All Along the Watchtower: The Global GeoSentinel Network
Q&A with David Hamer, MD (Part 1)
Davidson Hamer, MD is Professor of Global Health and Medicine at the Boston University Schools of Public Health and Medicine, co-lead of the climate change and emerging infectious diseases research core at the BU Center on Emerging Infectious Diseases, and an attending physician in infectious diseases and Director of the Travel Clinic at Boston Medical Center. Dr. Hamer is a board-certified infectious disease specialist and medical epidemiologist with particular interest in emerging arboviral diseases, tropical medicine, travel medicine, infection control, and antimicrobial resistance. Dr. Hamer has been involved in travel medicine for thirty years. From 2014 to 2021, Dr. Hamer served as principal investigator and, since September 2021, as Surveillance Lead, of GeoSentinel, a global surveillance network of 68 sites in 32 countries using returning travelers, immigrants, and refugees as sentinels of disease emergence and transmission patterns throughout the world. He also has been actively involved in enhanced screening for and treatment of patients with Chagas disease at Boston Medical Center. Dr. Hamer is currently the Scientific Program Chair for the American Society of Tropical Medicine and Hygiene and Section Editor of the American Journal of Tropical Medicine and Hygiene (global health and Chagas disease).
History & Mission
Tom Mahoney: GeoSentinel is now over 30 years old. What was the original problem it was designed to solve, and was the founding vision prescient given what we know today about how infectious diseases move?
David Hamer: GeoSentinel was founded at the CDC around 1995. The idea came from Marty Cetron, Division Director at the time. It was initially called Division of Quarantine and eventually became the Division of Global Migration and Quarantine, including the Traveler's Health Branch as a division. The project was in response to an Institute of Medicine report on concerns about emerging infectious diseases and the need for better surveillance. GeoSentinel was one of a couple of different projects; the Emerging Infections Network at IDSA was another launched at essentially the same time. GeoSentinel definitely has been prescient. It's been very useful over the last three decades to help identify outbreaks, as these have obviously continued to occur.
Mahoney: The network is described as using returning travelers, immigrants, and refugees as "sentinels of infection." Can you unpack that concept for us - why are travelers uniquely valuable as an epidemiological signal?
Hamer: That's an important question. This signal stems from the fact that in high-income countries like the United States, Canada, Western Europe, Japan, Australia, and many places where we have sites, there's very good diagnostic capacity. At least at more specialized tropical medicine centers, clinicians think about the kinds of diseases that might be circulating in certain parts of the world. By contrast, many lower income and even some lower-middle-income countries, don't have good surveillance systems in place, may not have adequate laboratory supplies to make specific diagnoses, and may miss outbreaks before they’ve progressed pretty far. Look at what's happened, for example, with Ebola in the last couple of months. This was a combination of failure to report the outbreak, inadequate surveillance, and a delay in diagnosis because of a lack of proper laboratory reagents. With this major outbreak and others, with travelers in particular, but also immigrants and refugees, when they move to a new place while they're infected, that location - if it is a GeoSentinel site - can make a diagnosis of a disease that may not have yet been picked up in the country of origin.
Mahoney: How has the network's governance and funding structure - the CDC cooperative agreement, ISTM partnership, and Public Health Agency of Canada support - shaped what GeoSentinel can and cannot do?
Hamer: Having GeoSentinel managed by a professional society, which is quite different than the standard academic model, has worked well. It's been beneficial to the CDC because they didn’t have to pay any overhead initially. Now they're paying a very low indirect rate, so that's allowed more money to go into project activities. There have been some challenges. For example, Public Health Agency of Canada (PHAC) hasn't played that great a role. They've given us a little bit of funding, and we can access Canada-specific data and have allowed PHAC to use this data for various analyses over time. Interestingly, when we looked at the data from our eight sites in Canada a couple years ago, we estimated that those sites accounted for 15-20% of all returning travelers seen in Canada. So even though we’re not nationally represented, the coverage is more extensive than in many other locations. Having a partner like the CDC in particular is really important when we need to reach out to either global health agencies, like the WHO or PAHO, or country-level agencies. If there's an outbreak identified, we don't have all those agency connections, but CDC does.
Mahoney: You've published on GeoSentinel's "past, present and future." In that arc, what has changed most fundamentally - the pathogens GeoSentinel tracks, the tools it uses, or the world in which it operates?
Hamer: I would say all three, but there has been an evolution in the kinds of core data collected. For example, for antibiotic susceptibility testing we added nine core pathogens from the World Health Organization Global Antimicrobial Resistance and Use Surveillance System (“GLASS”). That's been really interesting, because it's generated data over time and by geographic region for pathogens such as S. typhi and S. paratyphi, or enteric pathogens like Shigella. We've also done more enhanced surveillance studies with a specific set of questions related to a single pathogen or a single syndrome that have allowed us to collect more data and gather better insights than in either of these single lines of investigation. Perhaps the biggest change in recent years is that GeoSentinel is doing a lot more molecular epidemiologic studies, which really is the direction of the future.
Network Architecture & Surveillance Mechanics
Mahoney: GeoSentinel aggregates de-identified electronic patient data - demographic, clinical, and travel history - from expert clinicians at each site, enabling linkage of final diagnoses to geographic exposures. Walk us through the life of a case from the moment a patient walks into a GeoSentinel clinic to the moment that case potentially triggers a network-wide alert. What does that information flow look like in practice?
Hamer: There are different ways this might happen, but there are certain diagnoses we consider alarming, for a list of diseases we want to know about right away. Yellow fever is one of them, as are Ebola and Marburg. In the past, going back to around 2015-16, Zika virus was on this list, but then became so prevalent that we took it off the list. But if there's a report of an alarming diagnosis, after the identifying site sees it we try and get them to email us to say, for example, we've got a possible yellow fever case. The site sometimes communicates with us in advance, but if not, as soon as they enter that case in the database - and for this purpose we've got a very comprehensive diagnostic manual, with very specific definitions for probable versus confirmed cases - if it's one of those alarming diagnoses, it triggers an alarm consisting of an email to the GeoSentinel Office of the Principal Investigator, to our CDC collaborators, and then to the site itself.
This protocol allows us to know about the diagnosis immediately, but also to respond to the site and ask for more details if there are any questions. Over a decade ago, we had a “clinically suspect” category, but cases were often so suspect that we didn't trust the data, so we stopped using that category. So, we must have some degree of confirmation. In any case, once the site enters the diagnosis, the next step is interrogating the database - have we seen this disease recently in this place, or at all? Yellow fever was a unique example where we had seen no cases historically, and then suddenly we saw one case, looked at the database, saw there weren't any in the past, and then sent out an alert to all of our sites. We used to post these alerts on ProMed but now that service has gone behind a paywall; not as many people use it because it requires a subscription. We're now posting the alerts on BEACON, which is open access, disseminate the information, use AI approaches to try and gather more information about the disease which is the subject of the post, confirm that there is an outbreak, and double-check with other sources of information. Besides reporting to BEACON, we disseminate the information to all our sites. We also have affiliate members who don't routinely provide data, e.g. smaller-scale tropical medicine locations that are seeing returning travelers and occasionally will have seen the disease in question, so we alert them. We also collaborate with TropNet, a collection of tropical medicine sites within Europe. Some bigger TropNet centers are part of GeoSentinel, but several smaller ones are not. By undertaking all the alert steps I’ve just described, we end up finding out if anybody has seen the disease we are concerned about from any given location. If so, that may then generate an outbreak report, communication with public health authorities, and ability to launch a response.
Mahoney: One of GeoSentinel's stated strengths is physician-confirmed diagnoses versus, say, passive self-reporting or social media signals. How does this diagnostic validity advantage translate into downstream public health value - and what does it cost in terms of coverage and speed?
Hamer: As to the first part of the question, the advantage is better-quality data, because we have very strict definitions of how a disease can be confirmed or probable. Confirmed usually means, for example, PCR or antigen positive, or visualization of a parasite. By contrast, probable might mean we have just a single serology. But there's also a lot of clinical metadata associated with the diagnosis. We typically know where and when people traveled, time of exposure, and other key details like purpose of travel. Where a given traveler was when they became infected, and when they were there, can be very important information to public health authorities. So, we generate a lot more granular data than a lot of surveillance systems have, certainly compared to, say, social media, where you might see a post like “there's a bunch of people in my neighborhood that are coughing”. That doesn't really tell you what disease these people may have. Thus, we can give a much more specific diagnosis. One potential downside is that acting on our alerts requires physicians to be available, to be seeing the patients, and they need to have diagnostic capacity. There may also be a delay in diagnostic tests coming back, as depending on how many cases the site has and a variety of other factors, there may be a delay in submitting patient records into the database. This is an issue we've been trying to address in recent years, at least for certain target organisms. To your question about GeoSentinel economics, the overall network costs are relatively low in relation to the amount of data GeoSentinel generates, since the major costs are for overall management by the lead surveillance team and site incentives. In reality site incentives alone are insufficient to cover all site costs for generating and submitting data to GeoSentinel, so we depend substantially on the motivation, good will, and academic interest of the sites to work together to submit data that can be used identify outbreaks and to assess the evolution of risk for travelers by geographic region.
Mahoney: GeoSentinel now has over 400,000 traveler records since 1995. How do you manage data heterogeneity across sites and across 30 years - shifting variables, changing diagnostic technologies, different clinical definitions - while preserving longitudinal comparability?
Hamer: Some of this relates to our definitions of syndromes and diseases, which have stayed fairly similar over time. The reality is there have been some major changes in the range of variables we've collected, and sometimes definitions, which we must consider when we do a very long-term analysis. Many of the analyses we've done in recent years have been limited to the last 10 to 15 years of data rather than back to 1996, when data started being entered, often because we've already done smaller analyses. In the last 10 or 15 years, there have been a lot more sites, and the data volume is higher. But there are issues with comparability. Having standardized definitions helps, as well as quality control measures where we periodically evaluate data from different sites to try to fix systematic errors. We also do data cleaning when we finally do an analysis. If we're doing a retrospective large study, we will review, and sometimes must exclude, pieces of data. It really depends on what the context is. If a patient presents with a febrile illness, and we're considering a diagnosis of dengue, but the patient had concurrent malaria, then we might determine that we won't accept dual diagnoses in that case, but rather just focus on dengue alone. There’re various forms of cleaning, but sometimes there's just too much data missing. For example, the region or country of exposure may not be ascertainable. We try whenever possible to pin down the country of exposure, but the reality is, if somebody's traveled over a couple weeks, and they've been to multiple countries during the incubation period of disease, we can't say exactly where they contracted the disease. At best we can just say that they became infected in a given region. In sum, there are definitely some challenges with the data.
Alarming Diagnoses & Rapid Response Capability
Mahoney: The network has flagged sentinel events including dengue in Angola (2013), Zika in Costa Rica (2016), and yellow fever in Brazil (2018). Can you take us inside one alert episode - from the first unusual cluster of cases to the moment GeoSentinel notified public health partners - to illustrate how the rapid response loop functions? The 2017 Zika outbreak in Cuba is striking - GeoSentinel traveler data revealed far more cases among international visitors than Cuba's own domestic surveillance was reporting. What does that case tell us about political and structural gaps sentinel traveler surveillance can fill?
Hamer: The 2017 Cuban Zika outbreak is an exemplary situation where unfortunately, as in a couple of other instances, there's been incomplete reporting from the country due to restrictions. A couple of factors are at work here. In the first instance, government restrictions on release of information may be attributable to concern about the country’s international image, e.g., alerting the world they've got a big outbreak of dengue, Zika or chikungunya. But at the same time, there may also be limits in the country’s deployable diagnostic capacity, so a lot of probable or suspect cases of the disease lack actual laboratory confirmation, or there may be delays in diagnosis. I think with Zika in 2017, there was indeed substantial underreporting, but I can't say for certain which of the foregoing factors was responsible for that. The scientist who uncovered this outbreak was our collaborator, Nathan Grubaugh, at Yale, looking at patients going from Cuba to southern Florida. His lab was making many diagnoses of Zika a year after the big peak of Zika in the Caribbean. The authorities in Cuba had done a lot of spraying the year before, knocking out a lot of mosquitoes, resulting in less disease transmission. But they must not have sprayed the next year, and the mosquitoes – and Zika - surged back, thus showing sort of a late peak. But our GeoSentinel data, with high sensitivity, was able to identify that sort of secondary rise. If you looked at all the data across the Caribbean, you couldn't see the Zika signal, but when you focused on Cuba, suddenly you saw that there was a meaningful rise in Zika cases.
Mahoney: When GeoSentinel issues a rapid alert, in a situation like the 2017 Cuban Zika outbreak, which as you point out, is rather unique, what are the formal notification pathways - CDC, WHO, PAHO, ProMED? How coordinated is the response that follows?
Hamer: We are not always that deeply involved in the response. We often will just defer to the CDC. In contrast, some diseases, like African Trypanosomiasis (sleeping sickness), we report directly to WHO. In many countries we work in, sites are obligated to report to their national public health authorities. A great example, as you already mentioned, is what happened with the yellow fever outbreak in Brazil. In this situation, we identified a case, wrote to all the sites and quickly identified a couple more cases. But then one site had a patient dying in the ICU in Bucharest and presenting with jaundice. Although his skin was yellow and he had fever, the clinicians were thinking about all other possible pathogens, e.g. acute viral hepatitis and EBV, that could cause that pathophysiology. They had not thought about yellow fever. Once the physicians saw our alert and knew this patient had been in southern Brazil, they decided to test for yellow fever, and indeed, that is what the patient had. So, we were able to quickly compile a group of cases, and then share that with public health authorities, and with PAHO. Then CDC and PAHO realized they had a problem, as there were areas of Minas Gerais in Brazil that they didn't think were yellow fever endemic, all the way down to Rio, but they found that in an island offshore Rio called Ilha Grande, there were yellow fever cases. So, we had to change our vaccine recommendations. This led to a remapping of southern Brazil in terms of the risk of yellow fever and the need for vaccination, so overall this was a satisfying public health response. In the later steps of the response CDC, PAHO, and WHO all worked together in a coordinated manner.
Landmark Case Studies & Outbreak Characterization
Mahoney: GeoSentinel has characterized mpox epidemiology in a cross-sectional study (2023), Zika in Cuba, and chikungunya in Bali (2022), among others. Let’s talk about some landmark case studies and related outbreak characterization. Mpox was a case where GeoSentinel produced one of the defining early epidemiological characterizations, published in Lancet Infectious Diseases. What did the network's global, multi-site data reveal that single-country surveillance might have missed?
Hamer: The mpox case was a unique situation, where we broke our own rules to some extent. The requirement to be a case in GeoSentinel is you must have traversed an international border. But in this outbreak, we heard about it through a contact in the UK. We had an annual site directors meeting, and they said they just saw a case of mpox, and then we started looking, and suddenly it just exploded, with many sites that were seeing mpox cases. Not all the cases had traveled, but some had crossed an international border, so we quickly compiled the cases to try and describe their characteristics. This was a great example of taking advantage of our network to do large-scale analysis of a new outbreak very quickly and to try to characterize it clinically.
Mahoney: Are there cases where GeoSentinel raised an alarm and the public health response was inadequate or slow despite the signal - and what does that teach us about the interface between surveillance intelligence and political or institutional will to act?
Hamer: That's a tough question. I can think of a couple of situations during the Zika outbreak in the Western Hemisphere. As the disease spread from country to country, the initial case definition developed by the Council of State and Territorial Epidemiologists was a constellation of certain symptoms, like fever plus rash, for example, but you had to have been in a place with Zika in circulation. The problem was, what if you were in a new place that didn't have Zika yet? We had a patient seen by my colleague here in Cambridge, who had been to Costa Rica, which wasn't on the Zika endemic map yet. My colleague suspected Zika and was eventually able to confirm the diagnosis. We reported this to the CDC, who shared that with our colleagues in Costa Rica. Their reaction was this had to be a mistake as they didn't have any Zika there. We pointed out that our patient was in a specific area of Costa Rica. Two weeks later, the government announced an outbreak of Zika in that area. So, there was a delay in response in this situation because local authorities didn't believe our results. Something similar happened in Vietnam when we identified a Zika case, and again local public health officials denied that our diagnosis was possible. So indeed, in cases like these, local government denials of diagnoses can potentially lead to a delay in a public health response. Another problem occurs if we report back to a country that has limited resources. Government health officials may say that they have other higher priority diseases to contend with, and limited financial and surveillance resources, so they decline to do anything about a new disease we have identified.
Mahoney: Beyond acute outbreak detection, GeoSentinel has published deep characterizations of diseases like malaria and dengue stratified by geography, traveler type, and clinical presentation. How might this kind of descriptive epidemiology ultimately change clinical practice at the bedside, in your opinion?
Hamer: There are many different large case series that we've done over the years - disease-specific or global analyses - and the data from these studies are important for practicing clinicians. The data often makes its way into peer reviewed publications, for example, reviews of fever and returning travelers, or into chapters in books, as well as the CDC Yellow Book on travel medicine. The data provides continuing guidance for practitioners - seeing returning travelers, immigrants or refugees in their clinical practice - on what to expect from different parts of the world. In addition to helping clinicians with the epidemiology, we've also published more detail on clinical manifestations than they might find in other sources, for example a review a couple years ago on fever in returning travelers in the New England Journal of Medicine. Many citations in the article are GeoSentinel studies, which provide a lot of evidence useful for practitioners.
Mahoney: Dr. Hamer, on behalf of our readers, thank you for the generosity of your time and depth of your insight into how this network quietly helps keep the world safer, three decades on. What comes through powerfully is that GeoSentinel isn't just a data repository, but a dynamic early-warning system built on clinical instincts of physicians around the world, from the yellow fever case in Bucharest to the Zika signal your colleague caught in Costa Rica before the endemic map had even caught up. In an era of increasing global mobility and porous borders for pathogens, that kind of granular, physician-confirmed intelligence is exactly what keeps a local anomaly from becoming a global crisis.
Looking Ahead to Part 2: In the next installment of this conversation, Dr. Hamer will take us deeper into GeoSentinel's evolving frontlines, including the network's distinctive role in detecting and monitoring emerging arboviral threats, its contributions to tracking antimicrobial resistance across borders, the growing use of AI-supported tools in outbreak detection, and his outlook on where epidemiological surveillance is headed next.
About the Author:
Tom Mahoney, a 2024 Senior Fellow at the Harvard Advanced Leadership Initiative, is focused on global venture philanthropy initiatives to catalyze investment in development of breakthrough vaccines, therapeutics and diagnostics for infectious diseases. A career investment banker, technology entrepreneur and asset management senior executive, Tom is an Associate in Immunology and Infectious Diseases at the Harvard T.H. Chan School of Public Health; a member of the Advisory Board of EdJen BioTech, LLC, a developer of novel vaccines, and Virufy, a respiratory disease diagnostics platform; a Founding Sponsor of the Harvard Alumni Entrepreneurs Accelerator; and a member of the Venture Board of the Harvard HealthLab Accelerators, the Massachusetts Consortium on Pathogen Readiness, and the Council on Foreign Relations.
This Q&A has been edited for length and clarity.