How can in silico models truly shape public health decisions? Despite their power, a gap remains between modelling and policy. As future crises loom, we need models that are used early, explained clearly, and trustworthy.
In the first two articles of this special series on the role of in silico medicine during the COVID-19 pandemic, we explored how computer models can help detect emerging outbreaks, simulate pathogen spread, model treatments, and assess the outcomes of pharmaceutical and non-pharmaceutical interventions.
But now, five years after the peak of the COVID crisis, where do we stand in using computer models in response to infectious disease outbreaks?
To understand this, it helps to look back. The marriage between infectious disease control and mathematical modelling is not new. Back in 1760, Daniel Bernoulli, crafted a pioneering model of smallpox transmission and control. In the early 20th century William Hamer and Ronald Ross laid the foundations for modelling measles and malaria transmission.
Then, the digital age supercharged this tradition. One of the first computerised efforts emerged in a 1967 report by the World Health Organisation (WHO) that focused on malaria. The report, edited by George MacDonad, Caton Cuellar, and Cecil Foll, opened with a strikingly modern vision: “An extremely powerful tool for the design of eradication and control programmes, could be produced in the extension of dynamic studies by computer techniques”.
It was a visionary prediction. From the Atlas supercomputer mentioned in the WHO report, to the Amdahl 5850 mainframe (a then powerhouse equipped with 128MB Hard Drive and 64KB RAM) used to model HIV transmission, computer power has long exploded. Today, even the average laptop supersedes those machines and are capable of running simulations in minutes that would have once taken days.
But computing power and model complexity alone aren’t enough. A more persistent bottleneck lies in how models are–or aren’t– integrated into real-world decision-making.
A 2016 paper in the International Journal of Infectious Diseases offered a stark warning. Titled Bridging the gap between evidence and policy, the authors pointed out that public health policies for infectious diseases often relied more on expert opinion than structured analysis, even though now we have the means to do that.
Why? A combination of factors: the perception that computer models are impenetrably complex, the fear that they rest on too many assumptions, and, crucially, a lack of meaningful dialogue between those who need quick answers and those with the tools to provide them.
Computational models offer a series of advantages. First, they allow us to map complex systems that underlie disease transmission and intervention. Second, they provide a structured and transparent way to compare policy options– especially when running real-world trials would be ethically, logistically, or financially complex. Third, they can reveal where data is missing and also guide what we need to collect next.
Unlike gut instinct or reputation-driven guidance, models are reproducible, adaptable, and open to scrutiny. They are not perfect, but they are transparent about what we know, what we don’t, and what might happen next.
Still hurdles remain. Some models are so intricate that become unintelligible to anyone outside the modelling community. Few other computer models are designed without effective collaboration with those who are supposed to be informed, say the clinicians. The result? Scientific publications and policy briefs with complex graphs may sometimes fail to speak to what a clinician, healthcare decision-maker, or a ministry of health representative actually needs to comprehend and make a decision on.
Then, there’s uncertainty–an unavoidable part of any type of modelling, be it in vitro, animal or even trials in human subjects. Every projection comes with a margin of error. What matters is how transparently that uncertainty is communicated, and whether decision-makers are willing to embrace and act despite the residual uncertainty.
Social and cultural context matters as well. A model is only as good as the assumptions behind it. If those assumptions ignore local customs, historical dynamics, or practical constraints, the model’s prediction risks being not just wrong, but irrelevant.
So, what would a better system look like?
According to this paper, it would start with collaboration. Policymakers and modellers should work together from the outset. The questions must come before the code. Once models are built, they need to be explained– clearly, and in non-jargon terms that relate to real-world actions. In turn, decision-makers need to also collaborate, by providing feedback, allowing the model to evolve along with the real-world scenarios.
Has this happened since COVID-19?
To a degree, yes. During the pandemic, models informed lockdown strategies, vaccine development, rollouts, and much more. The field made huge strides in visibility and influence. But visibility isn’t the same as long-lasting change and adoption.
Five years on, systematic collaboration between modellers and decision-makers remains the exception, not the norm. Partnerships are often ad hoc, informal, and fragile. As the pandemic fades from daily headlines, so does the urgency that once was driving cooperation.
And that’s a problem.
Because the next threats are already there: Dengue fever is expanding its reach; avian flu lurks at the human-animal interface; drug-resistant bacteria pose a growing challenge; West Nile, Mpox, and Ebola haven’t gone away. Climate change, ecosystem disruption, and closer human-wildlife interactions are shaping the odds. It’s not a question of if there will be another pandemic outbreak– but when.
If we’re to stand a better chance next time, computer modelling must become more than an emergency measure. It must be fully integrated into how we prepare and act in the face of infectious disease outbreaks.
This requires a cultural shift in how modellers, healthcare practitioners, and decision-makers collaborate, formalising pathways of persistent and continuous collaboration. Because, when the next crisis comes, decisions will again have to be made quickly. And better decisions start with better models– used early, communicated clearly, and trustworthy across all stakeholders.
References:
