Primary care dementia data: data quality

The Primary Care Dementia Data publication series supersedes the Recorded Dementia Diagnoses series. Data for the period April 2022 – October 2022 was collected under both services, however, this data is not comparable. This is due to the retrospective application of codes to patient records and changes in patient registration, as well as differences in coverage and the specification of a number of the counts as detailed below.

Change in coverage

The proportion of GP practices that have accepted an invitation to participate in the Core GP Contract service, of which the dementia data is a part of, has risen from 85.6% in April 2022 to more than 98%, in April 2024.

Changes to indicator specifications

Some aspects of the indicator specifications have changed.

These are:

1. Change to quality service start date

A number of counts look back at activity from the start of the reporting year up to the reporting period end date (the last day of the month). In the Recorded Dementia Diagnoses publication, the reporting year ran from October to September, but now runs from April to March in the Primary care Dementia data publication. The reporting period used for activity against the following counts has therefore changed:

• Number of patients with a record of receiving a dementia assessment
• Number of patients with a record of an initial memory assessment
• Number of patients with a record of declining an initial memory assessment
• Number of patients with a record of a referral to a memory clinic
• Number of patients with a record of declining a referral to a memory clinic
• Number of patients with a record of receiving a medication review

2. Change to medication review counts

Due to an issue with the extraction, the count of patients with a record of receiving a medication review was under-reported in the Recorded Dementia Diagnoses publication series, and should therefore be treated with caution. These counts have been corrected for this new primary care dementia data publication series.

3. Change to dementia type breakdowns

For the 2024-25 service additional breakdowns for the following dementia types were included:

• Lewy Body
• Frontotemporal

These dementia types were previously included in the 'Other' dementia type breakdown. These breakdowns are available from June 2024; prior to this they were aggregated into the 'Other dementia type' group.

4. Addition of inconclusive breakdowns

For the 2024-25 service an additional 'inconclusive' breakdown has been added for the following measures:

• Dementia type
• Ethnicity
• Residence type

Where patients do not fall into a single category in each of the measures above, they either fall into:

• none of the categories or
• more than one of the categories,

they will now be counted in the new inconclusive category.

For example:

Patient A was recorded as having 'Alzheimer's disease' and 'Vascular dementia' and will be counted in both breakdowns. Now Patient A will only be counted once as 'Inconclusive dementia type'.

This change was introduced with the aim of reducing double counting of patients in the different breakdown categories.

5. Change in age restriction for patients with Mild Cognitive Impairment (MCI)

From April 2024 the count of patients with a diagnosis of Mild Cognitive Impairment (MCI) without a diagnosis of dementia is collected for patients aged 40+; prior to this it was collected for patients aged 65+ therefore these counts are not comparable.

From June 2024 these counts are published broken down by age (5 year age bands) and sex. Prior to this a single aggregate count of patients aged 40+ is available.

Data quality is generally considered to be very good for all GPES extracts.  Between 90 and 100 per cent of practices in England participate each month. 

Data collected is 100 per cent complete, as it is an automated extract GPES extracts data for all patients who have specified codes on their record.

Age/sex, ethnicity, dementia type and residential type breakdown totals

Due to the way age, sex, ethnicity, dementia type and residential type data are recorded in and collected from GP clinical systems, the sum of these breakdowns may not be equal. To support the calculation of the dementia diagnosis rate, the sum of the age/sex breakdown is used as the total dementia register value in this publication.

From 2024-25, updates to the extract specification means the sum of these breakdowns are now nearer to being equal.

They may still differ for the following reasons:

1. Where current monthly data for a GP practice is unavailable, the most recent data available is used (up to a maximum of 6 months); if the most recent data was from 2023-24 the discrepancies will be greater as the updated extract specification was not in use.   

2. Where a patients age and/or sex is unavailable they will not be included in the total count of patients with dementia, however, they will be included in the breakdowns for ethnicity, dementia type and residential type.

Dementia care plans/care plan reviews

April - September 2022 - due to an issue with the collection of data relating to dementia care plans/care plan reviews, the count of patients who have received a dementia care plan/care plan review is inaccurate and has been suppressed for this period.

April 2022 – March 2023 - due to an issue with the specification of the count which captures the number of patients who declined a dementia care plan/care plan review, this data is inaccurate and has been suppressed for this period

SNOMED code incorrectly used

February - March 2023 - As part of the January 2023 code release an additional SNOMED code was incorrectly added to the refset which was used to identify patients with a recorded dementia diagnosis. This code has now been removed from the refset however a small proportion of patients will have been incorrectly identified as having dementia in the published data.

Government Office Region (GOR) data

September - October 2023 - Government Office Region (GOR) recorded number of patients with dementia aged 65 and over does not sum to the National (England) count. There is a small difference due to the inability to map two practices to an appropriate GOR.

March – April 2024 - Government Office Region (GOR) recorded number of patients with dementia aged 65 and over does not sum to the National (England) count. There is a small difference due to the inability to map a single practice to an appropriate GOR.

Antipsychotic Prescribing

November 2023 - Data for the prescribing of antipsychotic medication may be inaccurate with counts being lower than expected. This is due to a delay in the updating of drug code reference sets in GP systems whereby new dm+d codes for various antipsychotic medications were made available for use in patient records. However, there was a delay in updating these reference sets within the GP systems for extraction purposes, therefore patients with these codes on their records will not have been captured in the publication.

The updated reference sets were implemented shortly after and data is accurate for December 2023 onwards.

List size data

If a practice's status changes to dormant, dementia data extracted prior to this change will still be included for historic months, however there will be no list size data available.

From the April 2024 publication, the data source for the practice list size data changed from the National Health Applications and Infrastructure Service (NHAIS) to the Patient Demographic Service (PDS). Further information can be found at the Patients Registered at a GP practice publication data quality information page.

Dementia type

December 2023 to March 2024 - counts of patients in the dementia type breakdowns are lower than expected due to an error whereby changes to the 2024-25 service were implemented early. Consequently, where patients who have a coded diagnosis of Lewy body dementia, Frontotemporal dementia or inconclusive dementia types would previously have been captured in the 'Other' dementia type breakdown, between December 2023 and March 2024 they were only captured in the total count of patients with dementia, and not in any of the dementia type breakdowns. This has led to a reduction of ~10,000 patients with a dementia type breakdown for each of the affected months.

This issue will be rectified from April 2024 onwards.

Data from the previous month is extracted via GPES on the first day of the next month.  This extraction period usually lasts around 7 days.  The aim is to publish within the next 7 days.  For example - June 2019 data was extracted between 1 July and 7 July 2019.  Analysis began on the 8 July 2019 with the aim of publishing it on 11 July 2019.

Data is available from our website. The publication includes:

• 1 .xlsx summary file
• 11 .csv data files
• 1 .xlsx data dictionary file
• 2 x supporting information .csv files (mapping and latest extract date)

All data included is similar and easily linked on GP Practice Code and or Sub ICB code.

Where current monthly data for a GP practice is unavailable, the most recent data available is used (up to a maximum of 6 months).

The most recent mapping data is produced each month to ensure data can be compared over time. In this release, where a practice closes within the reporting year, it is not assigned geographical mappings following its closure. A small amount of data will therefore be unmapped (unallocated)  in the time series files.

Comparisons with the data collected and published under the Recorded Dementia Diagnoses series is not recommended on account of the changes outlined in the Primary Care Dementia data publication series - associated changes' section above.

Updated terminology

From November 2023 the term 'early onset dementia' has been replaced with 'young onset dementia' in this publication to match internationally agreed terminology. There have been no changes to how these counts have been calculated and data is comparable to previous months where the 'early onset dementia' term was used.

Comments and questions can be made through various media including general enquiries by emailing [email protected] or by telephone 0300 303 5678

Regular consultation between NHS England and DHSC are carried out and adjustments to the publication made, based on their feedback. 

All costs are covered by NHS England on an annual basis with a new work package being agreed each year.

This publication is subject to a standard risk assessment prior to issue.  Disclosure control is implemented where it is judged necessary.

Detailed methodology specification documents and other supporting material are available on our website. 

The UK Statistics authority code of practice for statistics is followed regarding security and release of information prior to publication.