The Value of Disease Surveillance


RAMANAN LAXMINARAYAN, a senior research scholar at Princeton, is the founder and president of the One Health Trust, a Washington- and Bangalore-based research center.

Published January 23, 2024


The worst of the Covid pandemic is behind us. But after three-quarters of a billion (reported) cases, 20 million deaths and uncounted trillions of dollars in lost income, it is hardly time to declare victory and go back to denying the almost inevitable. Indeed, it is surely time to start preparing for the next pandemic – and, in particular, to start tackling the problem of detecting disease outbreaks earlier in order to minimize their impact.

Disease surveillance is at the heart of any good public health system. However, establishing and maintaining surveillance can be expensive, and it drains resources that could otherwise be used to directly improve health. Should we spend that marginal dollar on identifying and tracking unknown diseases, or in managing the clear and present dangers of global scourges ranging from tuberculosis to malaria?

Governments and donors don’t always see the value of investing in the former, where returns are less tangible. By contrast, the benefits of a new hospital or a vaccine distribution program can be counted in lives saved and misery spared. Even if decision makers are openminded about comparing lives saved today versus lives possibly saved tomorrow, how should they calculate outlays on surveillance? The answer depends on the value of surveillance – or to be more precise, the value of the information that surveillance systems provide. To a public health practitioner or policymaker, surveillance data are plainly helpful in knowing when and where an outbreak is occurring and formulating the most effective response. But that begs the question: how do we determine how much to invest in surveillance? Furthermore, are we interested in investments that can provide early warnings from many geographically distributed places, or do we value speed of detection? (Most of the time there is a trade-off.)

In the early 1960s, Ron Howard, a professor of management science and engineering at Stanford and the man who coined the term “decision analysis,” was helping an oil company decide whether to drill an exploratory well. The company could choose to drill without additional information, or first acquire seismic data about the area before deciding whether to drill. Howard and his team quantified the value of information by comparing the expected values (expected payoffs) of drilling decisions with and without the seismic data. They incorporated probabilities of finding oil and the associated economic payoffs into their analysis to calculate the expected utilities (benefits) of each decision option.

This value-of-information analysis was likely among the first formally conducted. In this particular example, the analysis showed that the potential gains from acquiring seismic information, in most instances, outweighed the costs.

The utility of surveillance information is dictated by whether public health officials can and will act if given sufficient notice of an outbreak. But in many countries, officials prefer to be seen as responsive to a crisis rather than preventing one.

These methods, which were popularized in the writings of Howard Raiffa, really originated centuries ago – notably, in the work of Thomas Bayes, an influential English mathematician and Presbyterian minister. Bayes’ theorem formalizes the process of updating probabilities based on new evidence and plays a crucial role in decision theory. By combining prior knowledge with observed data, Bayes’ theorem allows for the calculation of “posterior probabilities” that inform optimal decision-making.

What Determines the Value of Surveillance Systems?

Surveillance is one among many potential health interventions. To decide whether it is worth pursuing, we can evaluate the metrics that determine the value of information, which in turn can also help to identify which surveillance mechanisms provide the greatest value. Costs matter, too, of course. Costs range widely depending on the business model and sophistication of the platform.

The optimal investment in outbreak surveillance depends on three factors. The first is an assessment of the risk of an outbreak in the absence of surveillance information. A very low outbreak risk, all things equal, lowers the value of a surveillance platform. For this reason, it makes sense to combine surveillance for rare events – such as the emergence of a novel pathogen – with more routine surveillance for antibiotic resistance, influenza or other more commonly occurring pathogens.

The second factor is how actionable the surveillance information is, and whether health systems are equipped to use that information to improve health outcomes. Broadly, the less actionable, the less value it has. But the specifics help us design a surveillance system of greatest value.

Consider an example. Earlier this decade, an influential set of analyses was published on the odds of containing an outbreak of highly pathogenic avian flu. The analyses indicated that for successful containment one would have to have information about an outbreak within two weeks of the first instance of human- to-human transmission. So surveillance would only be actionable if it came within a defined period – in this case, a very short period.

A related question on actionability is whether it is even possible to contain the outbreak. Our reliance on airlines for long distance travel may make border restrictions and quarantines useful as barriers in the short term. But they can do little to contain global transmission in the medium term, except, perhaps, in cases like New Zealand’s response to Covid-19, where isolation was possible.

The utility of surveillance information is dictated by whether public health officials can and will act if given sufficient notice of an outbreak. This is a serious problem. In many countries, officials prefer to be seen as responsive to a crisis rather than preventing one, especially if they get little credit or public recognition for the latter.

The ability to use information depends on the capacity of the health system, of course. But the value of information may also depend on the kind of information being made available. For example, a well-functioning system of case notification for tuberculosis is of little value if systems are not in place to deliver treatment. However, predictive information may be particularly valuable when it raises awareness but doesn’t require actions with uncertain outcomes.

In the case of dengue fever, for example, the value of knowing what to expect may be high since it allows for reallocation of resources to the affected area at a fairly low administrative cost.

Laxminarayan Ramanan Disease Surveillance 2

The third factor in this calculation is how serious the outbreak will be and how much economic and health damage it will cause. All else being equal, it makes sense to spend more on detecting rapidly transmitted pathogens that could cause serious health and economic damage and less on those that are waterborne or foodborne and therefore spread less rapidly.

An Example of Calculating the Value of Information

Imagine a decision on how much to invest in a new surveillance platform for avian influenza, a highly transmissible virus with a high fatality rate. A human vaccine could be rolled out, but to be effective it would have to be implemented immediately after the first case of transmission from human to human. Producing and distributing a vaccine when there is no outbreak can be expensive. On the other hand, managing an outbreak when a vaccine has not been deployed can be catastrophic. We can illustrate the choice faced by policymakers using the hypothetical payoff table below. Let’s assume a 10 percent chance that a pandemic will occur in any given year. This is admittedly high, but this is an illustration, not the real deal.


The expected value of doing nothing is ($500 x 0.9 – $2,000 x 0.1) = $250. The expected value of vaccinating everyone is (-$1,000 x 0.9 + $400 x 0.1) = -$860. Therefore, the optimal strategy given prior probabilities is to do nothing because $250 is greater than -$500. And this, indeed, is why there is little preparation for pandemics. They are rare events, and control measures are expensive and make little sense to carry out without actually knowing if a pandemic is underway – by which time, it is too late.

Now imagine we have a choice of investing in an early warning system that can tell us immediately when a pandemic is barreling toward us. In this situation, we would vaccinate only when there is actually a novel outbreak underway. Otherwise, we may well waste a lot of money on vaccination of no value.

The payoff from this system is ($500 x 0.9 + $400 x 0.1) = $490. Therefore, the value of the surveillance system is the difference between the value associated with our best decision without the surveillance ($250) and the value with the surveillance system ($490), which works out to $240 (490 minus 250). One could see that if the cost of doing nothing when there is a pandemic goes up from $2,000 to $4,000, then the expected cost of doing nothing is $50. In that case, the value of surveillance goes up even more, to $440.

If the probability of a pandemic is 1 percent rather than 10 percent a year, then the economic value of surveillance goes down to $16. This is small but nearly 1 percent of the cost of the pandemic under a do-nothing strategy.

Unfortunately, these approaches to evaluating the optimal investment in surveillance have been available for decades but have yet to be deployed widely. Here, the enemy is a combination of ignorance on the part of policymakers and inertia on the part of decision analysts.

Laxminarayan Ramanan Disease Surveillance 4
Evolving Methods of Disease Surveillance

The classical method of disease surveillance involved looking for infected patients and then piecing together a picture of where the disease was spreading and in what time frame, as well as what the offending pathogen was. This is the method used by epidemiologists going back to John Snow, who identified the Broad Street communal water pump as the vector for cholera in London in 1854.

But there are other models. A longstanding approach is based on reporting by physicians. The most prominent system in existence is ProMED, a low-tech, mostly volunteer- run, online platform that relies on health workers around the world to contribute information on disease patterns they come across. ProMED was the first platform to report the 2003 SARS outbreak and was early to flag SARS-CoV-2, the pathogen behind Covid-19. There are other more sophisticated influenza tracking and sentinel surveillance platforms, but these lack the geographical reach of much simpler approaches.

With the expansion of internet access, other methods of tracking disease have emerged, including monitoring social media posts. This method has been used by researchers, as well as by portals, such as Google flu trends, that are based on web searches.

For years, the island nation of Taiwan has been tracking social media posts in China for potential signs of a recurrence of the 2003 SARS outbreak. Based on this tracking, the Taiwanese media monitoring unit detected social media mentions of an unidentified pneumonia outbreak in Wuhan in late 2019.

Although the original posts from China were swiftly deleted, screenshots had already been shared on a well-known online forum in Taiwan. And on this basis, Taiwan brought them to the attention of the World Health Organization. An advantage of social media tracking is its relatively low cost, and its ability to be conducted from afar. However, this method has the downside of needing confirmation before the information can be acted upon.

Another approach is wastewater surveillance, monitoring the sewage and wastewater in a community to gather information about the health of the population. This technique has been used for decades to detect the presence of infectious diseases (including polio), environmental contaminants and other substances in the water supply.

In recent years, wastewater surveillance has gained significant attention as a tool for detecting the local presence of Covid-19. By testing wastewater for the virus, public health officials can identify areas where it may be spreading even if local residents are asymptomatic or have not yet been tested. Wastewater surveillance involves collecting samples at various points in a community’s sewage system, such as at treatment plants or in the sewer network. These samples are then tested in a laboratory to detect the presence of offending substances, including viruses, bacteria and chemicals.

Although not as cheap as social media tracking, wastewater surveillance has the advantage of being able to pick up a range of pathogens, including new ones. The disadvantage is that one has to know what to look for. So a truly novel pathogen may escape detection in wastewater.

Are We Spending Enough?

The cost of the Covid-19 pandemic to the global economy ran into the trillions of dollars – even ignoring the intangible costs to families and victims. Thus surveillance can safely be assumed to be a good investment if the cost is manageable and the information has even a remote prospect of being actionable. Because surveillance information is a public good, the benefits and costs of health information are not reflected in its market price. However, it does have a social value that is possible to quantify, and this should guide the choice of investment as well as the appropriate scale and design of surveillance.

In this sense, a disease surveillance system is really no different than a weather forecasting system. And it seems clear that such a system should be funded at least at the scale of the National Weather Service (to the tune of about a billion dollars annually), or even surpass it. We are well short of that goal.

main topic: Public Health