Combined Epidemiologic and Entomologic Survey to Detect Urban Malaria Transmission, Guinea, 2018

Malaria incidence is generally lower in cities than rural areas. However, reported urban malaria incidence may not accurately reflect the level of ongoing transmission, which has potentially large implications for prevention efforts. To guide mosquito net distribution, we assessed the extent of malaria transmission in Conakry, Guinea, in 2018. We found evidence of active malaria transmission.


Study Site, Population, and Sampling Approach
All 5 communes of Conakry were entirely included in the consideration for study site selection with the exception of 3 islands off the coast, which are technically part of the city but rural in nature.
Routine malaria surveillance data from July-September 2018 were reviewed to identify the areas within each commune reporting the highest malaria incidences. Using these data, 2 geographically separate facility catchment areas were chosen within each of the city's 5 communes (10 urban areas overall) on the basis of elevated malaria incidence as well as input from local partners regarding feasibility and logistic concerns. Each catchment area selected served to provide the sampling units for all phases of the study involving community sampling (households, health facilities, and environmental evaluation for vectors). A similar procedure was conducted for the selection of 4 rural control sites within the rural prefecture of Dubréka just outside of Conakry.
At all chosen healthcare facilities, potential neighborhoods for household surveys were identified by soliciting input from healthcare workers and clinic officials to determine the specific locations contributing the most cases of malaria within the catchment area. The choice of final locations for study were the results of open discussions held between investigators and clinical staff, taking into account malarial disease burden, individual neighborhood accessibility, and overall feasibility.

Household Surveys
Within each study site, investigators randomly selected 30 households using a modified sampling method based on the WHO expanded program on immunization methodology (28). Surveys were administered at every third household. At each household, investigators used standardized questionnaires to collect information from the self-identifying head of household. Another adult member of the household was interviewed in lieu of the head of household if needed. All data, including number of household mosquito nets, number of nets hanging, and number of nets in good repair, were collected by participant response rather than surveyor observation. Surveys were designed to assess risk factors for local acquisition of disease as well as travel patterns of Conakry residents. Surveys were conducted and data were collected using electronic tablets loaded with SurveyCTO software version 2.50 (Dobility, https://www.dobility.com).
A maximum of 4 members per household were tested for the presence of Plasmodium falciparum-specific histidine-rich protein 2 (HRP2) antigen in the blood using rapid diagnostic tests (RDTs) (SD Bioline Malaria Ag P.f,; Abbott Laboratories, https://www.abbott.com). Children <5 years were prioritized for testing in order to participants with a limited travel history outside the city. If fewer than 4 children < 5 years were present in the household, testing was offered to other household members, first by offering testing to pregnant women, followed by randomization of older children and non-pregnant adults. If more than 4 children <5 years of age were present, 4 of them were chosen at random for screening. All individuals found to be positive for P. falciparum by RDT were given artemisinin-based combination therapy and instructions for use, in accordance with national guidelines.

Healthcare Facility Visits
Registers and routine surveillance reporting forms from each selected healthcare facility were reviewed. Investigators reviewed clinical registers from July-September 2018 to enumerate all-cause consultations, cases of febrile illness (defined as documented subjective fever and or temperature >38.5° C), cases of confirmed malaria, cases of malaria meeting WHO criteria of severity (Appendix Table 5) (29), and malaria-attributable deaths (death in the setting of documented severe malaria in the absence of alternative identified cause). These figures were compared to those reported to the NMCP as part of the monthly national malaria surveillance program.

Travel-Related Risk in Outpatients
Data regarding potential travel-related exposures were collected from persons seeking healthcare throughout Conakry from December 8-27, 2018. All patients visiting participating outpatient facilities for medical attention whose clinical workup included malaria testing were asked to disclose any travel outside the city within the 4 weeks before illness onset. Each facility collected data for 2 weeks using standardized paper forms. Clinical testing was performed in each facility according to its own diagnostic protocols and included a mix of microscopy and RDT. Data were subsequently transcribed to Microsoft Excel 2016 (Microsoft Corporation, https://www.microsoft.com) data files for analysis.

Potential Larval Habitat Enumeration
Mosquito larval surveys were conducted in Conakry by performing exhaustive searches for all accessible pools of standing water within delineated areas of roughly 50,000-90,000 m 2 . Search areas were chosen by soliciting healthcare worker and resident opinions on the area within each targeted neighborhood most heavily infested with mosquitoes, a method previously documented to be effective by Mwangungulu et al. (30) One survey field was defined for each commune of the city, resulting in a total of 5 urban sites for larval characterization (Appendix Figure 3).
Larval habitat surveys in Dubréka were performed using defined straight line transects to visualize all sources of standing water along the long axis of selected villages rather than within predefined areas. Subsequent, shorter transects were performed in each village at right angles to the initial path. Two of the 4 villages from Dubréka included in the overall study were chosen for potential larval habitat survey and characterization (Appendix Figure 4).
In all study locations, all collections of standing water were examined for the presence of mosquito larvae. Any mosquito larvae found were classified to the genus level by entomologists. Larvae were not quantified in either relative or absolute terms. Data were collected and stored on electronic tablets loaded with SurveyCTO software.

Adult Mosquito Collection
Adult mosquitoes were captured by human landing catches within each search area selected for larval habitat characterization. Mosquito collections took place continuously from 6:00 PM-7 AM on 2 consecutive nights at each study location. Collections were performed by 2 volunteers working simultaneously, 1 collecting mosquitoes outdoors and the other collecting mosquitoes inside the household of a neighborhood/village resident. Collectors were provided antimalarial prophylaxis free of charge. Adult mosquito collection sites were selected by entomologists to maximize potential nightly yield.
Adult mosquitoes were segregated by location of collection (indoors vs. outdoors) and hour of collection. All mosquitoes were subsequently sexed and morphologically identified to genus. Anopheles mosquitoes were morphologically identified to the species/group level.

Data Analysis
Data collected from the household surveys, healthcare facility visits, and entomological studies were subsequently analyzed using R version 3.5.0 (R Foundation for Statistical Computing, www.rproject.org). Data from each commune of Conakry were both statistically compared to one another and aggregated to compare Conakry as a whole against the rural control sites. Fischer's exact test was used for comparison of categorical variables, using an α of 0.05 modified using Bonferroni's correction for multiple comparisons when needed. Factors found to be statistically significant under the assumption of independence (i.e., no clustering) were used to build single variable mixed effects, random intercept models to assess the vigor of statistical significance while accounting for the spatial clustering of data.
Household survey data were additionally used to model P. falciparum antigenemia using generalized linear mixed effect models using the lme4 package in R (https://github.com/lme4/lme4). Predictors included in the multivariable, random intercept model were respondent age, city commune of residence, travel history, pregnancy status, and self-reported insecticide-treated net use.
Data collected from outpatient visits were used to calculate relative risks, 95% CI, and population-attributable fractions for potential travel-related exposures using Excel. Reported population attributable fractions were calculated using the formula (IP -Iu)/IP, where IP = total incidence of malaria at each health facility and IU = incidence of malaria among non-travelers at each health facility.

Ethical Considerations
Verbal informed consent was obtained from all participants of this study. Written informed consent form was obtained for each of those participating in the household survey and malaria prevalence screen by RDT. The signature of a parent or guardian was obtained for all children <18 years of age tested by RDT. The protocol for this study was reviewed and approved as a non-research activity by the Center for Global Health's Office of the Director at the Centers for Disease Control and Prevention (#2017-347a), as well as the Guinea Ministry of Health.
Healthcare-seeking behavior for febrile illnesses was largely similar across all study sites (Appendix Figure 6). Use of private medical facilities was more common in Conakry (16.7%, 67/402) than in Dubréka (3.0%, 6/198; p<0.001). However, there were no other significant differences in  Table 6; Appendix Figure 7).
The facility-reported data from both Dubréka and Conakry were found to be largely accurate when compared with registry records. All clinical sites in the study produced both surveillance reporting forms and clinical registers for comparison for all months requested. Of the 14 sites visited, 12 had reported malaria incidence data to within 10% of the true values listed in the clinical registers (Appendix Table 6, Appendix Figure 8). Notable discrepancies between the reported and registered sources of data were identified at 2 clinical sites. At both locations, both the reported number of consults and the reported number of malaria cases were incorrectly elevated during the period examined.

Potential Larval Habitat Enumeration
Potential mosquito larval habitats were sampled in 5 designated areas across Conakry and 2 linear village transects in Dubréka. A total of 187 potential larval habitats were identified within Conakry, and 28 were identified in Dubréka (Appendix Table 7; Appendix Figure 4). The types and proportions of larvae found in Conakry and Dubréka did not statistically differ from one another. No Anopheles larvae were found in Conakry; 1 Anopheles sp. larval habitat was identified in Dubréka. The most frequently identified mosquito larvae in both Conakry and Dubréka were Culex sp. (Appendix Table 7).

Heterogeneity of Transmission
Using the data we collected, we can calculate a crude approximation of the relative effectiveness of nets as a disease preventative in the locations studied. The ratio of nets needed to provide universal coverage to the number of locally acquired incident cases would be much higher in Conakry (11:1) as compared to Dubréka (1.2:1), where incidence is several-folds higher and risk of acquiring malaria is greater. Similarly, within Conakry itself, we noted considerable heterogeneity of the risk of locally acquired malaria. Using the PAR to adjust for the risk of malaria associated with travel outside the city, the incidence of locally acquired, clinically diagnosed malaria is 1.79-3.08 times higher in Kaloum than that in the other 4 communes of Conakry. This heterogeneity in risk of local malaria infection could have profound implications for the distribution of mosquito nets in the future. For instance, recalculating the nets required for universal coverage of the population of Kaloum and the number of estimated locally acquired malaria cases gives a ratio of 3.1:1. Thus, these data can be used to not only ascertain the need for net distribution in Conakry, but also to prioritize portions of the city over others in the event the campaign is unable to provide universal coverage. Of note, these calculations rely on a number of assumptions that our rapid study cannot ensure, thereby reducing their overall precision and accuracy.

Limitations
While our study quickly gathered actionable information crucial to the NMCP for use during its net distribution campaign, it does have important limitations. In order to achieve results in a timely fashion, some generalizability had to be sacrificed. The findings presented here summarize a brief snapshot in time, and as such, cannot be used to make broad generalizations of malaria transmission throughout the year. Although we were able to demonstrate the likely presence of ongoing autochthonous malaria transmission in Conakry, estimations of the relative contribution of local transmission to urban disease burden are subject to a seasonal variability that we are unable to document here.
In addition to the temporal restrictions on the generalizability of our findings, the study populations in both Conakry and Dubréka are not representative of the underlying populations overall because they were purposefully selected. Not only were those in the geographic areas of greatest risk consciously selected, but children <5 years were overrepresented in the household testing. Odds ratios and comparisons as calculated are thus not reflective of Conakry and Dubréka overall but are rather comparisons of the subpopulations of each in the areas with the greatest disease burden. We purposefully biased our sample to increase our ability to detect the presence of low levels of ongoing local malaria transmission at the expense of some degree of quantitative accuracy of the underlying population as a whole. More robust quantification of the risk of contracting malaria in Conakry requires more extensive sampling over a longer period.
Despite these limitations, our approach allowed us to describe the heterogeneity in urban disease burden and to provide an estimation of the risk of local disease acquisition in a relatively short period and within a relatively modest budget of $15,000. Moreover, in contrast to earlier studies, our study was conducted in the context of increasingly available routine data. Routine data were used to select study sites and were found to be generally reliable over the course of the study, thus providing evidence that routine data can be used for further monitoring of malaria risk in Conakry. The results of our study, while not broadly generalizable, yielded practical, actionable information in a timely fashion to directly influence use of disease prevention resources. 169 figure 1). Larval habitats in Dubréka sampled via transect, precluding calculation of density. ¶Numerators do not add to denominator due to larval habitats containing multiple genera.