Fuzzy Bayesian Network Model Weighs the Odds of Urban Public Health Crises

When a new pathogen begins to circulate in a dense city, the question that haunts public health officials is deceptively simple: how likely is it that this will become a major emergency? A new study published in Humanities and Social Sciences Communications tackles that question head-on, presenting a computational framework designed to estimate the likelihood of major urban public health emergencies before they fully unfold. The research, led by Lingmei Fu, Ge Shi and Mingguang Zhang of Nanjing Tech University, together with Yandi Zheng and Jinmei Wang of Wuhan University of Technology, applies an improved fuzzy Bayesian network to assess how a constellation of interacting factors can push an urban health threat from a manageable incident into a full-blown crisis. The team illustrates the method using the early COVID-19 outbreak in Wuhan as a case study, and their results point to a handful of governance factors that appear to carry disproportionate weight in determining whether an emergency escalates.

The core of the approach is the Bayesian network, a probabilistic modeling tool that has become a mainstay of risk analysis across engineering, environmental science and medicine. A Bayesian network represents a set of variables as nodes in a graph, with directed links encoding causal or influential relationships between them. Each node carries a conditional probability table that describes how likely the node is to be in a given state, given the states of its parents. Once the structure and parameters are specified, the network can perform what analysts call inference: given evidence about some variables, it computes updated probability distributions over the others. This makes Bayesian networks unusually well suited to problems like emergency likelihood assessment, where many factors interact, where evidence arrives piecemeal, and where decision-makers need to know not just which factors matter but how much they matter in combination.

Standard Bayesian networks, however, have a well-known weakness when applied to social and governance questions: they demand precise numerical probabilities, and such numbers are rarely available for phenomena like inter-agency coordination or the effectiveness of quarantine enforcement. This is where the fuzzy component of the new framework enters. Fuzzy set theory, introduced in the 1960s, allows variables to take graded membership rather than crisp binary states; a factor can be partly high and partly moderate, reflecting the vagueness inherent in human judgment. By combining fuzzy evaluation with Bayesian inference, the researchers built a hybrid model in which expert judgments expressed in linguistic terms, such as the difference between a measure being somewhat effective and very effective, can be translated into probability distributions that the network can process. The result is a model that captures uncertainty of two distinct kinds: the randomness of events themselves and the imprecision of the knowledge we have about them.

The improvement the authors introduce goes beyond simply bolting fuzzy logic onto a Bayesian skeleton. In assessing the likelihood of a major urban public health emergency, the relevant factors do not operate in isolation; they influence one another, forming webs of dependency that a naive model would ignore. The improved fuzzy Bayesian network explicitly accounts for these interrelationships among factors, structuring the network so that information can propagate through chains of influence rather than flowing only from independent inputs to a single output. The framework also provides a principled procedure for aggregating the opinions of multiple experts, so that the assessment does not hinge on the judgment of any single individual. According to the authors, these refinements make the assessment results more reasonable and more defensible than what conventional approaches produce, particularly in the governance-heavy context of urban public health.

To demonstrate the method, the team applied it to the COVID-19 pandemic in Wuhan, treating the outbreak as a test case for likelihood assessment within the overall framework of the public health governance system. The analysis drew on structured expert input, gathered under ethical approval from Nanjing Tech University with informed consent from all participants, to parameterize the network. When the model was run, it returned an unambiguous verdict: the likelihood of the COVID-19 pandemic in Wuhan was assessed as very high. That conclusion may seem unsurprising in hindsight, but the value of the exercise lies in the demonstration that a formal, transparent, reproducible procedure can arrive at such a verdict from structured expert knowledge, rather than relying on intuition or retrospective narrative. A likelihood score produced this way can be compared across scenarios, cities and time points in a way that expert intuition alone cannot.

Perhaps the most actionable findings concern which factors most strongly influence the likelihood of a major emergency. The model’s analysis identified three standouts: information sharing, coordination and cooperation among the various personnel involved in the response, and the effectiveness of quarantine and isolation measures. In the network’s structure, these factors sit upstream of the emergency outcome, and perturbing them produces measurable shifts in the computed likelihood. The implication is that the trajectory of an urban health crisis is not determined solely by the biological properties of the pathogen; it is shaped, to a substantial degree, by how well the governance system processes information and executes control measures. Cities that share data quickly across institutions, align the actions of medical, administrative and community personnel, and implement quarantine and isolation effectively can meaningfully reduce the probability that an outbreak becomes a major emergency.

The study also emphasizes timing. The authors report that strengthening public health governance and preventing escalation during the critical window, the early period in which an emerging threat can still be contained, is beneficial for reducing the likelihood of a major emergency. This finding aligns with a broader lesson of pandemic preparedness: the cost of intervention rises steeply as an outbreak grows, and the leverage available to decision-makers is greatest before exponential spread takes hold. A likelihood assessment tool of the kind developed here is precisely the sort of instrument that can be deployed in that window, giving officials a quantified estimate of risk while there is still time to act on it. The framework, the authors write, is intended to facilitate decision-making for abating and defusing major urban public health emergencies, not merely to grade them after the fact.

Methodologically, the work contributes to a growing movement in the risk sciences toward hybrid models that respect the messiness of real-world data. Purely statistical models struggle when historical data on rare emergencies are sparse; purely qualitative models struggle to combine evidence coherently. The fuzzy Bayesian approach occupies a middle ground, converting structured expert judgment into a form that supports rigorous probabilistic inference while preserving the graded uncertainty that experts actually express. The authors also situate their tool within the broader context of public health governance, arguing that preventing and controlling major public health emergencies is a prerequisite for ensuring high-quality urban development. In that framing, likelihood assessment is not a technical afterthought but a core function of urban governance, as essential to a city’s resilience as its hospitals or its infrastructure.

The research was supported by the National Natural Science Foundation of China and the Philosophy and Social Science Foundation of the Educational Department of Jiangsu Province, and it appears in an open-access journal, making the full model description available to researchers and practitioners worldwide. For urban planners, epidemiologists and emergency managers, the study offers both a tool and a message: the likelihood of a major urban public health emergency can be assessed quantitatively, the factors that drive that likelihood can be ranked, and the highest-leverage levers, information sharing, coordination and effective quarantine, are institutional rather than biological. As cities continue to grow denser and more interconnected, frameworks like the improved fuzzy Bayesian network may become standard equipment in the preparedness toolkit, turning the chaotic early hours of an outbreak into a window of quantified, actionable risk.

Subject of Research: Likelihood assessment of major urban public health emergencies using an improved fuzzy Bayesian network

Article Title: Likelihood assessment of major urban public health emergencies based on an improved fuzzy Bayesian network

Article References: Likelihood assessment of major urban public health emergencies based on an improved fuzzy Bayesian network. (n.d.). https://doi.org/10.1038/s41599-026-09227-z

Image Credits: AI Generated

DOI: 10.1038/s41599-026-09227-z

Keywords: Bayesian network, fuzzy logic, public health emergencies, urban governance, COVID-19, Wuhan, risk assessment, quarantine, information sharing, emergency preparedness, probabilistic modeling, epidemic response