-effectiveness ratio=£442,337/QALY, 95% CI=£417,933-£466,074, with 0% probability of cost-effectiveness at £20,000/QALY. The threshold intervention cost to achieve £20,000/QALY is £4.63 per recipient. Analysis will be updated when 6-month follow-up is collected.
Conclusion: The case study suggests U@Uni is unlikely to be cost-effective given current intervention costs and roll-out to other Universities would require substantially reduced cost per recipient. The model framework has enabled integrated, correlated, analysis of four key health behaviours. Given evidence of effect on some or all four behaviours, the framework can be applied to evaluate an enormous range of public health interventions/policies.