Oral Presentation 64th International Conference of the Wildlife Disease Association 2015

Agent-based modeling of respiratory disease transmission in chimpanzees to quantitatively evaluate the performance of syndromic surveillance  (#49)

Tiffany M Wolf 1 2 , Wenchun A Wang 3 , Dominic Travis 1 , Elizabeth Lonsdorf 4 , Thomas Gillespie 5 , Iddi Lipende 6 , Karen Terio 7 , Anne Pusey 8 , Beatrice Hahn 9 , Ian Gilby 10 , Carson Murray 11 , Randall Singer 1
  1. University of Minnesota, St. Paul, MN, USA
  2. Minnesota Zoo, Apple Valley, MN, USA
  3. University of Saskatchewan, Saskatoon, Saskatchewan, Canada
  4. Franklin and Marshall College, Lancaster, PA, USA
  5. Emory University, Atlanta, GA, USA
  6. Gombe Stream Research Center, Kigoma, Tanzania
  7. University of Illinois Zoo Pathology Program, Chicago, IL, USA
  8. Duke University, Durham, NC, USA
  9. University of Pennsylvania, Philadelphia, PA, USA
  10. Arizona State University, Tempe, AZ, USA
  11. George Washington University, Washington D.C., USA

Syndromic surveillance, or surveillance of general disease “types”, to study disease trends in wildlife is novel.  Accordingly, associated methodologies for outbreak detection and assessment of surveillance system performance are highly needed. Since 2004, syndromic surveillance has been employed in Gombe National Park, Tanzania, to collect data on several major disease syndromes affecting free-living chimpanzees. Our team utilized 9 years of syndromic data for a qualitative assessment of system performance and development of algorithms for respiratory disease outbreak detection. Here we describe the continuation of that effort with a quantitative assessment of surveillance sensitivity, or the probability that a respiratory disease outbreak is detected by the system in place. To do this, empirical data on community demographics, social contacts and frequencies of observation by surveillance were integrated with an agent-based, network disease model to simulate surveillance of respiratory outbreaks previously observed. Out of 2000 Monte Carlo simulations of disease introduction, 1064 outbreaks were produced, with mean duration of 5 weeks and mean cumulative incidence of 36 cases in a community of 60 individuals. Two algorithms of weekly outbreak detection were examined, one producing an outbreak signal when 2 or more cases were detected, and the other signaling when case prevalence exceeded 15.6% of those observed. Surveillance sensitivity was estimated as 66% (95% Confidence Interval: 63.1, 68.8%) and 59.5% (95% Confidence Interval: 56.5%, 62.4%) for weekly count and prevalence thresholds, respectively. In addition to differences between detection algorithms, differences were also observed in surveillance sensitivity between quarters of the year. Overall, disease model simulations revealed important temporal differences in outbreak characteristics, which are likely impacting surveillance system performance. Through this work, we were able to identify the best algorithm for respiratory outbreak detection and key strategies to improve syndromic surveillance performance in a free-living chimpanzee population.