As ever when I attempt to write a blog here I have stumbled across several topics I want to delve into further and I will to my best not to ramble on at tangents. What I plan to talk about is why, in our current work, we are proposing the use of agent based models (ABM) along side other modelling techniques (tangent number one: inconsistent use of terminology).
We are sure that ABM can have substantial value healthcare but we need to be able to explain why and how without intimidating people with the complexity and fine intricacies. The level of implementation of modelling methods is disappointingly low (tangent two), so we want to make our proposals as easy to use and see the benefits of as possible.
So, what are the benefits of using agent based modelling in general?
As with all modelling, ABM provides a quick, cheap and risk-free way to explore both reality and possibilities, compared to exploring and experimenting with real assets.
ABMs have a very flexible level of detail, as they are built from the bottom up. The details of decisions made are available as well as the statistics of all agents.
This bottom up approach also means that we can discover emergent phenomena, resulting from the behaviour and interactions between agents, some of which can be surprising.
The kinds of visualisations that can be created from ABMs make it a lot easier to comprehend than other modelling methods. The kinds of 2D and 3D graphics produced can be familiar and easily understood by the typical user.
The increased accuracy of ABMs mean that they can be used more reliably to make forecasts. Many predictions can be made for different ‘what if?’ scenarios.
“the duty of simulators ought to be first of all to the domain being simulated, not to a given simulation technology”
I’m sure we are a bit biased towards our agent based modelling approach but I certainly agree with this quote. If we are proposing ABMs for healthcare, we need to justify why it is a useful method for healthcare. And if it is such a useful method for healthcare than why is it not already being done?
A strong argument for using ABMs is that most of what make a person healthy is outside the healthcare system (80-90% of health is determined outside healthcare [https://www.local.gov.uk/health-all-policies-manual-local-government]). If we want to explore how to get people healthier we need to be looking, not so much at the patient’s health record but, at what they are doing day to day in their lives. Modelling the patients as individual agents makes it easier to identify where interventions can have the most effect.
The subject of why it isn't currently used is yet another tangent. There are reasons stemming from the current structure of health data and systems which require a separate post to address properly. Without getting into those intricacies however, what we can see is that ABM and other simulation methods are complex and scary to potential new adopters. The explanations and tools needed to implement ABM should be exciting and a joy to use, so that it what we are working to achieve.