Community Participation in Chagas Disease Vector Surveillance: Systematic Review

Background Vector control has substantially reduced Chagas disease (ChD) incidence. However, transmission by household-reinfesting triatomines persists, suggesting that entomological surveillance should play a crucial role in the long-term interruption of transmission. Yet, infestation foci become smaller and harder to detect as vector control proceeds, and highly sensitive surveillance methods are needed. Community participation (CP) and vector-detection devices (VDDs) are both thought to enhance surveillance, but this remains to be thoroughly assessed. Methodology/Principal Findings We searched Medline, Web of Knowledge, Scopus, LILACS, SciELO, the bibliographies of retrieved studies, and our own records. Data from studies describing vector control and/or surveillance interventions were extracted by two reviewers. Outcomes of primary interest included changes in infestation rates and the detection of infestation/reinfestation foci. Most results likely depended on study- and site-specific conditions, precluding meta-analysis, but we re-analysed data from studies comparing vector control and detection methods whenever possible. Results confirm that professional, insecticide-based vector control is highly effective, but also show that reinfestation by native triatomines is common and widespread across Latin America. Bug notification by householders (the simplest CP-based strategy) significantly boosts vector detection probabilities; in comparison, both active searches and VDDs perform poorly, although they might in some cases complement each other. Conclusions/Significance CP should become a strategic component of ChD surveillance, but only professional insecticide spraying seems consistently effective at eliminating infestation foci. Involvement of stakeholders at all process stages, from planning to evaluation, would probably enhance such CP-based strategies.


Introduction
Chagas disease still imposes a heavy burden on most Latin American countries, with about 10-12 million people infected by Trypanosoma cruzi [1,2]. Multinational control initiatives have since the early 1990s drastically reduced prevalence and incidence, mainly through insecticide-based elimination of domestic vector populations (blood-sucking bugs of the subfamily Triatominae) [3] and systematic screening of blood donors with highly sensitive serological tests [1,2,4,5]. In spite of these advances, vector-borne transmission is estimated to cause about 40,000 new infections per year [6]. Reinfestation of treated households by native vectors as the residual effect of insecticides vanishes is the most likely mechanism underlying such persistent transmission [7]. Similarly, outbreaks of acute Chagas disease have been attributed to the contamination of foodstuffs by infected adult (i.e., winged) triatomines that invade premises where food is processed or stored [8][9][10][11][12]. In Amazonia and other humid forest ecoregions, where the bugs rarely colonise inside houses, endemic, lowintensity transmission seems also mediated by adventitious, household-invading triatomines [13][14][15]. In addition, there is growing concern that insecticide-resistant vector populations, such as those detected in southern South America [16,17], may threaten effective disease prevention.
This rapid overview shows why sustained Chagas disease control is believed to require some sort of longitudinal, long-term surveillance system capable of detecting and eliminating household infestation foci [1,18]. Surveillance typically relies on the periodical inspection of households by trained personnel. Active vector searches are performed with or without the aid of chemical 'flush-out' agents such as low-dose pyrethroid dilutions, and infestation foci are eliminated by insecticide spraying when discovered [18].
However, detecting the vectors can be difficult, particularly when only small populations occur within or around households. In fact, vector colonies are expected to become rarer and smaller as control programmes proceed, and managers are progressively less prone to fund costly active surveillance resulting in few detection events. A number of vector-detection devices have been designed in an attempt to enhance surveillance; most consist of boxes that triatomines can use as refuges or of paper sheets or calendars where the typical faecal streaks of the bugs can be identified [19][20][21][22][23][24][25][26][27]. Such 'sensing devices' are placed within households or in annex structures and checked periodically for bugs or their traces, supposedly reducing the costs of surveillance while retaining adequate sensitivity [26][27][28][29].
Finally, and since the early vector control trials, there has been a perception that resident householders may have better chances of discovering bugs in their own homes than a visiting team searching the house for a few minutes every several months [30][31][32]. 'Community participation' in entomological surveillance gained extra momentum with the Declaration of Alma Ata [33,34], which ''…encouraged approaches to health care that incorporated community participation and community development'' (ref. [34], p. 1). Experiences involving community participation in Chagas disease control have been described in several settings across Latin America [18,30,31]; they seem to converge towards an encouraging overall picture, and the Chagas disease example has accordingly been praised in several subjective reviews (e.g., [35,36]).
However, the effectiveness of these diverse strategies for Chagas disease vector surveillance, including community participation, has not been thoroughly and objectively assessed at the continental scale. With the aim of filling this gap, we systematically reviewed the published evidence on this issue, tackling specifically the following major questions: (i) How common and important is the phenomenon of house reinfestation by triatomine bugs after control interventions?; (ii) How effective are different vector surveillance strategies at detecting infestation/reinfestation foci?; (iii) To what extent have community participation and empowerment been effectively promoted?; and, finally, (iv) Can available strategic options be condensed in overarching recommendations for surveillance that apply across the highly diverse ecological and social-cultural settings where the problem is present?

Methods
The review protocol is available upon request from the corresponding author. This review was carried out in the context of a collaborative project led by the Inter-American Development Bank, and was not formally registered. We searched Medline, ISI Web of Knowledge, Scopus, LILACS, and SciELO; the major query argument was ''Triatomin* AND (Control OR Surveillance)''. Searches retrieved records from 1948 to 2009, including additional documents identified by searching bibliographies and in the authors' records. This search strategy aimed at recovering documents describing vector control interventions, with or without surveillance, so that post-control reinfestation trends could also be assessed. Only documents describing field interventions aimed at the control and/or surveillance of domestic Chagas disease vectors were included in the full review process. Descriptive (nonintervention) reports, results of research with laboratory or experimental vector populations, expert reviews, and opinion or commentary pieces were either excluded or used only for the introduction and/or discussion.
We were particularly interested in comparing strategies involving institutional (by professional staff) or participatory surveillance. We also compared alternative methods for vector detection, including active searches, vector-detection devices, and community participation. Major outcomes included household infestation/reinfestation indices (or, in some cases, bug catches) and vector detection rates. Inclusion/exclusion of documents was assessed independently by ARdA and FA-F, and discrepancies resolved by consensus. Figure 1 presents the flow diagram of the review process. Data were independently extracted by ARdA and FA-F using predefined data fields inspired by the Guide to Community Preventive Services [37] (www. thecommunityguide.org) and including study quality indicators. FA-F revised data extraction results and resolved inconsistencies by rechecking the original documents. The following items were considered: (1) study classification (study design, intervention components, whether or not the intervention was part of a broader initiative, outcomes); (2) descriptive information, including (2.i) description of the intervention (what was done, how, where and by whom it was done, theoretical basis of the intervention, types of organisation involved, whether or not there was any intervention in a control group), (2.ii) study characteristics (place, time, population, settings, outcome measurement, whether or not there was a measurement of exposure to the intervention), (2.iii) results (primary results, sample and effect sizes), and (2.iv) applicability in settings other than the actual study one (direct and indirect costs, harms and benefits, implementation process, and whether the community participated at each stage of the process -design, pre-implementation, effecting, and evaluation); and (3) study quality, including quality of descriptions, sampling (universe, eligibility and selection of participants, sample size, potential sampling biases), effect measurements, data analyses (statistics, confounders, repeated measures or other sources of non-independence), and interpretation of results (rate of adherence, control and assessment of potential confounders and sources of bias). Relevant references and other details deemed important were also recorded. The protocol required extracting detailed demographic data about intervention and control or indirectly affected populations. Such information was however absent from or incomplete in most studies; this, together with the fact that the outcomes of primary interest refer to households, not individual people, led us to exclude these items from the protocol during the course of the review.

Author Summary
Blood-sucking triatomine bugs are the vectors of Chagas disease, a potentially fatal illness that affects millions in Latin America. With no vaccines available, prevention heavily depends on controlling household-infesting triatomines. Insecticide-spraying campaigns have effectively reduced incidence, but persistent household reinfestation can result in disease re-emergence. What, then, is the best strategy to keep houses free of triatomines and thus interrupt disease transmission in the long run? We reviewed published evidence to (i) assess the effectiveness of insecticide-based vector control, gauging the importance of reinfestation; (ii) compare the efficacy of programme-based (with households periodically visited by trained staff) and community-based (with residents reporting suspect vectors found in their homes) surveillance strategies; and (iii) evaluate the performance of alternative vector-detection methods. The results confirm that insecticide-based vector control is highly effective, but also that persistent house reinfestation is a general trend across Latin America. Surveillance systems are significantly more effective when householders report suspect bugs than when programme staff search houses, either manually or using vector-detection devices. Our results clearly support the view that long-term vector surveillance will be necessary for sustained Chagas disease control -and that community participation can substantially contribute to this aim. The often important morphological, ecological and behavioural differences among triatomine bug species [3], combined with the likely sensitivity of results to study-specific (methods, research team performance) and site-specific conditions (vector density, household building materials and structure), led us to avoid estimating meta-analytical summary effects from different reports. Inadequate design and/or reporting of several studies were further factors hindering meta-analysis. When enough information was given in the original reports, we nonetheless re-analysed data from studies comparing control strategies (in terms of household infestation rates) and vector detection techniques (in terms of detection rates). Whenever possible, we used McNemar's tests for correlated proportions [38], with odds ratios (OR) estimated as the ratio of discordant results. When independence of observations was likely, or in the absence of complete data on repeated observations, ORs were estimated from standard contingency tables [39]. Approximate OR 95% confidence intervals (95%CI) were calculated by assuming normality of log-odds [39]. The VassarStats online facility (http://faculty.vassar.edu/lowry/ VassarStats.html) and Microsoft Office ExcelH spreadsheets were used for the analyses.

Overall results
Database searches retrieved 1,342 candidate documents; elimination of duplicates yielded 858 unique records ( Figure 1 were also used for partial quantitative assessments, and the rest were considered as supplementary sources of qualitative information for the introduction and/or discussion. The spatial and ecological coverage of our review is represented in Figure 2. Only 11 randomised trials [40][41][42][43][44][45][46][47][48][49][50] were identified, with just one crudely assessing a community-based intervention [50] and four describing different aspects of the same trial [44][45][46][47]. Over half of the studies dealt directly or indirectly with different strategies for household-level vector surveillance. Interventions ranged from insecticide spraying (the most frequent) to educational activities, with a few studies describing alternative control approaches such as environmental management [51][52][53][54][55][56][57][58] or insecticide-treated materials [48,49,59]. Most studies measured intervention effects as reductions in household infestation rates (through entomological surveys) or as vector detection rates (through detection records). While the quality of the descriptions was generally adequate, analytical procedures were often dubious; for instance, albeit many studies describe results in which the same sampling units were assessed more than once (e.g., before-after, time-series) or by more than one method (e.g., vector-detection studies), only a few apply statistical tests suited for repeated measures or other sources of non-independence of observations.
Collaborative efforts involving both academic institutions and official public health agencies were common (,70% of studies), a typical historical trait of Chagas disease vector control [60]. Even though sustainability was discussed in several documents, detailed assessment of the costs (monetary and not) and potential unintended benefits and harms was rare. Forty-eight reports described some sort of 'community participation' in the intervention; however, none of them explicitly stated that participation took place at the design stage, and only three describe a participatory evaluation process [47,58,61]. In contrast, local residents helped carry out the intervention in 45 studies, mainly by reporting vectors caught in their homes; in 20, the community was also involved in the pre-implementation phase.

Control effectiveness and the role of surveillance
Since Carlos Chagas historic paper [62], vector control has become the cornerstone of primary Chagas disease prevention [60,63]. Pioneering attempts involved chemical (including cyanide gas) and physical means (including flamethrowers) [64]. The failure of DDT in controlling triatomines was followed by substantial optimism when HCH (lindane) proved successful in early trials in Brazil [65,66], Argentina [67], and Chile [68]. The effectiveness of insecticide-based control kept improving as new chemicals and better formulations, with longer residual effects and lower toxicity, were introduced [40][41][42]45,69,70]. Synthetic pyrethroids are now widely used and continue to be very efficient [71][72][73][74][75]; yet, recent research suggests that resistance may be widespread among some Triatoma infestans populations [16,17], and insecticides are less effective in peridomestic environments [43,76]. The top-quality report (in terms of sample size, design, and data treatment) we retrieved shows that peridomestic T. infestans foci reappear quickly after spraying (albeit with lower-density colonies) and that standard deltamethrin application with manual sprayers performs better than more sophisticated techniques [43]. Table 1 summarises the results of major reports on Chagas disease vector control [5,18,44,57,61,63,[71][72][73]. Overall, these studies unequivocally show that household insecticide spraying has successfully reduced infestation rates throughout Latin America, but also that reinfestation of dwellings by native vector species is common, spatially widespread, and temporally persistent. In many cases, the elimination of introduced populations was closely followed by the occupation of vacant niches by 'secondary' vector species, suggesting that the former had displaced the latter upon introduction [114,115].

Active bug searches versus participatory surveillance
For the purposes of our quantitative appraisal, we defined 'community participation' in Chagas disease vector surveillance as simply the involvement of local residents in reporting the presence of suspect bugs in their households. This narrow definition is justified by (i) the need to use some measure of effect size that is (at least qualitatively) comparable across studies, (ii) the fact that vector detection is the primary purpose of entomological surveillance, (iii) the fact than most 'participatory' experiences are limited to stimulating bug notification, and (iv) the principle of parsimony, whereby simpler approaches to surveillance, if they are shown to work, enjoy better chances of effectively translating into policy and practice. Table 2 shows the main results of studies quantitatively comparing the effectiveness of vector notification by residents with either active bug searches by control programme staff (the standard approach) or different vector-detection devices (e.g., 'sensor boxes') [32,85,107,[130][131][132].
With a few exceptions, notification by residents performs obviously much better than active bug searches at detecting infestation foci, although the effect seems to be somewhat smaller in the peridomestic environment [32,132] (Figure 3). Because notification costs less than active searches, these results are strong indication that it is probably much more cost-effective [20,116,133]. Vector-detection devices also seem to be largely outperformed by notification; the evidence is more limited in this case, but comparisons between detection devices and active searches (next subsection) suggest that notification by residents is also superior.

Vector-detection devices
Several 'passive' vector surveillance methods have been devised and tested over the years. As defined here, they differ from the traditional, 'active' surveillance approach in that control programme agents do not search the whole residence to determine whether it is infested; instead, they rapidly check for bugs (or their traces) in a 'detection device'. Table 3 summarises the main results of major comparative studies [20][21][22][26][27][28]130,132,[134][135][136][137][138][139][140]. In general, the sensitivity of vector-detection devices does not seem to be superior to that of active searches, but (i) both methods appear to complement each other, with only one of them revealing infestation in many instances (see also ref. [141]), and (ii) the costs of the passive approach are, in general, lower (but see ref. [28]).
Several studies with small sample sizes favour sensing devices, whereas the results of larger trials tend to show that they perform equally or worse than active searches (Figure 4). The evidence in relation to vector-detection devices remains therefore inconclusive, and further research is needed; below (Conclusions and outlook) we provide methodological suggestions to this end.

Discussion
In the long run, Chagas disease prevention will depend on keeping households free of T. cruzi vectors [60,116,142]. Insecticide-based control campaigns have been extremely successful, but there is compelling evidence that persistent reinfestation of a fraction of treated households is the pattern to be expected across Latin America; reinfestation, in turn, can result in disease transmission re-emergence [18,105,106,143,144]. These wellsupported findings clearly substantiate the view that long-term vector surveillance will be critical for the interruption of Chagas disease transmission [5,7,18,35,142,145,146].
Entomological surveillance primarily aims at detecting (then eliminating) household infestation foci; it thus allows for monitoring reinfestation trends in areas under control [5,92,94,95,[147][148][149][150][151]. This is of fundamental importance for both (i) eliminating residual foci of introduced species targeted for local eradication and (ii) keeping reinfestation by native species at levels below disease transmission thresholds [73,115,152,153]. We note, however, that 'native' vector species may be equally or more efficient than introduced ones at transmitting T. cruzi, and that even the most notorious 'primary' vectors, T. infestans and Rhodnius prolixus, are native (and reinfest treated households) [18,143,[154][155][156][157][158] in their original ranges. Thus, entomological surveillance has a major role to play in most of Latin America even after introduced vector populations have been eliminated; in areas under surveillance, rapid diagnostic tests could be used to discover residual or re-emergent transmission foci [142].
But in order to attain these goals, vector detection must be as effective as possible, and the evidence we have reviewed shows that available vector-detection techniques all work far from perfectly. What would be, then, the best strategy to meet the permanent challenge of detecting reinfestation? Our appraisal yields strong support to the view that notification of suspect vectors by residents is the most sensitive among the several detection approaches tested to date -and that it is also probably the cheapest. Furthermore, the difference in performance seems to widen as vector population density declines, which is the typical situation in post-control settings.
Such an austere 'participatory' strategy signals the minimum degree of community involvement required to effectively enhance surveillance: residents are just asked to report suspect insects found in their homes, and a response is mounted by professional staff, often related to decentralised health services [142,154,159,160] [4,35,162,163], but our review suggests that very simple interventions can be effective enough. Perhaps the main challenge here is to sustain community awareness in the face of even rarer infestation events; continuous education, a clearly defined channel for communication between residents and control agents, and an opportune response to any notification (including those involving insects other than triatomines) are probably the key to long-term success [35,73,152,159,[164][165][166]. This is not to say that more sophisticated approaches would not perhaps bring further benefits to people living under risk conditions. For instance, we found that most community-based experiences in Chagas disease vector surveillance are merely utilitarian, with little or no participation of the community in the design, planning, and evaluation of interventions. Effective involvement of all stakeholders along the whole process would no doubt foster true empowerment, and this could in itself result in improved health and living standards [33,34,[167][168][169][170][171]. Still, we underscore that, in the absence of adequate resources for comprehensive community-based programmes, stimulating vector notification by residents may suffice to boost the efficiency of entomological surveillance across highly diverse ecological and socio-economic settings.
Finally, our review revealed that there is plenty of room for improvement of both methodological and reporting standards in the Chagas disease control/surveillance literature. In many cases, the results were reported incompletely and/or confusingly, sometimes precluding data extraction; in several instances, the data in the text, tables, and figures were incongruent. Indeed, just a few of the reviewed studies followed high-quality designs (e.g., with some sort of randomisation) and used sound analytical approaches, particularly in relation to the non-independence of observations; these reports tended to rely on small sample sizes and/or have limited spatial scope. Apart from the obvious need for using adequate design and analytical procedures, several guidelines for good reporting practices are readily available (e.g., the STROBE statement [172]); researchers and journal editors share the responsibility of improving the standards of published reports on Chagas disease control and prevention.
Indeed, we believe that the main limitations of our review relate to the quality of the original reports, even if the breadth of our appraisal probably lightens the effects of individual study drawbacks. We did not test formally for publication bias, but deem it unlikely that any major study was overlooked; the possibility that such a bias exists should however be kept in mind when interpreting our results, particularly in relation to vectordetection devices. In an attempt to overcome possible study-level biases, we made every effort to extract and re-analyse the data in each document, without taking reported results at face value, but this does not alleviate design or data collection bias. However, we are confident that our main findings (that reinfestation by triatomines is common and widespread and that householder involvement in vector reporting enhances surveillance) are not bias-induced artefacts. We also note that our assessment focused on the initial stage of surveillance -the detection of infestation foci. The responses triggered by detection events, the monitoring of infestation trends, and the analysis and dissemination of epidemiological data are also essential components of disease surveillance [173], but their appraisal was beyond the scope of this review.

Conclusions and outlook
Entomological surveillance is and will remain crucial to contain Chagas disease transmission; yet, the zoonotic nature of the parasite's life cycle implies that eradication is unfeasible [1]. The enduring challenge of household reinfestation by locally native vectors can only be met by means of horizontal strategies -and these work better when the community takes on a protagonist role. Even very simple forms of participation, such as encouraging vector notification by residents, can substantially enhance the effectiveness of surveillance. Control programmes should therefore incorporate community-based approaches as a strategic asset from inception; such approaches must include a timely, professional response to every notification, and would very likely benefit from a strengthened focus on community empowerment.
It must finally be emphasised that, in practice, vector detection failures are unavoidable, particularly when bug population density is low [174]. It may then be argued that infestation rates are virtually always underestimated and that, because these rates are the foremost indicator used in decision-making [175], imperfect detection can seriously misguide Chagas disease control programme management. We consequently suggest that a critical area for future research relates to the reliable estimation of vector detection probabilities. This is somewhat more difficult in the absence of a 'gold-standard' technique, but by no means unworkable: repeated-sampling approaches [176][177][178] readily yield detection probability estimates (with confidence intervals) that can in addition be modelled as a function of covariates -such as, for instance, alternative detection methods, different fieldwork teams, different vector species, or physically diverse ecotopes. These approaches have been successfully applied in wildlife [179] Table 2 for details. doi:10.1371/journal.pntd.0001207.g003 Figure 4. Detection of Chagas disease vectors by vectordetection devices vs. alternative methods: estimated odds ratios and 95% confidence intervals. AS, active searches by vector control staff (ASfo, using a flushing-out agent; ASkd, using full insecticide application to 'knock-down' the bugs); DD, vector-detection devices (DDgn, Gó mez-Nú ñ ez boxes; DDmb, 'María' boxes; DDb, box; DDps, paper sheet; DDp, plastic boxes); (p), results in the peridomestic area; the reference number and sample size are indicated in parentheses; studies were ranked by mean effect size; effects are significant at the 95% level when the CI does not cross the dashed line; point estimate values .1 indicate a positive effect of the first method in the comparison; see Table 3   and disease ecology studies [180,181], and can also help enhance Chagas disease vector research [182].

Supporting Information
Abstract S1 Spanish and Portuguese translations of the abstract. (DOC) Checklist S1 PRISMA checklist. (DOC) List S1 93 documents submitted to full data extraction. (DOC) List S2 27 documents used in partial quantitative assessments but not submitted to full data extraction. (DOC)