The accuracy of the underlying data lies with the organisations providing the data. Missing or invalid data is removed from the datasets, and methods are continually being updated to improve data quality. This includes working with data providers to improve the completeness and coverage of the data.
As the report is based on a sample, the findings are weighted to present estimated results for the overall GP population. To do this, the population is stratified and appropriate weights are applied to minimise the effect of any bias in the composition of the sample.
Stratification methods have changed over the years, and detailed testing has always been carried out to assess the effect on the results. This testing indicates that changes in the weighting and stratification methodology do not result in a break in the time series.
Each country’s weighting factors are based on the strata (division of the population by age, gender, contract type, etc.), and the same set of weights are applied throughout, for all results except earnings distribution tables. Use of consistent weighting reduces complexity and potential risk of error. Standard errors are used where possible to assess the accuracy of the estimates. More detail is available in the Main Findings.
Because the results are based on samples with weighting applied, rather than the whole population, they are subject to sampling error and uncertainty. This is because using information from or about a sample of a population can never be as accurate as using the entire population.
Differences between groups and subgroups of GPs, or changes from one year to the next may not be statistically significant. Statistical significance is used in this report to illustrate the extent to which users can be confident that differences between compared results are not due to chance.
In addition, small GP populations for some subgroups mean that extreme values can have noticeable effects on the averages, and results in these instances may be subject to greater uncertainty.
Results shown in the GP Earnings and Expenses Estimates reports are rounded to the nearest £100 to show that the results are not exact. Similarly, population estimates are rounded to the nearest 50. Percentage changes and ratios are calculated using unrounded figures.
Figures are estimates and summarise how GP earnings and expenses have changed over time. However, it is important to note that they include both NHS and private work, and both full-time and part-time GPs are included in the sample.
Report population figures stated in the report are estimated and should not be regarded as the definitive GP populations; they will not be the same as those population estimates published in other NHS Digital reports. This is due to some GPs being excluded from the population for GP Earnings and Expenses Estimates for reasons listed in the methodology that accompanies the report.
All results received from HMRC are carefully checked before inclusion in the publication. In certain analyses, results are suppressed for groups with a small sample size, due to the sensitivity of earnings and expenses information. More detailed results are published in an accompanying Excel Time Series, an interactive web-based report and open data CSVs.
Any changes to the methodology or presentation of results are discussed by the Technical Steering Committee (TSC) prior to implementation. In addition, at each stage the report was further validated and quality assured by NHS Digital analysts unconnected with authoring the report.