Algorithm to predict asthma crisis events acceptable
The algorithm can successfully identify people who are at risk of A&E attendance, hospitalisation or even death due to their asthma
A publication in the British Journal of General Practice describes a study that designed a prediction tool to identify people with asthma who were at a high risk of having the severest form of asthma attack (an Asthma Crisis Event) – one which resulted in hospitalisation, A&E attendance or death. The algorithm can identify people who are five times more likely to have one of these severe attacks than those not at risk.
The researchers, from the National Institute for Health Research-funded ARRISA-UK (At-Risk Registers Integrated into primary care to Stop Asthma crises in the UK) study, wanted to develop and identify a prediction tool to create an “At-Risk Register” of people with an Asthma Crisis Event only using the information contained within primary care electronic health records. They are studying the effects of supporting practices on how to use this “At-Risk Register” as part of the ARRISA-UK study.
The study
Development of the algorithm
To develop the algorithm, the researchers used data from 650 primary care practices in the UK to identify patients with ‘active asthma’. The dataset comprised of data from the Clinical Practice Research Datalink (CPRD) between 2001 and 2012.
Variables considered as potential predictors for the at-risk algorithm included:
- Age
- Sex
- Smoking history
- Comorbidities
- Respiratory related medication
- Healthcare contacts and
- Blood eosinophil count
To develop the algorithm, the variables were measured against one or more asthma-related crisis events (A&E attendance, hospital admissions or death due to asthma) within a 12-month period.
A model was used to create ‘at-risk’ scores, which indicate the risk of an asthma-related crisis event for each patient in the dataset.
Validation
A dataset of patients from the Secure Anonymised Information Linkage (SAIL) databank who were registered at 340 general practices in Wales was used to validate the algorithm.
The algorithm correctly identified 28.5% of the asthma population most at risk, and 93.3% of those not at risk.
Implications for research and clinical practice
The algorithm developed in this study has the potential to save clinicians’ time and provide accurate real-time assessments of patients’ risk. It can be used to generate alerts or prompts to identify patients at high risk of asthma crisis events when their electronic healthcare records are accessed so that care can be targeted appropriately. Also, it does not require patients to attend appointments with their clinician.
Further research is needed to explore some of the findings from this study, such as low BMI. Additionally, how important social and behaviour determinants can be incorporated into this algorithm should be considered. These are not currently captured in primary care electronic healthcare records.
Each year in the UK an average of 1500 people die (on average, 3 a day) and 93 000 are hospitalised due to asthma. A total of 5.4 million people in the UK are currently receiving treatment for asthma: 1.1 million children (1 in 11) and 4.3 million adults (1 in 12). Identifying those most at risk means that care can be targeted towards them. Although risk factors for asthma attacks are known and there are other algorithms, they frequently require knowledge about patient characteristics such as adherence to medicine or asthma symptoms. This is time-consuming and has to be done individually but now it can be done automatically.
ARRISA-UK study
The algorithm is currently being used in the ARRISA-UK study to validate the role of at-risk asthma registers in primary care.
Find out more about the ARRISA-UK study.
Professor Andrew Wilson is Chief Investigator on the ARRISA-UK study and was involved in this publication. He said:
In our previous study that investigated the use of At-Risk Registers, we had to identify the people at risk manually. We can now do this with the click of a button.
Dr Samantha Walker, Director of Research and Innovation at Asthma UK, was also involved in this study. She said:
This research presents an exciting alternative to the current one size fits all approach to preventing asthma attacks by identifying those most at risk of having one. Despite the current evidence that people who have been previously hospitalised are at increased risk of another life-threatening attack, two out of three asthmatics do not receive a potentially life-saving follow up. If this research is to be implemented effectively, we need to see better data sharing across the NHS, linking GP and secondary care systems to prevent people with asthma falling through the cracks as they are now.
Read the paper
This publication is available from the British Journal of General Practice
Cite as
Predicting asthma-related crisis events using routine electronic healthcare data
Michael Noble, Annie Burden, Susan Stirling, Allan Clark, Stanley Musgrave, Mohammad Al Sallakh, David Price, Gwyneth Davies, Hilary Pinnock, Martin Pond, Aziz Sheikh, Erika Sims, Samantha Walker, Andrew Wilson
British Journal of General Practice 15 June 2021; BJGP.2020.1042. DOI: 10.3399/BJGP.2020.1042
Image: BlackJack3D via Getty Images