The big challenge in social care is how to support a growing elderly population with reducing funds and fewer carers.
In less than 10 years, the number of people over 75 is estimated to grow by 50 per cent. Added to that, there is an increasing need to support people with a range of complex disability requirements. Plus, there is growing pressure to clear hospital beds as quickly as possible, meaning people will need recovery time at home with varying levels of support.
There is a big black hole in the funding for social care and the ability to recruit enough people to provide the care. An ageing population and more demanding care issues with younger age groups mean that the need for care is ever-increasing.
All of this would be challenging enough, but on top of this we are faced with a unique political/economic environment, with spiralling gas and electricity costs, the highest level of inflation for decades, and of course Brexit having resulted in a drop in the care workforce.
In other words: we are looking at a perfect storm.
Undoubtedly, technology can form part of the solution. The latest technology that everyone is talking about is AI, but what is it, and how can it help us now?
To understand this, it might be helpful to break this down into some key questions:
Typically, we respond to an issue such as a fall when it happens. After a fall has happened, the cost really grows – the human cost in terms of loss of confidence, losing independence, long recovery times, and the financial costs of ambulances, hospitals, care recovery, and so on.
But what if we focused on preventing the fall in the first place? Figure 1 shows how the costs grow as the incident process progresses.
So, the earlier we can identify the issue, the lower the personal and financial cost.
The key thing to understand is that everyone is different and everybody has a different living pattern. So, if we could identify changes in these patterns, we may be able to anticipate an issue before it happens.
Carebeans is looking at the use of monitoring and sensors to build digital doubles of people, defining their ‘normal’ patterns and identifying deviations. Based on growing knowledge, the aim is to predict issues early in the process rather than when an incident happens. This is where a ‘digital twin’ comes in. The digital twin is a pattern formed by monitoring the normal patterns of a person over a period of time. Changes to these patterns can identify an issue early, prompting early intervention to prevent an incident before it happens.
A simple example of this may be:
Of course, as the amount of data collected is enormous, the quality of the data is paramount.
Undoubtedly sensors are a key component. There are literally thousands of monitors and sensors on the market. Many have an early novelty value which inevitably declines over time.
Individually, such sensors may be of some use, but the value really increases if the right sensors are used in conjunction to form patterns. For example, following the UTI story:
Individually they are of limited value and any changes may be irrelevant, but together they form a detailed and complex picture.
Of course, sensors are not the only data inputs that are important. There are many other factors that need to be recorded, such as:
The question is: how can all of this information be processed, and how can divergence from the ‘digital twin’ be identified? Ultimately, how can the early stages of an incident be identified and the issue mitigated?
So, here we are – we have installed a number of sensors, we are recording multiple factors, and we have built the patterns for a digital twin from these sensors. We must now implement a learning process.
To do this we need to identify deviations from the digital twin, and then ask ourselves:
Machine learning is the gradual refinement of early issue identification and the best actions to take to mitigate the impact. Taking in thousands of pieces of data and analysing millions of pieces of historical data is the job of a myriad of complex algorithms. The greater the amount of historic data, the more accurate the algorithms can be in predicting the most likely patterns and hence the more effective the early intervention. Thus, the depiction of this system as a loop that never ends (see Response Feedback Loop diagram).
Following on from the learning process detailed above, we can determine a practical implementation (see Technology Enabled Care diagram). The core of the process is the person being cared for as represented by their care plan. In this case it is not just a document, it is:
So, with the care plan at the centre, let us run through a simple example:
Next time this happens, the issue progress and outcomes can be accurately predicted. Therefore, when the pattern starts forming, we can predict the issues earlier and provide earlier interventions.
As we know, the earlier in the process we can address issues, the lower the personal and financial cost, and here lies the real opportunity. In time, the overall costs of managing the issues will decrease, providing more funding for proactive care. This may be a simplistic view, and it will take interoperability between the NHS, councils, private social care, and software innovators.
The initial reaction to this question is to suggest the NHS or councils, but what is really needed here is innovation, uncluttered with bureaucracy, consultation, and consultants. This can come later when the concept is well proven.
The people who are going to drive this process are care providers who are looking to grow their business without the matching growth in staff numbers (i.e. delivering more with less), and there are a wealth of software companies in the social care arena who are continually bringing innovation to the market. This is going to require collaboration between these two parties, trialling, learning, and improving, without immediate expectations.
So why are the care providers the right people to make this happen? Well fundamentally because they know the person they are looking after better than other parties, except perhaps family members, and they are close at hand and can respond quickly. Plus, care providers have the skills training and knowhow, and they have a need to offer a high level of care with limited resources.
As demonstrated by the image below, the closer the proximity, the greater the capability and depth of knowledge about the person – and the greater the care impact.
This is a complex process with algorithms and sensor technology that will take years to deliver their full value. The great news is that work is underway, and with the current focus from private social care, software providers, and academic research, the AI dream is becoming a reality.
Article written in Sept 2023
Copyright © 2024 Carebeans All Rights Reserved Carebeans
The Innovation Centre, Sci-Tech Daresbury, Keckwick Lane, Daresbury, Cheshire WA4 4FS. Tel. 01925 386800