How to Conduct Retrospective Studies in the Absence of Confirmatory Diagnostics: An Example from a Study of Feline Herpesvirus (FHV) in Cheetahs (Acinonyx jubatus)
Abstract
A common challenge in conducting retrospective epidemiologic studies is incomplete confirmatory diagnostic information to aid in the classification of animal disease status. If cases are limited to those with confirmed diagnostics alone, many true positives would be missed.1 Similarly, inclusion of all individuals with clinical signs but lacking confirmatory diagnostics may result in significant misclassification error.1 Results could also be biased if reasons for diagnostic confirmation differ across confirmed and non-confirmed animals.1 To address the potential for such biases, systematic quantitative methods for identifying clinically compatible (CC) individuals should be used. A population-level study on the epidemiology of feline herpesvirus (FHV) in 322 cheetahs housed in six zoos used a combination of scholarly literature, expert opinion, and exploratory multiple correspondence analysis2 to determine the distribution of clinical signs among 35 laboratory-confirmed (LC) cases of FHV. A final case definition for clinical FHV was then developed, ensuring that the distribution and grouping of signs identified in the LC cheetahs were mirrored in the 61 identified CC cases. The inclusion of both LC and CC cases created a sensitive case definition that is effective for both disease surveillance and developing lists of diagnostic differentials.1 This study not only highlights the importance of confirmatory diagnostics, which are often lacking in routine case investigations, but also demonstrates methodology that can be used to address diagnostic deficiency in retrospective studies. Although limitations exist, such methods should help improve accuracy when developing case definitions based on non-specific clinical signs or unknown syndromes.
Acknowledgments
The authors thank the Morris Animal Foundation for funding this study, and all participating institutions and personnel for the contribution of data. Specifically, we thank Dr. Michael Barrie and Columbus Zoo and Aquarium, Dr. Holly Haefele and Fossil Rim Wildlife Center, Dr. Suzan Murray and Smithsonian National Zoological Park, Dr. Randy Junge and St. Louis Zoo, and Cyd Shields Teare and White Oak Conservation Center. Additional thanks goes to Dr. Victoria Fields for her assistance with data abstraction.
Literature Cited
1. Rothman KJ, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia, PA. Lippincott Williams & Wilkins. 1998.
2. Sourial N, et al. Correspondence analysis is a useful tool to uncover the relationships among categorical variables. J Clin Epidemiol. 2010;63:638–646.