Ongoing CRIS projects: dementia
New Mind 2
In previous work, we utilised the UK-CRIS, (Case Record Interactive Search) a database of pseudonymised mental health records from 12 NHS Mental Health Trusts, to investigate the effectiveness of two types of medication in the symptomatic treatment of dementia (Vaci et al., 2020) using data from two UK Mental Health NHS Trusts (Oxford Health and Southern Health). In the sample of over 7000 patients, we show that medication prescription stabilises cognitive performance for a period of 2 to 5 months. The New Mind 2 study expands our previous work, by including diagnosis of depression and related medications, as well as symptoms and well-being measures that frequently follow these two diagnostic outcomes. Similarly to the previous work, once the information extraction systems are developed and data is structured, we plan to explain and interpret the data using frequently used statistical methods in medical research.
Prediction of death, nursing home use and admission in the older population using routinely collected data
In this work we seek to provide evidence for the real-world effectiveness of two main medication classes aimed at slowing down the cognitive decline in people with Alzheimer’s disease. The medication classes were acetylcholinesterase inhibitor (AChEI) such as Donepezil, Rivastigmine or Galantamine and another class of Memantine. We used methods from natural language processing to develop an information extraction system to convert Clinical Record Interactive Search (CRIS) medical records into a structured table and where each observation, such as prescription of medications, had the date of prescription) . Therefore, we can create a timeline of the treatment with medications for each patient Some patients may have been treated subsequently with multiple medications. Eg. a patient started on Memantine, but later was switched to Donepezil, followed by Rivastigmine. We also extracted the information that indicates and measure s the level of cognitive ability in people with dementia. This gives us an unprecedented possibility to investigate how prescription and intake of dementia-related medication influences the cognitive ability and helps the patients. The usual pattern that we observe is that before taking any medications, patients decline rapidly in their cognitive abilities, becoming cognitively slower and impaired. Once the medication is prescribed, we observe stabilisation of this cognitive decline or even improvement of cognitive ability. In other words, medication preserves the cognitive ability of patients for some time, ranging from 2 to 6 months depending on the type of medication and the type of Alzheimer’s However, this effect is not sustained , as the cognitive performance starts to decline again, once the initial effect of medication declines.
Validation of UK Biobank Data for Mental Health Outcomes: A Pilot Study Using Secondary Care Electronic Health Records
UK Biobank (UKB) comprises a large data resource with more than 500,000 participants and with a wide variety of variables, such as demographic, lifestyle, environmental and health information for the assessment of various life-threatening and disabling conditions, including mental health disorders. However, the applicability of UKB and its relevance in a clinical setting and the assumptions required have not been sufficiently and systematically defined. Here, we present the first study to authenticate UKB’s applicability and relevance by using secondary care mental health data with linkage to UKB from Oxford – Clinical Record Interactive Search (CRIS) focusing on comparison of demographic information, diagnostic outcome, medication record and cognitive test results.
We applied a natural language processing model to extract information embedded in unstructured text from clinical notes and attachments. UKB data comprises a large variety of different data modalities, such as genomics, imaging data, self-reported outcomes, physiological time-series from wearable devices and many more, however, the UKB data are lacking follow-up records, whereas CRIS offers a longitudinal and detailed clinical picture with more than ten years of observations.
The linkage of both data sources, UKB and UK-CRIS, represents untapped potential for comprehensive research in mental health, synergistically complementing each other in various data modalities and to allow more robust research in mental health. Specifically, it is now possible to combine precise information from UKB, such as data from wearable devices, imaging and genetic data with accurate and clinically validated longitudinal data from UK-CRIS, including diagnoses, treatments, administered medications, psychological sessions and information from specialised clinics.
Enabling earlier and more accurate diagnosis in Alzheimer’s disease (AD)
iASiS is an EU funded project that seeks to pave the way for precision medicine approaches by utilising insights from patient data. It aims to combine information from medical records, imaging databases and genomics data to enable more personalised diagnosis and treatment approaches in two disease areas – lung cancer and Alzheimer’s disease. In the above study (study title), will be analysing data of Dementia and Alzheimer’s patients from Oxford’s UK-CRIS.
Patients with emergency admission to hospital are increasingly older and have a variety of underlying medical conditions. They are also at higher risk of adverse outcomes such as falls and disability this is known as frailty as indicated by increased vulnerability and dependency on others. We know that these patients are at high risk of future dementia, particularly those who have new or worsened confusion (so-called “delirium”). However, not all hospitalized patients develop dementia and some go on to recover fully but at present, we do not know which patients will do well or badly. This makes it difficult to provide individualized care. In the current study, we will use existing electronic datasets from the Oxford University Hospitals NHS Foundation Trust (OUHFT) and the Oxford Health NHS Foundation Trust (OHFT) to identify the factors which predict future dementia in patients admitted to hospital.
Using OUHFT electronic patient records on patients aged >65 years, we have created the Oxford Cognitive Comorbidity and Ageing Research database (ORCHARD). ORCHARD contains a wealth of information on clinical diagnoses, “observations” (measurements of blood pressure, breathing rate, temperature and heart rate), illness severity, blood testing, and X-rays or brain and body scans together with markers of frailty including confusion, poor nutrition and risk of pressure sores. To find out which ORCHARD patients go on to develop dementia on follow-up after admission to the OUHFT, we shall link ORCHARD with the OHFT Clinical Record Interactive Search (CRIS) database containing mental health outcomes. We shall then do data analysis to find out the factors that are linked to high dementia risk. Ultimately, we hope to be able to identify a few key risk factors that together can be used to stratify patients into low, medium and high risk groups. This information could then be used to help provide better care and information to patients and to select patients for follow-up in memory clinic or for trials to prevent dementia or further deterioration in thinking and memory.
Care factors in early and late-stage dementia disease
This research programme aims to investigate behaviours that can potentially preserve cognitive performance once the dementia diagnosis has been established. In previous work, we utilised the UK-CRIS, (Case Record Interactive Search) a database of pseudonymised mental health records from 12 NHS Mental Health Trusts, to investigate the effectiveness of two types of medication in the symptomatic treatment of dementia (Vaci et al., 2020) using data from two UK Mental Health NHS Trusts (Oxford Health and Southern Health). In the sample of over 7000 patients, we show that medication prescription stabilises cognitive performance for a period of 2 to 5 months. The New Mind 2 study started expanding this work by including diagnosis of depression and related medications, as well as symptoms and well-being measures that frequently follow these two diagnostic outcomes. In this project, we plan to work on extraction and statistical modelling of additional information that can potentially explain different trajectories of cognitive declines after the diagnosis of dementia. We initially plan to focus on early and middle life modifiable behaviours, such as indication of better educational and occupational attainment, as well as lifestyle factors, such as physical activity, smoking, alcohol consumption and obesity. Extraction of this information would give us an unprecedented possibility to investigate whether some of these factors relate to more favourable cognitive outcomes. In addition, it would also allow us to examine how these factors change the likelihood of responding to the dementia-related medications.
Background: The use of artificial intelligence in brain imaging can help us to understand the underlying causes of mental illness and dementia, and support clinicians in to provide better treatment. Alzheimer’s disease represents a substantial and looming public health crisis, affecting roughly half a million people and one in fourteen older adults in the UK. Older adults with AD tend to have complicated health problems, and roughly half have three or more chronic conditions. They are at elevated risk for avoidable hospitalisations for chronic conditions, hospital readmission, urinary tract infection, inflammation, and other problems, which may be preventable with timely and effective care.
Data from routinely collected brain images could help guide care and support the early identification of patient needs. Clinical neuroimaging (“brain scans”) are collections of images showing the structure of a person’s brain obtained from Computed Tomography (commonly called CT or CAT scans) or Magnetic Resonance Imaging (often referred to as MRI scans). They require expert review where a radiologist combines what they’ve been told about the patient (symptoms, signs and clinical history) with their interpretation of the images. The result is a written clinical report which is delivered to the requesting clinician that describes the radiologist’s interpretation of the images in the specific clinical context. In mental health, these written clinical reports are very often stored in the patient’s electronic health record to help the current (and future) clinicians make treatment decisions. The “raw” imaging data itself ie. the scans themselves is not stored in the EHR because it is usually very large and always requires interpretation by an expert radiologist.
Though the written reports of brain scans may be contained in patients’ clinical notes, the usefulness of this readily available written information has not been tested. Natural language processing (NLP) refers to a group of methods for analysis of text content, which could be used to automatically explore untapped, yet critical data on relating to brain imaging found in clinical notes.
Aims: In this feasibility project we will explore the amount and usefulness of information on brain scans recorded in clinical notes for AD patients in Oxford Health. We will use NLP tools to search clinical notes in the CRIS database for imaging information. We will also explore the availability of actual scan reports, which are sometimes attached to patient records. If we find evidence that there is enough useful information on brain scans in electronic patient records, we will use this feasibility study as justification to seek funding for a larger project to develop methods of automatically extracting this scan data using more advance NLP algorithms.
Page last reviewed: 25 November, 2021