Imperial College London

Academic Institute

Erik Mayer

Unknown

No

NIBDAPC_2022_0014

Blood clots, also called venous thromboembolisms (VTE) occur as either a clot in a deep vein, usually an arm or leg (Deep vein thrombosis (DVT)) or a clot that has broken off and travelled to the lungs (pulmonary embolism (PE)). They can happen to anybody and can cause serious illness, disability, and in some cases, death. Even though it causes a significant number of deaths and disability in the UK and worldwide, VTE is preventable and treatable if discovered in time. Identifying who developed a serious blood clot during their stay in the hospital is an important step in managing and preventing the illness, financial costs, and deaths associated with it. The current methods of detecting VTE depend largely on administrative data available after the patient is discharged. This method is known to have a number of drawbacks. The medical billing codes appear much later after a patient is discharged and are often not dated precisely. It makes it challenging to differentiate between events that occurred before the hospitalization and those that were acquired during the hospital stay. This makes the timely surveillance of the blood clots both inefficient and inaccurate. The detection and estimate of the actual number of patients who developed VTE during their hospital stay can be significantly improved by using the clinical narrative text available as part of the electronic health records. The results of imaging, such as ultrasounds, chest CT scans etc, that identify VTE are summarized in free-form text reports. While it is easy for human experts to identify an event by reading these manually, it is time-consuming and costly. This project proposes to apply advanced analysis techniques to detect instances of VTE from this free text data available digitally. Automating parts of this process could help reduce the time and cost significantly and help clinicians to manage the risk and treatment of VTE more efficiently in acute care hospital settings.

Hospital-acquired venous thromboembolism (VTE) covers VTE that occurs in hospital and within 90 days after a hospital admission. It is a common and potentially preventable problem but accounts for thousands of deaths annually in the NHS, and fatal pulmonary embolism remains a common cause of in-hospital death. Treatment of non-fatal symptomatic VTE and related long-term conditions is associated with a considerable cost to the health service. Without improvements and use of better data-driven processes to identify the incidence of VTE, the number of people affected by VTE can be expected to increase. People admitted to hospital or mental health units have varied risk factors for VTE. Although anyone can develop a blood clot, over half of blood clots are related to a recent hospitalization or surgery. A more accurate estimate of hospital-acquired VTE will allow us to improve and support monitoring and prevention of its occurrence. It will also lead to a better assessment of the public health burden of VTE by providing more accurate picture of the health and economic impact of VTE, including identifying high-risk groups and settings. It can also inform the development of improved monitoring tools to measure the success of prevention activities by tracking and monitoring trends in HA-VTE occurrence over time. The current methods used to measure safety events around VTE rely primarily on incident reports, which detect only a small proportion of events. Data-driven methods can be designed to automatically detect errors of omission, such as patients who are overdue for medication, monitoring, patients who lack appropriate surveillance after treatment, and patients who are not provided with follow-up care after receiving abnormal laboratory or radiological tests results. Combining different sources of data could potentially help detect some of the incidents of in-hospital VTE that are often overlooked and under-reported. Automating the process of detecting events through statistical processing of narrative text data, already present in the electronic health records could greatly reduce the time and cost required for monitoring VTE rates.

Public Health Research

17/10/2022

24/06/2022

De-Personalised

Not applicable

Not applicable

One-off

17/10/2022

TRE