Carers Census: results 2018 to 2019

2. De-duplication of Carers Census Records

Unpaid carers can be supported by more than one local service and may therefore appear in the systems of more than one data provider. To ensure that carers are not being double counted as a result different organisations returning information on the same people, the figures presented in this report refer only to records that have been de-duplicated.

2.1 De-duplication process

First, instances where a data provider had returned more than one record for the same carer were removed. Then a de-duplication ID was created for each remaining record by combining month and year of birth, datazone and gender. In cases where the de-duplication ID was not unique, further analysis of the data was carried out to identify where those records with the same de-duplication ID referred to different carers.

It was assumed that if multiple records submitted by a single data provider had the same de-duplication ID but different record IDs (e.g. ‘Carer 1’ and ‘Carer 2’), that these records referred to different carers. In cases where one provider submitted on behalf of others but returns were also received from those providers being submited on behalf of (as was the case in Argyll & Bute) or where the same system was used by mulitple providers (e.g. Carer Centres run by VOCAL), a single record was taken for each carer.

Records where month and year of birth or datazone were missing were removed since de-duplication was not possible without those indentifiers. This affected 15% of the records submitted.

As a result of the de-duplication process outlined above, 68% of the records submitted were included in the final data analysis.

In future years, we intend to link the Carers Census data with the National Records of Scotland’s population spine, which contains the personal identifiers of everyone in the Scottish Census, in order to obtain an accurate number of individual carers from the information submitted. 

2.2 Analysis of duplicate records and records unable to be de-duplicated

The de-duplication process removed 10,780 records from the dataset (32% of records submitted). Further analysis was carried out on these records in order to ascertain if certain areas or groups of carers were impacted more than others.

Effects of de-duplication by local area

Some areas were more impacted than others by the de-duplication process. The table below shows the areas impacted the most by the de-duplication process. All records for carers living in Orkney were removed due to missing month and year of birth.

Note: Area is based on carers’ residential postcode rather than location of data provider.

Effects on equality groups

The de-duplication process does not appear to have had a disproportionate effect on any particular group of carers.

Around 20% of records were removed for each of the age groups: 0 – 18 years, 18 – 64 years and 65+ years. Equal proportions of records for male and female carers were removed through the de-duplication process (31% for both males and females). 

There is slightly more variation across ethnic groups, though the proportion of each ethnic group removed varied between 18% and 28% (not including the groups ‘Not Known’ and ‘Not Disclosed’). The proportion of records removed for each deprivation decile varied between 22% and 29%, but there was no clear trend.