|Data Series||Revision arrangements|
|Annual reported injury statistics (RIDDOR)||Data published for the most recent year is given a 'p' status (provisional), and a year later is released as final (without p or f notation). For provisional data this allows for late reports and where necessary additional quality checks. No further revisions are made to finalised data (unless there is a methodological improvement or errors are found). Publication of data in this manner attempts to strike a balance between producing data promptly (initially as provisional), and accurately (as final, a year later with minor adjustments).
The changes are usually small, increasing by no more than about 1% for non-fatal injuries. Percentage changes between provisional and final for fatal injures can vary more, partly because of the much smaller numbers involved. Finalised fatal numbers can go down as well as up, in the range +/- 2.5%, mainly because all fatalities are investigated in detail, with some talking considerable time to conclude.
|Annual reported injury rates (RIDDOR)||Since the HSS data release in November 2011 the Annual Population Survey (APS) has been used as the denominator for RIDDOR injury rates (previously, the Workforce in Employment (WiE) series and the BRES survey were used). As with the previous surveys, the Office for National Statistics (ONS) is the provider of the APS data. The analysis and interpretation of these data are the sole responsibility of the HSE. Revisions typically go back two years and will be incorporated into the RIDDOR injury rates at the time of the next scheduled release of injury numbers. Occasionally, more historical revisions are made to the employment series. When these revisions occur, HSE statisticians will make a judgement as to whether they materially impact on the RIDDOR injury rate series and decide whether to incorporate them into the rates.|
|Mesothelioma deaths||Deaths from Mesothelioma are marked as provisional for two years and will be revised due to late registrations to ensure the series is as complete and consistent as possible. On publication of figures for the latest year revisions to data for earlier years will be made such that the total for the previous year will typically increase by around 10-15 deaths and the total for the year before that by 1 or 2 deaths.|
|Enforcement data||At first release, enforcement data (notices and prosecutions) are marked as provisional. The following year, data are finalised and no further revisions made. (unless errors are found). The level of revision is minor, typically, being less than 1%.|
|Quarterly, in-year reported deaths (RIDDOR) data||This series covers work-related fatalities that are reportable under RIDDOR, and are updated on a quarterly basis. A consequence of providing detail that is as up-to-date as possible is that many of the cases listed are still under investigation. Therefore, these statistics are released on a provisional basis, and the quarterly figures will be revised during the year up until the time that the annual figures are finalised.|
|THOR||THOR records reports of cases of work-related illness from members of the THOR surveillance network. There is a system of regular checks with reporters throughout the year, and an annual process of completeness and logic checks before the data are released. Occasionally, specific analyses can identify data errors in individual THOR cases after initial publication. When found, these are corrected immediately in the THOR database, and at the next annual update in the HSE published tables.|
|LFS||HSE follows LFS revision practice to the extent that it materially affects estimates from the health and safety questions. When there is the potential impact, an assessment of the scale is made and a judgement about the need for revision is made.|
|HSE Cost model||Each year HSE produces a revised back series (to 2004/05) to express costs for each year in the current year prices. This is to enable reliable comparisons between years. The revised back series in current year prices will also include revisions to any of the model inputs or assumptions.|