System dynamics modelling of layered climate–health vulnerability: structure, calibration and empirical validation

When scientists try to understand how climate change will hurt human health, they often start by looking at three things: how much dangerous weather a place faces, how sensitive the people there are to that weather, and how well the community can adapt to it. This traditional view helps identify which populations are most at risk before a disaster strikes. However, this approach has a blind spot. It treats vulnerability as a static condition, like a fixed score on a test, without fully explaining how that risk actually turns into sickness or death once a storm hits. In the real world, the strength of a local hospital, the availability of doctors, and the specific way a disease spreads can dramatically change the outcome. A community might look very vulnerable on paper, yet suffer fewer deaths because its health system is strong enough to absorb the shock. Conversely, a similar community with a weaker health system might see far worse results. To plan effective programs, experts need to see not just the risk, but the entire journey from the weather event to the final health impact.

A new study by researchers in Kenya and international health organizations attempts to map this entire journey. Instead of relying on a single, blended score, they built a layered model that separates the underlying risk from the systems that manage it. They began with the standard measure of vulnerability, which combines exposure to hazards, human sensitivity, and the ability to adapt. They then added a second layer representing the health system's ability to buffer or soften the blow. Finally, they added a third layer that accounts for the specific biological rules of how diseases spread in response to environmental changes. By stacking these layers, the researchers created a system that mimics how vulnerability actually unfolds in time, moving from a baseline condition to a managed state, and finally to a realized health outcome.

The team tested this model using thirty-seven years of data from Kenya, covering the period from 1987 to 2023. They wanted to see if adding these extra layers made their predictions match what actually happened in the real world. When they looked only at the traditional vulnerability score, it explained about half of the variation in actual health impacts. It was a decent starting point, but it left a lot of the picture unclear. When they introduced the health system's buffering capacity into the model, the accuracy jumped significantly, explaining eighty percent of the variation. This showed that knowing how well a health system can function during a crisis is crucial for understanding the final outcome. The model became even more precise when they included the specific biological relationships between the environment and disease, reaching an accuracy of eighty-six percent. This progression suggests that the traditional view captures the starting point, but only by including the health system and the biology of disease can we truly understand the final result.

The researchers also looked closely at how the system behaves as the underlying risk increases. They found that the relationship is not a straight line where more risk always leads to proportionally more harm. Instead, there is a tipping point. When the baseline vulnerability is low, the health system is effective at holding things together, and the impact remains manageable. However, once the vulnerability crosses a certain threshold, the system's ability to cope begins to falter, and the harm starts to amplify rapidly. This transition happens in a specific range of vulnerability, where the system shifts from being buffered to being overwhelmed. The model captures this shift, showing that the same level of risk can lead to very different outcomes depending on where the system sits relative to that tipping point.

To ensure their findings were solid, the team ran thousands of computer simulations, slightly changing the numbers and weights in their model to see if the results would fall apart. They found that the model remained stable and reliable even when the inputs were uncertain. The different parts of the model—underlying risk, health system strength, and disease biology—all contributed to the final picture without any single part dominating the others. This balance suggests that the model represents a complex, interconnected system rather than a simple calculation driven by one factor. The researchers also examined the errors in their predictions and found a pattern: the model tended to overestimate harm in the middle range of risk and underestimate it at the extremes. This pattern itself provided valuable information, confirming that the system's response changes depending on the severity of the situation.

The study concludes that treating climate-health vulnerability as a layered process offers a much clearer view than traditional methods. By keeping the underlying risk, the health system's capacity, and the biological reality of disease separate but connected, the model provides a transparent way to see where interventions will work best. If a region has high underlying risk but a strong health system, the focus might be on maintaining that system. If the system is weak, the priority shifts to strengthening it to prevent the risk from turning into a crisis. This approach does not replace the old ways of measuring risk but builds upon them, adding the necessary steps to understand how risk becomes reality. The work demonstrates that to protect people from climate change, we must look beyond the weather and the population's sensitivity to include the very real, very human capacity of our health systems to hold the line.