For issue 56:3’s Editor’s Choice, Associate Editor Silke Bauer explains why Wolf et al.’s model for syndromic surveillance presents an important first step in supporting park managers to better understand and manage wildlife diseases.

The selected Editor’s Choice article is Optimizing syndromic health surveillance in free-ranging great apes: The case of Gombe National Park by Tiffany M. Wolf et al.

The health status of wildlife populations presents increasing concerns, not only to detect outbreaks of infectious diseases, notably those with zoonotic potential, but also for the viability of wildlife populations themselves. One tool for the early detection of outbreaks is syndromic surveillance, where the number or proportion of individuals observed with signs of disease (syndromes) is recorded, and if they exceed some predefined threshold, this is classified as an outbreak.

GM_snot - Christopher Walker
A female chimpanzee (Pan troglodytes schweinfurthii) monitored by the Gombe Ecosystem Health Research team demonstrating signs of respiratory illness characterized by rhinorrhea. Photo by Christopher Walker.

Since syndromic surveillance is a non-invasive monitoring tool, it can be the method of choice for detecting outbreaks of infectious diseases in wild populations. However, how ‘good’ is syndromic surveillance in detecting an outbreak while avoiding false alarms? Or more specifically, what is its sensitivity – the probability that an outbreak is detected – and its specificity –its resistance to false alarms? And how do sensitivity and specificity vary with other factors such as outbreak size, season or behavioural changes?

As evaluating syndromic surveillance in the field will be challenging at least and often hardly possible at all, Wolf and colleagues employed a theoretical approach and tested syndromic surveillance with an agent-based epidemiological network model. Their original motivation was to evaluate the existing syndromic surveillance in chimpanzees in Gombe National Park (Tanzania). The chimpanzees’ population has declined over the past decades and disease appears to have contributed to this decline as more than 60% of known deaths were associated with clinical illness.

Wolf et al.’s model simulates the epidemiology of respiratory disease in the social network of chimpanzees. Individuals in the model can transition between epidemiological states – from susceptible -> exposed -> infectious -> recovered. Particularly the step from ‘susceptible’ to ‘exposed’ is highly dependent on contacts between infected and susceptible individuals, and thus, determined by social structure and fission-fusion dynamics within a group. Wolf et al. included the heterogeneity of contacts with a social network model, also considering changes in the network’s structure over the seasons.

On top of epidemiology and social network comes the surveillance – the model simulates whether an individual was observed given the current surveillance rates and, if it was observed, whether it was in state ‘infectious’. Surveillance rates were varied in virtual experiments to test whether increased surveillance effort would improve surveillance sensitivity for a variety of outbreak sizes.

Collecting data - Tiffany Wolf
Researchers in Gombe National Park collecting health data on the presence/absence of clinical signs of disease from individual chimpanzees as part of the Gombe Ecosystem Health syndromic surveillance system. Photo by Tiffany Wolf.

The results provided several important insights: generally, using weekly count thresholds (i.e. absolute numbers for infected individuals observed) for outbreak detection worked better than a prevalence threshold (i.e. their relative numbers). Furthermore, larger outbreaks were more likely to be detected than smaller ones, and outbreak size varied markedly between seasons. Particularly during the dry season, outbreaks were smaller and thus, less likely to be detected. Additionally, with the scarcer food in the dry season, chimpanzees change behaviour and their group’s social structure, both of which make it even more difficult to detect outbreaks.

Perhaps not overly surprising, intensifying surveillance effort increased surveillance sensitivity from 53% to >70% (depending on outbreak size), but this applied only to the weekly count threshold and not to the prevalence threshold. Thus, one important and easy-to-implement management recommendation for a better detection of outbreaks would be to increase surveillance effort. Even if it would be logistically (or resource-wise) difficult to increase overall surveillance effort, surveillance could be targeted to periods with higher probabilities of outbreaks or to high-risk individuals.

Wolf and colleagues have developed a highly useful evaluation instrument for syndromic surveillance in a system with seasonal differences in epidemiology that make disease detection challenging. Although certainly only a first step in applying the model, their results can help park managers to better understand the dynamics of diseases in seasonally varying social networks and to target and focus surveillance effort for more effective disease detection.

The full article, Optimizing syndromic health surveillance in free-ranging great apes: The case of Gombe National Park by Wolf et al. is free to read as the Editor’s Choice for issue 56:3 of Journal of Applied Ecology.