A systematic review of published literature on mosquito control action thresholds across the world

Background Despite the use of numerous methods of control measures, mosquito populations and mosquito-borne diseases are still increasing globally. Evidence-based action thresholds to initiate or intensify control activities have been identified as essential in reducing mosquito populations to required levels at the correct/optimal time. This systematic review was conducted to identify different mosquito control action thresholds existing across the world and associated surveillance and implementation characteristics. Methodology/Principal findings Searches for literature published from 2010 up to 2021 were performed using two search engines, Google Scholar and PubMed Central, according to PRISMA guidelines. A set of inclusion/exclusion criteria were identified and of the 1,485 initial selections, only 87 were included in the final review. Thirty inclusions reported originally generated thresholds. Thirteen inclusions were with statistical models that seemed intended to be continuously utilized to test the exceedance of thresholds in a specific region. There was another set of 44 inclusions that solely mentioned previously generated thresholds. The inclusions with “epidemiological thresholds” outnumbered those with “entomological thresholds”. Most of the inclusions came from Asia and those thresholds were targeted toward Aedes and dengue control. Overall, mosquito counts (adult and larval) and climatic variables (temperature and rainfall) were the most used parameters in thresholds. The associated surveillance and implementation characteristics of the identified thresholds are discussed here. Conclusions/Significance The review identified 87 publications with different mosquito control thresholds developed across the world and published during the last decade. Associated surveillance and implementation characteristics will help organize surveillance systems targeting the development and implementation of action thresholds, as well as direct awareness towards already existing thresholds for those with programs lacking available resources for comprehensive surveillance systems. The findings of the review highlight data gaps and areas of focus to fill in the action threshold compartment of the IVM toolbox.


Introduction
There are about 3,500 identified mosquito species in different geographic/climatic regions in the world. Some species act as human disease vectors while other species cause only biting nuisance problems. Vector-borne diseases account for more than 17% of all infectious diseases, causing more than 700,000 deaths annually [1]. More than 80% of the global population is at risk of vector-borne diseases, with mosquito-borne diseases (MBDs) being the largest contributor [2]. MBDs such as dengue, Zika, chikungunya, West Nile virus (WNV), eastern equine encephalitis (EEE), western equine encephalitis (WEE), and malaria cause major public health issues across the world. The global burden of malaria has been estimated to be 219 million cases, with more than 400,000 deaths every year [1], and reported 627,000 deaths in 2020 alone [3]. More than 3.9 billion people in over 129 countries are at risk of contracting dengue, with an estimated 96 million symptomatic cases and an estimated 40,000 deaths every year [1]. WNV is the most common MBD in the United States [4]. A total of 21,869 confirmed or probable cases of WNV disease were reported across all 50 states during 2009-2018 [5]. Chikungunya and Zika have caused the average yearly loss of over 106,000 and 44,000 disability-adjusted life years (DALYs) worldwide, respectively, between 2010 and 2019 [6]. There is a significant economic burden of MBDs as well. The direct global costs of malaria have been estimated to be at least US $12 billion per year [7]. The estimated total global cost of dengue was US $8.9 billion in 2013 [8]. The average cost of WNV in the United States was equivalent to an approximate average of $56 million per year during 1999-2012 [9,10]. There is no specific drug or vaccine available for MBDs such as dengue, chikungunya, etc. [11,12]. Control of vector populations is often the best way, if not the only way, to control the transmission of these diseases. On the other hand, although the major transmission of diseases such as malaria and lymphatic filariasis is controlled by mass drug administration (MDA), the vectors of the diseases are still a human nuisance. For all those reasons, mosquito control is still an integral constituent of the public health infrastructure of many countries.
There are many illustrious instances in history that mosquito control has effectively reduced disease burdens; reduction of malaria in the 1950s-1960s in some countries (in the Americas, Europe and Asia), dengue and yellow fever in the 1950s-1960s in the Americas, dengue in the 1970s-1980s in Singapore and dengue in the1980s-1990s in Cuba [13]. However, rapid and aggressive changes in mosquito population dynamics imposed by changing influential factors such as the environment, climate, geography, demographics and social culture, have been making mosquito control more challenging in the modern-globalized world. Development of resistance against all major classes of insecticides, and a shortage of effective insecticide innovation, registration, and supply have become another threat to mosquito control [14]. Compounding all those complexities, mosquito control is often reactive [14] as it is implemented only in direct response to already increased risk based on vector or disease surveillance results in the present time [15]. The effects of such control programs would be too late to prevent undesirable levels of human-mosquito contact, and thus the potential for disease explosions. One of the key elements lacking in many mosquito control programs is the use of reliable preemptive indicators to initiate proactive control operations before mosquito populations or disease risks are elevated to undesirable levels. It is widely accepted that Integrated Vector Management (IVM) and Integrated Mosquito Management (IMM) programs that make reliable evidence-based control decisions and that utilize appropriate control tools, can effectively reduce mosquito abundance and disease risk [16]. Therefore, early warning systems with evidence-based action thresholds that trigger interventions at the optimal time are now becoming critical in operational mosquito control. In addition to being more effective in controlling mosquito populations, such systems would help ensure prioritized allocation of resources, manage insecticide resistance, and hence help manage overall program logistics.
Understanding the relationship between influential factors and the dynamics of mosquito populations and disease transmission through properly planned surveillance systems is fundamental in establishing evidence-based mosquito control action thresholds. The structure of surveillance systems and therefore the data collected by different mosquito control programs are widely varied due to differences in geography, climate, economy, and other logistic factors. Therefore, there could be different action thresholds for the same species or species groups in different geographic/climatic regions as well as in the same region. Given the critical importance of the use of action thresholds in mosquito control, we systematically reviewed the published literature to identify different mosquito control action thresholds existing across the world and associated surveillance and implementation characteristics. The findings will give insights to mosquito control program managers and others, who anticipate developing and establishing their own action thresholds, to plan appropriate surveillance programs. Otherwise, this review will be helpful for those that wish to utilize already developed action thresholds due to the inability to establish comprehensive surveillance systems, especially the routine collection of mosquito data.
from March 2020 to July 2021 using two main public search engines, Google Scholar (GS) and PubMed Central (PMC). The search terms were chosen to encompass a broad range of major vector species, diseases, surveillance/threshold measures, and analogies for "threshold". Primary search queries involved various combinations of key terms using (i) major vector mosquito genera ("Anopheles", "Aedes", "Culex"), followed by AND "mosquito" AND "[] threshold" ([]: risk, epidemic, disease, tolerance, action), (ii) [(disease name) control" (disease name: malaria, dengue, zika, chikungunya, West Nile virus) followed by AND "threshold" /"action threshold". (iii) "mosquito control", "mosquito abundance", "evidence based" with "threshold"/"action threshold"/"tolerance threshold", and (iv) "control" followed by AND "mosquito" AND some analogous "threshold" terms that were discovered at the initial planning and search stage such as "trigger"/"warning". For GS searches, all key terms were put in quotation marks to ensure a feasible number of search results (i.e. no more than several hundred). Quotation marks were also used for PMC searches; however, several PMC searches were conducted without quotation marks to instead broaden the search net after receiving any output of "no results found". Searches were independently conducted by two human reviewers (VSA, MRS) with separate search terms until all planned combinations were exhausted and no new links appeared on the first listed page of results, considered as our point of saturation. Search process and the inclusion/exclusion criteria are summarized in Fig 1. Any disagreement at each stage was resolved by in-person discussions.
Direct control thresholds/operational targets to keep mosquito populations below unaccepted levels of biting nuisance or disease transmission, risk level indicators to guide mosquito control, defined pest levels, and indicators of seasonal mosquito activity were identified as mosquito control action thresholds in this review. Retrieved publications were divided into three inclusion categories to reflect the various circumstances an action threshold was implicated in them; (i) "generated thresholds"-publications with an originally generated action threshold, (ii) "generated-models"-publications with a statistical model that involved a computable trigger/risk assessment with input of dynamic data to indicate if the area under study is over a threshold or within a risk level, and (iii) "mentioned thresholds"-publications with clear mention of a previously generated action threshold/s when the cross-references cited in them were not eligible to be included in (i) or (ii). Within those three categories, two types of action thresholds were identified: "entomological thresholds"-the threshold directly indicates mosquito abundance without connecting it to an epidemiological indication (EI) of a disease (could be associated with either vectors or non-vector nuisance mosquitoes), and "epidemiological thresholds"-the threshold is directly related to an EI (transmission, outbreak, epidemic). Calculation methods of EIs greatly differed between publications, some calling for flexibility dependent on the choices of local government bodies, others using specific calculations. Definitions for EI that were defined and/or justified in publications are included alongside action threshold descriptions in supplemental data tables (S1, S2 and S3 Tables). Other data points extracted during the full-text screening included publication type (scientific journal, guidelines, theses/dissertations, etc., geographic origin of the threshold/s (continent, climatic region), spatial unit of analysis (study area size), spatial unit of implementation (suggested/planned area associated with threshold), origin of data (self-sampled data, previously recorded data), time span of used data, data frequency (temporal resolution used data), mosquito species in concern, disease in concern, variables used in the threshold, types of mosquito data (e.g. larval indices, adult trap counts), data frequency, associated proactive lead time from the exceedance of the threshold to its indication, self-assessment of the validity of the threshold, and intended use of the threshold.

Results
GS and PMC produced 1,559 and 2,322 search results, respectively, between the two human reviewers and search query combinations. The quantitative value of unique search results was revealed by the removal of replicates through the compilation of all citations from recorded search queries. Replicate overlap between the two search engines was minimal and the total number of initial unique search manually screened for abstracts was 1,485. The abstract screening reduced the initial unique 1,485 publications to 233 publications for the next screening (i.e., reading the full text) and data extraction process. Combined with 4 cross-references found from publications that mentioned previous thresholds and that met all the inclusion criteria, the final screening identified 87 publications to be included in the review (inclusions, hereafter) (Fig 2).
There were 30 inclusions with "generated thresholds" (hereafter GTs, S1 Table) [18-47], 13 inclusions with "generated-models" (hereafter GMs, S2 Table) [48-60], and 44 inclusions with "mentioned thresholds" (hereafter MTs, S3 Table)  . Some of the MTs were evidencebased generations of previously (before 2010) published studies while a few did not discuss how they were generated. As per the respective inclusions, they were still in use and therefore, worthwhile looking into available relevant data points. Outcomes of extracted data points were presented as percentages of all inclusions or GTs/GMs and MTs as appropriate.

Publication type
The majority of inclusions were retrieved from scientific journals (89.7% [78/87]

Geographic origin of the threshold
Inclusions with thresholds were identified in broadly defined two climatic regions: tropical/ sub-tropical (

Targeted mosquito species and diseases
Aedes aegypti Linn. and/or Aedes albopictus Skuse were the most reported mosquito species In parallel with the most reported mosquito species, dengue was the most reported disease ( [58]. Zhang et al. [57] stood out because the climate and epidemiological surveillance parameters of one city were modeled to predict the extension of disease transmission risk to a neighboring city. There was a unique GM that approached utilizing social media platform data -Twitter tweets [55].

Statistical procedures used in GMs
All GMs incorporated a threshold as the dividing value for a binary classification system, i.e., the model output-estimated mosquito count [ [59,60] or developed a risk-assessment framework involving stochastic simulations [49]. A series of inclusions of the same author group detailed the advancing development of a Bayesian spatio-temporal hierarchical model to predict dengue risk in Brazil by microregion [52,53], later tailored to proactively forecast risk for the 2014 World Cup [54], and finally validated after the tournament's conclusion [64]. The advanced model was also implemented for the island country of Barbados [51]. Interestingly, the advanced model incorporated a pre-defined threshold of cases and a probability threshold that could be optimized to reduce false warnings while aiding policymakers to effectively allocate control resources and activities, a priority also discussed in another GM [49].

Surveillance and implementation characteristics
Associated surveillance and implementation characteristics reported by GTs/GMs (n = 43) were analyzed. Analysis of MT data were included whenever available. Species-wise analysis was not appropriate due to the very low number of inclusions of other species except Ae. aegypti/Ae.

Intended use of the threshold
Two broad categories of intended uses were identified from the inclusions. Direct intentions included those that stated direct utility for initiating mosquito control operations by setting an operational target threshold to keep populations below unaccepted levels of biting nuisance/ disease transmission and others that determined different risk levels to guide a mosquito control response and/or effective resource allocation. Indirect intentions involved inclusions that mentioned mosquito/vector control at least once but discussed the utility of the action threshold within the wider expanse of a public health response without a specific discussion of focused implementation in mosquito control. The

Discussion
Although the establishment of action thresholds has become critically important to achieve the maximum cost-effective benefit of mosquito control interventions, there is no formulated guidelines on the development of thresholds in different environmental settings. This review attempted to identify publications with mosquito control action thresholds across the world during the last 10 years and the basic surveillance and implementation characteristics associated with those thresholds. We hope the findings of the review will help guide mosquito control program managers and any other personnel who are interested in developing/using action thresholds to be included in their programs.
The 87 inclusions of the review were categorized into three broad categories; originally generated thresholds, generated models, and mention of previously generated thresholds. General information and surveillance/implementation characteristics of the three categories were examined to understand the structure of their surveillance systems.
More than 90% of the inclusions were from scientific journals/conference proceedings indicating the scientific/academic interest on developing action thresholds. However, the very low number (6.8%) of inclusions from program guidelines/technical reports could not be considered due to the scarcity of thresholds being used practically in mosquito control programs. There could be established thresholds that have already been in the specific IMM toolboxes of mosquito control programs but not disseminated in publications or in publications that fall into our inclusion criteria. We have planned and already initiated to identify those action thresholds through individual discussions with program managers/authorities in another study.
Overall, 77% of identified thresholds were "epidemiological thresholds" and thus, indicate that the need to establish mosquito control action thresholds has been driven mainly by the transmission dynamics of mosquito-borne diseases rather than by biting nuisance. The majority of epidemiological thresholds here were to indicate transmission of disease although there is a chance of overlapping as the used epidemiological indication terms were not based on strict definitions. It is evidence for the need of more promising proactive mosquito control as discussed by Eisen et al. [105] rather than waiting for the transmission to transform into outbreaks or epidemics. Although mosquitoes are widely distributed in different climatic regions in the world, they thrive in regions with warm temperatures, humid conditions, and high rainfall, and thus tropical and sub-tropical areas are ideal for their survival [106]. Most of the identified action thresholds (>70%) being generated in tropical and sub-tropical climatic regions indicate their relevance to the burden of mosquitoes and mosquito-borne diseases in those regions, compared to others. The fact that the highest number of thresholds, were originated in Asia and associated most with dengue and dengue vectors should have a link with recent increases in the incidences of major dengue outbreaks and epidemics reported mainly in the Asian region [107,108]. However, recent studies show the geographical distribution of Ae. aegypti and Ae. albopictus is now extensive in all human inhabited continents; Europe and North America [109], South America [110], Oceania [111,112], and Africa [113].
Overall, the majority of inclusions have used mosquito data in thresholds. However, some of the recently developed entomological thresholds have moved on to use climate data, mainly temperature, rather than using mosquito data itself as done previously (before 2010). It is a good step forward to proactively inform the need to initiate mosquito control and thus help bring about more effective prevention of unacceptable levels of mosquito abundance. Despite controversy on the association with the number of disease cases [36,41,[114][115][116][117][118][119][120][121][122], the majority of epidemiological thresholds for dengue and chikungunya were generated using Stegomyia larval indices.
The data collection duration is a substantial characteristic in the process of action threshold development. A sufficient time span would allow identifying any time sensitive variations in the identified correlations, thus in the generated action thresholds. Therefore, thresholds developed using historic data would be more reliable than those generated using only current data or one-time data. Davis et al. [123] suggested the inconsistent reliability and questionable validity of the thresholds generated in the study by using <3 years data would be refined with additional years of data. Using <5 years' historic data in their study, Bowman et al. [124] suggested that data of a greater number of historic years would have generated more reliable results. Other studies which reported alert thresholds for malaria based on disease case numbers discuss the need of at least 5 years of historical data [125,126]. The majority of studies reviewed here being used carried out over more than 5 years have allowed the possible temporal variations to be considered and compromised. Studies conducted in less than 2 years might not be able to catch time sensitive variations in the variables used thus making the threshold less reliable to use over the years. Likewise, mosquito abundance and disease transmission dynamics may change over time due to many reasons such as environmental/climatic changes, human mobility, etc. Most of the "mentioned thresholds" being developed more than ten years ago might need assessments and necessary updates before using in current mosquito control programs. Notably, almost all identified studies have used external data sources mainly from relevant government entities to generate thresholds. The use of external data sources requires careful selection and matching of data in the temporal and spatial units in concern to generate reliable thresholds. The distribution of mosquitoes and hence the mosquito-borne diseases are clustered in nature based on different ecological niches they use [127]. However, the demarcation of spatial units of those studies is based on man-made administrative boundaries rather than on ecological or geographical determinants. Therefore, the use of a particular action threshold in the spatial unit of its origin is more reasonable than using it in an area with different geographical, climatic, environmental, demographic, socio-economic, and other relevant characteristics. Stegomyia house index (HI) of less than 1% indicates no risk of a dengue outbreak in Brazil [128] but outbreaks occurred in Singapore when the national overall HI was less than 1% [129]. Researchers have shown that the geographical level heterogeneity affects the performance of thresholds with the thresholds developed for smaller units performing best [120]. The thresholds varied among different spatial units such as different districts in the same country [37,41] as well as among different climatic zones in the same country [39]. Romiti et al. [130] report a significant difference in Ae. albopictus development threshold temperature among municipalities. In contrast, some studies report the same threshold for the seasonal emergence of Ae. albopictus in two countries [18,47]. Bowman et al. [124] report the predictive ability of certain meteorological and epidemiological alarm variables of dengue across all countries studied. Hence, there could not be universally reliable thresholds and there is no definitive fact that the same threshold cannot be used elsewhere. The majority of the publications analyzed here have action thresholds intended to be used in the same spatial unit as its origin. However, locations with similar determinant characteristics but yet unable to collect required surveillance data would need to consider using an already developed appropriate threshold after confirming with initial environmental assessments/validations.
Temporal resolution of data collection and analysis plays an important role in the development of reliable action thresholds. Smaller scale resolutions (e.g. weekly data) would reveal more precise relationships between variables compared to higher scale resolutions (e.g. monthly/yearly data) [131]. Realistically, the available resources and other logistics will be deterministic factors of the temporal resolution of data collection. However, most of the studies included in this review achieved at least a monthly resolution.
One of the main purposes of establishing mosquito control action thresholds is to provide sufficient time for planning, organizing, and implementing appropriate control activities at the optimal time. Sufficient lead time from the established threshold to the exceedance of its indication would help mosquito control program managers to plan, organize and implement appropriate control measures for effective reduction of mosquito populations. Entomological thresholds seem to have shorter lead time (mentioned in only 2 studies) than epidemiological thresholds. Epidemiological thresholds which used climatic variables have longer lead times than those using mosquito data variables. Hii et al. [132] discuss the importance of the optimal lead time which would give sufficient time for successful implementation of mosquito control. They suggest that an average of 2 months with maximum 3 months lead time is typically required for effective mitigation of dengue transmission in Singapore in both non-epidemic and epidemic years. The development of a reliable mosquito control action threshold/s is a very complex process due to the interplay of several tangential factors, such as frequency and amount of virus importation, and herd immunity [133] and human-mosquito contact rates [134]. However, those thresholds should be accurate enough not to provide too many false warnings which would lead to issues such as wastage of resources and development of insecticide resistance. Validation of generated thresholds is important before establishing to confirm their performance in operational settings. Validation studies have confirmed the effectiveness of thresholds generated in the same study [36] as well as previous thresholds [49]. After field validation of previous thresholds by Sanchez et al. [120], the performance of overall neighborhood Breteau Index (BI) � 1.3 was denied and the most straight forward performance of BI max � 4 was confirmed as an indicator for predicting dengue transmission [48]. MacCormack-Gelles et al. [51] showed that the HI threshold imposed by the Brazilian Ministry of Health was not in compliance with increasing dengue risk in Fortaleza, Brazil. Most of the thresholds included in this review still need to be field validated, before establishing to inform the need of initiation of mosquito control activities.

Limitations
Pertaining to the restrictions of available time and resources, the systematic review has several limitations which might have eliminated some important publications: (i) the literature search was limited only to two search engines (Google scholar and PubMed), (ii) the search was limited to 2010-2021, (iii) language restricted to English only, and (iv) grey literature not used.

Conclusions
• Identification of (i) different action thresholds in existence for different mosquito species and different related diseases in different countries (S1,S2 and S3 Tables), and (ii) the characteristics associated in generating them will help interested mosquito control program managers and others to better plan their surveillance systems for required data collection.
• The results give an insight to the possibility of moving on to climatic variables rather than mosquito data variables itself to generate action thresholds which would give longer lag time for the required preparations to implement control activities.
• The review uncovers (i) the gap in the literature on action thresholds for mosquito nuisance, other mosquito-borne-diseases than dengue in Asia, mosquito-borne diseases in other geographic regions and (ii) the need to update previous action thresholds in space and time.  Table. Generated models for mosquito control action thresholds ("generated-models") found in publications included in the review.