General occupancy model for systemic diseases

Zoonotic diseases have received attention in recent decades due to the emergence of pathogens resulting in epidemics and pandemics with substantial implications for public health, global economy, and agriculture industry. About 75% of emerging infectious diseases are zoonotic in origin, and wild animals are often their primary reservoirs. Examples of such pathogens include those causing tuberculosis and brucellosis, with implications for agriculture industry; and Ebola virus and SARS-CoV-2 causing disease outbreaks with grave economic and public health implications. Zoonotic and other wildlife diseases also have substantial conservation implications because they have been implicated as the cause of population decline of many wildlife species, including the African lion Panthera leo (attributed to canine distemper virus) and the African wild dog Lycaon pictus (caused by rabies). Thus, quantitatively rigorous surveillance of wildlife diseases and knowledge of factors affecting their emergence/reemergence is important, and sometimes mandated by public health, agricultural or conservation authorities.

Disease etiology is often poorly understood for many pathogens, yet it can have consequences for estimating prevalence. For example, many pathogens infect their hosts through multiple routes, and they can cause systemic diseases where pathogens infect, and are shed through, multiple tissues (hereafter, multi-tissue diseases) with little or no understanding of the sequence of infection. An animal can be in one of several possible states of infection depending on the type or number of tissues affected. Furthermore, infection of one tissue by the pathogen might depend on the infection status of other tissues. There exist no clear guidelines regarding definition of infection for systemic diseases or how many tissues are to be sampled; the choice of tissue(s) to be sampled can strongly influence pathogen detection and can lead to underestimation of prevalence if the pathogen is present in tissues other than the one being tested. Furthermore, diagnostic methods used may fail to detect the pathogen even when it is present. A rigorous disease surveillance program must attempt to address these sources of potential bias.

We consider a situation where s = 1, 2, …, S tissues within a host may be infected by a pathogen. At any given time, the pathogen may be present in a tissue (true disease state = 1) or not (true disease state = 0). In our most general model, each potential set of infected tissues in an organism is defined as that organism’s disease state. Thus, for S tissues, 2S states are possible. For example, an animal can be infected in all four tissues (oral, blood, nasal and genital; S = 4; true disease state = 1111), or none are infected (true disease state = 0000), or a combination of ≥2 tissues are infected; this scenario leads to a 24 = 16 possible true disease states (Table 1). Although the presence of viral DNA is indicative of current or recent past infection, we assume that a sample that tested positive for viral DNA contains the virus

In a collaberative project with disease ecologists and statisticians, I have developed general occupancy model which accounts for imperfect detection. I have implemented this model on Pseudorabies virus. Code to implement general occupancy model for systemic diseases