The power of a carer's intuition - AI in social care
With winter pressures mounting, reducing front door demand at hospitals has never been more important. Social care providers have an increasing responsibility to prevent avoidable hospitalisations and alleviate pressure on acute providers.
Data and the use of artificial intelligence (AI) are opening up new opportunities to help care providers better understand which patients are at risk and intervene earlier to avoid deterioration and unnecessary admissions.
The data opportunity
Monitoring patients and their conditions in care settings can provide early warning of potential risks and enable care workers to take preventative action to avoid further deterioration. Doing this successfully requires good quality patient data in care settings, and tools that support care workers to make decisions about how to intervene.
Typically, identifying those at risk in care settings relies on inpatient, outpatient, and accident and emergency data from secondary care, as well as general practice. However, there are additional data sources hidden in social care which may provide further insight; namely, the daily ‘intuitive’ observations of a carer.
Carers are uniquely placed to perceive when their patient's behaviour changes; their daily interactions with patients over time result in an enhanced intuition to gauge when something isn’t ‘quite right’. This could be something as simple as a change in appetite, mood, or sitting position. Providing these observations can be routinely and accurately collected, and this intuition can be used to provide an early warning of patient risk.
Plugging qualitative insights into AI
But how do we take this intuition and plug it into an AI model? Carer observations are collected in the form of structured or unstructured visit notes. AI techniques can then be used to uncover the relationships between different observations, learn which data and patterns are indicative of a future hospitalisation, and then make predictions about a patient's risk.
A small change in sitting position, appetite, sleep, or speech may not seem cause for concern. But AI can reveal relationships between these variables which, when combined with other historical patient information, can signal risk of deterioration that might otherwise go unnoticed.
In this way, AI can help to extract additional insight from the data by revealing what might happen next and inform decisions about what to do.
For example, when a patient is flagged as high risk, the care provider might decide to change schedules to deploy more specialised care to individuals in need. AI is already being used in this way to help inform complex operational decisions. For example, a selection of hospitals in Wales are using AI to manage patient flow by forecasting estimated discharge dates when individuals are admitted. This provides operational and clinical staff with more confidence and forewarning, such that they are able to manage care and plan for discharge more efficiently.
Looking after the carer workforce
One challenge is that the value of observational data relies on the relationship between carer and patient. More time spent looking after the same individual leads to sharper intuition and improved observations.
It is no secret that recruitment and churn are major issues in social care, particularly in home-based settings. Research from NCFE notes the turnover rate within social care is 31%, compared to the UK average of 15% across all employment sectors. Early experiments suggest carers with longer tenures could improve the quality and, therefore, value of observational data, reinforcing the importance of retention.
With the reliance on accurate data input from carers, a question remains as to how accurate AI will be in the event of high staff turnover.
Currently, the technology can be used to address potential data quality issues via their ability to flag and spot anomalies. However, advancements beyond that are unclear, and truly personalised care and accurate observations depend on the consistency of the one primary carer.
Enabling accurate demand forecasting
In addition to informing intervention decisions, this insight can also be shared with acute providers to give them forewarning of demand for beds. Not all hospitalisations will be preventable, but having a better understanding of demand could also help hospitals better manage patient flow, particularly in the winter periods.
Similar prediction techniques will also be critical for the success of virtual wards. The opportunity to prevent unnecessary admissions and predict incoming demand could prove to be a useful tool in tackling NHS winter pressures.
Social and domiciliary care have a critical role to play in helping manage demand for acute services. Yet, to ensure these under-resourced settings are not burdened further, technology can help by extracting hidden value from social care's most important asset - carers and their intuition.
The ability to capture and delve into carer intuition could help ensure demand is better managed and distributed more effectively. Most importantly, it could support people to remain well, and stay out of hospital, for longer.
About the author
Megan Lathwood is an engagement manager and project delivery lead, working within the health and life sciences business unit at Faculty. Lathwood helps clients understand how their organisation can adopt AI and ML effectively, and works with data scientists and engineers to deliver applied solutions across the public and private sectors.
