What types of surveillance exist?

OLIVER MORGAN: In this presentation, I’ll provide an overview of different types of surveillance and how they are used to detect disease outbreaks. Surveillance systems capture information to understand disease trends in the general population. However, surveillance data only reflects a subset of cases that are actually occurring. This is because at each step in the surveillance process, some individuals or patients are missed. For example, not all people in the general population who are infected with the pathogen may become ill. Only a subset of ill people will seek care, of which some may have a laboratory specimen taken. Some samples might not be tested due to sample collection problems, and conclusive confirmatory testing may not be possible for all samples.

The result is that surveillance data reported to health departments will only be a subset of the actual disease occurring in the population.

Because of the time required for sick individuals to seek care and the information about their illness episodes to be reported to a health department, there is often a lag between when an outbreak starts and when it is detected. In this figure, we can see the time taken to discover outbreaks that occurred between 1996 and 2015 in each of the regions of the World Health Organization. We can see a clear reduction in the time taken to identify outbreaks, which is good. The median time to detect outbreaks was 20 days with a 95% confidence interval, a measure of uncertainty, ranging from 16 to 25 days. The median time to laboratory confirmation was 36 days.

Surveillance systems have many uses of which only one is detecting outbreaks.

Other uses of surveillance data include: understanding disease trends in geographic distributions; generating hypotheses for future study; and evaluating and planning disease control programmes, among other uses. It is therefore important that a broader approach is used to detect outbreaks rather than using surveillance data only. So what is a different way to detect outbreaks? Epidemic intelligence is an extension of the traditional surveillance approaches that we discussed previously. Epidemic intelligence has its origin in the 1960s when WHO established the first Epidemiologic Surveillance Unit. The term “epidemic intelligence” was defined by Alex Langmuir from the US Centres for Disease Control and Prevention (CDC), in 1971.

In the 1990s, two broad types of surveillance were defined, indicator- and event-based surveillance, which we will come on to shortly. The European Centre for Disease Prevention and Control definition of epidemic intelligence is “the systematic collection and collation of information from a variety of sources, which is then validated and analysed.” We include data from as many sources as we can, covering a wide range of surveillance data types. I will review some of these shortly. But the key message here is that epidemic intelligence uses multiple data sources.

These data need to be processed before they can be analysed,

which includes: de-duplication; linkage with other data sources; and geotagging, which is categorising data by geographical location.

The process data are reviewed to detect signals that may indicate the occurrence of a disease outbreak.

Each signal then needs to be verified by health professionals on the ground. This step may also include additional laboratory or field investigations. The verification step is very important to rule out signals that are not real outbreaks, such as those caused by data reporting errors.

Further analysis of verified outbreaks is then needed to fully understand the outbreak.

A risk assessment for each outbreak event is conducted as part of the decision-making process about how to respond.

The occurrence of verified outbreaks and the risks that they pose must then be reported to decision-makers.

For each verified event, a decision must be taken about how to respond. Responses can range from continued observation of the event through to operational actions and immediate control measures. At the WHO, we perform this epidemic intelligence cycle every day to detect and respond to disease outbreaks around the world.

Earlier, I mentioned that there are many different surveillance data sources. Some of the most frequently used are listed in this table with links to examples in the right-hand column. Briefly, indicator-based surveillance, sometimes called “notifiable disease reporting systems”, capture data about patients presenting at health care facilities. Event-based surveillance detects, mentions or reports aggregate data from events, such as reports in the media about the occurrence of an outbreak. Sentinel surveillance uses a limited number of reporting sites, usually clinics or laboratories, to report the occurrence of illness among the population in specific geographic areas. Participatory surveillance is a newer type of surveillance whereby members of the public or communities report illness events.

Many of these systems use mobile phone apps to facilitate interactions with members of the public. Laboratory networks can be used to capture diagnosed cases. This is especially helpful when conducting surveillance for diseases that need specialist laboratory testing. Countries can also report the occurrence of events using the International Health Regulation systems of WHO. Health services data can be used to understand outbreak events. For example, an increase in the number of patients presenting to emergency departments may indicate the occurrence of an outbreak. During humanitarian crises, special surveillance systems are often established to enhance the capability of public health authorities to detect outbreaks. Research studies may also produce data useful for surveillance, such as serologic surveys for certain diseases.

There are many other non-health sector data that can be used, such as meteorological data.

An important element of developing epidemic intelligence is the creation of new electronic tools. In countries with fragile health systems, it may be difficult to implement electronic tools, thereby limiting the ability to use epidemic intelligence. The Epidemic Intelligence From Open Sources initiative, led by WHO, has created a new event-based surveillance tool. This tool gathers data that are available electronically and processes them so that analysts can identify signals of disease events. EIOS is particularly good at scanning media reports of the occurrence of outbreaks. It is also able to process data from other surveillance systems, such as national reportable disease systems.

Here is an example of some data from EIOS that shows the frequency of media reports over a 24-hour period for some events. We can clearly see how we can use media reports to detect signals that an outbreak may be happening. The EIOS platform is also particularly helpful when considering diseases in both humans and animals, frequently referred to as “One Health”. Here we see an example of a report about Rift Valley fever that is relevant to both public health practitioners as well as veterinary practitioners.

To effectively use tools for epidemic intelligence, surveillance data needs to be digitised. In many parts of the world, surveillance data are still collected by paper and then manually entered. This is a time-consuming and error-prone process.

An increasing number of tools are being developed for electronic reporting of surveillance data. One of the most popular at the moment is DHIS2, which is hosted by the University of Oslo.

For surveillance during public health emergencies and humanitarian crises, WHO has developed the Early Warning (EWARS) in a box system. This is a kit including phones and servers that can be deployed anywhere in the world. A specially designed software system that is preloaded onto the devices enables large surveillance networks to be set up within days of an emergency.

Signals about possible outbreaks can also come from members of the community. Community-based surveillance is an approach that harnesses the local knowledge of communities to identify outbreaks early. Members of the community are often trained and provided with supplies so they can report suspected outbreaks to public health authorities. Technological advances in laboratory diagnostics means that laboratory data are increasingly available for surveillance. Portable molecular testing platforms and new rapid diagnostic tests means that near-patient testing can occur even in remote locations. In 2018, at WHO we processed around 700,000 pieces of global surveillance information each month. From these pieces of information, we identified about 7,000 signals of possible health events.

This led to about 300 investigations to verify signals, and of those events investigated, we conducted 10 risk assessments each month. At WHO, between the year 2000 and 2018, we detected and verified 4,600 events. About half of these were detected using event-based surveillance and nearly one-fifth using indicator-based surveillance, which you will recall, is the reporting of patients presenting at health facilities.

In conclusion, it is important to note that surveillance systems only capture a subset of illness that occurs in a population. There is always a time lag between the occurrence of an outbreak and the detection via surveillance system. However, this can be reduced by making the surveillance systems more efficient and by using new technologies. The epidemic intelligence approach reviews multiple data sources in a cyclical process to identify signals of disease outbreaks, which needs to be verified and responded to. In parts of the world with fragile health systems, it may be difficult to apply the electronic tools needed for effective epidemic intelligence.