March 2024 data Contracted WTE Investigation and fix
1 Introduction
The purpose of this paper is to set out the issues relating to reporting WTE as at 31 March 2024.
2 What the issue is
From 01 April 2024, NHS Scotland Agenda for Change (AfC) staff saw a reduction in their working week (or conditioned hours) from 37.5 hours to 37 hours. AfC staff are all non-medical and non senior management staff groups. It includes nurses, staff in Allied Health Professions (AHPs; including physiotherapists, occupational therapists, etc) and administrative staff, among other staff groups. For part-time staff, the contracted hours (the number of hours someone is contracted to work per week) was reduced pro-rata.
Each month, NES extracts information from Scottish Workforce Information Standard System (SWISS), a reporting database, for people employed in NHS Scotland on the last day of the previous month. A slight delay is used between the census date and the extract date to allow late updates to the data. The data extracted includes the conditioned and contracted hours for each employment, which are used to calculate the Whole Time Equivalent (WTE) for the official statistics: simply, WTE is contracted hours / conditioned hours.
The process to update the contracted and conditioned hours in SWISS to reflect the reduction in working week was carried out between the 31 March 2024 census date and the extract date, and due to the way this information is stored in SWISS, this update overwrote the previous contracted and conditioned hours. This meant that the NES extract for data as at 31 March 2024 included the contracted and conditioned hours data after the update, rather than the information present on 31 March 2024 before the reduced working week came into effect.
Moreover, there was an error in the update process for contracted and conditioned hours data in SWISS. We expected to see all conditioned hours reducing from 37.5 hours to 37 hours and all contracted hours reducing by 0.5 hours pro rata. Instead, the reduction in conditioned hours in SWISS happened for 93,365 employments, with the remaining 77,817 employments left unchanged at 37.5 hours. For some employments, the contracted hours had reduced but the conditioned hours had remained at 37.5 hours: for example, 38,479 employments had a conditioned hours of 37.5 and a contracted hours of 37, which was many more than the 84 employments in this situation the previous month.
From the extracted data, which contained these errors, the WTE calculated for AfC staff in March 2024 was 143,926.4, down from 144,165.3 WTE in December 2023.
After investigating the issue alongside Atos and NSS, it was discovered the error was due to the full employee records not feeding through to SWISS from the Workforce Intelligence Repository (WIR) database. This issue only affected SWISS, which is a reporting database, and did not affect the live HR and payroll systems that feed SWISS.
3 How we approached this
In effect, we are missing the contracted and conditioned hours information as at 31 March 2024. A report prepared by Alma Economics for the Office for National Statistics details various approaches for dealing with missing data.
Of these methods, we implemented the Last Observation Carried Forward (LOCF) where the last known observed value is carried forward and used in place of the missing data. We used this in combination with the Next Observation Carried Backwards (NOCB) for cases where there is no last known observed value. LOCF was prioritised because February WTE is more likely than April WTE to represent March WTE, because the change in working hours came into effect in April.
We chose this approach because WTE rarely changes from one month to the next. For example, 97.8% of employments in March 2023 had the same WTE as in February 2023, and in separate work modelling March 2023’s WTE based on data from February to April 2023, February and April’s WTE were very strong predictors of March’s WTE (and the strongest predictors in models that also looked at age, sex, job sub family and health board).
The results of the combined LOCF/NOCB approach are as follows.
There were 171,182 AfC pay numbers in March 2024. A pay number denotes a single employment. A member of staff can have multiple employments, whether in the same board or across multiple boards.
Where a pay number was present in both February and March 2024 (i.e. continuing employment), the February WTE was carried forward (LOCF). This was done for 170,062 pay numbers (99.35% of the total number of pay numbers).
Where a pay number was not present in February but present in March and April 2024 (i.e. new employments from March that continued into April), April WTE was carried backwards (NOCB). This was done for 1,105 pay numbers (0.65%).
For the remaining 15 pay numbers (0.01%) that were only present in March 2024 (and not in February or April 2024), the WTE was not updated.
After this, the WTE for all AfC staff in March 2024 was calculated as 144,778.4.
4 Impact assessment
To assess the impact of using the NOCF/LOCB methodology, we applied it to the previous year’s data (February - April 2023) and compared the calculated WTE for March 2023 to the actual WTE for March 2023. This showed that the calculated WTE for March 2023 for AfC staff would have been 140,098.4 WTE, compared to the reported value of 139,961.9 WTE, a difference of 136.5 WTE (0.1%).
The chart below shows the percentage difference between using the LOCF/NOCB methodology compared to the reported value for the previous 10 years. March 2023 has the largest difference between the reported value and the calculated value. There are smaller differences over the period covered by the COVID-19 pandemic, possibly due to fewer people changing hours (or if they do change them, increasing their hours). The calculated WTEs tend to be an overestimate (the % difference is above 0%; the range of differences from 2013 to 2023 is -0.01% to 0.1%). On average, the calculated value is 0.04% higher than the reported value.
In recent years, staff have been more likely to reduce their hours rather than increase them. Between February and March 2023, 1,446 employments had a reduction in WTE and 840 had an increase, while between February and March 2022, 689 employments had a reduction in hours and 523 had an increase.
Given the above, the calculated WTE for March 2024 is likely to be a small overestimate of the WTE of NHS Scotland as at 31 March 2024. In addition, one drawback of using the LOCF/NOCB approach is that, by definition, there will be no variation in WTE between periods where the same data is used (i.e. February and March, March and April).
5 Sickness Absence
Sickness absence is calculated as the hours lost divided by the total contracted hours. When the Sickness absence report was run for March 2024, the total contracted hours had partially updated, but the total hours lost was constant.
We have again used Last Observation Carried Forward on this data for each Health Board, using the absence data for each Health Board from 01 March 2023 - 29 February 2024, instead of 01 April 2023 - 31 March 2024 which has been used in previous years.