Last week I went off to flood-bound Exeter, for a stimulating two day conference led by Martin Pitt at Peninsula Medical School. It was designed to bring together clinicians, managers, and patients, with researchers practising those strange sciences of systems modelling and simulation. These techniques have been under used in health, but there was a palpable sense of excitement over the two days that this was an approach whose time had come.
This is not new—health planners in the 1950s were using primitive modelling methods for booking outpatient systems. But latest techniques embrace the complexities of health and social care, the uncertainties and the multiple interests of commissioners, and the range of providers and patients. It is no longer—and perhaps never has been—a two dimensional numerical exercise.
We heard inspiring stories of how the particular techniques of operational research had been brought to bear on tricky NHS problems. These include using queuing theory to allocate and share scarce specialist mental health assessment slots between teams; applying stochastic modelling techniques to predict ambulance response times and plan rosters; using scenario planning to allocate capacity between medical, surgical, and cardiac beds on “service lines” in paediatric intensive care; and using system dynamics to re-model the entire unscheduled and emergency care systems in one locality.
There was a great presentation from Paul Harper, using software animations to illustrate the dangers of planning capacity on averages. If you fail to build in variability, a given in most systems dependent on human behaviour, your estimated average wait of 30 minutes in a walk-in centre becomes two hours. Check out his youtube presentation. This made me think of a brilliant book I read recently on the dangers of relying on “common sense” by the US engineer turned sociologist, Duncan Watts. A common sense planner would schedule outpatient waits based on average times from reception, to work-up with a nurse, to seeing a doctor. This would be wrong. A quote by Watts—“the whole trick is to know what variables to look at and then know how to add”—could itself be an epigraph for operational research.
One of the best parts of the two day event was a sandpit exercise where small groups of service leaders and operational researchers quickly worked up bids for new projects. These were pitched to the room, dragon den style. The outputs were impressive—from using location analysis to site diagnostic services across one region, to modelling how best to implement NICE guidelines for DVT care.
I ended the day talking to a paediatrician who had stumbled on the event, with no prior knowledge of systems modelling, and was inspired to get analytic help when making a business case for a new specialist epilepsy nurse and pathway redesign. There is a tension though between the very applied, local problem-driven analytics, and a more lasting body of knowledge. Sally Brailsford (mathematician, turned nurse, turned health modelling academic) had pointed to the paradox—we have a huge body of evidence, but few generaliseable outputs. She had identified 1008 individual papers on re-modelling emergency department flows. Were all these necessary? How can we learn from the best? As well as the embedded local analysts within a health organisation focused on particular problems, we need high quality research studies to generate national learning, by testing and validating models and carrying out robust evaluations of impact.
And so a long return from Exeter, with rather trying transport arrangements given the flood damage. During the discussion, some had raised the old argument that healthcare was just too complex to lend itself to mathematical techniques. The same of course used to be said of weather forecasting, where predictions of more than three days were notoriously inaccurate. But today’s weather modelling techniques, using historic data from multiple sensors, and understanding the interplay of solar activity, land masses, water temperatures, and wind flow are much better. Applied to health, techniques such as system dynamics can build in uncertainties (such as patient preferences) and variability (patient and clinician behaviours), with more sophisticated understanding of interactions (through network analysis and other) to predict more accurately how services might be used and savings could be made. Scenario planning can also present various “what-ifs” to integrate strategic uncertainties—a given in the NHS—into the planning process. Numbers themselves are not enough. But at a time of ever tighter financial pressures, can we afford to ignore the weathermen?
Tara Lamont has worked for over twenty years in health services research, audit, and patient safety. She currently works for the National Institute for Health Research and is an honorary fellow at the University of Warwick, but blogs in a personal capacity.