2.1. Study design

A systematic scoping review is a research method that involves a structured and comprehensive synthesis of existing literature on a specific topic, using predefined criteria and transparent procedures. Our approach follows the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol developed prior to commencing data collection and analysis, guided the study’s design and methods (S2 File). Briefly, this may be considered a hybrid approach with characteristics of a systematic review (predefined search string, inclusion and exclusion criterion; formal screening process), and a scoping review (search strategy based on refinement of initial exploratory search; focus on current practice in uncertainty visualisation rather than effectiveness of a given method of uncertainty visualisation; qualitative approach to outcome data extraction). The aim was to examine the current literature on communicating uncertainties in geospatial maps related to climate-sensitive infectious diseases and their implications for policymakers.

2.2. Data sources and search strategy

A three-step search strategy was employed to ensure a comprehensive and holistic review of the literature. First, we conducted an initial search in the PubMed database to identify a preliminary set of articles representing relevant literature. Review of these articles and their reference lists were used to establish the final search strategy detailed in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Search strategy.

https://doi.org/10.1371/journal.pclm.0000707.t001

Next, we systematically searched peer-reviewed articles in two databases: PubMed and Web of Science, selected for their extensive coverage of biomedical and environmental health literature pertinent to our topic.

The search strings outlined in Table 1 were applied across both databases, in “all fields” (specific syntax was TS= for Web of Science, and [tiab] for PubMed). Using filters, we limited our search to articles published in English (specific syntax was LA=(English) in Web of Science, and English[lang] for PubMed). There were no restrictions on publication date. The search included all papers available up to April 2025.

Finally, we conducted a hand search of the reference lists of included papers to identify additional literature that might have been missed during the database searches.

Relevance to the Asia-Pacific region was inferred by explicit mention of terms used for countries, subregions, and broad regional labels commonly used in the relevant infectious diseases literature pertaining to Pacific Island countries and territories, Australia, and New Zealand (Table 1, line 1). South Asia, Southeast Asia, and South Asia were not included in the search terms to limit the number of records to something feasible for a scoping study.

2.2.2. Study selection process.

The screening process is summarized in Fig 1. First, results from the two databases were combined, and duplicate records were removed. Titles of all remaining records were screened based on the inclusion and exclusion criteria, with any disagreements resolved through consensus. The same process was applied to the screening of abstracts. Following this, full-text screening was conducted based on the inclusion of GIS maps in the research study. Additionally, to identify relevant studies not captured in the database searches, the reference lists of the studies included for data extraction were also screened.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram of manuscript inclusions/exclusions.

https://doi.org/10.1371/journal.pclm.0000707.g001

2.2.3. Data extraction and analysis.

Data from all included studies, were extracted by one reviewer (AJ) with 15% blinded double screening verification from EIW, and randomised checks by AL, with any disagreements resolved through consensus. Ambiguous cases (e.g., where low image quality of a given figure may obscure whether a visual feature was intended to be blurred or sharp; or where presence of absence of a given form of ambiguity was unclear) were flagged by the extractor and resolved through discussion with EIW and AL. As this was a qualitative scoping review, a fully featured data extraction template was not applied; rather, we applied a series of a-priori data features (Table 2) and inductively identified recurring types of uncertainty (Table 3) to guide extraction of salient characteristics of included figures.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Information extraction.

https://doi.org/10.1371/journal.pclm.0000707.t002

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Identified recurring types of uncertainty.

https://doi.org/10.1371/journal.pclm.0000707.t003

A critical qualitative analysis was then conducted on all full-text studies containing maps, based on the above criteria, to evaluate the communication of uncertainties in maps and implications for policymaking.

3.1. Overview of included studies

There were 2,826 results from searches. After removing duplicates, applying the inclusion and exclusion criteria, 135 studies from the original search underwent full text screening. The original search yielded seven reviews, from which an additional 28 papers of possible relevance not included in the search results were identified. Following full text screening, 66 papers were included for data extraction. (S1 File).

All included studies focused on the Asia-Pacific region, consistent with the geographic scope of this systematic scoping review. The time period covered by these studies ranged from 1986 to 2025, with 96.6% published within the last 20 years, reflecting a growing research interest in this field in recent decades.

Among the included studies, the most commonly investigated climate-related infectious diseases were malaria, dengue fever, scrub typhus, hand, foot and mouth disease, and chikungunya virus. Malaria was particularly well-represented with multiple studies focusing on its spatial and temporal distribution across Papua New Guinea, Indonesia, and other Asia-Pacific regions. Dengue fever and other arboviral diseases, such as Ross River virus and Barmah Forest virus, were frequently examined in studies from Australia and Southeast Asia.

Other infectious diseases studied included leptospirosis, influenza A virus, soil-transmitted helminths, and scrub typhus, highlighting the diverse range of climate-sensitive infectious diseases addressed using geospatial mapping in the Asia-Pacific context. This diversity underscores the broad relevance of GIS-based uncertainty communication across multiple public health challenges.

3.2. Target audience and policy relevance

Approximately 52 out of 66 studies (approximately 79%) clearly mentioned policymakers, public health authorities, or health practitioners as key audiences for their research findings.

These studies contributed to policy relevance by informing targeted interventions and resource allocation and enhancing disease surveillance accuracy through improved spatial prediction models to support control and elimination programs. They also provided actionable insights that support national or regional disease elimination goals. They offered evidence to update or develop new public health policies and control strategies and enabled rapid identification and response to outbreaks through refined spatial and temporal risk mapping. The diseases addressed where policy relevance was emphasized include malaria, dengue, chikungunya, scrub typhus, hand-foot-mouth disease, visceral leishmaniasis, influenza, and soil-transmitted helminths.

3.3. Data source

Among the 66 studies included for full-text screening, the majority (53%) relied exclusively on secondary data sources, including national surveillance databases, electronic health records, government reports, and remote sensing datasets. Approximately 23% of studies collected primary data directly through methods such as field surveys, sentinel surveillance, clinical interviews, and laboratory-confirmed case detection. The remaining 24% of studies employed a mixed approach, combining both primary data collection and secondary data sources.

Uncertainty in the source of data was classified into low, moderate, and high categories. Thirty-five studies using only secondary data generally exhibited low uncertainty, reflecting standardised collection protocols and established databases. Conversely, fifteen studies based on primary data collection often had higher uncertainty due to variability in data quality, potential reporting biases, poor validity and reliability of research tools and methodological differences. Sixteen mixed data studies typically demonstrated moderate uncertainty.

Common challenges across the included studies included underreporting of cases, missing or incomplete data, inconsistent surveillance methods, and, in some instances, reliance on older datasets where data collection methods were less accurate and reliable. These factors contributed to inconsistencies and potential biases in the data, affecting the quality of geospatial analyses.

3.4. Temporal considerations

The timing of sampling in the included studies showed considerable variability, which impacted the ability to capture seasonal and temporal disease dynamics accurately. Twenty studies had limited sampling periods focusing on peak seasons or specific months, which may bias results by missing seasonal variation. For instance, Hu et al. [22] collected data only for May 2008, justified by the peak case count that month, but this approach risks missing important seasonal fluctuations in hand, foot, and mouth disease (HFMD) incidence. Similarly, in scrub typhus studies, sampling was often skewed towards peak transmission months such as October and November, possibly underestimating cases during off-peaks (e.g., as in Xin et al. [23] and Wu et al. [24]).

Conversely, about 15 studies employed long-term continuous or monthly data collection over several years, which better captured seasonal patterns and temporal trends. For example, Ding et al. [25] analysed 10 years of continuous monthly data for HFMD, allowing a robust understanding of seasonal effects. However, Dewan et al. [26] on visceral leishmaniasis lacked explicit discussion on seasonal variability despite known climatic influences on transmission, indicating a temporal sampling limitation.

Eight studies noted delays between illness onset and diagnosis, especially in rural or high-risk areas, which could blur temporal clustering and affect data reliability. Qian et al. [27] highlighted that rural-urban differences in reporting timeliness could lead to variations in observed transmission patterns.

Allen et al. [28] sampled just before a major chikungunya epidemic in Dhaka, which may not represent risk dynamics during the epidemic itself. This provided valuable baseline data but potentially limited the generalizability of findings to other periods.

Regarding spatial scale, studies ranged from national or provincial-level analyses to fine-scale district, community, or household-level studies. Approximately 50% of studies utilized aggregated data at national or regional scales, providing broad geographic coverage but often with less spatial precision. Around 27% of studies conducted analyses at district or sub-district levels, allowing identification of localised transmission patterns. A smaller subset of about 23% of studies used geo-referenced household or village-level data to achieve high spatial resolution, although such fine-scale data posed challenges regarding data completeness and consistency.

Overall, the variability in sampling timing and spatial scale across studies represents a key challenge in accurately characterising disease dynamics and their environmental drivers.

3.5. Visual depiction of uncertainty

In addition to data source and temporal considerations, 3 broad categories of recurring uncertainty in maps were identified from qualitative analysis of the included 66 studies (table 3).

The majority of the included studies did not explicitly represent uncertainty visually on their maps. Instead, uncertainties were often only mentioned descriptively in the text or in figure captions.

Seventeen studies acknowledged uncertainty qualitatively in the text but did not provide quantitative measures such as confidence intervals or error margins (e.g., Reid et al. [30]). This lack of precise quantification limits the ability to assess the reliability of the spatial estimates.

Eleven studies reported statistical measures of uncertainty, such as standard errors or p-values in the textual descriptions but failed to incorporate these metrics directly onto the spatial visualisations [22,31]. Without visible representations of these uncertainties on the maps themselves, policymakers may misinterpret spatial patterns as more precise or reliable than they truly are, potentially overlooking areas where spatial heterogeneity is uncertain or poorly supported by the data.

Nine studies included uncertainty details in figure captions without visualising them within the figures or maps themselves (e.g., as in Hasan et al. [32], Owada et al. [33], Qian et al. [27]). This disconnects between textual information and spatial representation may result in decision-makers neglecting uncertainty when interpreting the maps for policy formulation.

Only three studies effectively integrated explanatory visual representations of uncertainty into their maps, using techniques such as Bayesian Decision Network (BDN) modelling, spatio-temporal modelling, shading gradients or spatial fuzziness to denote areas of higher uncertainty [34–36]. Such clear depictions are valuable for policymakers, enabling them to identify regions with less reliable data and thus adopt more cautious or targeted interventions.

Seven studies, including Zellweger et al. [37], Edenborough et al. [38] and Akter et al. [39], used aggregated spatial units prone to the Modifiable Area Unit Problem, potentially masking fine-scale variability. No maps displayed clear signs of uncertainty arising from poor visual clarity that could not be better attributed to spatial or statistical uncertainty. Temporal uncertainties were also found in the studies, but these were rarely visualised within spatiotemporal maps (e.g., an omission seen in Fahmi et al. [40], and Trewin et al. [41]). For instance, Trewin et al. [41] used data from 1887-2016. This study reported significant temporal gaps in surveillance, especially during the mid-20th century, and changes in data collection methods from sporadic, paper-based reports to systematic digital surveillance. Additionally, aggregating data across such a broad time period without accounting for methodological differences. The lack of explicit temporal uncertainty depiction in the figure could mislead policymakers regarding the stability and robustness of observed trends over time.

3.6. Concordance between observed and reported uncertainties

The previous section focused on the visual representation of uncertainties in maps. It is equally important to consider how uncertainties are communicated in the accompanying text. Over a third of studies (n = 26, 39%) didn’t report any uncertainties in their research papers (e.g., as in Murphy. et al. [42], Tewara et al. [43], Viennet et al. [44]). In many of these cases, key details such as confidence intervals, error margins, or sampling limitations were missing from both the text and the maps. This lack of transparent reporting means that although uncertainties existed, they were not clearly communicated to readers. Such omissions can mislead policymakers into assuming that the data is more precise and reliable than it truly is, which could lead to overconfident or potentially flawed decisions.

Thirty-seven studies reported some uncertainties in text (and/or figure captions), but these were incomplete (e.g., seen in Hasan et al. [32]). Not all relevant uncertainties were mentioned by the authors, as they acknowledged sampling limitations but failed to report spatial or temporal uncertainties.

Taken together, these findings indicate that most spatial epidemiological studies do not employ visual methods to convey uncertainty. This insufficiency may foster misplaced confidence in map-based data, increasing the risk of misguided public health decisions. Therefore, transparent and explicit visualization of uncertainty is critical to support evidence-informed, prudent policymaking.

4.1. Toward better uncertainty communication

Geospatial scientists, decision scientists, epidemiologists, policy experts, and behavioural scientists should work together to ensure that uncertainty is not just technically calculated but also meaningfully interpreted and applied [45,53]. A fruitful first step would be to increase efforts to raise policymaker awareness of the value of understanding uncertainty when prioritising the information sources and focusing interventions and further data collection. There are some examples of this. Degeling et al. [54] outline of how such training may be used to enhance policymaker decision making in disease surveillance, and Wiggins et al. [55] present a decision uncertainty toolkit to support decision making during public health emergencies. Once the value of uncertainty is understood, the value of visualising uncertainty may be more salient to policymakers. This could be supported by an expectation from journals, reviewers, and agencies commissioning reports that uncertainty will be routinely integrated into policy-focussed reporting, ideally within visuals as well as text. Involving public health stakeholders from the outset ensures maps meet both the informational needs and cultural expectations of decision-makers [56,57].

There is substantial variability in individual’s decision performance when using maps [58]. Usability testing, where policymakers interact with maps and provide feedback, can reveal how uncertainty representations such as shaded risk regions, error bars, confidence interval overlays, fuzzy boundaries, probabilistic shading, and ensemble-map variability layers are interpreted. This may require repeated specific co-design, exemplified by Scholz & Lu [46], who emphasise the importance of context-specific maps that explicitly explain uncertainties, modelling assumptions, and data limitations to decision-makers. As a potential way forward, Hadjimichael [59] proposes an exploratory modelling and decision-making framework that integrates scenario discovery with stakeholder engagement. This approach allows uncertainties to be visualised in a decision-relevant manner, enabling policymakers to navigate trade-offs without being overwhelmed by complexity. Such structured participation can support more transparent and actionable geospatial communication in public health, and help to mitigate the endemic tendency toward incredible certitude [15] in policy. An excellent example of this approach, and consequently useful recommendations for best practice is Ślusarski and Jurkiewicz’s (2019) [60] exploration of depicting positional, attribute, and temporal uncertainty in Poland’s Database of Topographic Objects.

4.2. Strengths and limitations

This systematic scoping review has several strengths, as well as limitations. A major strength of this work lies in its inclusive approach which allowed us to capture a broad spectrum of study types, modelling approaches, and uncertainty communication practices. Building on prior reviews, which tend to focus on text or visuals in isolation, we assessed the representation of uncertainty in text, figure captions, and maps. This level of granularity enabled us to identify subtle omissions such as confidence intervals reported in text but not present in figure captions or displayed on maps. Another notable strength is the wide temporal coverage: we examined all eligible literature from the earliest available records to 2025. However, we included only English-language publications, which may underrepresent research from non-English-speaking countries. Our inclusion criteria was geographically restricted to the Asia-Pacific region, and findings may not be generalizable to other regions with different health system structures or data availability. Relatedly, this manuscript has focussed on uncertainty from a specifically Western epidemiological lens, and using a largely qualitative approach. Objectivity in data and visualisation practice is to some degree socially constructed [61,62]. As Crampton [63] notes, maps embed social values and uncertainties as well as data uncertainties. This may have been compounded by our approach of relying on a single extractor for most data with a randomised double screening of fifteen percent of publications, which is common practice for systematic scoping reviews, but may have introduced some subjectivity. Future research should consider this social context alongside inclusion of policymakers and experts in efforts to improve depiction of uncertainty in practice. A source of insight for this may be citizen science data practises, such as the integration of insights from large “macro” datasets with citizen input on the “micro” level in order to increase data accuracy and usefulness for policy as described in Soltani et al. [64]. Finally, although we limited the scope to climate-sensitive infectious diseases, the explicit focus on uncertainty in visualisations expands the generalisability of results to other health domains, such as non-communicable diseases or injury surveillance.