Recording of dementia among UK women: a comparison of primary care and hospital admission records

Tuesday, 19 August 2014
Exhibit hall (Dena'ina Center)
Anna Brown, MS , University of Oxford, Oxford, United Kingdom
Mary E Kroll, PhD , University of Oxford, Oxford, United Kingdom
Oksana Kirichek, MS , University of Oxford, Oxford, United Kingdom
Gillian K Reeves, PhD , University of Oxford, Oxford, United Kingdom
Jane Green, MD , University of Oxford, Oxford, United Kingdom
Valerie Beral, MD , University of Oxford, Oxford, United Kingdom
INTRODUCTION:  There is little reliable evidence on modifiable (lifestyle) risk factors for dementia; epidemiological studies are typically small and ascertainment of outcomes may be incomplete.

METHODS: The Million Women Study is a cohort of 1.3 million UK women recruited through the NHS Breast Screening Program during 1996-2001, and offers the opportunity to study dementia in a large prospective study with long and virtually complete follow-up through record linkage to National Health Service medical records.  The cohort has linkage to death and hospital admissions records in England (Hospital Episode Statistics), which may be incomplete, since some women with dementia may not be admitted to hospital or die from the condition. 

A sample of the cohort has recently been electronically linked to primary care data using the Clinical Practice Research Datalink (CPRD). This provides primary care records for about 8% of the population of England.  In this linked dataset we have compared records of dementia in hospital admissions and in primary care data.

Dementia diagnosis codes were taken to be ICD-10 F00,F01,F02,F03,G30 in hospital records, and the equivalent READ/OXMIS codes in CPRD. 

RESULTS: A random sample of 10,000 participants in the cohort were matched to CPRD data and for 821 (8.2%) women primary care records were obtained.

A preliminary analysis of the data shows that the proportion of women with any mention of dementia was 1.1% in primary care records, and 0.8% in hospital admissions. 

For women with a record of dementia in both primary care and hospital records, the first mention of dementia occurred in their primary care records and was on average approximately 2 years earlier.

The entire cohort is now being linked to CPRD data and comparisons of the recording of different subtypes of dementia will be presented.

CONCLUSIONS: Linkage of primary care and hospital admission data is feasible in England and should provide important information for epidemiological studies of dementia.