Operationalizing metagenomic data from environmental surveillance for one health decision-making in live animal markets: Findings from a multisectoral workshop in Cambodia

Abstract

Background

Live animal markets (LAMs) are recognized as hotspots for zoonotic disease emergence. Environmental surveillance (ES), particularly when paired with metagenomic sequencing, offers an advanced and actionable approach to pathogen detection in high-risk settings. However, the complexity of metagenomic data and the lack of user-friendly communication tools hinder its integration into routine public health decision-making.

Methods

We conducted an exploratory qualitative participatory workshop study with descriptive analysis. A three-day multisectoral workshop was held in Phnom Penh, Cambodia, in May 2024, bringing together stakeholders from health, agriculture, and environment sectors to explore how metagenomic ES data can be visualized, understood, and applied. Through simulation exercises, surveys, and interviews, the workshop evaluated user preferences for data formats, thresholds for action, and decision-making strategies.

Findings

In total, 52 participants attended the workshop and ten completed semi-structured interviews. Participants discussed their preferred familiar visualizations (bar, pie, and line charts) and intuitive color-coded thresholds (e.g., traffic-light schemes). While digital dashboards were welcomed, analog, printer-friendly formats remained essential due to infrastructure constraints. Key barriers to ES integration included limited bioinformatics capacity, lack of inter-ministerial coordination, and minimal ES prioritization at the provincial level.

Interpretation

Metagenomic ES data can inform public health actions when visualization tools are tailored to end-user needs and embedded in multisectoral governance. This exploratory participatory workshop generated preliminary stakeholder-informed insights and an initial draft roadmap for future implementation planning in Cambodia. Further expert-led and funded work is needed to validate visualization tools, pathogen-specific thresholds, escalation pathways, and operational use under real-world surveillance conditions.

1. Introduction

Environmental surveillance (ES) refers to the systematic collection and analysis of environmental samples, such as from surfaces, air, water, or animal waste to detect the presence of pathogens. Unlike clinical surveillance, which depends on symptomatic cases, ES can identify pathogens preemptively, offering early warnings of outbreak risks [1]. This approach is especially relevant in live animal markets (LAMs), where humans, domestic animals, and wildlife converge in close contact. Such environments create ideal conditions for zoonotic spillover, particularly affecting market workers and consumers with direct exposure to animals [2], [3].

ES has proven useful for monitoring pathogen circulation in high-risk environments such as LAMs. In Bangladesh, for example, a weekly ES program conducted from 2018 to 2020 collected fecal and offal swabs from live bird markets to monitor avian influenza viruses (AIVs), particularly H5 and H9 subtypes. The results revealed year-round AIV circulation, with 37.8% of samples testing positive and frequent co-detection of H5 and H9 strains. High-risk zones within markets, especially slaughtering areas, showed elevated detection rates, demonstrating how ES data paired with spatial analysis can guide targeted interventions [4]. In Cambodia, another study using aerosol samplers during periods of high and low AIV circulation demonstrated that AIV RNA was detected in 100% of air samples collected during the high circulation period, with A/H5N1 and A/H9N2 viruses isolated from 50% of air samples. These findings underscore the risk of airborne transmission to market workers and reinforce the utility of ES for early detection and planning [5].

ES offers a low-cost, non-invasive approach for monitoring pathogen circulation in LAMs and enables preventative or corrective action during the early stages of disease emergence [6]. Such systems can significantly strengthen early warning capacity and reduce the risk of zoonotic outbreaks. Similar surveillance strategies have been successfully applied to detect antimicrobial resistance, poliovirus, and most recently, SARS-CoV-2, demonstrating their effectiveness in diverse public health contexts [7], [8], [9]. When paired with technologies like metagenomic sequencing, ES enhances our ability to detect a broad spectrum of pathogens and respond swiftly, ultimately improving outbreak preparedness and public health outcomes [10], [11], [12].

Metagenomics enables the detection of a wide range of pathogens, including those unknown or unculturable, by sequencing all genetic material present in an environmental sample [13]. This untargeted sequencing approach provides a comprehensive snapshot of microbial communities, including viruses, bacteria, fungi, and parasites [14]. In high-risk settings like LAMs, where pathogens often exist in complex mixtures and may be difficult to isolate, metagenomics offers a powerful alternative to traditional diagnostics. By combining high-throughput sequencing with bioinformatics, microbial signatures can be identified, classified, and quantified, making this approach a critical tool for early-warning systems and disease surveillance [15].

While integrating metagenomics into ES offers significant benefits, it also generates large and complex datasets that require specialized bioinformatics tools, trained personnel, and reliable computational infrastructure [16]. These demands pose challenges for routine public health use, especially in resource-limited settings. Nevertheless, when effectively analyzed, metagenomic data can guide the design, implementation, and evaluation of targeted interventions, making it a powerful asset for early detection and response.

We hypothesized that tailoring data visualizations to stakeholder preferences and integrating metagenomic with syndromic surveillance would enhance the interpretability and actionability of ES data in LMIC settings. To explore this, we convened a multisectoral workshop in May 2024 in Phnom Penh, Cambodia, focused on the challenges of interpreting and applying metagenomic ES data. The event brought together stakeholders from key ministries, including health, agriculture, and environment, to examine how metagenomic ES outputs could be better communicated and used in One Health (OH) decision-making. The workshop assessed baseline understanding, captured preferences for data formats, evaluated communication tools, and considered the value of syndromic surveillance in complementing ES. This study draws on those outcomes to identify practical enablers and barriers to integration of ES into national surveillance systems and explores how visual and syndromic tools can support early risk assessment and zoonotic outbreak preparedness in LAMs and other high-risk environments. This is a novel methodology to systematically explore how metagenomic data from environmental surveillance can be translated into stakeholder-friendly formats and linked to decision-making pathways in LAMs. While environmental surveillance and metagenomics have been previously applied to pathogen detection, their integration into participatory, multisectoral exercises for communication and action planning remains largely unexplored. This article describes how Cambodia convened stakeholders to engage with complex surveillance data and co-develop approaches for improving OH coordination in LAMs settings.

2. Methods

2.1. Pre-planning workshop

The team conducted an exploratory participatory workshop study with descriptive analysis centered on a three-day multisectoral consultative workshop (Phnom Penh, Cambodia; 15–17 May 2024) to examine how stakeholders interpret and apply metagenomic ES data within a One Health framework, with a focus on visualization and decision-making. The manuscript is reported in accordance with the Standards for Reporting Qualitative Research (SRQR); the completed checklist is provided in the Supplementary Material 1: Table S1. The programme and materials (simulation exercises, interview guides, and facilitation tools) were co-developed over six months by a joint team from public health, veterinary, and environmental backgrounds within the research group, and were informed by insights from a prior national One Health Zoonotic Disease Prioritization (OHZDP) workshop in Cambodia that guided the selection of priority pathogens and market-related risk scenarios [17]. To standardise scenarios and reduce confidentiality and anchoring biases, all exercises used fictional datasets mirroring plausible ES signals. The fictional datasets used in the exercises were developed to reflect plausible pathogen patterns and market conditions based on prior environmental surveillance experience from Institut Pasteur du Cambodge (IPC), prior publications [5], [18], [19], [20], [21], [22] and expert input from the study team. Fictional datasets were used to standardise scenarios, protect confidentiality, and reduce anchoring bias.

2.2. Participants

Participants were nominated by the Ministry of Health (MoH), Ministry of Agriculture, Forestry and Fisheries (MAFF), and Ministry of Environment (MoE) based on their technical expertise in surveillance, laboratory diagnostics, and data interpretation. Ministries were invited to nominate two or more representatives with relevant expertise, but the final number of nominees depended on each institution's availability and internal selection process. Because invitations were issued at the institutional level and representatives were nominated internally, there was no fixed individual-level invited denominator. Selection aimed to ensure broad representation across human, animal, and environmental health sectors, in line with a One Health approach. They were organized into five cross-sectoral working groups, each facilitated by a trained moderator, note-taker, and rapporteur to support consistent engagement and data capture.

2.3. Data source and collection

Data sources included pre-workshop questionnaire responses, Mentimeter polling, facilitated group discussions, rapporteur summaries, note-taker records, group exercise outputs, post-workshop questionnaire responses, workshop evaluation forms, and semi-structured interviews. Mentimeter polling was used during selected exercises to capture individual responses in real time, followed by facilitated group discussion. Response denominators varied across Mentimeter and survey items because not all participants were present for, or responded to, every activity. The full workshop agenda, Mentimeter questions, pre- and post-workshop questionnaires with descriptive results, semi-structured interview guide, and workshop evaluation form are provided in Supplementary Material 2.

2.4. Workshop process

The workshop was held over three consecutive half-days, and was designed to progressively assess participants' understanding, interpretation, and application of metagenomic ES data within a One Health framework (Fig. 1). Five structured sessions combined presentations, simulation exercises, and group activities to gather both quantitative and qualitative data on participant preferences, learning outcomes, and multisectoral coordination strategies. Fictional datasets were used in all exercises to standardise scenarios and minimize bias. Participants were intentionally organized into five cross-sectoral groups and used tools such as Tableau and Mentimeter to engage with data in real-time. To minimize sectoral and hierarchical imbalances, particularly between ministries such as MoH, MAFF, and MoE, participants were intentionally distributed into mixed cross-sectoral working groups, with senior and junior staff separated to promote open dialogue. International partners were assigned across groups to prevent disproportionate influence from any single institution. Before each simulation, facilitators introduced explicit “rules of engagement” to create a low-stress, no-fault environment: differing viewpoints were expected and welcomed; decisions made during exercises did not represent institutional positions; and the purpose of the activity was to explore multiple options and collectively identify feasible solutions for LAM challenges using environmental surveillance. These safeguards, together with neutral fictional data, helped create conditions where disagreements on prioritization could be voiced, examined, and constructively reconciled. Facilitators from IPC supported all activities to ensure consistency and accurate capture of feedback. Members of the facilitation and analysis team had expertise in ES, metagenomics, public health, veterinary health, and OH implementation.

Fig. 1.

The timeline of the consultative workshop held in Phnom Penh, Cambodia, May 15–17, 2024.

2.5. Session 1: orientation and framing

Participants received an introductory presentation outlining workshop objectives, the concept of ES, and the role of metagenomics in early warning systems. A pre-workshop questionnaire was administered to assess baseline understanding of ES, pathogen detection, and data interpretation across sectors.

2.6. Session 2: introduction to ES and metagenomic data

Technical presentations covered environmental sampling strategies (e.g., air, surfaces, water), the fundamentals of metagenomic sequencing, and examples of ES application in LBMs. Participants discussed the challenges of interpreting complex data outputs and the importance of cross-sectoral communication to support outbreak detection and response.

2.7. Session 3: data visualization and scenario simulations

Participants engaged in hands-on simulation exercises using fictional datasets to evaluate different visualization formats (bar charts, pie charts, line graphs, heatmaps). They explored how data presentation influenced understanding and decision-making and were asked to identify preferred formats and color schemes for risk communication. Individual responses were captured through Mentimeter, followed by group discussion and synthesis.

2.8. Session 4: thresholds, coordination, and interventions

This session explored how ES data thresholds (e.g., rising pathogen levels) could trigger interventions. Participants reviewed different ways to visualize thresholds and discussed how ministries would coordinate actions across sectors. Structured group tasks identified challenges related to coordination, infrastructure, and resources. Participants also proposed realistic mitigation strategies based on pathogen trends.

2.9. Session 5: evaluation and post-workshop assessment

A post-workshop survey measured changes in participants' confidence and understanding of ES and metagenomic data interpretation. Qualitative feedback was collected on the utility of different visualization tools and perceived barriers to ES implementation. This session also explored future capacity-building needs and institutional actions to integrate ES into national OH surveillance systems.

The workshop was conducted over three days. Day 1 focused on basic data visualization exercises using different formats and colors. Day 2 introduced more complex data and visualization features to signify action. Day 3 addressed use cases, key barriers and advantages, and proposed ways forward. The final goal was to translate insights into actionable recommendations and usable outputs.

2.10. Interviews

To complement the group-based exercises, semi-structured interviews were conducted with a subset of workshop participants to explore institutional perspectives on ES and capacity for metagenomic integration. Participants were invited to volunteer through a pre-workshop Microsoft Forms survey, and purposive sampling was used to ensure representation across ministries and organizational levels. Each interview lasted approximately 30–45 min. Ten individuals participated in qualitative interviews, including representatives from the MoH, MAFF, MoE, Duke-NUS Medical School, USAID, and other partner organizations. Interviews were conducted between May 13 and May 17, 2024, using a guided protocol focused on current ES practices, perceived barriers, capacity-building needs, and strategies for multisectoral coordination. Responses were thematically analyzed to triangulate findings from the workshop discussions and highlight context-specific constraints and opportunities.

2.11. Data analysis

Qualitative data were analyzed using a framework thematic approach. An initial coding framework was developed from the workshop objectives and exercise structure, including visualization preferences, interpretation of metagenomic data from ES, thresholds and action triggers, multisectoral coordination, reporting needs, and roadmap requirements. Additional inductive codes were added when recurrent issues emerged during data review. Note-taker records, rapporteur summaries, group outputs, open-ended Mentimeter responses, pre and post-workshop questionnaire responses, and interview notes were consolidated into an analytic matrix. Three researchers (F.E.S., K.S. and S.P.) independently reviewed and manually coded the dataset. Coding differences were discussed and resolved by consensus, and themes were refined by comparing patterns across workshop notes, polling data, evaluation responses, and interviews. Mentimeter and survey data were summarised descriptively and used to contextualize qualitative findings; no statistical analysis was planned or conducted.

2.12. Evaluation

A post-workshop evaluation was conducted using a structured Microsoft Forms questionnaire to assess participant satisfaction, perceived learning outcomes, and the relevance of workshop content and methods. The evaluation questionnaire is provided in Supplementary Material 2, Table S5. The survey included four Likert-scale questions covering content clarity, session structure, facilitation quality, and overall utility. An open-ended section invited suggestions for improving future training and identified additional topics for capacity development. The evaluation results were analyzed to inform interpretation of participant engagement, usability of visualization tools, and priorities for future One Health surveillance initiatives.

2.13. Ethical approvals

Ethical approval for this study was obtained from the National Ethics Committee for Health Research of the National Institute of Public Health, MoH, Cambodia (Reference No. 336 NECHR). Participants received study information and provided oral consent before participating in workshop activities, surveys, and interviews; including audio-recording and attribution preferences where applicable. Participation was voluntary, and participants could decline any question or withdraw from any activity at any time without consequence. Participants were informed that their contributions reflected individual professional perspectives and did not represent formal institutional or ministerial positions. Responses were de-identified, stored securely, and reported in aggregate; direct quotations are used only when consistent with participant consent and when they do not identify participants or institutions.

3. Results

3.1. Participants

A total of 52 participants attended the workshop, representing a diverse cross-section of stakeholders from human, animal, and environmental health sectors (Fig. 2). This included 13% from MoH, 15% from the MAFF, and 17% from the MoE. Additional participants came from international partners, including the Food and Agriculture Organization of the United Nations (FAO), World Health Organization (WHO), United States Agency for International Development (USAID), EcoHealth Alliance, and Duke-NUS Medical School. This multisectoral representation ensured alignment with the OH framework and provided a range of perspectives on environmental surveillance and data interpretation.

Fig. 2.

Proportion of Participants (n = 52) by Sector, May 2024.

3.2. Key findings per section

3.2.1. Exercise 1: data visualization preferences

Participants were presented with fictional metagenomic ES data from LAMs and asked to evaluate four visualization formats (heatmap, longitudinal bar chart, horizontal bar chart, and pie chart) based on clarity, interpretability, and perceived utility (Fig. 3). Mentimeter responses indicated a strong overall preference for longitudinal bar charts (23/41 votes, 56.1%), followed by heatmaps (13/41 votes, 31.7%). Although pie charts were rated as easy or very easy to interpret by 81.3% (26/31) of respondents, they received only 4.9% (n = 2/41) of votes when participants selected their preferred format, suggesting that while familiar and simple, they were considered less informative for decision-making. Participants cited five reasons for their choices: completeness of information, clarity, ease of understanding, visual comparability, and familiarity. In the second part of the exercise, color use in data presentation was explored. A majority preferred primary color schemes, particularly yellow–red gradients for heatmaps, to represent pathogen levels and thresholds. Darker shades were associated with risk; lighter shades with safety. Practical considerations also emerged: participants emphasized black-and-white printability and monochromatic palettes, especially for communication with higher-level decision-makers accustomed to printed reports. Group discussions reinforced that familiarity and simplicity were central to stakeholder preferences. Participants highlighted the need for clear legends, numeric data, and descriptive labeling to support interpretation. While more advanced visualization types (e.g., PCoA, Sankey diagrams, radar plots) were mentioned, these were generally seen as better suited to technical audiences. Participants also suggested that digital dashboards are viewed positively, however, infrastructure and digital-literacy limitations necessitated offline formats, especially in provincial/field settings.

Fig. 3.

Visualization formats used to present pathogen detection across sample types and live animal market areas during the exercise 1. Note: All data shown are fictional. The data presented are entirely fictional and were developed solely for the purpose of the workshop exercise. These simulated datasets were designed to represent environmental surveillance signals from LAMs, allowing participants to evaluate visualization preferences and decision-making approaches during exercise 1. (a) heat map of pathogen levels across sample types in the LAM, (b) longitudinal bar chart of pathogen levels across sample types in the LAM, (c) bar chart horizontal of pathogen levels across sample types in the LBM, and (d) pie chart of pathogen levels across sample types in the LAM.

3.3. Exercise 2: visualization of temporal trends

This exercise evaluated how stakeholders interpret changes in pathogen prevalence over time using different visualization formats. Participants reviewed four graph types—stacked bar charts, line graphs, area charts, and pie charts—based on fictional longitudinal ES metagenomic data (Supplementary Material 1: Fig. S1). Mentimeter polling showed that stacked bar and line graphs were consistently rated as the most effective: 69.7% (23/33)found stacked bar charts easy/very easy, and 63.6% (21/33)did so for line graphs. Area charts were considered the most difficult to understand, with 40.7% (11/27) rating them hard/very hard. In discussions, line graphs were seen as ideal for tracking monthly trends and presentations, while stacked bar charts were preferred for conveying absolute numbers and proportional changes across large datasets. Pie charts were considered more appropriate for brief, high-level summaries for non-technical audiences, whereas area charts were viewed as confusing and better suited for scientific publications. Participants also suggested combining graph types to serve both technical and policy audiences, depending on the decision-making context. Effective temporal visualization was considered useful for outbreak preparedness by highlighting unusual patterns or seasonal spikes, although some cautioned that metagenomic ES may detect many pathogens, underscoring the need for prioritization criteria to avoid misinterpretation or false alarms. Overall, the exercise reinforced tailoring temporal data displays to target audiences and framing action triggers within a OH surveillance system.

3.4. Exercise 3: the influence of seasonal and cultural events on pathogen detection

This exercise examined the relationship between seasonal cultural events and pathogen trends in poultry wash water samples from two LAMs, one urban and one provincial. Participants reviewed year-long fictional surveillance data visualized as heatmaps and bar charts to identify temporal peaks in pathogen abundance and to explore how events like Khmer New Year and the Water Festival may influence risk patterns. Most participants accurately identified elevated levels of avian influenza A/H5N1 in the urban market (Market A) during January–February and April, and higher levels of avian paramyxovirus in the provincial market (Market B) during March and November. These variations were linked by participants to differences in bird sourcing practices and movement across the country. Participants found heatmaps and longitudinal bar charts most useful for visualizing seasonal patterns and prioritizing intervention windows. Area charts were again noted as difficult to interpret, while pie charts were considered more effective for quick summaries but limited for trend analysis. Risk mitigation strategies proposed by participants included targeted disinfection before and after holiday periods, enhanced biosecurity protocols, segregation of symptomatic birds, improved waste disposal, and hygiene education for vendors. Participants emphasized that pathogen trends must be interpreted in relation to market location, seasonal demand, and movement dynamics. This exercise underscored the value of environmental surveillance as a context-aware decision tool within a OH early warning system.

3.5. Exercise 4: threshold interpretation and multisectoral action planning

This exercise assessed how participants interpret surveillance trends and determine when pathogen levels should trigger a coordinated response. Using fictional longitudinal data visualized through line and bar graphs, participants were asked to evaluate whether changes in pathogen detection represented actionable risk signals. They then identified appropriate mitigation strategies and coordination mechanisms across sectors. Group discussions revealed consensus on the importance of clear, standardized thresholds for triggering action. Participants emphasized that such thresholds must be tailored to the local context, visually intuitive, and supported by predefined protocols. Institutional barriers were consistently noted, including unclear response mandates, weak cross-ministerial communication, and challenges in interpreting complex metagenomic outputs. Participants recommended that effective implementation would require not only visualization tools, but also formal guidance on inter-ministerial coordination and escalation pathways. This exercise highlighted the need for clearly defined decision points embedded in OH frameworks to translate metagenomic ES data into real-time action. The summary of the objectives, content, and outputs for each workshop exercise is in Table 1.

Table 1.

Summary of the objectives, content, and outputs for each workshop exercise.

Exercise	Objective	Content	Output
Exercise 1	Determine participant data visualization preferences using mock data from routine surveillance in LBMs	Basic data visualization, including different chart types and color schemes	•Participants consistently showed preference for longitudinal bar charts, while they noted that pie charts were easiest to interpret. •Most participants preferred primary color schemes and a yellow-red gradient for heatmaps. •Dark colors were preferred for indicating risk, while light colors indicated safety. •There was a request for black-and-white printing options due to high color printing costs. •Group discussions highlighted the importance of clear legends, descriptions, and including data and numbers in visualizations.
Exercise 2	Understand temporal variations from routine environmental surveillance data from LBMs over time	Type of graphs interpreting changes in pathogen levels over time	•Participants found stacked bar charts and line graphs to be the most effective for this purpose. They appreciated these graph types for their clarity in showing temporal trends and facilitating comparisons over time. Stacked bar charts were particularly valued for managing large datasets, while line graphs were favored for presenting and comparing monthly variations. •In contrast, participants mentioned that area charts are more complex and harder to interpret. Additionally, they recognize the importance of utilizing metagenomics data from LAMs to enhance decision-making and potentially serve as an early-warning system for future outbreaks.
Exercise 3	Detailed analysis across two markets during Festival periods	Different types of graphs regarding pathogen levels in poultry wash water samples from two markets, one in the province and another in the city, taking into account seasonal and cultural influences	•Market A had higher detections of avian influenza A/H5N1 in January–February and April, whereas Market B had higher detections of avian paramyxovirus in March and November. •Participants noted that pathogen detection varied during events such as Khmer New Year and the Water Festival. •Heatmaps and longitudinal bar charts were identified as effective methods for visualizing the data over time. •Regarding risk mitigation measures, participants recommended regular disinfection, enhanced biosecurity practices, and proper waste management, emphasizing the importance of considering location and timing factors.
Exercise 4	Assessing intervention impact and actions within the ministry	Setting thresholds for detecting high pathogen levels and evaluating intervention effectiveness	•Participants preferred using shaded line charts with danger zones highlighted. •Regarding risk mitigation measures, priorities identified included public education, hygiene practices, public awareness campaigns, proper waste management, and using personal protective equipment. •They stressed the importance of informing health workers in warning zones and coordinating risk mitigation actions in danger zones. •Each Ministry clarified its role in the LAMs: MoH for public health, MoE for environmental sampling and waste management, and MAFF for animal health and veterinary actions. Participants emphasized the need for improved coordination among ministries, suggesting involvement from the Ministry of Commerce and Ministry of Economy and Finance. They proposed establishing a rapid response hotline, creating decision-making guidelines for LAMs, and integrating them into the One Health Technical Working Group. •During discussions on monitoring and evaluation, participants highlighted the importance of continuous surveillance and post-intervention assessments.

LAMs = Live Animal Markets, MoH = Ministry of Health, MoE = Ministry of Environment, MAFF: Ministry of Agriculture, Forestry and Fisheries.

3.6. Post-workshop evaluation of learning outcomes

At the conclusion of the workshop, participants completed a post-session questionnaire to assess changes in understanding, confidence, and perceived utility of environmental surveillance and metagenomic data. Results indicated a marked increase in participants' confidence in interpreting complex data outputs, particularly when supported by appropriate visualization tools. While many acknowledged that metagenomic data remain technically challenging, most reported improved comprehension and recognized the potential value for early warning and risk assessment. The exercise reinforced the importance of tailoring outputs to user expertise and demonstrated that targeted capacity-building, coupled with user-centered communication strategies, can enhance the practical integration of metagenomic ES into national OH surveillance systems. The pre- and post-workshop questionnaire items, response formats, and descriptive results are provided in Supplementary Material 2, Table S3.

3.7. Key informant interviews: feasibility, barriers, and opportunities

Semi-structured interviews with ten key informants, including representatives from the MoH, MAFF, MoE, Duke-NUS, USAID, and international NGOs, were conducted to contextualize workshop findings and assess broader institutional readiness for integrating metagenomic ES. Interviewees consistently cited the need for strong political will and formalized cross-ministerial collaboration as prerequisites for success. Technical capacity and sustained funding were identified as critical barriers, particularly at the subnational level. Respondents also noted that while metagenomic data offers significant promise, current outputs are too complex for routine operational use without substantial simplification and standardization and that multisectoral data sharing is critical for rapid, coordinated outbreak response. The need for clear communication tools, risk thresholds, and decision protocols was emphasized. Several informants stressed the importance of engaging provincial authorities in surveillance efforts to support decentralized early warning systems. Despite systemic challenges, one MAFF interviewee explained, “In the Ministry… they don't know about ES now, but we would need to explain it to them, and they will be interested… We need to do the research on this to see ourselves here in our context,”, highlighting ES and metagenomic sequencing as valuable additions to Cambodia's broader OH surveillance architecture.

3.8. Roadmap

Following the consultative workshop, an initial draft roadmap was developed to organize the priorities, barriers, and implementation requirements identified by participants. This draft roadmap was developed after the workshop and should be interpreted as an initial draft to guide future implementation planning, rather than as a completed, costed, or formally endorsed implementation plan. Detailed thresholds, escalation flows, dashboard templates, ministry ownership, and costing require future expert-led, government-led, and funded work before implementation. The roadmap outlines an initial theory of change in which Cambodia strengthens ES through multisectoral coordination, capacity building, and the contextualized application of metagenomic data. It envisions ES as a cross-cutting component of national health security embedded in existing OH structures, risk communication strategies, and public health response mechanisms. It identifies key action areas such as stakeholder engagement, development of national protocols, visualization and interpretation tools, and policy alignment. More specific elements, such as detailed timelines, designation of responsible ministries, budget and resource needs, and appropriate legal or regulatory instruments, were beyond the scope of this workshop and will require future, dedicated planning processes with national stakeholders. In this sense, the roadmap should be understood as an early strategic framework that can inform, rather than replace, subsequent expert-led implementation planning. While still preliminary, this roadmap represents a starting point for a structured and actionable framework for translating the workshop findings into operational and policy priorities. A draft National One Health Strategy (2025–2029) is still under development in Cambodia, and several themes identified in this workshop, such as strengthened coordination, laboratory capacity, and data integration, are broadly consistent with its proposed directions. Formal alignment will require future multisectoral processes once the Strategy is officially adopted. Its full application will require further refinement through continued engagement with ministries and partners, and alignment with national priorities. Nonetheless, its development marks an important step toward operationalizing the insights generated through participatory exercises. Applying this roadmap in practice can help ensure that ES, particularly when coupled with metagenomics, becomes an actionable and sustainable tool within Cambodia's OH surveillance system.

4. Discussion

ES, particularly when powered by metagenomic sequencing, offers a promising complement to traditional case-based detection systems in LMIC settings. When tailored to user needs, these tools can provide early warnings for zoonotic threats emerging at high-risk interfaces such as LBMs. This workshop demonstrated that stakeholders from multiple sectors can engage with complex metagenomic data, provided it is visualized clearly, grounded in local context, and linked to actionable decision thresholds.

Metagenomic data from ES is increasingly recognized as a complement to case based and syndromic surveillance, offering population level signals that can precede clinical detection and support earlier, targeted responses particularly in resource constrained settings. Recent guidance and reviews emphasize that wastewater and other environmental matrices, when integrated into existing decision cycles, can be a scalable and cost-effective adjunct for multi-pathogen monitoring and preparedness [29], [30].

Across participatory exercises, participants showed strong preferences for simple, familiar visualization formats. In Exercises 1 and 2, stacked bar charts and pie charts were consistently rated as easiest to interpret, with 70% of respondents selecting stacked bar charts as “easy” or “very easy” to understand. Color-coded gradients (e.g., red-yellow-green) were widely supported to convey risk. These findings align with existing literature on the optimal design of risk communication and educational outreach information, which highlights the importance of visual salience and familiarity in driving comprehension [28]. While enthusiasm for digital dashboards was high, participants emphasized that printable, low-tech outputs remain critical, especially in settings with limited internet or printing capabilities [26], [27]. This emphasis on analog-compatible reporting is an important implementation lesson: digital dashboards may support analysis and data review, but surveillance outputs may also remain usable in paper-based, offline, or infrastructure-limited workflows. To address challenges such as red–green color-vision deficiencies and the loss of interpretability in black-and-white printing, we suggest including brief textual descriptions within graphs to support more inclusive interpretation.

Exercise 3 revealed the importance of situational context in interpreting environmental data. Participants correctly identified pathogen peaks around cultural festivals such as Khmer New Year and the Water Festival and attributed them to increased poultry movement and market density factors known to elevate spillover risk. Visual tools like heatmaps and longitudinal bar charts were seen as valuable for highlighting such temporal trends, especially when paired with contextual markers. Participants also proposed pragmatic mitigation strategies, targeted disinfection, improved hygiene, vendor training, showing how ES data, when contextualized, can inform tailored responses. These mitigation strategies align with evidence that targeted interventions in live bird markets (e.g., temporary closures/rest days, upgraded hygiene/biosecurity) can reduce avian influenza virus contamination and risk, with effectiveness contingent on duration and implementation fidelity. [31], [32]

Threshold setting and cross-sectoral response planning emerged as key operational challenges in Exercise 4. Although participants recognized signals of increased risk, many noted the lack of standardized criteria or inter-ministerial protocols to act upon them. These constraints, unclear lines of authority, underdefined triggers, and fragmented data systems, are consistent with previous OH implementation studies. Yet many participants offered potential solutions: simplified threshold guidance, routine joint reviews of surveillance outputs, and more inclusive coordination structures that incorporate provincial authorities.

These findings also reinforce that the main barriers to operationalizing One Health surveillance are not only scientific or technical, but also socio-political and institutional. Participants could engage with metagenomic environmental surveillance data when they were presented clearly, but translating those outputs into action depends on coordination, mandates, resource sharing, communication channels, and trust across sectors. Although recommendations such as formalized coordination, data-sharing protocols, and joint decision-making mechanisms are well recognized in One Health implementation, they remain difficult to operationalise in practice [33], [34], [35], [36]. In Cambodia, as in many settings, responsibilities for sampling, reporting, risk communication, and intervention are distributed across ministries and administrative levels. Therefore, future implementation will require more than technical tools or visualization templates; it will require agreed authority to act, clear escalation pathways, designated focal points, sustainable resources, and routine multisectoral review processes that can function under operational pressure.

By the end of the workshop, post-exercise surveys (Exercise 5) and interviews indicated a measurable shift in participants' confidence and willingness to engage with metagenomic data. While initial responses reflected unfamiliarity, final reflections emphasized increased comfort and perceived utility, especially when visualizations were labeled, intuitive, and clearly tied to response frameworks. These gains support findings from similar training initiatives that link hands-on data interpretation to increased institutional uptake [24], [25].

When combined with syndromic surveillance, which refers to detection of early health signals in humans and/or animals, and supported by multisectoral coordination, metagenomic ES has the potential to strengthen early warning capabilities for zoonotic disease threats in LMICs. This workshop reinforced that surveillance is not only about generating data, but also about making those data usable, interpretable, and timely. The integration of user-centered design, context-aware thresholds, and cross-sectoral engagement is critical for translating ES into real-time risk assessment and response.

Metagenomic ES will not immediately replace existing systems, but it can fill critical blind spots. As tools and data pipelines mature, embedding these approaches within a OH surveillance architecture offers a path toward earlier detection, smarter prioritization, and stronger resilience against emerging infectious diseases [9], [23].

The contribution of this work lies in moving beyond technical demonstration of metagenomics toward its practical application in policy and multisectoral coordination. By explicitly testing visualization preferences, thresholds for action, and coordination mechanisms with Cambodian stakeholders, this study provides a initial roadmap for making metagenomic surveillance data actionable in LBMs settings. Such participatory approaches remain rare in the literature and represent an important step toward embedding advanced surveillance tools into real-world OH decision-making.

4.1. Limitations

Limitations of this study include that the findings are based on a single workshop and a small number of key informant interviews, and no inferential statistical analyses were conducted. The findings should therefore be interpreted as preliminary, context-specific, although they may inform future work in similar One Health and LAM settings. The strength of the approach lies in the depth of engagement, cross-sectoral representation, and scenario-based learning, which yielded rich qualitative insights into stakeholder preferences, perceived barriers, and implementation needs. However, the workshop primarily captured institutional and technical perspectives. Market workers, vendors, and community members, who are directly affected by surveillance findings and market-level interventions, were not directly included. Future work should engage these groups to ensure that risk communication messages, surveillance outputs, and proposed interventions are acceptable, feasible, and actionable in market settings.

Follow-up research is needed to assess transferability across Cambodian provinces, other market systems, and comparable countries with different governance structures, resources, and surveillance capacities. Adoption in other settings will also depend on operational and political factors, including ministry mandates, authority to act on ES signals, data-sharing mechanisms, subnational capacity, financing, and the feasibility of implementing market-level interventions under political and economic constraints.

The use of fictional data helped standardise discussion, reduce confidentiality concerns, and avoid anchoring participants to sensitive real surveillance findings. However, it also created a simulation gap. Participants' choices in a low-stakes workshop may differ from their risk tolerance during an actual crisis, when political and economic pressures, uncertainty, reputational risk, and urgency may affect decision-making. This study did not test how the proposed visualizations, communication formats, or threshold concepts would perform under real outbreak conditions.

The workshop also identified key operational requirements. Limited bioinformatics capacity remains a major barrier: future implementation defines who will analyze raw sequencing data, maintain pipelines, ensure quality control, interpret outputs for non-specialist users, and fund computational infrastructure. Similarly, the study identified the need for evidence-based, context-specific thresholds, but could not determine what level of avian influenza A/H5N1 detection should trigger a market-level intervention. This remains a central operational question that the study highlights but cannot yet answer. These issues require longitudinal surveillance data, expert review, costing, and prospective evaluation before implementation. The roadmap should therefore be interpreted as an initial draft developed after the workshop to guide future planning, rather than as a completed, costed, or formally endorsed implementation plan.

5. Conclusion

Effectively communicating complex environmental surveillance metagenomic outputs requires more than accurate data. It demands user-centered design, clearly defined thresholds, and coordination across the human, animal, and environmental health sectors. This workshop demonstrated that multisectoral stakeholders in Cambodia could engage with, interpret, and identify potential actions based on metagenomic ES data when supported by clear visualizations and contextually relevant tools.

Participants consistently emphasized the importance of simplicity, accessibility, and familiarity in visual formats, along with the need for dual digital and analog delivery methods. They highlighted real-world barriers such as infrastructure limitations and coordination gaps but also proposed practical solutions, from standardized action thresholds to improved cross-sector communication.

Findings from this workshop suggest that tailoring surveillance outputs to user preferences, embedding them within OH coordination mechanisms, and combining them with complementary tools like syndromic surveillance may help enhance the interpretability and practical utility of ES outputs in low-resource settings [37] [6]. These insights provide a foundation for integrating metagenomic surveillance into national early warning systems, not as a replacement, but as a vital extension of existing approaches. By specifying how end-users prefer to see and act on complex outputs and translating those preferences into an initial, context-anchored roadmap, this study offers an initial pathway for operationalizing metagenomic ES at LAMs in Cambodia and, with adaptation, across comparable settings in the Western Pacific region.

Future efforts should prioritize ongoing capacity building, decentralized engagement, and the co-development of operational protocols to turn data into action. As zoonotic threats continue to emerge at complex human–animal–environment interfaces, investing in inclusive, responsive, and practical surveillance systems will be essential to preparedness.

Research in context

Evidence before this study

Environmental surveillance (ES) has increasingly been recognized as a tool for detecting pathogens in high-risk settings such as live animal markets (LAMs). Previous studies in Asia have demonstrated the value of ES for tracking avian influenza viruses, poliovirus, antimicrobial resistance, and SARS-CoV-2. However, most research has focused on pathogen detection and technical outputs rather than on how complex metagenomic data can be communicated to decision-makers. We searched PubMed and Google Scholar for articles published in English up to May 2024 using the terms “environmental surveillance”, “metagenomics”, “zoonoses”, “live animal markets”, “One Health”, “data visualization”, and “decision-making”. We found limited published evidence describing participatory, multisectoral workshops that tested how metagenomic ES data could be visualized and integrated into public health decision-making in low- and middle-income country (LMIC) settings.

Added value of this study

There are minimal previous examples of participatory exercises to apply metagenomic ES data in a multisectoral workshop setting in LMIC and their potential use for One Health decision making in LAMs settings. By convening stakeholders from health, agriculture, and environment ministries as well as other stakeholders in Cambodia, we explored preferences for data visualization formats, thresholds for action, and coordination mechanisms. Participants highlighted the importance of simple, familiar formats (bar, pie, and line charts), intuitive color codes, and offline printer-friendly outputs, while also recognizing the potential of digital dashboards. Barriers identified included limited bioinformatics capacity, weak cross-ministerial coordination, and insufficient prioritization of ES at the provincial level. The study generated a preliminary roadmap for integrating ES into Cambodia's One Health early-warning systems.

Implications of all the available evidence

Our findings show that the usefulness of metagenomic ES data depends not only on technical accuracy but also on user-centered communication and governance structures. Tailored visualization, context-specific thresholds, and formalized inter-ministerial coordination are crucial for transforming complex datasets into actionable signals. Embedding participatory approaches within national surveillance frameworks can accelerate the integration of ES into One Health systems, strengthening early detection and preparedness against zoonotic threats in Cambodia and similar resource-limited settings.

Data access and verification

F.E.S. and E.A.K. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the analysis. Fictional datasets and visualization source templates used during the exercises are available from the corresponding author upon request.

CRediT authorship contribution statement

Frida E. Sparaciari: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Karen Saylors: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis. Malen Chan: Writing – review & editing, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation. Sofia Perez: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Cadhla Firth: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Paul F. Horwood: Writing – review & editing, Supervision, Project administration, Methodology, Funding acquisition, Conceptualization. Erik A. Karlsson: Writing – review & editing, Visualization, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Funding

This study was funded by the Gates Foundation (INV-049293). The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript. The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material 1

Supplementary material 2

Data availability

The workshop agenda, Mentimeter questions and response options, pre- and post-workshop questionnaires with descriptive results, semi-structured interview guide, and workshop evaluation form are provided in the Supplementary Material 2. Full fictional ES/metagenomic datasets, visualization source templates, and dashboard mock-ups used during the workshop exercises are available from the corresponding author upon reasonable request. Individual workshop responses, raw interview materials, and identifiable participant data will not be shared publicly because of confidentiality considerations and the potential risk of re-identification.