Disease surveillance: no silver bullet

In thinking about the strengths and limitations of different surveillance systems — from the perspective of infectious disease forecasting in my case, but the point applies more generally — it becomes clear that there is no silver bullet. No one surveillance system will tell us everything we need to know in order to understand the current impact of a disease on a population, or to predict the future impact of a disease. The surveillance pyramid model illustrates this in a very nice and clear manner:

There is a sequence of events that must occur, and conditions that must be met, in order for someone to become visible to a surveillance system. This sequence proceeds in order from the bottom to the top of the pyramid.

1. Firstly, they must be susceptible to infection!
2. They must be exposed to the pathogen and become infected.
3. They must then develop symptoms; this can depend on factors such as their immune status, co-morbidities, and so on.
4. The symptoms must be sufficiently severe that they choose to seek health care. If the symptoms are particularly severe, it may purely be a matter of having access to health care.
5. The clinician must then choose to collect a specimen for testing.
6. The specimen must be viable (which can depend on how it was collected, and on how long it has taken the patient to seek health care) and produce a positive test result.
7. This positive result must then be reported in an accurate and timely manner.

Every one of the above events filters out more and more of the infected individuals, and so the end result is better thought of as the tip of a much larger "iceberg" of disease.

And, even more important than this matter of the visible proportion, is the fact that biological, immunological, and behavioural factors determine which individuals are visible to this system. These layers also sit between the quantities that are explicitly represented in mathematical models of infection (at the bottom of the pyramid), and the disease surveillance data (at the top of the pyramid).

These factors are subject to unknown biases that can vary markedly over time. The behavioural factors are also subject to conscious and unconscious biases and are surely influenced by external factors such as media coverage and peer opinion. Changes in any of these factors, independent of any change in disease incidence or prevalence, will affect the surveillance data. And this is the critical point.

No matter how good your surveillance system is, the data it obtains are heavily influenced by external factors that are themselves subject to complex and unknown dynamics.

In other words, there is no silver bullet. We need multiple types of data to give us an idea about what is happening in each layer of the surveillance pyramid, if we are to accurately interpret our surveillance systems and truly understand the burden of disease.

P.S. I am not the first person to think about disease surveillance in this way and come to these conclusions! But I did have fun talking about this problem at the 12th Australian Influenza Symposium.