Public innovation in the real world

The Work Foundation’s seminar on public service innovation was redeemed by a powerful presentation by Geraint Hywell Lewis on the apparently arcane subject of predictive risk modelling.  Lewis was a  public health doctor in Croydon interested in how to identify the people at the greatest risk of hospitalisation who with earlier (and much cheaper) health interventions could avoid or reduce hospital admissions.

Finding the people who spend a lot of time in hospital is easy.  But time spent in hospital in the past is not a good predictor of time needing to be spent in hospital in the future.  Hospital treatment works, and people get better.  The King’s Fund has developed two models to help plan services – Patients at Risk of Re-hopitalisation (PARR)  and the Combined Predictive Model (CPM).  The former is limited in the way its name suggests – it only covers people who have a history of hospital stays.  The latter is more broadly based:

The Combined Predictive Model integrates accident and emergency, inpatient, outpatient and GP data sources to predict risk of admission to hospital across an entire patient population. It builds on the work undertaken to develop the Patients at Risk of Re-hospitalisation (PARR) models but, because it uses additional primary and secondary care data sources, the combined model is able to identify individuals along the whole continuum of risk as opposed to just those who have already experienced a recent hospital admission.

Lewis’s innovation was to use this tool to develop ‘virtual hospital wards’  – where his view of the essence of a hospital ward is that there is (i) a range of specialist professionals all of whom (ii) work from a single set of patient’s notes and (iii) meet daily – described in more detail in a case study written at an earlier stage of the work.

Looking at this from outside the healthcare system, there were two particularly striking points:

  • This is clearly an extensible approach.  The model already takes account of some social factors, and it presumably wouldn’t be hard to extend or adapt it to social care needs more generally
  • It shows that the most needy do not necessarily consume the greatest total share of resources.  In the crude graph on the right (reproduced from memory from one of Lewis’s slides), it is the middle group where the total cost of care is highest and (probably) where the application of predictive risk approaches may have the greatest benefit both to service users and service providers.  Again, it seems intuitively that that may well be relevant in other areas – and at the very least is a phenomenon to look out for.

All of that is good stuff – but at another level, this account misses the point of why this was such a powerful presentation.  The combination of new ways of looking at, and so of understanding, people’s needs, new ways of thinking about how those needs might better be addressed and, above all, immediate and practical exploration of how best to make it work is a model we should all be emulating.