AIMS
This study aims to develop a machine learning model to estimate an electronic frailty index using routinely collected test results from hospital records.
BACKGROUND
There is a need to identify patients at risk of poor outcomes when they are first admitted to hospital. Frailty (described as a general weakness and inability to cope with physical challenges) is a known factor which is associated with poor outcomes, a greater length of stay and more hospital-associated complications or adverse events. Identifying those who are frail is helpful to identify situations that require early targeted clinical intervention, and to guide discharge planning and social care in order to lower the readmission and mortality rates.
Frailty is defined as a health state related to the ageing process in which multiple body systems gradually lose their in-built reserves. Identifying people as being frail when they come into hospital enables healthcare teams to provide additional support for these often vulnerable adults, and ensures discharge planning considers any extra support needed in the community. We identify frailty using complex scoring systems, which often are very subjective. This means that different healthcare teams can come to very different assessments about the same person.
Currently, assessments of frailty are made subjectively by healthcare professionals, and studies suggest that different clinical staff can make different diagnoses of frailty. If frailty could be assessed more accurately by automatically calculating frailty in an objective manner, it would reduce clinical variability of frailty scoring and support more targeted diagnoses of frailty.
RESEARCH
In this project, the researchers plan to use information that’s already collected in the hospital records to build a computer model that can predict how frail someone is. They will then see if this model can help predict if someone might have problems during their hospital stay, like falls, infections, or needing to stay in the hospital for a long time.
If the model works well, the researchers will be able to use it in the hospital to help healthcare teams quickly figure out how frail someone is when they first come in. This work will allow routine frailty assessment at admission and improve the accuracy of clinical judgement that will be beneficial to patients and efficiency of the busy acute care system.
PATIENT INVOLVEMENT
The researchers have spoken to patients admitted to hospital along with carers and family members discussing frailty assessments, how they are carried out and their complexity. Patients and family members were supportive of the study.
APPROVAL
This project was supported unanimously by the PIONEER Data Trust Committee.
This work is led by Xiang Li (Lily), Data Scientist at University Hospitals Birmingham NHS Foundation Trust.
