Report comment

10:58 – The SHMI technical group (leading the methodology for the SHMI) do publish the SHMI with 2 control limits (including the 99.8% one used in the guide) as they have been recognised to both have their use.

For the Hospital Guide, all mortality indicators have been banded using 99.8% control limits. For the SHMI, we have used the 99.8% control limits as published by the Health & Social Care Information Centre at https://indicators.ic.nhs.uk/webview. The use of 99.8% control limits is in line with advice from the Dr Foster Unit at Imperial College London to the SHMI technical group and is consistent with those used for other mortality indicators published in the Hospital Guide.
SHMI figures by trust are published with two banding systems. One applies 99.8% control limits, consistent with our approach to all the other indicators in the Guide. The second banding system identifies a smaller number of outliers on the grounds that the data is over dispersed. In simple terms, the second approach assumes that the extent of variability between trusts is too high to be explained by variation in quality of care and applies additional statistical processes to set a higher benchmark for outlying status. The Dr Foster Unit at Imperial College has advised that, in their view, the degree of dispersion across the data does not warrant this further adjustment and that, since the data is being used as a ‘screening test’ to identify where it is more likely that quality of care may be an issue, it is not appropriate to adjust the data on the assumption that variability is not being caused by quality of care.

This explanation was fully explained to the trusts prior to publication of the Guide.