Could dementia be detected from UK primary care patients’ records by simple automated methods earlier than by the treating physician? A retrospective case-control study

Abstract

Background: Timely diagnosis of dementia is a policy priority in the United Kingdom (UK). Primary care physicians receive incentives to diagnose dementia; however, 33% of patients are still not receiving a diagnosis. We explored automating early detection of dementia using data from patients’ electronic health records (EHRs). We investigated: a) how early a machine-learning model could accurately identify dementia before the physician; b) if models could be tuned for dementia subtype; and c) what the best clinical features were for achieving detection. Methods: Using EHRs from Clinical Practice Research Datalink in a case-control design, we selected patients aged >65y with a diagnosis of dementia recorded 2000-2012 (cases) and matched them 1:1 to controls; we also identified subsets of Alzheimer’s and vascular dementia patients. Using 77 coded concepts recorded in the 5 years before diagnosis, we trained random forest classifiers, and evaluated models using Area Under the Receiver Operating Characteristic Curve (AUC). We examined models by year prior to diagnosis, subtype, and the most important features contributing to classification. Results: 95,202 patients (median age 83y; 64.8% female) were included (50% dementia cases). Classification of dementia cases and controls was poor 2-5 years prior to physician-recorded diagnosis (AUC range 0.55-0.65) but good in the year before (AUC: 0.84). Features indicating increasing cognitive and physical frailty dominated models 2-5 years before diagnosis; in the final year, initiation of the dementia diagnostic pathway (symptoms, screening and referral) explained the sudden increase in accuracy. No substantial differences were seen between all-cause dementia and subtypes. Conclusions: Automated detection of dementia earlier than the treating physician may be problematic, if using only primary care data. Future work should investigate more complex modelling, benefits of linking multiple sources of healthcare data and monitoring devices, or contextualising the algorithm to those cases that the GP would need to investigate.

Barriers and facilitators to the adoption of electronic clinical decision support systems: a qualitative interview study with UK general practitioners

“…At the beginning of the interview, a hypothetical example of a dementia early detection tool was introduced. This was chosen because there is a large body of work on developing automated dementia risk prediction work, some of which focuses on using primary care data [ 20 , 37 – 42 ]. However, prediction or detection of dementia is still controversial because of a lack of treatment available which makes any different to the disease trajectory [ 43 ], because of the high risk of false positives [ 44 ], and due to the patient’s “right not to know” [ 45 ].…”

Cognitive Performance at Time of AD Diagnosis: A Clinically Augmented Register-Based Study

“…The largest European register-based studies report mean age at the time of AD diagnosis to be 80-83 years (Tolppanen et al, 2016;Zakarias et al, 2019;Ford et al, 2020;Ponjoan et al, 2020). Based on age-specific dementia incidence rates (Prince et al, 2015) and Finnish population predictions for 2029 [Official Statistics of Finland (OSF) 2019], the majority (71%) of new AD diagnoses will occur among those aged 70-89 years, followed by those aged >90 years (21%); the phenomenon is expected to be even more pronounced during the 2030s.…”

“…An increasing number of studies are based on data from very large registers. In the AD field, register-based data have been used, e.g., to investigate incidence and prevalence of AD/dementia (Tolppanen et al, 2016;Kivimäki et al, 2018;Zakarias et al, 2019;Ponjoan et al, 2020), and AD classification (Ford et al, 2020). Registry-based data have also been used to evaluate medication use, healthcare service use (Tolppanen et al, 2016), and quality of diagnostic processes (Zakarias et al, 2019).…”

A systematic review of clinical health conditions predicted by machine learning diagnostic and prognostic models trained or validated using real-world primary health care data

“…In 22 studies of this ICD-10 classifications addressing six health conditions [28,45,[89][90][91][92][93][94][95][96][97]119,81,120,121,[82][83][84][85][86][87][88], the involved population were from eight countries, mainly the US and the UK (n=14). These studies were published since 2013 with the highest number of studies in 2020 (44.4%).…”