Quantifying and Addressing Bias Associated with Imperfect Observation Processes in Epidemiological Studies

Most epidemiological studies aim to describe, analyse and explain disease data which have usually been generated by imperfect observation (or detection) processes. Indeed, disease surveillance systems often capture only a fraction of the true epidemiological situation. Also, gold-standard diagnostic tests are rare, meaning that false positive and false negative results may be a problem. Further complications arise when dealing with elusive populations (for example hard-to-reach groups of people, or free-ranging wild animals) because many aspects of such populations are never observed. Finally, for social and economic reasons, people such as farmers may be reluctant to report suspicions of, or seek diagnosis of, particular diseases. These imperfect observation processes represent sources of bias which may have profound impacts on our understanding (and therefore management) of important diseases in animal and human populations.

Historically, these imperfect observation processes were considered as nuisances, the consequences of which were at best discussed, and at worst ignored. In recent years, methodological developments such as the application of capture-recapture techniques, and the dawning realisation of the potential value of interdisciplinary approaches, are beginning to allow us to account for and sometimes explain these observation biases. Despite these significant advances, we believe that effort is still required in order to persuade epidemiologists that imperfect observation processes should be accounted for in epidemiological studies, or at the very least discussed.

The aim of this Research Topic is to bring together and showcase research expertise, ideally from multiple disciplines, to present the latest developments in quantifying and addressing imperfect observation processes. For example, we welcome original manuscripts that describe:
- The consequences of not accounting for imperfect observation processes when making epidemiological inference;
- The quantitative estimation of diagnostic test performance in the absence of a reliable reference test;
- The use of mathematical models for addressing imperfect detection processes when modelling infectious disease dynamics;
- The evaluation of the effectiveness of surveillance systems for the detection of diseases or other entities;
- The transmission of pathogens in elusive populations, such as free-ranging wildlife or hard-to-reach human populations;
- The beliefs, attitudes and practices affecting the likelihood of disease presence (or the suspicion thereof) being reported by people such as farmers, animal owners and the public. This category could also include the factors affecting individual people from seeking diagnosis or treatment for disease.

Note that we strongly encourage the submission of manuscripts from a wide range of fields (including veterinary and medical epidemiology, mathematical biology, ecology, engineering and related areas). Manuscripts that utilize systems approaches are also greatly encouraged and welcome. We hope that this will help identify techniques that are used in one field that have potential to be exploited to address pressing disease challenges in another field. All submissions should illustrate an innovative methodology (or application of existing methods to a novel field), or they should compare new/alternative methodologies for addressing imperfect observation issues with traditional approaches to illustrate potential advantages, to advance our collective understanding of the potential impact of imperfect observation processes in epidemiological studies.