New research shows that routinely collected social care records can help predict which people will need hospital admission or who may die in the next one to three years.
The findings suggest there is great potential in linking health and social care information to support more proactive, anticipatory care.
The study, published in the npj Health Systems journal, analysed pseudonymised records from 27,590 adults receiving social care in Oxfordshire, representing around 90 per cent of people receiving care in the county, to investigate whether they could predict future health and care needs.
The research was supported by the National Institute for Health Research Biomedical Research Centre: Oxford and UK Research and Innovation’s Engineering and Physical Sciences Research Council (EPSRC).
Using machine learning methods, the researchers analysed information such as age, health conditions, care assessments, service use, care plans, occupational therapy support and recorded hospital admissions.
They found that these data could predict future hospital admission and mortality with meaningful accuracy, with the best-performing models being around 90% correct when predicting mortality.
Predicting future care plan requirements was more challenging, reflecting the complexity and variability of people’s social care needs. This support does not depend only on a person’s health, but also on a number of other factors not necessarily recorded in social care data, such as whether they have family to support them, their housing and financial circumstances, local service availability and their own personal preferences.
Importantly, however, the researchers found that changes in care needs could emerge before evidence of subsequent clinical deterioration.
One of the authors, Dr Xiao Gu, Senior Research Associate in Medical AI at the University of Oxford’s Institute of Biomedical Engineering (IBME) and lead author, said: “We believe the approach outlined in this study could help health and social care services identify people who may need support sooner, before problems become more serious. This approach could support earlier reviews, better care planning and more joined-up working between health and social care teams.
“We foresee a future where well designed and evaluated data systems could help professionals identify people whose needs are changing and allow them to put care plans in place before the person’s condition deteriorates. These systems also have the potential to allow health and social care resources to be targeted more effectively.”
Another author, Dr Rebecca Nourse of the Nuffield Department of Primary Care Health Sciences, added: “As populations age and increasing numbers of people live with multiple long-term conditions, healthcare systems face the challenge of moving beyond the management of individual diseases towards understanding and responding to changing health and care needs over time.
“Better linkage of routinely collected health and social care information could provide primary and community care teams with a more complete picture of those trajectories.”
The research team stressed that these tools should support, not replace, professional judgement. They also said that predictive models of this kind needed further evaluation before they could be used to guide individual care decisions.
Given that the models were only tested in one local authority area, they said they should be validated in other areas, and future research should also examine whether these predictive models work equally well across people of different age, sex, ethnicity, disability or deprivation.