National Statistical

Healthy life expectancy is an important measure to understand not just how long people are living, but how long they are living in good health, at national and local level. It could be for example that length of life is similar in two places, but in one of those two, individuals could expect to spend more years of their later life in poor health. As we publish our latest healthy life expectancy figures for England and Wales, ONS’s Greg Ceely explains why we have used a new methodology to produce the data.  

Measuring healthy life expectancy (HLE) paints a picture of how society’s health is changing over time, and the challenges different population groups may face. The data provides insight for healthcare planners and policymakers, allowing them to better understand health needs and demands on services. In the spirit of continuous improvement, we are always assessing the strength of our statistics. With that in mind, we have developed a new approach for estimating healthy life expectancy which creates more reliable data and deeper insights at local level.  

How did we calculate it for previous years?  

We produce our healthy life expectancy data using a combination of death registrations information, and survey responses from the Annual Population Survey (APS). This survey provides information on how a person sees their own health, which we classify into “good” and “not good” health to combine with the life expectancy which we calculate from the mortality data. We produce healthy life expectancy by local authority, but reducing survey sample sizes have made results at that level less robust, especially when we break down results into different age groups and sex to measure their self-reported health.

Past results have shown large differences in health within local areas over time, in part due to the impact of the pandemic. Understanding whether these differences are temporary, or indicative of a longer trend is crucial for policy makers and health care planners. Will some areas, for example, take longer to recover from a fall in healthy life expectancy? And within that, it may be that certain population groups are more impacted than others. With these important questions in mind, we sought to develop a new method for estimating good health prevalence to apply in healthy life expectancy measurement, which would provide greater understanding about what is driving changes in health.  

A new approach 

Our objective was to create a simple model capable of providing plausible distributions of good health by age and sex across all local areas, suitable for reliably estimating healthy life expectancy at birth and at age 65. Statistical modelling techniques allow us to overcome some of the limitations of not having the necessary sample sizes previously available. 

We have continued to use APS as our source data, now adjusting it using logistic regression modelling. This technique allows an outcome (‘good’ or ‘not good’ health) to be predicted using the influence of other factors, and can include interactions between these other factors. Logistic regression allows the relationship between age and health in a local area to be different depending on your sex and where you reside in the country.  

More robust results 

The new model produces a more plausible distribution of health by age in local areas, while having very little effect on the distributions at national level over and above the previous method. In general, the new method reduces scores in areas which previously had very high HLE, and increases them in areas with very low estimates, reducing the range of results. These areas with highest and lowest results previously also had quite volatile results, meaning they could change a lot between years, which suggests our new method is an improvement. When ranking local authorities’ HLE results in both old and new methods, the orders are similar. Thirteen of the previous top 20 local areas for HLE are also in the new top 20, and 14 of the previous bottom 20 are also in the new bottom 20. We have updated our data back to 2011-13 with the new method, so that comparisons over time can be made with our latest results. 

We have developed these methods in discussion with our statistical partners and health bodies including the Department for Health and Social Care, the Office for Health Improvement and Disparities, Welsh Government, Public Health Wales, National Records Scotland, Northern Ireland Statistics and Research Agency, the Department of Health Northern Ireland and ONS Local. Keeping these users informed of our changes ensures our new method produces results suitable for their purposes. 

What’s next? 

A methodology article detailing our chosen techniques and how results compare to the previous method has been published alongside our latest results. While our analysis to test the effect of the new method on HLE was restricted to England, Wales and Northern Ireland, the new method will also be applied when estimating HLE for Scotland’s council areas, facilitating a UK wide local area publication due in 2025.  

We are exploring alternative options to change the method in the longer term, which may look similar to our latest approach, but will aim to future-proof this statistic against further changes to the source data. As ever, if you have any questions or comments on the latest results or methodology, you can contact us at health.data@ons.gov.uk.