Can markers of dementia progression be derived from primary care electronic health records?

Conference:

SAPC ASM 2019 - Exeter

Talk Code:

P2.21

Presenter:

Professor Carolyn Chew-Graham

Twitter:

@CizCG

Co-authors:

Paul Campbell, Trishna Rathod-Mistry, Michelle Marshall, James Bailey, Carolyn A. Chew-Graham, Peter Croft, Martin Frisher, Richard Hayward, Rashi Negi, Swaran Singh, Athula Sumathipala, Nwe Thein, Kate Walters, Scott Weich, Kelvin Jordan

Author institutions:

Keele University, Midlands Partnership NHS Foundation Trust, University College London, University of Warwick, University of Sheffield

Problem

Understanding the progression of dementia is central to its management. Electronic health records (EHR) from primary care provide a potentially rich resource to measure progression and to identify markers of that progression. Dementia research using EHR has already produced information on factors that can identify those at risk of dementia onset, but no study has yet examined the feasibility of determining the progression of dementia post-diagnosis. The aim of this study was to investigate whether a set of credible markers of dementia progression can be identified from routine EHR primary care data and grouped into larger domains relevant to dementia.

Approach

Findings from a systematic review and an initial expert consensus group led to the identification of a set of possible markers of progression potentially identifiable within EHR. These markers related to areas such as cognition, neuropsychiatric symptoms, daily functioning, frailty, and care provision. An EHR database (CiPCA) of consultation records from 9 general practices in North Staffordshire in the UK (annual population ~90,000) was used to test the frequency of recording of these markers. Patients with a recorded diagnosis of dementia were compared to age, gender and practice matched (1:1) controls on presence of Read codes related to these markers. We also performed a hypothesis free analysis to determine other codes that were associated with dementia but not identified from the review or initial consensus group. A final consensus group exercise then confirmed markers and their aggregation into high-level domains based on the findings.

Findings

There were 2714 individuals identified with dementia. Interim analysis yielded 57 individual markers housed within 15 domains, with 78% of these markers recorded more frequently within the dementia group. The developed domains (e.g. cognitive function, neuropsychiatric, daily functioning, frailty, care provision) were recorded with sufficient frequency to suggest their potential as EHR indicators of progression. Comparing the dementia cohort to the matched cohort for frequency of recorded domains (per 100 person years) revealed differences, for example: cognitive function (15.4 vs 3.8), neuropsychiatric symptoms (12.3 vs 6.2), frailty (8.8 vs 2.8), care provision (28.3 vs 18.2).

Consequences

This study suggests that EHR are able to capture some of the domains and specific markers which have been identified elsewhere as potentially important as markers of progression and outcomes for persons with dementia. The next stage of this research will carry out analysis within a larger UK national primary care EHR dataset to establish longitudinal patterns of progression following a dementia diagnosis. This research has the potential to provide clinically useful information to identify individuals with dementia at risk of more rapid progression, and can provide a readily available method that may be useful as an outcome measure in future research (e.g. trials).

Submitted by:

Paul Campbell

Funding acknowledgement:

This work is supported by a grant from the Dunhill Medical Trust (RPGF 1711/11) to Professor Kelvin Jordan. The views and opinions expressed within this abstract are those of the authors and not necessarily the views of the Dunhill Medical Trust.