Population Health Management (PHM) is key to unlocking the shift from reactive treatment to preventative care. While readiness for PHM varies across the country, NHS Bedfordshire, Luton and Milton Keynes (BLMK) ICB is progressing several pilots designed to enhance care for specific patient cohorts and reduce A&E and GP attendances.
PHM uses linked data sets from primary, secondary and community care as well as social care where possible. Our role is to work with ICBs to analyse millions of data records and present information back in ways that are accessible and meaningful for clinicians and their teams. We are working with BLMK ICB to enable them to use PHM to segment their population and develop proactive interventions based on a detailed understanding of patient needs.
BLMK ICB is approaching the final stages of a pilot programme using PHM to support care for the 67,833 patients with diabetes across the system. In what has been a clinically-led project, we’ve co-designed a dashboard for practices with a multidisciplinary team that gives GPs, pharmacists and community nurses the most relevant information to help them manage their patients. The dashboard – which combines linked data sets and is refreshed daily – allows practices to proactively monitor activities such as whether care processes are being completed, when diagnostic data was last updated, and correlate that with other elements such as the medication patients are receiving and their hospital usage. The live data is particularly useful in monitoring those most at risk of hospital attendance and enabling practices to proactively contact patients and provide support to reduce this risk. The dashboard is due to roll out to all practices early this year.
BLMK is also using PHM to maximise the potential impact of social prescribing by using data analysis to identify which patients are most likely to benefit from and engage with these approaches. At the top level, we are using aggregate data to help BLMK understand existing and predicted use of social prescribing. We’re also drilling down to specific patient cohorts to identify who could benefit from social prescribing based on a mixture of profile data and expert input from social prescribing link workers. Multiple factors such as socio-economic status, long term conditions, past uses of health services, body mass index etc. are scored and applied to identify those most amenable to social prescribing. Although currently still in pilot phase, the aim is that this data will be used to contact patients proactively to offer options which can improve health and wellbeing and ultimately reduce GP and hospital usage.
The second phase of the social prescribing pilot is training machine learning to apply weighting to the different factors, using historic data, to help link workers make more informed decisions.
COPD and machine learning
Machine learning is also being used in BLMK’s PHM project to reduce the risk of admission to hospital among patients with Chronic Obstructive Pulmonary Disease (COPD). There are over 16,000 patients with COPD across the system with respiratory problems accounting for 11% of total emergency hospital admissions. The first phase of this project is using machine learning to help score individual patients’ risk considering 120 different variables, resulting in an accessible dashboard which predicts those most at risk of admissions due to exacerbation of COPD. The dashboard enables clinicians to quickly identify and support those at highest risk.
The second phase involves developing a survival analysis model for COPD patients to help improve quality of life and improve how care is monitored. Survival analysis uses statistical data (rather than clinical) to predict how long patients are likely to have before the next event happens. Although COPD is a progressive disease, research suggests that there is room for delaying adverse outcomes of the condition with proactive interventions. After a second hospitalisation for severe COPD exacerbation, rate of subsequent admission is three times higher than after a first hospitalisation1. This suggests a significant window of opportunity after the first hospitalisation in which to intervene, for example with structured self-management, to prevent a subsequent severe exacerbation. This new model will help equip clinicians with tools to identify when and what types of interventions can have best effect at patient level.
Although PHM has been talked about for some time, its use in frontline care is still relatively new and it will take time for the benefits to be fully reflected in patient outcomes. Linking data sets is not easy and some areas are more advanced than others, but even limited data can provide a strong starting point. With resources in primary care stretched more than ever, ICBs and PCNs have much to gain by gaining a deeper understanding of their population needs and using analytical tools and dashboards to identify and implement proactive interventions which will have the greatest impact for their communities and reduce pressure on health services.
- Suissa S, Dell’Aniello S, Ernst P Long-term natural history of chronic obstructive pulmonary disease: severe exacerbations and mortality.Thorax 2012;67:957-963.
By Prasanth Peddaayyavarla, Head of Data Science at NHS Arden and GEM Commissioning Support Unit