University of Edinburgh

Dr Beatrice Alex

Advanced Care Research Centre

Other

DL_2021_009

People in the UK are living longer, and more people now live with multiple conditions. Healthcare focuses on single diseases, and is often not good at dealing with many of the common problems of later life. For example, there are many reasons that people fall or get confused or have problems with their bowels or bladder. In electronic health records, these problems are almost always documented in ‘free-text’ (ie the typed notes that doctors or nurses make). Identifying who has this type of problem is therefore difficult because different doctors or nurses will use different words to describe them (‘confused’, ‘mixed-up’, ‘delirious’, ‘not themselves’, ‘cognitively impaired’). We will use data science methods that can automatically turn free-text into carefully-defined categories (eg all of the words in the last brackets are ‘delirium’). This will support research to understand how common these problems are and what causes them, and healthcare improvements.

Historically, healthcare has focused on diseases and discrete conditions, but this often does not capture key elements of ill-health. Geriatric syndromes are a set of conditions with major impact which multiple diseases can contribute to, and include delirium, falls, frailty, dizziness, fainting, and incontinence. Their presence is a strong predictor of poor outcomes, but many are also amenable to intervention if detected. There are also contributions to poor outcome from non-medical context such as living alone and lack of social support. However, recording of geriatric syndromes and social context in coded data is poor. Recent work in the US has shown that data extracted from free text electronic health records using natural language processing (NLP) identified far more people with falls (69.7% of identified falls were only recorded in free-text), visual impairment (86.6%), and lack of social support (99.8%) than coded data. Therefore, there is a great need to use automated approaches to identify these important syndromes from free text clinical notes such as GP referral letters, inpatient notes, discharge summaries and other free-text such as radiology reports (eg to identify fractures). Better understanding of geriatric syndromes is needed to support intervention development.

Other

01/04/2022

De-Personalised

TRE