Jul 05, 2021

How machine learning can transform home care

homecare
MachineLearning
Analytics
Data
Nick Weston
4 min
How machine learning can transform home care
Nick Weston, Chief Commercial Officer at Lilli, explains how machine learning and behavioural analytics are set to transform home care

Right around the globe, organisations that provide home and social care need more advanced technology to remodel provision in the face of enormous pressures.

Over-stretched workforces and budgets are struggling with ageing populations that live longer with chronic conditions. Into this mix has come the pandemic, adding the costly and unpredictable complications of long Covid.

The financial pressures are unrelenting. At the end of September 2020, the care sector in England estimated it faced £6.6bn in extra costs as a result of the pandemic and only four per cent of social care directors were confident their budgets would cover statutory duties. Although the government has subsequently made significant cash injections since then, including £2.9 billion for social care and hospital discharge, the financial constraints remain. 

Added to these pressures are the significant workforce shortages caused by low pay and the immediate after-effects of Brexit. With government spending likely to be further tightened, forward-looking social services chiefs know they must find more innovative ways to deliver care, especially in-home care and reablement, where budgets are often only second in size to those for residential and nursing provision.

Primarily this will require more advanced use of data and technologies such as machine learning (ML) to improve the quality of care and provide better outcomes for patients while making far more efficient use of limited resources. Although machine learning can often seem an impersonal technology, it is emerging as a major force in the transformation of care through its ability to learn from masses of patient data, spot patterns and flag up deviations. In a home care setting, this has huge potential.  

ML-driven solutions can analyse data from any kind of device or sensor in a patient or client’s home, learning their patterns of behaviour. The source of this data could be a Fitbit-type wearable device monitoring heart rate, or other sensors monitoring movement or the client’s use of power and domestic appliances. Elderly people, alongside those with learning disabilities or chronic conditions, tend to have established routines from which they rarely deviate, making it possible to establish a behavioural baseline.

What analytics solutions do, is to spot when an individual’s behaviour moves away from the norm, flagging up warning signs so care providers can intervene early and take appropriate action. That could initially be a phone call or a visit to establish why a patient is behaving differently, offering the chance of early intervention and the option of taking preventive measures. 

Organisations can tailor provision to the individual, creating a “flightpath” and setting thresholds appropriate to their condition that is not rigid. The accuracy of the data and the insights extracted reduces unnecessary and costly call-outs and visits while providing firm evidence on which to base decisions about care and resource allocation. Whereas patients may be reluctant or unable to discuss symptoms of a new or emerging problem, the insights from the data provide caregivers, clinicians and managers with firm evidence they could never obtain otherwise. 

Designed to be discreet, cloud-based ML solutions work on a wide range of unobtrusive devices, including smartphones, removing the need for large, unpopular medical hardware or highly intrusive monitoring technology. Such applications offer a far more sophisticated, preventive approach than traditional hardware-based, reactive systems that rely on alarms or use impersonal, rigid rules that cannot adapt to individual behaviour and result in false alerts. 

Rather than waiting for critical incidents, such as trips and falls, to flag deterioration, ML can spot the minor changes in behaviour that could indicate a new symptom. It is then open to care professionals to decide whether intervention is required. A simple example would be less frequent use of taps, kettle or toilet, indicating a potential problem with hydration, even though the client may be maintaining mobility. 

The advantage of ML is its ability to analyse vast amounts of data and extract insights specific to an individual and their condition. The information is designed to be meaningful and easy-to-understand for service providers and caregivers. But a more sophisticated platform will also format the information in a way that potentially makes it usable right across health and social care organisations – a vital requirement as organisations seek to create more unified care pathways and networks.

Systems such as this are already undergoing trials within the NHS where they are reducing care costs, through reduced visits by carers and lower rates of hospital admission. A trial of an ML behavioural analytics solution in Dorset is saving in the region of £4,000 per person annually through reduced visiting. 

Hardware-agnostic, with low costs of implementation ML-driven behavioural analytics systems, are certain to play a much bigger role in streamlining care provision and producing better outcomes for patients at a lower cost. Organisations need to be open to these innovations at times of financial and resource scarcity. Just as importantly, though, advances in technology such as behavioural analytics enable people to remain independent for far longer, which is a great gain in human dignity. 

Share article