Department of Health & Social Care
Using the index of multiple deprivation
Introduction
To measure the effect of deprivation on a particular indicator, areas are generally assigned a deprivation score, or rank, using the English indices of deprivation. This is often done at the smallest geography for which the indices are available, which is lower layer super output areas (LSOAs).
LSOAs with similar levels of deprivation are frequently grouped together to form deprivation deciles or quintiles, for example, and we often wish to measure how the extent of inequality between these groups has changed over time.
There have been 6 versions of the English indices of deprivation, released between 2000 and 2019 (the next version is planned to be 2025), so any analysis needs to begin by choosing which is the most appropriate to use. This guidance informs the selection of a version, or versions, of the indices to use for analysis either for a particular time period or over time. Its aim is to improve consistency between outputs produced both within OHID and by other organisations producing health-related indicators using the indices of deprivation.
Although many other measures of deprivation are available, IMD is used consistently within OHID. This is because it is more frequently updated than indicators based only on Census data, and covers many aspects of deprivation. Its use with health data is sometimes criticised because it includes a health domain. This is, however, quite a small component of the index. The income domain alone is proposed by some as an alternative, but this is based just on means-tested benefits, so is a very narrow definition of deprivation. There is clearly no one correct method, but the overall IMD is the official measure of deprivation, it covers a range of aspects of deprivation and it provides a degree of consistency for monitoring over time.
The assigning deprivation categories page gives guidance on how to group individual areas into categories such as deprivation quintiles or deciles.
Guidance for measuring trends in inequality by deprivation
Use the version of the indices which most closely aligns with the time period of the data
It is possible to measure trends in deprivation-based inequality using a single version of the indices to define deprivation for the whole time period being looked at. This has the advantage of defining deprivation in a consistent way, which means that any observed change in trend is not the result purely of a change in definition.
Doing this, however, is problematic in local areas where there have been significant changes in levels of deprivation within the area over time. For example, an apparent narrowing of inequality over time between deprivation quintiles, based on a single version of the indices, could reflect a genuine narrowing of inequality. It could, however, also mean that the quintiles are no longer made up of similarly deprived LSOAs, that is, the deprivation quintiles have become less homogenous over time, thereby masking inequality between more and less deprived areas. An alternative is to measure the inequality in each time period using the version of the deprivation indices which most closely aligns with the time period of the data. This reduces this issue and means that the indicator more accurately reflects inequality within each area at a given time period.
Consider the time period of data used to construct the version of the indices
The versions of the indices are named according to the year in which they were published, but they are not based on data for that year.
Since each version of the indices is made up of around 40 indicators, there is no single time period on which each version is based, as data availability for each indicator differs. This means that it is less clear when each version should be used.
The time period of data used to construct each of the indices should be taken into account when considering which version to use for trend analysis. Table 2 sets out the versions available since 2004, with the time periods on which the majority of the data are based.
Use older versions of the indices with caution
Using different measures of deprivation based on the time period being examined introduces the possibility that any changes observed in the indicator are due to differences in the way deprivation is measured.
The versions from 2010, 2015 and 2019 are not identical in their construction but the differences between them are small so they can be compared over time. The statistical release for the 2019 indices notes that “The Index of Deprivation 2019 has been produced using the same approach, structure and methodology for the Index of Deprivation 2015 and previous releases. Keeping a consistent methodology in this way does allow relative rankings between iterations to be compared over time.”
IMD2010 was published based on 2001 LSOA geographies. OHID have produced a set of IMD2010 figures adjusted to 2011 LSOA boundaries but have not done this for earlier versions of IMD.
Older versions of the indices should be used with caution since there may be differences in the methods and indicators used in their construction, meaning they are not comparable with more recent versions. In addition, they may also be based on different geographical boundaries, for example, 2001 LSOA boundaries rather than 2011 LSOA boundaries.
Use the same version of the indices within aggregated time periods
When analysis is produced for 3- or 5-year aggregated time periods, data could be assigned to deprivation categories based on each single year, then aggregated to 3- or 5-year periods, potentially using the most appropriate version of the indices for each single year. However, doing this is potentially problematic for users of the data, such as local authorities who may need to identify and target the LSOAs in a particular quintile or decile.
It is therefore recommended that data are aggregated into the 3- or 5-year period first, then deprivation categories assigned using a single version of the indices.
Recommended approach
The principles set out above can be applied to any deprivation-based groups, for example those based on groups of LSOAs, and those based on groups of local authorities.
Table 1 provides guidance on which version of the indices should be used for each time period, based on the principles above. It assumes that each version of the indices will be mostly based on data from two years before the year of its release. This is largely consistent with the ONS approach for measuring trends by deprivation over time. There are many different ways in which time periods could be assigned to versions of the indices, and this table provides guidance only. In some situations, it may be appropriate to select alternative versions of the indices for measuring trends. It is important to note that the scores on different versions of the indices are not comparable but ranks can be compared across versions.
| 5-year aggregate | 3-year aggregate | Single year | Financial year ending | Recommended version of the indices |
|---|---|---|---|---|
| 2001 to 2005 | 2001 to 2003 | 2001 | 2002 | 2010 |
| 2002 to 2006 | 2002 to 2004 | 2002 | 2003 | 2010 |
| 2003 to 2007 | 2003 to 2005 | 2003 | 2004 | 2010 |
| 2004 to 2008 | 2004 to 2006 | 2004 | 2005 | 2010 |
| 2005 to 2009 | 2005 to 2007 | 2005 | 2006 | 2010 |
| 2006 to 2010 | 2006 to 2008 | 2006 | 2007 | 2010 |
| 2007 to 2011 | 2007 to 2009 | 2007 | 2008 | 2010 |
| 2008 to 2012 | 2008 to 2010 | 2008 | 2009 | 2010 |
| 2009 to 2013 | 2009 to 2011 | 2009 | 2010 | 2010 |
| 2010 to 2014 | 2010 to 2012 | 2010 | 2011 | 2010 |
| 2011 to 2015 | 2011 to 2013 | 2011 | 2012 | 2015 |
| 2012 to 2016 | 2012 to 2014 | 2012 | 2013 | 2015 |
| 2013 to 2017 | 2013 to 2015 | 2013 | 2014 | 2015 |
| 2014 to 2018 | 2014 to 2016 | 2014 | 2015 | 2015 |
| 2015 to 2019 | 2015 to 2017 | 2015 | 2016 | 2015 |
| 2016 to 2020 | 2016 to 2018 | 2016 | 2017 | 2019 |
| 2017 to 2021 | 2017 to 2019 | 2017 | 2018 | 2019 |
| 2018 to 2022 | 2018 to 2020 | 2018 | 2019 | 2019 |
| 2019 to 2023 | 2019 to 2021 | 2019 | 2020 | 2019 |
| 2020 to 2024 | 2020 to 2022 | 2020 | 2021 | 2019 |
| 2021 to 2025 | 2021 to 2023 | 2021 | 2022 | 2019 |
| 2022 to 2026 | 2022 to 2024 | 2022 | 2023 | 2019 |
| 2023 to 2027 | 2023 to 2025 | 2023 | 2024 | 2019 |
| 2024 to 2028 | 2024 to 2026 | 2024 | 2025 | 2019 |
The table aligns with the approach generally used by ONS for use of the 2015 and 2019 versions of the indices. OHID have proposed using the version 2010 for all time periods back to 2001. This is different from the ONS approach, which generally also uses the 2004 and 2007 versions in their trend analysis.
The use of older versions of the indices is not recommended because, although the approaches to producing all versions of the indices are similar, the older versions have not been published based on 2011 LSOA boundaries.
Analysis may already have been published by OHID which does not use this recommended approach. As long as the version of the indices used is clear, earlier analysis will not necessarily be adjusted.
Boundary changes also need to be considered, particularly for groups of local authorities. Trends should not be presented where there have been boundary changes. For example, deciles based on pre-April 2019 boundaries for one time period should not be presented as a trend with deciles based on April 2019 boundaries.
The 2000 version of the indices is also available from the National Archives. This version was only published at electoral ward level, not for LSOAs.
Analysis of the effect of using a different version of the indices for measuring trends in inequalities
To help assess the effect of moving from using one version of the indices to another, analysis was undertaken using the slope index of inequality in life expectancy as an example. Values for 2016 to 2018 were calculated using both IMD2015 and IMD2019, and the results compared.
A summary of the findings of this analysis has been published on the Fingertips website, including charts showing the effect in each area (PDF).
Page last updated: August 2024