Population estimates & GP registers: why the difference?

How many people live in England? The most recent ONS estimates (from mid 2015) put the number at 54.8 million. However, at the same point in time, NHS Digital estimates that the number of patients registered with GP practices in England was 57.1 million – 2.3 million more (4%). This gap increased by 88,000 between 2014 and 2015.

What causes this difference, and why is it increasing?

It’s not possible to say for certain. There are a few possible reasons, which aren’t mutually exclusive:

  • Over-counting in GP practice registers
  • Under-counting in population estimates
  • Different definitions of who counts as ‘resident’ in the country

The NHS is attempting to audit its GP registers and remove ‘ghost patients’. But is it possible that some of the difference reflects a genuine discrepancy? ONS’s process for estimating internal migration relies heavily on health service data, including changes in where people are registered for GP services. However, as ONS admit, there are some aspects of population change that this won’t necessarily detect.[1]

What light can we shed on this issue, and how does the discrepancy between GP registers and population estimates vary across the country?

Where are the differences largest?

The difference between GP registered populations and ONS’s population estimates varies a lot across England.[2] The table and map below show where the differences are largest – at both ends of the scale. Positive percentages show local authorities where the GP register is higher; negative percentages show areas where ONS’s estimated population is higher.

 

 

 

Which areas have higher GP registered populations than the ONS estimate?

Several places where GP populations are much higher than ONS estimates are cities with a high number of university students. This implies that some students aren’t being picked up in the ONS population estimates, but are nevertheless registered with GPs in the area. In the table above, Oxford and Cambridge are the most obvious candidates for this explanation. Other cities with a higher GP registered population, like Norwich and Leeds, also have a large student population. If we look at small areas within local authorities (LSOAs), most of the largest differences between GP populations and ONS estimates are in student areas (with Warwick, Cambridge and Kingston providing the largest discrepancies in small areas).

But it’s unlikely that all the variation is explained by student populations. As the more detailed map at the end of this blog shows, there is a large swathe of west and south-west London where almost every small area has a higher GP registered population than ONS estimates. It’s possible – though, of course, not certain – that this partly reflects a systematic underestimate of the number of people living in places like Hammersmith & Fulham, Brent, Ealing, Merton and Wandsworth.

Questions should also be raised about places like Peterborough and Boston, which we already know have had relatively high inward migration, and who are also among the areas with the highest discrepancy between GP registered populations and ONS’s estimates. Previous population estimates have turned out to underestimate the scale of population change between censuses, particularly in some local areas.

It’s very likely that some of the differences are down to double-counting or ‘ghost’ patients in GP registers – e.g. people who have moved house or left the country without de-registering, or died. There have been reports of the NHS trying to address this issue by removing people from the register if they haven’t contacted their GP in five years. Also, short-term migrants aren’t included in ONS’s population estimates, but some might be registered with GPs, so they could be a factor.

And which areas have lower GP registered populations than the ONS estimate?

Richmondshire (North Yorks) and Forest Heath (Suffolk) are the only places with a GP population more than 10% smaller than ONS’s estimated population. Both cases are probably explained by the presence of a large military population – Catterick Garrison in Richmondshire, and RAF Mildenhall in Forest Heath. These facilities house a substantial number of residents who are included in population estimates but not registered with a local NHS GP practice. Of the list above, Rutland, Exeter, Gosport and Hambleton are also home to military installations.

The other negative outliers are found on the Welsh border – Forest of Dean, Shropshire, and Herefordshire – where a proportion of the population are probably registered with a Welsh GP practice rather than an English one.

Focus on London, Birmingham and Manchester

The three maps below show the same data, but for small areas (LSOAs) with a population of around 1,500. The colour scale is the same as the national map above. As mentioned previously, the London map shows that a large portion of west and south west London has a higher GP registered population than the numbers estimated by ONS. By contrast, some areas in central London have a much lower GP registered population – which may reflect that people living in these areas are less likely to register with an NHS GP than those living elsewhere.

Only built-up areas are shaded with a colour on these maps – land that is outside of built-up areas (e.g. large parks, agricultural land, empty land) is shaded grey.

 

 

 

 

Notes on the data

NHS Digital’s data on GP practice populations shows, for each GP practice, which LSOA its patients live in. For example, the data shows that over 99% of Brownlow Medical Group’s patients live in Liverpool – but a handful live in neighbouring Knowsley and Sefton. The analysis here adds up the total number of people living in each area who are registered with any GP practice, and compares this number with the ONS’s estimate of the resident population of that area.

21,500 patients in July 2015 didn’t have a location of residence recorded. These aren’t disproportionately in particular local authority areas.

15,000 patients of English GP practices are resident in Wales. However, the data suggests that at least as many English residents as this use Welsh GP services: the total ‘deficit’ between the population estimate and the GP register in Herefordshire, Shropshire and Forest of Dean is almost 19,000.

Explore the data yourself

This file includes data at local authority and LSOA level, collating information from NHS Digital and ONS: GP populations data (Excel 1.56 MB)

[1] ONS has an ongoing project to estimate the population based on administrative data alongside census-based estimates.

[2] The ‘GP registered population’ of a local authority is the total number of people registered with any GP practice (even those outside the area) who are recorded as living in that local authority. In other words: it’s what GP practices, between them, think the registered population of that local authority is. Keep in mind, too, that not everyone in the country will be registered with a GP practice.