Study protocol for a scoping review of Lyme disease prediction methodologies

Abstract

Introduction

In the temperate world, Lyme disease (LD) is the most common vector-borne disease affecting humans. In North America, LD surveillance and research have revealed an increasing territorial expansion of hosts, bacteria and vectors that has accompanied an increasing incidence of the disease in humans. To better understand the factors driving disease spread, predictive models can use current and historical data to predict disease occurrence in populations across time and space. Various prediction methods have been used, including approaches to evaluate prediction accuracy and/or performance and a range of predictors in LD risk prediction research. With this scoping review, we aim to document the different modelling approaches including types of forecasting and/or prediction methods, predictors and approaches to evaluating model performance (eg, accuracy).

Methods and analysis

This scoping review will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Review guidelines. Electronic databases will be searched via keywords and subject headings (eg, Medical Subject Heading terms). The search will be performed in the following databases: PubMed/MEDLINE, EMBASE, CAB Abstracts, Global Health and SCOPUS. Studies reported in English or French investigating the risk of LD in humans through spatial prediction and temporal forecasting methodologies will be identified and screened. Eligibility criteria will be applied to the list of articles to identify which to retain. Two reviewers will screen titles and abstracts, followed by a full-text screening of the articles’ content. Data will be extracted and charted into a standard form, synthesised and interpreted.

Ethics and dissemination

This scoping review is based on published literature and does not require ethics approval. Findings will be published in peer-reviewed journals and presented at scientific conferences.

Correspondence to Daniel Szaroz; [email protected]

STRENGTHS AND LIMITATIONS OF THIS STUDY

• This scoping review will synthesise information based on Lyme disease (LD) spatial prediction and temporal forecasting, including methodologies, predictors, outcomes and evaluation approaches, providing guidance in terms of methods and improving inter-study comparability.

• The broad search strategy will retrieve LD studies with forecasting and predictive methodologies published since 2000.

• The comprehensive search strategy includes five electronic databases for peer-reviewed literature and article bibliographies, as well as grey literature and manual searches.

• This review does not include a formal quality appraisal of the studies included.

• The scope of this study is limited to humans and excludes acarological and animal host risk mapping, which are part of the transmission cycle for LD; additionally, inclusion is limited with respect to time (since 2000) and language, which may result in relevant articles being excluded from the review.

Introduction

The causative agent of Lyme disease (LD), a ‘spirochete’ bacteria, was first isolated in 1982 on Shelter Island from the Ixodes scapularis tick. The bacteria was named Borrelia burgdorferi after its discoverer.^{1 2} Today, at least four common Ixodes species^{3 4} acquire the spirochete from host reservoirs as they mature through their lifecycle from larvae to nymphs and to adult ticks. Ticks can infect humans when they seek a blood meal by protruding the skin of a competent host for at least 36 hours, hereby transferring the Borrelia spirochete to the host.^{5 6} The abundance of tick populations correlates with the host’s natural habitats, making wooded areas and dense vegetation ideal for habitat establishment.⁷ Microhabitats that offer insulation provided by snowpack, leaf litter or both allow ticks to survive in soils through the harshest winters, maintaining geographical habitat boundaries.⁸ In recent years, humans have been increasingly exposed to high tick densities mainly due to encroachment in forests and animal habitats as well as through the fragmentation of landscapes by deforestation and agriculture.^{9 10} These fragmented landscapes have permitted highly efficient reservoir hosts for B. burgdorferi, such as the white-footed mouse to thrive, leading to more infected ticks.¹¹ In addition, milder winters and warmer summers, driven by higher temperatures and climate change, favour tick reproduction and survival.¹² Specifically, climate and environmental change favours tick population growth, gains in tick territorial latitudes and increases in the seasonal activity of ticks.¹³

Analytic studies on LD have used different quantitative approaches to characterise the disease risk or risk of the spread of LD at a population level. The literature reveals that some Bayesian spatiotemporal models or mechanistic math models have been used to model either disease occurrence in humans or tick species density.^14–16 In some cases, generalised linear mixed models have been combined with stochastic, spatially explicit diffusion models to predict the first case of human LD in a given region.¹⁷ These studies lead to an improved understanding of LD risk along with LD risk mapping work.¹⁸ For example, an early spatial analysis has demonstrated that I. scapularis is steadily expanding northward largely due to ecological shifts and host availability in the northern hemisphere.¹⁹ This trend is particularly noticeable as rising temperatures facilitate the northward shift of tick populations and the geographic range expansion of LD.²⁰ While I. scapularis is expanding in two North American regional hotspots, the Northeast and the upper Midwest (including central Canada), I. pacificus, a closely related tick, is the primary vector to humans for LD in the Western coastal regions of North America, like British Columbia, Washington, California and Oregon.^21–23 To date, no I. scapularis populations have been established in Western Canada and the USA. This distinction also underscores the importance of comparing the modest range expansion of the primary vector in the West, against the substantial expansion of the primary vector in the East.²² Overall, range expansion of vector and reservoir host species has led to the emergence of LD, in new regions and other temperate regions of the world.^{16 24 25}

Rationale

The objective of this scoping review is to characterise the approaches used for LD risk prediction in humans and identify possible gaps in the literature. Recommendations will also be made to improve the comparability and methodological rigour of LD risk prediction studies. While a large corpus of available reviews is based on LD prevention, diagnosis and treatment,^26–32 only a few have exclusively focused on disease prediction methods.^33–35 This proposed scoping review is novel given its focus on LD spatial prediction and temporal forecasting methods, including types of predictors, prediction/forecasting methodologies and evaluation approaches of model performance (eg, accuracy). This will be the first scoping review to provide a comprehensive overview of the existing LD risk prediction and forecasting methodologies, including model performance and predictors description.

Methods and analysis

The review will be conducted per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Review guidelines, PRISMA extension checklist for scoping reviews (online supplemental annex 1).³⁶ The protocol has been informed by a librarian and followed the recommendations of the work of Arksey and O’Malley.³⁷ A version of this protocol was registered in the Open Science Framework with the same authors (Registration DOI, https://doi.org/10.17605/OSF.IO/FGZQ9). The estimated start and end dates of the study will be between January 2023 and June 2024.

Research objective

The main objective of this scoping review is to report the characteristics of temporal forecasting and spatial prediction studies used to estimate the human risk of LD. The review will also describe the types of predictors, methodologies and evaluation approaches of performance (eg, accuracy) used in LD prediction and forecasting. This review will also provide recommendations to improve the methodology, transparency and comparability of LD risk prediction and forecasting studies.

The specific research questions are:

• What spatial prediction/temporal forecasting methods have been used to estimate the risk of LD?

• What are the predictors that have been included in these studies?

• What are the model evaluation methods used in these studies and what are the metrics used to evaluate a model’s accuracy?

Information sources: licensed databases and identification of studies

The review will target peer-reviewed literature focusing on temporal forecasting and/or spatial prediction methodologies for LD risk in humans to answer our research questions. The search strategy will be conducted in multiple peer-reviewed databases such as MEDLINE (OVID), EMBASE (Elsevier), CAB Abstracts (OVID), Global Health (OVID) and SCOPUS (Elsevier). The overall search strategy will use Medical Subject Headings (MeSH) and key terms that will be combined in Boolean and proximity operators. Preliminary searches of the literature were conducted prior to November 2022, to determine if previous work existed on this topic and to evaluate the feasibility of conducting a review on predictive methodologies of LD. With the assistance of a librarian, keywords and MeSH terms were devised to fit under the following three concepts: (1) relating to LD, (2) studies with a spatial predictive and/or temporal forecasting scope and (3) the recipient hosts are humans, by direct infection with ticks. This process was inspired in part by similar search strategies from another scoping review related to LD.³⁸ The complete search strategy can be found in the online supplemental annex 2.

Hand-searching will be performed on the reference lists of included studies to search for additional articles. A grey literature search will also be conducted manually, as guided by the Canadian Agency for Drugs and Technologies in Health Grey Matters³⁹ tool for reviews. Also, grey literature sources and websites of relevant authorities (eg, Food and Agriculture Organization, WHO, Pan American Health Organization and World Organization for Animal Health) will be searched.

Eligibility: inclusion criteria

Our review will encompass studies that have spatially predicted and/or temporally forecasted the human risk of LD at a population level. Specifically, we will include studies that fit into these three categories:

• Spatial prediction are methods that predict at unknown locations (prediction points), based on known values at observed locations. This process leverages the spatial relationships between variables to predict the probability of disease occurrence, incidence/prevalence or risk at prediction points.⁴⁰

• Temporal forecasting includes methodological approaches that use information on the temporal distribution of observed data (oftentimes series data) to forecast the probability of disease occurrence, incidence/prevalence or risk in the future or beyond the observation period.^{41 42}

• Spatiotemporal predictions are methods that predict the probability of disease, incidence/prevalence or risk at prediction points, beyond the observation period, or both, through the simultaneous analyses of the spatial and temporal distributions and patterns of the data.^{43 44}

We will include statistical, compartmental and machine/deep learning models studies if they meet the inclusion criteria. All included models will require to have an outcome as human LD risk (eg, probability of occurrence, number of cases, incidence or prevalence) that was the result of a spatial prediction or temporal forecasting model at an aggregate spatial (eg, postal code level) and/or temporal scale (eg, weekly). We will consider studies from a variety of geographic regions, without limiting our scope to a specific region.

Eligibility: exclusion criteria

The exclusion criteria will be articles with outcomes of host reservoirs or entomological risk. In addition, articles that did not spatially predict or temporally forecast human risk but rather estimated associations of risk (explanatory associations) or did not predict LD risk at an aggregate scale (eg, individual-level risk) will be excluded. Studies about non-human subjects will also be excluded. Reviews and commentaries will be excluded. Clinical prediction of tests and serology for patients will also be excluded. The review will be restricted to full-text articles available electronically in English or French and published since the year 2000.

Selection criteria and process

A two-step process will be used to select articles. In the first step, the abstract and title review will be screened by two authors (DS, CXRG). This will be performed using the Covidence platform.⁴⁵ The screening process will pilot test 50 articles prior to starting the screening process to improve harmonisation between the screeners. In the second step, full articles will be reviewed and the reason for the exclusion will be recorded if the study is excluded. Disagreement regarding exclusions will be discussed among the reviewers and a third reviewer will be consulted to resolve any discordance. Studies included in the second step will be classified into four types.

Data extraction

Data from the articles selected for inclusion will be extracted and organised into Covidence and backed up in a Microsoft Excel spreadsheet. Key elements will be extracted from the included articles. These are: title, author, year of publication, study objectives, study population, outcome, predictors (including measurement of predictors), methodology used for temporal forecasting, spatial prediction and/or spatiotemporal prediction, spatial scale and temporal resolution as appropriate, types of model evaluation methods used (eg, accuracy measures). A data-charting form will be collaboratively created to ensure that all relevant information is extracted and will be pilot tested on a subset of included articles. Two reviewers will independently extract the data and then compare their findings.

Summarising and reporting results

A quality assessment of articles will be limited to the discussion as we do not expect to perform an overall quantitative assessment given that this is a scoping review.³⁷ Results will be summarised in table format with common characteristics presented for all studies. The different prediction and forecasting studies will then be grouped by methods or based on prediction objectives (spatial, temporal, spatiotemporal). Tables with methods will include more detailed elements for prediction modelling based on best practices.⁴⁶ A descriptive summary will be included in the results and further expanded on in the discussion. The findings and recommendations will help identify trends and gaps in a growing methodological field of research with the goal of improving methodological transparency and comparability between studies.

Patient and public involvement

None.

Ethics and dissemination

This research aims to provide a comprehensive overview of existing LD risk prediction and forecasting studies to provide insight for researchers regarding methods and evaluation approaches and how guidance improves comparability between studies. This scoping review will compile and collate data from published research studies using publicly available sources and therefore does not require ethics approval. We plan on disseminating the results from the scoping review through publication in a peer-reviewed journal and in conference presentations.

Ethics statements

Patient consent for publication

Not applicable.

Footnote

X @ClaudiaXimena10

Contributors DS drafted the protocol and MK and KZ both assisted in the revisions of the draft. DS designed the search strategy following the advice of the institutional librarian. CXRG contributed to discussions, commented on ideas and reviewed versions of the protocol. KZ and MK revised and approved the final paper.

Funding This research was conducted as part of the Canadian Lyme Disease Research Network - CLyDRN project administered by Queen’s University and supported by Grant No. NLD-160482 under the Canadian Institutes of Health Research.

Competing interests None declared.

Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

References

¹ Burgdorfer W, Barbour AG, Hayes SF, et al. Lyme disease-a tick-borne spirochetosis? Science 1982; 216: 1317–9. doi:10.1126/science.7043737

² Burgdorfer W, Gage KL. Susceptibility of the black-legged tick, Ixodes scapularis, to the lyme disease spirochete, Borrelia burgdorferi. Zentralbl Bakteriol Mikrobiol Hyg A 1986; 263: 15–20. doi:10.1016/s0176-6724(86)80096-7

³ Yang LH, Han BA. Data-driven predictions and novel hypotheses about zoonotic tick vectors from the genus Ixodes. BMC Ecol 2018; 18: 7. doi:10.1186/s12898-018-0163-2

⁴ Kilpatrick AM, Dobson ADM, Levi T, et al. Lyme disease ecology in a changing world: consensus, uncertainty and critical gaps for improving control. Phil Trans R Soc B 2017; 372: 20160117. doi:10.1098/rstb.2016.0117

⁵ Falco RC, Daniels TJ, Vinci V, et al. Assessment of duration of tick feeding by the scutal index reduces need for antibiotic prophylaxis after Ixodes Scapularis tick bites. Clin Infect Dis 2018; 67: 614–6. doi:10.1093/cid/ciy221

⁶ Radolf JD, Strle K, Lemieux JE, et al. Lyme disease in humans. Curr Issues Mol Biol 2021; 42: 333–84. doi:10.21775/cimb.042.333

⁷ Clow KM, Leighton PA, Ogden NH, et al. Northward range expansion of Ixodes scapularis evident over a short timescale in Ontario, Canada. PLoS One 2017; 12: e0189393. doi:10.1371/journal.pone.0189393

⁸ Volk MR, Lubelczyk CB, Johnston JC, et al. Microclimate conditions alter Ixodes scapularis (Acari: Ixodidae) overwinter survival across climate gradients in Maine, United States. Ticks Tick Borne Dis 2022; 13: 101872. doi:10.1016/j.ttbdis.2021.101872

⁹ Allan BF, Keesing F, Ostfeld RS. Effect of forest fragmentation on lyme disease risk. Conservation Biology 2003; 17: 267–72. doi:10.1046/j.1523-1739.2003.01260.x

¹⁰ Larsen AE, MacDonald AJ, Plantinga AJ. Lyme disease risk influences human settlement in the wildland-urban interface: evidence from a longitudinal analysis of counties in the Northeastern United States. Am J Trop Med Hyg 2014; 91: 747–55. doi:10.4269/ajtmh.14-0181

¹¹ Simon JA, Marrotte RR, Desrosiers N, et al. Climate change and habitat fragmentation drive the occurrence of Borrelia burgdorferi, the agent of Lyme disease, at the Northeastern limit of its distribution. Evol Appl 2014; 7: 750–64. doi:10.1111/eva.12165

¹² Ogden NH, Lindsay LR. Effects of climate and climate change on vectors and vector-borne diseases: ticks are different. Trends Parasitol 2016; 32: 646–56. doi:10.1016/j.pt.2016.04.015

¹³ Bouchard C, Dibernardo A, Koffi J, et al. N increased risk of tick-borne diseases with climate and environmental changes. Can Commun Dis Rep 2019; 45: 83–9. doi:10.14745/ccdr.v45i04a02

¹⁴ Chen H, Stratton HH, Caraco TB, et al. Spatiotemporal Bayesian analysis of Lyme disease in New York state, 1990-2000. J Med Entomol 2006; 43: 777–84. doi:10.1603/0022-2585(2006)43%5B777:sbaold%5D2.0.co;2

¹⁵ Jiruše M, Machek J, Beneš V, et al. A Bayesian estimate of the risk of tick-borne diseases. Applications of Mathematics 2004; 49: 389–404. doi:10.1023/B:APOM.0000048119.55855.65

¹⁶ Kotchi SO, Bouchard C, Brazeau S, et al. Earth observation-informed risk maps of the Lyme disease vector Ixodes scapularis in central and Eastern Canada. Remote Sensing 2021; 13: 524. doi:10.3390/rs13030524

¹⁷ Bisanzio D, Fernández MP, Martello E, et al. Current and future spatiotemporal patterns of Lyme disease reporting in the Northeastern United States. JAMA Netw Open 2020; 3: e200319. doi:10.1001/jamanetworkopen.2020.0319

¹⁸ Ogden NH, St-Onge L, Barker IK, et al. Risk maps for range expansion of the Lyme disease vector, Ixodes scapularis, in Canada now and with climate change. Int J Health Geogr 2008; 7: 24. doi:10.1186/1476-072X-7-24

¹⁹ Wilson ML. Distribution and abundance of Ixodes scapularis (Acari: Ixodidae) in North America: ecological processes and spatial analysis. J Med Entomol 1998; 35: 446–57. doi:10.1093/jmedent/35.4.446

²⁰ Ogden NH, Maarouf A, Barker IK, et al. Climate change and the potential for range expansion of the Lyme disease vector Ixodes scapularis in Canada. Int J Parasitol 2006; 36: 63–70. doi:10.1016/j.ijpara.2005.08.016

²¹ Eisen RJ, Eisen L, Beard CB. County-scale distribution of Ixodes scapularis and Ixodes pacificus (Acari: Ixodidae) in the continental United States. J Med Entomol 2016; 53: 349–86. doi:10.1093/jme/tjv237

²² Hahn MB, Jarnevich CS, Monaghan AJ, et al. Modeling the geographic distribution of Ixodes scapularis and Ixodes pacificus (Acari: Ixodidae) in the contiguous United States. J Med Entomol 2016; 53: 1176–91. doi:10.1093/jme/tjw076

²³ Eisen RJ, Eisen L, Ogden NH, et al. Linkages of weather and climate with Ixodes scapularis and Ixodes pacificus (Acari: Ixodidae), enzootic transmission of Borrelia burgdorferi, and Lyme disease in North America. J Med Entomol 2016; 53: 250–61. doi:10.1093/jme/tjv199

²⁴ Diuk-Wasser MA, Vannier E, Krause PJ. Coinfection by Ixodes tick-borne pathogens: ecological, epidemiological, and clinical consequences. Trends Parasitol 2016; 32: 30–42. doi:10.1016/j.pt.2015.09.008

²⁵ Hvidsten D, Frafjord K, Gray JS, et al. The distribution limit of the common tick, Ixodes ricinus, and some associated pathogens in North-Western Europe. Ticks Tick Borne Dis 2020; 11: 101388. doi:10.1016/j.ttbdis.2020.101388

²⁶ Richardson M, Khouja C, Sutcliffe K. Interventions to prevent Lyme disease in humans: a systematic review. Preventive Medicine Reports 2019; 13: 16–22. doi:10.1016/j.pmedr.2018.11.004

²⁷ Lantos PM, Wormser GP. Chronic coinfections in patients diagnosed with chronic Lyme disease: a systematic review. Am J Med 2014; 127: 1105–10. doi:10.1016/j.amjmed.2014.05.036

²⁸ Brites-Neto J, Duarte KMR, Martins TF. Tick-borne infections in human and animal population worldwide. Vet World 2015; 8: 301–15. doi:10.14202/vetworld.2015.301-315

²⁹ Marques A. Chronic Lyme disease: a review. Infect Dis Clin North Am 2008; 22: 341–60. doi:10.1016/j.idc.2007.12.011

³⁰ Borchers AT, Keen CL, Huntley AC, et al. Lyme disease: a rigorous review of diagnostic criteria and treatment. J Autoimmun 2015; 57: 82–115. doi:10.1016/j.jaut.2014.09.004

³¹ Applegren ND, Kraus CK. Lyme disease: emergency department considerations. J Emerg Med 2017; 52: 815–24. doi:10.1016/j.jemermed.2017.01.022

³² Poland GA. Prevention of Lyme disease: a review of the evidence. Mayo Clin Proc 2001; 76: 713–24. doi:10.4065/76.7.713

³³ Lippi CA, Ryan SJ, White AL, et al. Trends and opportunities in tick-borne disease geography. J Med Entomol 2021; 58: 2021–9. doi:10.1093/jme/tjab086

³⁴ Fischhoff IR, Keesing F, Ostfeld RS. Risk factors for bites and diseases associated with black-legged ticks: a meta-analysis. Am J Epidemiol 2019; 188: 1742–50. doi:10.1093/aje/kwz130

³⁵ Ozdenerol E. GIS and remote sensing use in the exploration of Lyme disease epidemiology. Int J Environ Res Public Health 2015; 12: 15182–203. doi:10.3390/ijerph121214971

³⁶ Tricco AC, Lillie E, Zarin W, et al. PRISMA extension for Scoping reviews (PRISMA-SCR): checklist and explanation. Ann Intern Med 2018; 169: 467–73. doi:10.7326/M18-0850

³⁷ Arksey H, O’Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol 2005; 8: 19–32. doi:10.1080/1364557032000119616

³⁸ Jd G, Young I, Harding S, et al. A scoping review of Lyme disease research relevant to public health. CCDR 2018; 44: 243–56. doi:10.14745/ccdr.v44i10a03

³⁹ CADTH. Grey matters: a practical tool for searching health-related grey literature. 2022. Available: https://www.cadth.ca/grey-matters-practical-tool-searching-health-related-grey-literature

⁴⁰ Zhu A, Lu G, Liu J, et al. Spatial prediction based on third law of geography. Annals of GIS 2018; 24: 225–40. doi:10.1080/19475683.2018.1534890

⁴¹ Chan T-C, King C-C. Surveillance and epidemiology of infectious diseases using spatial and temporal Lustering methods. Infectious Disease Informatics and Biosurveillance 2011; 27: 207–34. doi:10.1007/978-1-4419-6892-0

⁴² Kapitány-Fövény M, Ferenci T, Sulyok Z, et al. Can Google trends data improve forecasting of Lyme disease incidence? Zoonoses Public Health 2019; 66: 101–7. doi:10.1111/zph.12539

⁴³ Wang J, Chen C, Zheng Z, et al. Predicting high-dimensional time series data with spatial, temporal and global information. Information Sciences 2022; 607: 477–92. doi:10.1016/j.ins.2022.06.021

⁴⁴ Watson SC, Liu Y, Lund RB, et al. A Bayesian Spatio-temporal model for forecasting the prevalence of antibodies to Borrelia burgdorferi, causative agent of Lyme disease, in domestic dogs within the contiguous United States. PLoS One 2017; 12: e0174428. doi:10.1371/journal.pone.0174428

⁴⁵ Covidence. Better systematic review management. 2021. Available: https://www.covidence.org/about-us-covidence/ [Accessed 11 Oct 2021 ].

⁴⁶ Poldrack RA, Huckins G, Varoquaux G. Establishment of best practices for evidence for prediction: a review. JAMA Psychiatry 2020; 77: 534–40. doi:10.1001/jamapsychiatry.2019.3671