Optimizing surveillance for early disease detection: Expert guidance for Ostreid herpesvirus surveillance design and system sensitivity calculation

https://doi.org/10.1016/j.prevetmed.2021.105419Get rights and content

Highlights

  • Experts identified surveillance traits important for OsHV-1 early detection.

  • Observation frequency, guidance, incentives, advocates, and risk-based focus were key.

  • Strong industry-government partnership appears crucial to effective implementation.

  • Methods introduce a simple approach to observational surveillance sensitivity estimation.

  • Results hold value for both early detection design and disease freedom assessment.

Abstract

To keep pace with rising opportunities for disease emergence and spread, surveillance in aquaculture must enable the early detection of both known and new pathogens. Conventional surveillance systems (designed to provide proof of disease freedom) may not support detection outside of periodic sampling windows, leaving substantial blind spots to pathogens that emerge in other times and places. To address this problem, we organized an expert panel to envision optimal systems for early disease detection, focusing on Ostreid herpesvirus 1 (OsHV-1), a pathogen of panzootic consequence to oyster industries. The panel followed an integrative group process to identify and weight surveillance system traits perceived as critical to the early detection of OsHV-1. Results offer a road map with fourteen factors to consider when building surveillance systems geared to early detection; factor weights can be used by planners and analysts to compare the relative value of different designs or enhancements. The results were also used to build a simple, but replicable, model estimating the system sensitivity (SSe) of observational surveillance and, in turn, the confidence in disease freedom that negative reporting can provide. Findings suggest that optimally designed observational systems can contribute substantially to both early detection and disease freedom confidence. In contrast, active surveillance as a singular system is likely insufficient for early detection. The strongest systems combined active with observational surveillance and engaged joint industry and government involvement: results suggest that effective partnerships can generate highly sensitive systems, whereas ineffective partnerships may seriously erode early detection capability. Given the costs of routine testing, and the value (via averted losses) of early detection, we conclude that observational surveillance is an important and potentially very effective tool for health management and disease prevention on oyster farms, but one that demands careful planning and participation. This evaluation centered on OsHV-1 detection in farmed oyster populations. However, many of the features likely generalize to other pathogens and settings, with the important caveat that the pathogens need to manifest via morbidity or mortality events in the species, life stages and environments under observation.

Introduction

Variants of the Ostreid herpesvirus 1 (OsHV-1), including the reference strain and microvariants, are best known as causal agents of mortality events of consequence for Crassostrea gigas (Pacific oyster) populations in many locations globally since the early 1990s (EFSA AHAW Panel, 2015; Castinel et al., 2015; Arzul et al., 2017; Burge et al., 2018; OIE, 2019). Microvariant genotypes of OsHV-1 were first identified in 2008 associated with large-scale C. gigas mortality in France (Segarra et al., 2010), and are distinguished from the reference genotype (Davison et al., 2005) by sequence variation in select regions of the genome (OIE, 2019). OsHV-1 variants have also been associated with events of lesser impact (Burge et al., 2006; Cáceres-Martínez and Vásquez-Yeomans, 2013; Bai et al., 2015). Though typically linked to large-scale losses in C. gigas, microvariants have been found in apparently healthy C. gigas, both during post-outbreak recovery (e.g., in Australia, Evans et al., 2017) and alongside mortality in other species (e.g., in Asia, Bai et al., 2015). Failure to apply efficient detection systems, as well as movement controls, contributed to dissemination throughout many affected regions (Carnegie, 2012; Carnegie et al., 2016; Fuhrmann et al., 2019). As such, early detection and timely response are critical to the goal of curtailing damage otherwise associated with rapid spread. However, the pathogen’s short incubation period and rapid transition from extremely low to high prevalence, alongside a tendency for spatiotemporal clustering of infection (Paul-Pont et al., 2013), make OsHV-1 difficult (and costly) to detect prior to a large outbreak using conventional surveillance modalities (Whittington et al., 2019).

Pathogen surveillance in aquaculture is largely built around intermittent submission of samples for laboratory testing (termed “active surveillance” when coordinated externally). While active surveillance can offer very strong evidence of disease absence at test-time, it is less well-suited for early detection of incursions that can occur at any time or location (Vennerström et al., 2017). Often designed to support translocation of stock for production or trade, active surveillance can be patchy in both space and time. Though bolstering active systems (via increased frequency, for example) is one way to improve early detection capacity, it may not be the most efficient. Rather, surveillance built to address both disease freedom and early detection objectives may need to couple periodic testing with observational systems for the most cost-effective designs (Cameron et al., 2020). Risk-based surveillance, targeting animals with the highest risk of exposure, infection, or consequence, may further improve detection capability, for both active and passive systems (Oidtmann et al., 2013).

Frequent observation of animal appearance, behavior, or mortality (here termed “observational surveillance”, also known as “passive surveillance” if led by the producer) can signal abnormalities in population health. Mortality investigations, for example, were key to the initial diagnosis of OsHV-1 μvar in France (Lupo et al., 2014). When observers are versed in disease recognition and networked with health professionals and response agencies, unexplained morbidity or mortality can trigger disease investigations, targeted laboratory diagnostics, and subsequent reporting. Similarly, the absence of reports from these same networks can provide baseline assurance of disease absence in the monitored systems if trust and mutual confidence are present between stakeholders (World Bank, 2010). However, the degree of assurance generated will vary by pathogen, population, and system, and its quantification requires an estimate of the sensitivity of observational surveillance.

Surveillance system sensitivity (SSe), the ability of a system to identify disease or infection in a population, is similar in construct to diagnostic test sensitivity (Se) but focused on populations rather than individual animals. Used in disease freedom calculations, SSe can influence decisions regarding disease control, animal movement, and trade. However, SSe for observational surveillance is notoriously difficult to estimate well, and SSe geared to early (rather than any) disease detection is rarely discussed (Cameron et al., 2020). Several methods are available, e.g., using capture-recapture (Lupo et al., 2012) or mixed qualitative/quantitative (Limon et al., 2014) analyses, to identify traits predicting observational surveillance success. However, data, time or resources required for empirical analyses of observational surveillance systems are often lacking. Therefore, SSe is more commonly derived from estimation of a series of subjective probabilities that are not necessarily data-driven: probabilities that the pathogen will cause clinical signs, the clinical signs will be noticed by a producer, the producer will report the disease, and so on. Expert elicitations, in contrast, offer a replicable mechanism to estimate epidemiologic parameters like SSe when empirical field data are limited (Gustafson et al., 2018). If structured to identify system traits that predict detection capacity, results can be used as guidance for surveillance design. Further, by replicating case-control studies and generating likelihood ratios (LRs, epidemiologic measures of association), the results are portable, informing models that compare the value of different systems or designs.

Expert elicitation accuracy is driven by several processes including whether experts have first-hand knowledge of the topic, whether experts have an opportunity to clarify the wording/meaning of identified factors, and whether elicited parameters are count rather than probability-based (Kynn, 2008; Burgman et al., 2011; McBride et al., 2012). Certain types of group processes, e.g., the integrative group process (IGP) used here, also minimize “group think” (or unjustified peer influence) by eliciting responses from individuals independently prior to group discussion (Gustafson et al., 2003). Finally, group discussion provides a forum for self-calibration as experts with specific knowledge have a chance to share experiences that may, or may not, influence independent revisions by the rest of the group.

Here we describe an expert elicitation to identify and weight factors predicting the capacity of observational surveillance for the early detection of OsHV-1. The expert-identified traits are considered key to early detection and thereby offer guidance to producers and governments aiming to improve surveillance. Results also parameterize a model to compare (1) the sensitivity of different surveillance system designs toward early detection, and (2) the confidence these systems can generate toward disease freedom. For simplicity, we focused this elicitation on the pathogen OsHV-1, regardless of genotype. However, many of the tenets, as well as the estimation process, likely generalize to other pathogens of concern for oyster production systems. Similarly, factors focus on oyster farming sites and their surveillance systems, but the resulting model should also generalize to geographic regions.

Section snippets

Terms and assumptions

We use the term “factor” throughout to describe possible predictors (risk factors or protective factors) of a surveillance system’s early detection capability, and the term “trait” to classify response or sub-categories of the identified factors. Factor describes a general concept (e.g., color) and trait describes its sub-category (e.g., blue, green, orange, etc.). We use the term “detection” to refer to the full sequence of observing, notifying, diagnosing, and reporting a case to the proper

Panelist participation

The twelve invited panelists included 3 industry representatives from the United States, and 9 shellfish health professionals from the U.S. (5), Mexico (1), Australia (1) and France (2). Two of the panelists (one from the U.S. and one from Mexico) invited an additional expert each to provide shared input, bringing the total number of experts involved in discussion and review to fourteen. The twelve core panelists completed the interview and ranking steps, all twelve completed the weighting

Discussion

As aquaculture evolves to meet global food security challenges, the resulting intensification and diversification of production systems may heighten the emergence and international spread of aquatic pathogens (Burge et al., 2014; Feist et al., 2019; King et al., 2019). Aquaculture must balance the pathogen risks inherent to natural settings with the costs and complexities of altered rearing environments that support their bioexclusion (Lafferty et al., 2015). Similarly, surveillance should

Declaration of Competing Interest

There are no conflicts of interest to declare for this article (“Optimizing surveillance for early disease detection: Expert guidance for Ostreid herpesvirus surveillance design and system sensitivity calculation”) focused on optimizing surveillance for the early detection of an oyster pathogen.

Acknowledgements

We would like to acknowledge the expertise, time, and energy of panelists (K. Humphrey, T. Sawyer) that contributed greatly to the content of the study but elected not to participate as co-authors in the presentation of results. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. This is contribution 4031 of the Virginia Institute of Marine Science, William & Mary.

References (51)

  • P.A. Martin et al.

    Demonstrating freedom from disease using multiple complex data sources 1: a new methodology based on scenario trees

    Prev Vet Med.

    (2007)
  • B. Oidtmann et al.

    Risk-based methods for fish and terrestrial animal disease surveillance

    Prev. Vet. Med.

    (2013)
  • M. Rana et al.

    Investigating data-driven approaches to understand the interaction between water quality and physiological response of sentinel oysters in natural environment

    Comput. Electr. Agric.

    (2020)
  • A. Segarra et al.

    Detection and description of a particular Ostreid herpesvirus 1 genotype associated with massive mortality outbreaks of Pacific oysters, Crassostrea gigas, in France in 2008

    Virus Res.

    (2010)
  • S.C. Ugalde et al.

    Analysis of farm management strategies following herpesvirus (OsHV-1) disease outbreaks in Pacific oysters in Tasmania, Australia

    Aquaculture

    (2018)
  • R.J. Whittington et al.

    Long-term temporal and spatial patterns of Ostreid herpesvirus 1 (OsHV-1) infection and mortality in sentinel Pacific oyster spat (Crassostrea gigas) inform farm management

    Aquaculture

    (2019)
  • E.A. Aalto et al.

    Models with environmental drivers offer a plausible mechanism for the rapid spread of infectious disease outbreaks in marine organisms

    Sci. Rep.

    (2020)
  • C.A. Burge et al.

    Mortality and herpesvirus infections of the Pacific oyster Crassostrea gigas in TB, California, USA

    Dis. Aquat. Org.

    (2006)
  • C.A. Burge et al.

    Climate change influences on marine infectious disease: implications for management and society

    Ann. Rev. Marine Sci.

    (2014)
  • C.A. Burge et al.

    Ecology of emerging infectious diseases of invertebrates

  • M.A. Burgman et al.

    Expert status and performance

    PLoS One

    (2011)
  • J. Cáceres-Martínez et al.

    Diseases, parasites and mortality episodes of commercially important oysters in Mexico and their production implications

    Ciencia Pesquera

    (2013)
  • A.R. Cameron et al.

    Quantification of the sensitivity of early detection surveillance

    Transbound. Emerg. Dis.

    (2020)
  • R.B. Carnegie

    Contemporary issues in molluscan health: challenges and opportunities

    Proceedings of the OIE Global Conference on Aquatic Animal Health Programmes: Their Benefits for Global Food Security (Panama, June 2011)

    (2012)
  • R.B. Carnegie et al.

    Managing marine diseases in the context of regional and international commerce: policy issues and emerging concerns

    Philos. Trans. Biol. Sci.

    (2016)
  • Cited by (0)

    View full text