Without doubt, one of the greatest legacies of the Five Year Forward View will be the profusion of new care models that are currently taking shape up and down the country.
From Primary and Acute Care Systems (PACS) to Multispecialty Community Providers (MCPs) and from Enhanced Health in Care Homes (EHCHs) to Acute Care collaborations (ACCs), each of these models has at its heart the Triple Aim of improving the health and wellbeing of local communities, providing a better experience of care for patients, and delivering lower per capita cost for the taxpayer.
To achieve the Triple Aim, the leaders of a new care model will first want to understand the health needs of its population, then analyse the quality, equity and efficiency of the care that is currently being provided, before identifying opportunities for improvement. Some early questions for these leaders to ask themselves are therefore:
- What data will we need to collect, collate, disseminate and publish?
- How will we obtain and use these data legally, securely and ethically?
- How will we use data to understand the needs and experiences of our population, to target preventive care and other initiatives most efficiently, and to monitor improvements in the Triple Aim?
To use data in these ways, the new care models will need to invest in new resources such as patient-level population datasets and capacity command centres; but they will also require skills and expertise that have traditionally been found on the commissioning, rather than the provider, side of the health service – skills such as actuarial analysis, predictive modelling, population health analytics and cost accounting.
In my previous role before joining NHS England, I helped establish a chain of Accountable Care Organisations (ACOs) in the United States. Although the context and culture were markedly different, these ACOs were likewise established to meet the Triple Aim, and they did face many of the challenges and opportunities currently being encountered by the vanguards in England.
Within each of our American ACOs, one of our first tasks was to establish two datasets:
- a planning dataset, which we used to understand the health needs and health care utilisation of the population, and to identify opportunities for the ACO improve the quality and efficiency of care. For example, we used this dataset to identify wasteful duplications in care (e.g. patients having the same test ordered multiple times), patients being offered low-value treatments (e.g. knee washouts), and unwarranted variations in care across the ACO (e.g. over-provision of chemotherapy at the end of life, and under-provision of discretionary surgery in different patient groups). We used this dataset to compare each clinician’s practice against national benchmarks and we fed this information back to them. We also trawled this dataset to identify Triple Fail events (i.e. costly, unpleasant, low-quality events such as unplanned hospital admissions or patients who received a preference-sensitive treatment without having first been offered decision support). Finally, we used this dataset to build a suite of predictive models designed to determine every patient’s risk of experiencing a range of different types of Triple Fail event.
- an operational dataset, which we used to track the quality of care being delivered by the ACO, to determine each patient’s eligibility for different types of preventive care, and to evaluate and improve our population health strategy. For example, we used this dataset to keep track of emergency bed days and average length of stay for the population, and to monitor gaps in care as specified by Medicare (e.g. an adult who had not had their blood pressure measured in the past two years). Every clinician in the ACO was prompted by this database if the patient in front of them had a gap in their care, and every clinician could instantly update the database once they had closed that gap. This too was the dataset upon which we ran our suite of predictive models to forecast which patients were at risk of different types of Triple Fail event, so that high-risk patients could be offered tailored, preventive care aimed at mitigating that risk. Finally, we used this dataset to refine our ACO’s services, for example through A/B testing our website and app, and by continually refining our impactibility models, which we used to predict which patients would benefit most from different types of preventive care.
In my current role, I am leading the ‘Care Model Design’ workstream of the new care models programme. The vanguards participating in this programme have huge ambitions to use data and analytics to improve the health of their populations; however, many vanguards say they would value some additional support in this area. In particular, they are keen to learn from each other, as well as from related initiatives such as the Integrated Care Pioneers and Test Beds, and from international best practice.
For this reason, my team and I are establishing two new networks to support the vanguards in an open and collaborative way: an information governance network and a population health analytics network. Vanguards will be able to use the latter to seek rapid peer review of their plans, to test and refine different hypotheses, and to become a more intelligent customer of data and analytics services. We hope in time that it will attract a broad mix of people from each of the different types of vanguard, as well as experts in the field of data and data science, whether from academia, the private, voluntary or public sectors.
This new network will focus on three aspects of population health analytics:
- Technical – what data flows are needed; issues of data storage and access; information governance etc.
- Analytical – which analyses to undertake; how to use analytical insights to improve clinical care etc.
- Cultural – using data and analytics ethically to shape and refine the vanguard’s strategy; how to become a more intelligent customer of analytical services etc.
If you are working in a vanguard or would like to support the vanguards through this network, you can register your interest using this email: email@example.com
Geraint Lewis is the Chief Data Officer at NHS England and an Honorary Clinical Senior Lecturer at University College London. He trained in medicine at the University of Cambridge and holds a Masters degree in Public Health from the London School of Hygiene and Tropical Medicine. Geraint began his career in acute and emergency medicine, working at hospitals in the UK and Australia over an 11-year period. After completing his higher specialist training in public health medicine, he was appointed Senior Fellow of the Nuffield Trust (an independent health policy think-tank), then as Senior Director for Clinical Outcomes and Analytics at Walgreens in Chicago, before returning to the UK to take up his current post.
A fellow of both the Royal College of Physicians of London and the UK Faculty of Public Health, Geraint is the lead author of the postgraduate textbook Mastering Public Health and has published over 40 peer-reviewed articles in journals in including Health Affairs, JAMA, Milbank Quarterly and the BMJ. Geraint was a 2007 Harkness Fellow in New York, during which time he received the National Directors’ Award at the U.S. Department of Veterans’ Affairs. In 2008 he was the “overall winner” of the Guardian Newspaper’s public service awards. In 2011, he was awarded the Bradshaw Lectureship of the Royal College of Physicians of London. Previous recipients include Sir Liam Donaldson, Dame Sheila Sherlock, and Sir Magdi Yacoub. More recently, he has served as an external adviser to the World Bank, and he leads the Care Model Design work-stream of NHS England’s New Care Models Programme.