OBJECTIVES: General population health transition data are frequently unavailable in a suitable format to support evaluation studies, including cost-effectiveness, health policy and intervention modelling. We aimed to: (1) develop a representative synthetic English population capturing health status at small area level, and (2) apply three key multi-state transition modelling methods for estimating life expectancies (LEs) and healthy life expectancies (HLEs): the Sullivan life table method, continuous-time and discrete-time Markov models.

METHODS: Gompertz distributions were fitted to 2017-2019 age-specific health status and mortality data from the Office for National Statistics (ONS) to generate a synthetic population. A continuous-time multi-state model was fitted to the data to estimate HLEs and LEs. Finally, discrete-time transition probabilities estimated from the continuous-time model were used to create an Excel-based Markov model, providing an alternative method of estimating HLEs and LEs. All estimates were compared with HLEs and LEs at English local authority level in 2017-2019, estimated by the ONS using the Sullivan method.

RESULTS: The different methods tested replicated ONS estimates well:

1. The synthetic population showed similar distributions of ages at illness and death as ONS-reported estimates.
2. Differences in mean HLEs and LEs (synthetic population vs ONS-estimated) were small (HLEs: -0.1% for men, 1.8% for women; LEs: 0.1% and 0.3%), and differences for most local authorities were smaller than +/-10%.
3. Differences for the discrete- vs continuous-time HLEs and LEs were also similar (-0.7–0.7%).

CONCLUSIONS: We have generated a synthetic general population at England local area level suitable for use in evaluation studies, adopting an approach which could equally be applied to populations with chronic diseases or to other countries. The continuous-time and discrete-time Markov models fitted to the synthetic population both replicated ONS Sullivan life table method estimates of HLEs and LEs well, offering flexibility for researchers.