Patients who suffer from Type 2 Diabetes often also develop cardiovascular diseases. The treatment of patients with these multiple conditions often disjointed. It has been recognised that existing electronic healthcare records provides unique opportunities to understand and improve care for such patients. New analytical tools are required to help the NHS to map patient journeys, from symptoms to diagnosis and treatment, to allow interventions at key points, preventing or slowing the development of new and concurrent health conditions.
This study aims to provide a holistic view of how patients with diabetes and related co-morbid conditions such as cardiovascular diseases are being treated using linked electronic health records from CPRD. By using machine learning methodologies which consider of sequential nature of treatment pathways, we aim to discover reoccurring patterns within the patient journeys which may lead to clinical outcomes such as unplanned hospitalisations and development of co-morbid conditions.
Early prevention and reduction of secondary referrals is a key target in NHS’s Long Term Plan. In the care for Type 2 Diabetes (T2DM) and its associated co-morbid conditions such as Cardiovascular Disease (CVD), the disjointed nature of care for patients with multiple morbid conditions has been identified as a key inefficiency which leads to otherwise avoidable secondary resource spending. Existing electronic healthcare records (EHR) allows us to provide a holistic view of these care pathways and allow the care system planners to identify the areas of inefficiencies in order to improve outcomes for patients.
In this study, we wish to conduct a retrospective hypothesis generating study focusing on the patient population targeted by the NHS Digital’s CVDPREVENT audit, patients who have confirmed diagnosis of cardiovascular disease, type 2 diabetes and related comorbidities in hypertension, familial hypercholesterolemia and other hyperlipidaemias, chronic kidney disease (grades 3 to 5), non-diabetic hyperglycaemia and atrial fibrillation, in the period between April 2003 to Oct 2020.
We aim to provide a holistic view of the treatment pathways for the patient population in relation to the above-mentioned diseases, to identify whether the differences in how patients interact with services in the health system can affect their clinical outcomes, such the number of unplanned hospitalisations and the number of patients go on to develop of comorbid conditions within the scope of CVODPREVENT audit.
At the same time, we wish to test the feasibility of applying machine learning techniques to predict future clinical outcomes based on the longitudinal records in the CPRD data. If proven to be successful, this could open doors to future studies looking at the possibility of predicting potential impact of any proposed changes to the treatment pathways to outcomes such as hospitalisations, which could be useful for the NHS system planers.
The primary outcomes we are interested to measure in our study are: number of hospitalisations, both planned and unplanned; number of patients who would go on to develop new co-morbid conditions within the scope of CVDPREVENT audit.
Please refer to the Exposures/Outcomes/Covariates section for their detailed definitions.
Lianheng Tong - Chief Investigator - Boehringer Ingelheim GmbH
Lianheng Tong - Corresponding Applicant - Boehringer Ingelheim GmbH
Aviva Mazurek - Collaborator - Boehringer Ingelheim GmbH
Jeevan Kumar - Collaborator - Boehringer-Ingelheim International GmbH
Naj Rotheram - Collaborator - Boehringer-Ingelheim - UK
Patrick Keeler - Collaborator - Boehringer Ingelheim GmbH
Sander Martins Gonçalves - Collaborator - Boehringer Ingelheim GmbH
Sander Martins Gonçalves - Collaborator - Boehringer Ingelheim GmbH
Patrick Keeler - Collaborator - Boehringer Ingelheim GmbH
Sohomjit Ganguly - Collaborator - Boehringer Ingelheim GmbH
HES Admitted Patient Care;HES Outpatient;ONS Death Registration Data;Patient Level Index of Multiple Deprivation Domains;Practice Level Index of Multiple Deprivation Domains