Criteria for selecting sentinel unit locations in a surveillance system for vector-borne disease: A decision tool

Objectives With vector-borne diseases emerging across the globe, precipitated by climate change and other anthropogenic changes, it is critical for public health authorities to have well-designed surveillance strategies in place. Sentinel surveillance has been proposed as a cost-effective approach to surveillance in this context. However, spatial design of sentinel surveillance system has important impacts on surveillance outcomes, and careful selection of sentinel unit locations is therefore an essential component of planning. Methods A review of the available literature, based on the realist approach, was used to identify key decision issues for sentinel surveillance planning. Outcomes of the review were used to develop a decision tool, which was subsequently validated by experts in the field. Results The resulting decision tool provides a list of criteria which can be used to select sentinel unit locations. We illustrate its application using the case example of designing a national sentinel surveillance system for Lyme disease in Canada. Conclusions The decision tool provides researchers and public health authorities with a systematic, evidence-based approach for planning the spatial design of sentinel surveillance systems, taking into account the aims of the surveillance system and disease and/or context-specific considerations.


Contact email for participation of experts
Below is the contact email which was sent to 11 experts to ask for their participation in validating the decision tool.
Dear all, For those of you who don't know me, I am a PhD student in Dr Patrick Leighton's lab. My thesis looks at sentinel surveillance, in the context of vector-borne diseases, and more specifically for Lyme disease.
In the process of designing a surveillance system for vector-borne diseases, selecting the best sentinel sites can be challenging. In the context of my PhD, I have created a decisional tool which aims to help users to choose relevant criteria which can be used to objectively select such sites. The output of the decisional tool is a list of criteria, which can be used to determine where in space these sentinel sites should be located e.g., used as part of a multi-criteria decision analysis.
The tool was created with information gained from conducting a literature search, and data compiled to create a logical tool. It follows on from a previous article: However, this tool needs to be validated by experts (academic and public health) to ensure its relevance, functionality, and completeness, before its publication.

Introduction to the questionnaire
Below is the description of the task which was provided to experts for validation of the decision tool: Hello, Thank you for taking the time to complete this validation step for a new decisional tool.
In the process of designing a surveillance system for vector-borne diseases, selecting the best sentinel sites can be challenging. In the context of my PhD, I have created a decisional tool which aims to help users to choose relevant criteria which can be used to objectively select such sites. The output of the decisional tool is a list of criteria, which can be used to determine where in space these sentinel sites should be located e.g., used as part of a multi-criteria decision analysis.
The tool was created with information gained from conducting a literature search, and data compiled to create a logical tool. It follows on from a previous article: However, this tool needs to be validated by experts (academic and public health) to ensure its relevance, functionality, and completeness, before its publication.
I have created a short questionnaire (20-30 mins) to allow for its assessment. Please use the pdf sent to you by email in order to answer the questions.
Once again, many thanks and don't hesitate to get in touch with any questions,

Original decision tool provided to experts
The following decision tool ( Figure 1) was provided in pdf form to experts. It was used to answer the questions within the online survey. D1 Consideration of administrative boundaries D2 Random distribution of locations within the surveillance zone E1 Consideration of the geography E2 Presence of appropriate ecology for the presence of the vector E91 Consideration of climate I1 Selection of locations previously used in surveillance programs I2 Selection of locations previously used in scientific studies I3 Previous public health interventions carried out within the sentinel unit locations I4 No previous public health interventions carried out within the sentinel unit locations I5 Variation in public health interventions carried out across the sentinel unit locations L1 Consideration of logistical constraints (e.g., traveling distance, access) L2 Voluntary enrollment of sentinel unit locations L3 Stakeholders' preferences, suggestions, or recommendations L4 Presence of specialists or a specialist center within or near the sentinel unit location L5 Presence of adequate communication facilities within or nearby the sentinel unit location P1 Selection of the sentinel unit locations to maximize the population surveyed P2 Consideration of population demographics (e.g., age, gender, socioeconomic status) P3 Stability of human population (no immigration / emigration) P5 Presence of a specific type of human activity (e.g., fishing, hunting, wild mushroom picking) P6 Consideration of the patient demographics from participating the health clinics (e.g., age, gender, socioeconomic status) R1 Presence of documented risk of disease, based on human case data R2 Variation in degree of risk of the disease between the sentinel unit locations R3 Presence of documented risk of disease, based on presence of appropriate disease vectors R10 Presence of documented risk of disease, based on the evidence of vector-human contact 1 criteria codes were kept from previous publication to keep these constant across our work In North America, Lyme disease (LD) has been identified a priority VBD for public health due to its ongoing emergence, accelerated by climate change. In eastern Canada, the vector tick species, Ixodes scapularis, was first studied in the 70s at Long Point, Ontario and its range has shown ongoing expansion since this time. With Canada's vast territory, and the heterogeneous spread of tick populations in space, sentinel surveillance could allow a cost-effective, nationwide surveillance strategy to monitor the environmental risk of LD and track the risk through time. Hence, LD surveillance at a national level will be used as a case example to illustrate the application of the criteria selection tool in the determination of sentinel unit locations. Such a surveillance system will be constructed by the Canadian Lyme Disease Research Network (CLyDRN) in the form of the Canadian Lyme Sentinel Network (CaLSeN). The aim of this system will be to follow trends in Lyme disease risk over time across Canada by evaluating the environmental risk of LD through active field surveillance of ticks. The sentinel unit will be a sentinel region, a geographical unit of 100km diameter around a population center. There will be minimally one sentinel region per province, and thus, the criteria selection tool will be used to determine which locations would be the most appropriate to act as sentinel regions at the provincial level.

Past information
Although there has been past active surveillance done in most Canadian provinces, they are no sentinel regions established for more intensive surveillance initiatives. Currently, there are no planned public health interventions as part of the surveillance system.

Risk
The primary aim of the sentinel surveillance system will be to follow trends in LD risk exposure in the environment. In the context of LD, although it was deemed a notifiable disease in 2009 in Canada, the human data case is owned by individual health boards and the geographic scale of the location of acquisition varies greatly between provinces. Due to these circumstances, data from passive surveillance is more easily accessible and remains the earliest signal of environmental risk of LD. Thus, for the system, the risk associated with the presence of the vectors will be used as a criterion for the selection of sentinel regional location.

Environment
Habitats vary greatly among and within provinces -from urban, barren land, wetlands, cropland, needle leaf forests, grassland, etc. (Canada Land Cover 2015). Many of these habitats will not be appropriate for the establishment of Ixodes spp. ticks, which require mixed or deciduous forests. Thus, this heterogenous land cover should be considered, and the presence of appropriate habitats for the establishment of vector populations should also be retained from the criteria selection tool. Furthermore, as the primary aim of the surveillance system is to follow disease trends, we can also select the criterion in which climatic features are considered, in this case temperature is the most relevant variable.

Population
Human population density varies greatly across provinces, with most of the population concentrated in urban centers. Thus, population density will be considered as a selection criterion. However, as LD remains a relatively rare disease in Canada, with an incidence of 2.7 cases / 100 000, population demographics or behaviors will not be considered.

Distribution
To ensure a minimum level of geographic representativeness, at least one sentinel region will be selected within each province, and up to three sentinel regions in the larger provinces that report higher incidences of LD, such as Ontario and Québec.

Logistics
As the system will cover a large geographical area, and that the system is intended to be maintained for many years to provide a longer temporal series, logistical aspects including costs of functioning are important to consider. Costs will vary with travel distance between research base and the sentinel region location. Materials, communication, and laboratory test results will stay constant regardless of the sentinel region location. Thus, to minimize costs and save time, travel distance between the sentinel region and the nearest research base should be considered during the selection of sentinel region location.
To summarize this case study, the criteria which researchers should use to select sentinel region location at the provincial level include: 1) There is documented risk of disease due to the presence of appropriate vector disease within the sentinel region (use of passive surveillance data) 2) Climate: considering temperature in the form of accumulated degree days.
3) The ecology of the sentinel region is appropriate for the presence of the vector (presence of mixed or deciduous forests) 4) Selection of the sentinel region to maximize the population reached within the units of the study zone 5) Logistical constraints (e.g., traveling distance) are considered when selecting sentinel regions 5. Questions within the questionnaire The questionnaire comprised of a total of 14 questions. It would depend on the surveillance objective: to detect the vector/pathogen/risk area or estimate the risk of transmission to humans. It also depends on the point of view e.g., a citizen who lives or walks somewhere versus public health who wants to know where there will be more cases 3 Yes 4 Yes 5 Yes

Yes
The influx of people in a region for work or tourism should also be considered. A region could have a low population but be very touristic and represent a risk for the population 11 Yes 7. For category 4 (Population), population particularities e.g., demographics, human activity, etc., are mentioned. The statement is broad in order to remain flexible and relevant for a variety of VBDs. Do you think this is appropriate and adequately formulated? (If not, please explain) 2 Yes Yes Preference of local public health authorities (however this is included within the logistic criterion) 11 No 9. Category 6 (Logistics) has been included at the end of the decision tool in order to ensure that the surveillance system is sustainable. Do you agree with this step? (If not, please explain)

Yes
Cost, long-term management of the surveillance system, standardization of the protocol (data collection, laboratory analyses, data analyses, information key actors) 3 Yes 4 Yes 5 Yes

Yes
Evaluation of the impact level of each of these logistical aspects on the sustainability and functionality of the system. Ideally, do not make too many comprises relating to these criteria, to prevent deviating from the initial objective, unless this has an important impact on the quality of the system and the data. Overcome these logistical difficulties by finding alternative or finding measures to decrease these impacts, if possible. Incidence rate or abundance of ticks high enough to be able to meet the objectives (except early warning system). For example, if the number of ticks collected or the number of human cases is low, this could prevent trends from being identified due to small numbers and statistical uncertainties. 11 Yes 11. Is the decisional tool clear and self-explanatory? Please write in the text box any suggestions to make the tool clearer or easier to use. Complex to understand, especially for someone who has never done surveillance site selection. The case study helps to understand: good idea. A presentation of the tool or an explanatory document could also have helped. At the end of the process, a question remains: which criterion(s) should be prioritized if we cannot find a sentinel unit corresponding to all these criteria? What do you suggest? Does the order of the criteria reflect the weight of the criteria? Could be translated in French (reviewer was francophone) 11 No I just have some general formatting suggestions. It would be great if you could move the legend on the same page as the schematic, as it would make it a lot easier to go around it. Also, legend items do not have to be in alphabetical order and listing them in the order they are mentioned makes more sense. Also, avoid skipping numerical labels for easier reading. For example, we have R1 to R3 but skip to R10 after, so in this case I would switch the label R10 to R4, if this makes sense. 12. Is the flow of the decision tool logical? Please comment in the text box any suggestions on how to optimize the sequence of the tool. • Some comments about criteria which are included elsewhere (e.g., relative to the distribution of the risk of disease, or relating to public health authority preference) • Changed according to iterative inspection of realist-type review, as not many comments from reviewers The variation in the environment is judged significant by the investigators c Early warning system d It is also relevant to consider potential important population influx e.g., from tourism, occupational reasons e Human activities which influence exposure to vectors / vector-borne diseases