In August 2021 NHS Digital revised the entire General Practice workforce time series back to September 2015, making several methodological changes to the way in which the workforce numbers are calculated.
-
Fully-estimated records (where a practice has not provided any valid data for a staff group) - were removed. Partial estimates (where the practice provides a record for an identifiable individual but does not include information about their contracted or working hours) - are still made and included in outputs.
-
Ad-hoc locums - were removed from the main tables as the move to monthly data extracts meant the headcount figures would not be comparable with previous quarterly collections and retaining them in the main tables would lead to an unavoidable break in the time series
-
Zero hours contracts – where working or contracted hours were recorded in the reporting period, staff are now counted in both the FTE and headcount figures, where neither working hours or contracted hours are recorded, staff are not counted in either.
Following stakeholder feedback regarding the impact removing the estimates may have had on the rate of the change in parts of workforce, NHS Digital committed to reviewing the removal of fully estimated records. All other changes remain in place.
Fully-estimated records were calculated according to perceived levels of local need for each job role and were based upon national staffing patterns which relied on all practices and CCGs having an equal need for staff in every job role. The creation of Primary Care Networks (PCNs) in 2019 introduced significant changes to the national primary care workforce. PCNs work collaboratively with general practices and other health and social care providers and share staff – primarily in the DPC group – across the network. This means that it is increasingly likely that PCNs will employ DPC staff and that fully estimated records will inflate figures for this particular staff group.
NHS Digital have reviewed the removal of fully- estimated records. As a result of the findings, estimates have been reinstated using an improved methodology, and the entire time series has been revised to include these estimates.
This page provides a summary of the review and the resulting methodological changes. Full details can be found in the Methodological Change Notice.
The review aimed to answer two main questions:
-
Was the removal of these estimates and redefinition of the historical timeseries the most appropriate solution for ensuring an accurate and consistent timeseries?
-
How appropriate would it be to estimate for missing staff, given near complete data and a changed primary care landscape with the formation of Primary Care Networks, who may now be employing these staff - particularly those in the Direct Patient Care staff group?
Question 1
Testing the estimation methodology on historical data
This was done in two ways.
-
The England average job role to registered patient ratio was calculated based on all supplied data. This ratio was then used to calculate an estimated headcount and FTE for each job role at each practice, and the results compared to the submitted data. This analysis was repeated on only half the records to further test the methodology. Calculating CCG-level job role to registered patient ratios were then investigated to see whether they would produce better results than an England-level figure.
-
Practices in March 2016 who supplied no data for one of the staff groups and therefore had estimated records, were tracked to a more recent period to see whether staff do appear in these groups in the later period. A comparison was then made between the estimated data and the ‘real’ data.
Question 2
The appropriateness of estimating for missing staff
Our Data Liaison Service contacted practices who do not currently submit data for some or all staff groups. This includes practices who have no records at all within NWRS for a staff group, and others who do have staff records but they either all fail validation, or all relate to staff members who have been recorded as leavers.
Questions put to these practices were designed to ascertain whether they genuinely do not have staff members in each of the staff groups they did not submit data for.
Testing the estimation methodology indicated that estimating for staff groups apart from DPC was appropriate despite likely over-estimation and using the STP based ratios improved the reliability. For the DPC staff group the methodology was deemed not to be reliable.
Results from the feedback and discussions with practices indicated that there are a significant number of practices that do not have DPC staff and therefore we should not estimate. Equally, about the same number reported they needed to update their data for this staff group. For other staff groups, it is more difficult to draw firm conclusions due to the smaller numbers engaging however for the majority of those practices who did engage, their records needed to be updated.
Alternatives to estimating for missing staff groups were also investigated including estimating for missing practice rather than staff groups. Another alternative considered was adding information on the percentage of practices included in the analysis for each staff group in published tables to add context. This reduces the impact of increases in data quality and participation on the scale of change and demonstrates where data are missing.