Background: The transmission of malaria through population inflows from highly endemic areas with limited control efforts poses major challenges for national malaria control programmes. Several multilateral programmes have been launched in recent years to address cross-border transmission. This study assesses the potential impact of such a programme at the Angolan–Namibian border. Methods: Community-based malaria prevention programmes involving bed net distribution and behaviour change home visits were rolled-out using a controlled, staggered (stepped wedge) design between May 2014 and July 2016 in a 100 × 40 km corridor along the Angolan–Namibian border. Three rounds of survey data were collected. The primary outcome studied was fever among children under five in the 2 weeks prior to the survey. Multivariable linear and logistic regression models were used to assess overall programme impact and the relative impact of unilateral versus coordinated bilateral intervention programmes. Results: A total of 3844 child records were analysed. On average, programme rollout reduced the odds of child fever by 54% (aOR: 0.46, 95% CI 0.29 to 0.73) over the intervention period. In Namibia, the programme reduced the odds of fever by 30% in areas without simultaneous Angolan efforts (aOR: 0.70, 95% CI 0.34 to 1.44), and by an additional 62% in areas with simultaneous Angolan programmes. In Angola, the programme was highly effective in areas within 5 km of Namibian programmes (OR: 0.37, 95% CI 0.22 to 0.62), but mostly ineffective in areas closer to inland Angolan areas without concurrent anti-malarial efforts. Conclusions: The impact of malaria programmes depends on programme efforts in surrounding areas with differ - ential control efforts. Coordinated malaria programming within and across countries will be critical for achieving the vision of a malaria free world. Keywords: Malaria, Angola, Namibia, Trans-Kunene malaria initiative, Cross-border, Spillover Background deaths worldwide in 2016, much of which occurred Between 2010 and 2016, global malaria incidence among children under-five [ 1]. declined substantially from 76 to 63 cases per 1000 peo- The majority of the current malaria mortality burden is ple at risk . However, the burden of malaria remains borne by countries in sub-Saharan Africa, where falcipa- high in many areas, with an estimated 445,000 malaria rum malaria is most common [1, 2]. Malaria control in many parts of this region has been challenging because of the stability and intensity of malaria transmission , emerging insecticide resistance , and increasing popu- lation movements between high- and low-endemic areas *Correspondence: email@example.com Department of Global Health and Population, Harvard T.H. Chan School [5–7]. of Public Health, 655 Huntington Avenue, Boston, MA 02115, USA Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Khadka et al. Malar J (2018) 17:224 Page 2 of 13 Human mobility across national borders is particu- malaria prevalence among children under-five was larly challenging from a logistical and political perspec- reported to be 1% or less in Cunene and Namibe prov- tive. Most malaria programmes are run through national inces, which was in stark contrast with the neighbour- offices, which try to optimize resource allocation and ing province of Cuando-Cubango where prevalence was impact within countries, but tend to have limited capac- reported to be approximately 38% . Access to preven- ity to affect or implement malaria programmes in areas tive measures such as insecticide-treated nets, indoor outside their borders. Even though several cross-border residual spraying, and malaria treatment was particularly malaria control programmes have been launched in low in the two Angolan provinces in the TKMI. recent years [8–10], little is known regarding the effec - The three TKMI regions in Namibia have sustained tiveness of these programmes in general, and evidence on levels of malaria receptivity and account for a major- the relative impact of coordinated cross-country efforts ity of malaria cases reported in the country [13, 14]. In remains largely lacking. contrast, the south of the country is malaria free while This study assesses the effectiveness of a cross-bor - the incidence of malaria in the central regions is low der community-based malaria prevention programme . Overall, malaria incidence and mortality has been launched at the Angolan–Namibian border in 2012. The decreasing in Namibia, which is largely attributable to programme was launched as part of the Trans-Kunene increased distribution of long-lasting insecticide-treated malaria initiative (TKMI), which was an agreement bed nets (LLITNs), and improved access to malaria treat- between the governments of Angola and Namibia to ment . develop, among others, an evidence base for cross-bor- der malaria control strategies and facilitate the sharing Community‑based malaria prevention intervention of technical and scientific information between the two The malaria prevention programme was conducted by countries to strengthen malaria transmission and con- community-based volunteers and involved the distri- trol initiatives. To allow for a rigorous evaluation of the bution of LLITNs and behavior change programming. programme, a controlled, staggered (stepped wedge) roll- The volunteers did not receive any monetary incentives; out plan for a 100 × 40 km corridor along the border was instead, they were given an end-of-year food basket, developed and implemented between 2014 and 2016. The TKMI t-shirts, and light refreshments on the day of net average programme impact was estimated in a first step, distribution. followed by an estimate of the extent to which simulta- For bed nets, community volunteers first compiled a neous cross-border programming modified the relative listing of all inhabited structures in programme villages impact of the malaria prevention programme. independent of construction type (households). Follow- ing this, they delivered and assisted with the installation Methods of rectangular, World Health Organization Pesticides Study population Evaluation Scheme (WHOPES) approved LLITNs in all The malaria prevention programme was implemented listed households. Each household received one LLITN in all villages in a pre-specified 100 × 40 kilometre cor- per sleeping space. Informal sleeping spaces outside ridor along the Angolan–Namibian border in the Trans- households were not considered during distribution Kunene region. The Trans-Kunene region comprises as all families sleep inside their homes at night because of two Angolan provinces—Cunene and Namibe—and of proximity to the land they work in. The distributed three Namibian regions—Kunene, Ohangwena, and LLITNs have an expected lifespan of 3–4 years and Omusati. Target districts for the malaria prevention have been demonstrated to retain effectiveness up to 20 programme was jointly decided upon by the Ministry of washes . Systematic reviews have also shown that Health in Angola and Namibia. Following this, the inter- insecticide treated nets can reduce incidence of Plasmo- vention corridor was defined by the study team under dium falciparum malaria episodes by 50% . the assumption that most cross-border infections would Behaviour change programming involved community- occur in this spatial area as movement across the border based volunteers making monthly visits to households to happens primarily on foot or bicycle. The intervention provide guidance on the use and maintenance of LLITNs. corridor encompassed Cunene province in Angola, and In addition, they also provided important malaria pre- Ohangwena and Omusati regions in Namibia. Additional vention information such as strategies for eliminating file 1 highlights the corridor in the Trans-Kunene region standing water. in which the programme was implemented. The malaria prevention programme followed a two- The two Angolan provinces in the TKMI represent phased block rollout schedule between 2014 and 2016. areas of lower and unstable malaria endemicity in com- During Phase I, which occurred between 2014 and 2015, parison to the rest of the country . In 2015–2016, all 35 Namibian villages in the intervention corridor Khadka et al. Malar J (2018) 17:224 Page 3 of 13 as well as 50% of purposely selected areas in Angola to the distribution of bed nets. Most of the questions in the received the treatment. As illustrated in Additional file 2, surveys were adapted from the Demographic and Health the Angolan corridor was divided into four approxi- Surveys, with some additions made by the study team. No mately equally sized zones. For the Phase I rollout, two of formal validation was done for the additional questions. these zones were selected to receive the treatment while Each survey consisted of a general household and the remaining zones were used as control areas. There knowledge module and an under-five child module were no buffer zones between these areas as the primary focused on fever incidence and treatment. Due to budg- objective of this study was to measure treatment effect etary and logistical constraints, malaria status of children spillovers. Instead, rollout was blocked, so that we could under-five was not measured or verified by rapid diag - directly measure spillovers in Phase I villages based on nostic tests (RDTs). A primary respondent, typically a whether they bordered areas targeted in Phase II. female household member, provided responses to ques- The rollout of bed nets was centrally controlled and tions in both modules. Of the 2184 surveys attempted coordinated across both countries. Thus, during Phase over the three rounds, 97% were completed. The most I, the delivery and installation of nets occurred at the common reasons for non-completion were refusals and same time on both sides of the border. Villages that did temporary unavailability of families. not receive the treatment in Phase I received it in Phase II, which was conducted between 2015 and 2016. In both Outcome measures phases, most bed nets were distributed and installed The primary outcome of interest was caregiver-reported before the peak rainfall months of February and March, child fever. As part of all surveys, respondents were asked which roughly coincides with the period of highest to first list all children in their household under the age malaria burden in the TKMI area. of five, and then indicate fever episodes and treatment- seeking in the 2 weeks preceding the survey. Sampling The secondary outcomes analysed were malaria knowl - A random sample of 64 villages in the 100 × 40 km cor- edge, LLITN utilization among children under-five, and ridor was selected for evaluation prior to programme household LLITN ownership. Malaria knowledge was rollout. The sampling of villages differed by country: in evaluated by asking each respondent 20 true/false ques- Namibia, a complete list of all 35 village names in the tions on the nature, consequences, and treatments for intervention corridor was provided by the local Min- malaria. For the empirical analysis, the percentage of istry of Health and a random sample of 26 villages was correct responses was computed and converted into a selected from this sampling frame to get a sample size of z-score to facilitate interpretation of the estimated coeffi - at least 250 households. In Angola, the study team con- cients. To determine child LLITN utilization, respondent ducted an independent listing of all 38 villages in the reports of whether each listed child slept under a LLITN intervention corridor at the beginning of the study. Due on the night before the survey were used. Household to the larger study team on the Angolan side, all 38 vil- LLITN ownership was defined as the number of LLITNs lages were surveyed. owned by a household and was determined based on All study villages in the intervention corridor were respondent reports as well. The complete English version situated in rural areas, with most villages comprising of of the survey questionnaire is included in “Appendix 1”. 50–100 households. In each of the 64 study villages, a complete household listing was made prior to each sur- Statistical methods vey round. A random sample of 10% of households was Summary statistics of key household, respondent, then surveyed in each village in each round. and under-five children characteristics as well as pri - mary and secondary outcomes of interest at base- line were computed. To assess the effectiveness of the Data collection malaria prevention programme, data across the three All 64 study villages were surveyed three times over the waves were pooled and multivariable linear and logis- evaluation period. Additional file 3 illustrates the roll- tic regression models that controlled for survey-round out of the evaluation surveys and the TKMI interven- fixed effects were estimated. Additionally, the overall tions. The first survey was conducted between May 2014 implementation period was classified into an imme - and July 2014 (baseline) prior to intervention launch. The diate period, i.e., the time period within 1 month of second survey was conducted after Phase I in September intervention implementation, and a 1-year follow-up 2015 (midline). The final survey was conducted after the period. Programme impact was evaluated over both completion of Phase II in July 2016 (endline). Additional timeframes as well. To determine if programme impact file 4 contextualizes the rollout of the surveys in relation Khadka et al. Malar J (2018) 17:224 Page 4 of 13 on child morbidity varied by age or gender, sub-group estimated. For the stratified regressions, the 20 km width regression analyses were conducted. of the Angolan half of the corridor was divided into four Two distinct sources of variation were used to assess roughly equally sized strata, and households within 5, 10, the extent to which programme impact depends on 15, or 20 km of the border were considered separately in coordinated cross-border programming (spillover each model. effects). For Namibian villages, variation in programme To account for correlated outcomes within households implementation in Angola during Phase I was exploited and study villages, standard errors in all regression mod- to conduct a standard difference-in-differences analy - els were estimated using Huber’s cluster-robust variance sis, interacting Phase I impact in Namibia with Phase I estimator . This procedure allows correct inference programme activities in Angola. To allow for both lin- independent of model specification. Missing responses ear and multiplicative interactions, linear probability were excluded from the calculation of descriptive statis- and logistic regression models were estimated. In both tics and coefficient estimates in the regression models. models, the main hypothesis tested was that Namib- ian villages adjacent to Phase I Angolan villages would Software experience greater programme impact than villages All statistical analyses were conducted in Stata/MP 15.0 adjacent to Phase II Angolan villages at midline. Fig- . ArcMap 10.3.1 was used to create maps and calcu- ure 1 illustrates this comparison. late kilometre distances between villages . For Angolan villages, spatial variation in Phase I treat- ment status was not available since all Namibian villages Ethical considerations received treatment in Phase I. However, since malaria This study was approved by the Institutional Review programmes were largely absent north of the TKMI cor- Board at the Harvard T.H. Chan School of Public Health, ridor in Angola, programme villages close to Namibia the Ministry of Health and Social Services, Republic (with extensive malaria control) can be compared with of Namibia, and the Comité de Ética do Ministerio da Angolan villages closer to the rest of Angola (with limited Saude, Ministério da Saúde, República de Angola (Ethics programme efforts). To evaluate if distance to the bor - Committee of the Ministry of Health, Ministry of Health, der modified programme impact in Angola, a three-step Republic of Angola). process was followed: first, kilometre distance between the border and village centroids was computed; second, Results data were restricted to Phase I Angolan villages; and A total of 3844 child records across 2126 household sur- third, linear and logistic regression models interacted veys were analysed. with distance as well as stratified regression models were Fig. 1 Map demonstrating the grouping of areas in Namibia for the difference-in-difference analysis. The horizontal line in the middle of the map represents the Angola–Namibia border. The labels Cunene, Omusati, and Ohangwena represent the Angolan province and Namibian regions encompassed by the intervention corridor. The double lines represent major road networks in the programme area. Dark gray highlights areas in Namibia that are adjacent to Phase I Angolan areas. Light gray highlights areas in Namibia that are adjacent to Phase II Angolan areas Khadka et al. Malar J (2018) 17:224 Page 5 of 13 Baseline evaluation countries. At baseline, the average household had nine Table 1 presents key baseline characteristics of house- individuals of whom two were children under 5 years of holds, respondents, and under-five children across both age. Sixty percent of respondents were female and, on average, were 56 years old and had completed education up to the fourth grade. In terms of study outcomes at baseline: 23% of children Table 1 Characteristics of households, survey under-five were reported to have had fever in the 2 weeks respondents, and children under-five at baseline (2014) prior to the survey; the average household owned 0.77 Angola LLITNs and had 14% of sleeping spaces covered with nets on and Namibia the night before the survey; 14% of children under-five were Household characteristics reported to have slept under a LLITN on the night before Number of households 740 the survey; finally, respondents correctly answered 60% of Mean household size (N = 739) 9.01 (6.21) the malaria knowledge questions. Additional file 5: Table S1 Mean number of children under-five (N = 740) 1.91 (1.58) shows heterogeneity in outcomes at baseline by village. Mean number of LLITNs in household (N = 713) 0.77 (1.39) Mean number of LLITNs used last night (N = 713) 0.60 (1.13) Mean number of sleeping spaces in households 4.54 (2.56) Programme impact evaluation (N = 735) Table 2 presents results from multivariable logistic regres- Percentage of sleeping spaces covered with bed net 14% sion models estimating average programme impact on on night before the survey (N = 704) child fever. Column 1 shows the overall estimated impact; Percentage of households that had visitors from across 43% columns 2–5 show estimated impact by age and gender. the border stay the night in the last month (N = 727) The programme reduced the odds of fever by 54% Percentage of households with individuals who have 44% stayed overnight across the border in the last month (aOR: 0.46, 95% CI 0.29 to 0.73) over the 2 year inter- (N = 732) vention period. Among children under 2 years, the pro- Respondent characteristics gramme reduced the odds of fever by 71% (aOR: 0.39, Percentage of female respondents (N = 727) 60% 95% CI 0.23 to 0.65). In comparison, among children over Mean age of respondents (N = 733) 56 (20) 2 years, the programme reduced the odds of fever by 47% Mean highest grade achieved in school (N = 715) 4 (4) (aOR: 0.53, 95% CI 0.30 to 0.93). However, the difference Percentage of correct responses to malaria related 60% in treatment effect between the two age groups was not questions (N = 740) statistically significant. Programme impact also appeared Under-five children characteristics to be larger among female children by 67% (aOR: 0.33, Mean age of under-five children (N = 1311) 2 (1) 95% CI 0.20 to 0.55); once again, treatment effect differ - Percentage of under-five children who slept under 14% ences by gender were not statistically significant. LLITNs last night (N = 1328) Table 3 presents multivariable linear regression results Percentage of under-five children with fever in last 23% 2 weeks (N = 1295) estimating programme impact on secondary outcomes of interest. On average, the intervention increased the Standard deviations in parentheses Table 2 Programme impact on respondent-reported fever among children under-five Outcome Fever episode in 2 weeks prior to survey Sample All children under‑five Age < 2 years Age ≥ 2 years Male children Female children (1) (2) (3) (4) (5) Treated 0.464*** (0.294 to 0.731) 0.387*** (0.230 to 0.649) 0.531** (0.304 to 0.929) 0.649 (0.340 to 1.239) 0.332*** (0.199 to 0.553) Constant 0.292*** (0.242 to 0.353) 0.415*** (0.324 to 0.532) 0.243*** (0.193 to 0.305) 0.312*** (0.249 to 0.390) 0.285*** (0.225 to 0.360) H : equal impact NA p = 0.08 p = 0.462 Observations 3750 1049 2612 1898 1783 Multivariable logistic regression results showing average programme impact on under-five child fever (Column 1), fever episodes among children under 2 years (Column 2), fever episodes among children between 2 and 4 years (Column 3), fever episode among male children (Column 4), and fever episode among female children (Column 5). All models control for survey-round fixed effects 95% confidence intervals shown in parentheses are based on Huber’s cluster robust variance estimator *** p < 0.01, ** p < 0.05, * p < 0.1 p-values based on a pooled linear model with an interaction term between age and treatment (Columns 2 and 3) and sex and treatment (Columns 4 and 5), respectively Khadka et al. Malar J (2018) 17:224 Page 6 of 13 Table 3 Programme impact on secondary outcomes Outcome (1) (2) (3) Under‑five child slept under LLITN Household LLITN ownership Knowledge score (z‑score) on the night prior to the survey Treated 0.568*** (0.464 to 0.672) 4.357*** (3.761 to 4.953) 0.352** (0.0335 to 0.670) Constant 0.144*** (0.0934 to 0.194) 0.769*** (0.551 to 0.986) − 0.279*** (− 0.434 to − 0.124) Observations 3788 2093 2126 R-squared 0.347 0.338 0.065 Multivariable linear regression results showing average programme impact on LLITN utilization among children under-five (Column 1), household LLITN ownership (Column 2), and malaria knowledge z-scores (Column 3). All models were estimated using Ordinary Least Squares regression models. Although not displayed, all models control for survey-round fixed effects. 95% confidence intervals shown in parentheses are based on Huber’s cluster robust variance estimator *** p < 0.01, ** p < 0.05, * p < 0.1 likelihood of LLITN utilization among children under- complementary Phase I malaria programme efforts in five by 57% points (β = 0.57, 95% CI 0.46 to 0.67), the Angola was not statistically significant. In contrast, Namib - average number of LLITNs owned by households by 4.36 ian villages exposed to coordinated Angolan programmes (95% CI 3.46 to 4.96), and the average malaria knowledge experienced an additional 17% points (β = − 0.17, 95% CI scores by 0.35 standard deviations (95% CI 0.03 to 0.67). −0.337 to − 0.003) decrease in child fever at midline in Additional file 6 disaggregates the programme’s impact comparison to unexposed villages. This corresponds to an on the primary and secondary outcomes of interest into approximately 62% additional reduction in the odds of fever. the immediate term and over a 1-year follow-up period. Additional file 7 shows unadjusted village level mean The immediate impact of the programme closely mir - differences in child fever prevalence between baseline rored the overall programme impact across all four out- and midline for Namibian villages disaggregated by areas comes of interest, while second year effects were mixed. exposed and unexposed to coordinated cross-border Table 4 shows the first set of spillover results using data programme efforts. While these results are primarily from Namibia. In comparison to baseline, decline in fever descriptive, villages exposed to coordinated efforts in prevalence at midline among Namibian villages without Angola appear to have had a higher likelihood of experi- encing significant fever prevalence declines in compari - son to villages unexposed to coordinated efforts. Table 4 Difference-in-differences analysis assessing treatment effect modification among Namibian villages between baseline and midline Table 5 Interaction between programme impact in Angola Outcome Fever episode in 2 weeks prior to survey and distance to Namibian border Linear probability Logistic model Outcome Fever episode in 2 weeks prior to survey model Linear probability Logistic regression (1) (2) model model Post − 0.053 (− 0.165 to 0.698 (0.338 to 1.444) (1) (2) 0.059) Post × complementary − 0.170** (− 0.337 to 0.377 (0.135 to 1.049) Treated (midline) − 0.112** (− 0.200 to 0.447*** (0.244 to 0.817) angolan programme − 0.003) − 0.0233) effort Distance − 0.000580 (− 0.00802 to 0.996 (0.955 to 1.040) Observations 706 706 0.00686) R-squared 0.039 Treated* distance 0.0111*** (0.00384 to 1.078*** (1.032 to 1.127) 0.0184) Multivariable regression results from a difference-in-differences analysis. Column Baseline 0.211*** (0.121 to 0.302) 0.268*** (0.160 to 0.449) 1 and Column 2 show results based on a linear probability model and logistic regression model respectively. Although not shown in the table, the models Observations 1469 1469 control for an indicator of Namibian villages adjacent to Phase I Angolan areas R-squared 0.013 at baseline. Constant terms from the two models are also not shown in the table. Data for this analysis is restricted to Namibian households surveyed at baseline Multivariable regression results from analysis for treatment effect modification and midline in Angola. Results based on linear probability model are shown in Column 95% confidence intervals shown in parentheses are based on Huber’s cluster 1. Results based on logistic regression model are shown in Column 2. 95% robust variance estimator confidence intervals shown in parentheses are based on Huber’s cluster robust variance estimator. Sample is restricted to Angolan villages receiving treatment *** p < 0.01, ** p < 0.05, * p < 0.1 between baseline and midline All Namibian villages were treated between baseline and midline, which *** p < 0.01, ** p < 0.05, * p < 0.1 means that the post indicator captures both time and treatment effects Khadka et al. Malar J (2018) 17:224 Page 7 of 13 Table 6 Programme impact in Angola, stratified by distance to Namibian border Outcome Fever episode in 2 weeks prior to survey Sample (1) (2) (3) (4) < 5 km of the Namibia border 5–10 km from the border 10–15 km from the border > 15 km from the border Treated 0.368*** (0.217 to 0.624) 0.962 (0.342 to 2.704) 1.156 (0.691 to 1.934) 1.835* (0.976 to 3.450) Constant 0.299*** (0.167 to 0.535) 0.203*** (0.101 to 0.407) 0.290*** (0.252 to 0.334) 0.245*** (0.105 to 0.572) Observations 461 439 445 124 Logistic regression analysis demonstrating treatment effect on child morbidity within 5 km of the Angola–Namibia border (Column 1), between 5 and 10 km of the border (Column 2), between 10 and 15 km of the border (Column 3), and beyond 15 km of the border (Column 4). 95% confidence intervals shown in parentheses are based on Huber’s cluster robust variance estimator. Sample is restricted to Angolan villages receiving treatment between baseline and midline *** p < 0.01, ** p < 0.05, * p < 0.1 Table 5 shows results from the second spillover test. In than areas benefitting from local programmes only. The terms of Phase I treatment effects in Angola, the inter - results are starker for Angola, where there were impres- vention reduced the odds of fever by 55% (aOR: 0.45, 95% sive reductions in fever prevalence in areas close to CI 0.24 to 0.81) on average at the border. For every kilo- Namibian villages benefitting from programmes, but no metre increase in distance from the border, the additional health improvements at all in areas closer to Angolan vil- change in treatment effect on child fever is reduced by lages not benefitting from any programme. approximately 8% (aOR: 1.08, 95% CI 1.03 to 1.13). These rather large local spillovers make sense from a bio - Stratified regression results presented in Table 6 fur- logical and public health perspective as they likely reflect ther illustrate these rather large interaction effects. both vector and human population movements between The intervention was highly effective in areas close to villages in the programme corridor. At baseline, 44% of the Namibian border with an estimated odds reduc- households reported having had at least one individual stay on the other side of the border in the month before the tion of 63% (OR: 0.37, 95% CI 0.22 to 0.62) within areas survey. Similarly, 43% of households at baseline reported 5 km of the border. However, the programme did not having had guests from the other country during the same have any significant impact in areas more distant from time period. Such a degree of cross-border mobility seems Namibia (i.e., closer to non-treated Angolan areas). For areas more than 15 km from the Namibian border (and, natural given the absence of major barriers as well as the therefore, right at the border of non-treated Angolan shared history and culture on both sides of the border. The villages), fever prevalence increased between baseline results from this study may, therefore, be relevant to many and midline despite programme rollout. other border regions which likely demonstrate similarly high levels of cross-border human population movement. A high degree of cross-border human mobility does, Discussion however, pose a major challenge to malaria programming The results presented in this paper have yielded two since neighbouring country’s preferences are unlikely to main insights. First, community-based malaria preven- align. The Angola-Namibia setting is almost ideal to illus tion programmes appear to remain highly effective in - reducing child morbidity, at least in areas with seasonal trate this: while Namibia has almost eliminated malaria malaria and relatively little intervention coverage in the in most parts of the country other than the regions bor- recent past. This study shows that over the 2-year inter - dering Angola and Zambia, malaria is still endemic in vention period, a relatively simple community volunteer- most parts of Angola, and is particularly common in the based prevention programme which distributed LLITNs Northern parts of the country. From an Angolan perspec- and had monthly behavior change home visits before the tive, the burden of malaria is relatively low in the South; peak rainfall months of February and March reduced the thus, such areas are not a primary target of Angolan odds of child fever by more than 50% overall. efforts. The opposite is true for Namibia, which focuses Second, and more importantly, this study shows that most of its efforts on the northern regions of the country. the effectiveness of community-based malaria pro The results presented in this study suggest optimal effec - - grammes strongly depends on concurrent efforts made tiveness may only be reached if efforts are coordinated in neighbouring areas. The estimates presented in this across the border. Since countries with a high malaria bur- paper suggest that Namibian areas at the border ben- den generally have little incentive to focus on areas border- efitting both from local programmes and simultaneously ing low transmission countries, coordinated cross-country implemented programmes in adjacent Angolan areas efforts will likely have to be developed and supported by experienced more than twice the reductions in fever external stakeholders. A general switch in focus towards Khadka et al. Malar J (2018) 17:224 Page 8 of 13 regional rather than national efforts, therefore, seems . Further research will be needed to disentangle the advisable from a political and donor perspective, even impact of the two intervention components. though such regional efforts would undoubtedly require increased coordination and monitoring efforts to ensure Conclusions overall accountability. Encouragingly, there has already The World Health Assembly and Roll Back Malaria cam - been an increasing focus on regional programming to paign have endorsed the vision of a malaria free world, decrease cross-border transmission, especially in Southern aiming for a 90% reduction in malaria incidence and Africa: for instance, the Elimination 8 (E8) secretariat has mortality by 2030 [4, 20]. The WHO Global Technical supported projects to increase access to testing and treat- Strategy identifies cross-border collaboration as a key ment through malaria surveillance posts at strategic loca- mechanism by which the vision of a malaria free world tions at the borders shared by the eight E8 member states. can be made a reality . The results presented in this This study has several limitations. The most immedi - paper provide strong evidence for the importance of ate and obvious limitation is that it does not use directly increased cross-national collaboration and coordina- confirmed cases of malaria among children as the primary tion of anti-malaria efforts, particularly in settings with outcome. Malaria cases were not recorded as it was infea- high levels of cross-border human mobility. While more sible within the study budget to license community health research is needed to understand the relative contribu- volunteers in using RDTs and have them visit households tion of different cross-border interventions, this study with high frequency. While fever is the most common shows that even simple interventions can be extremely manifestation of malaria, it can be a symptom of many effective in reducing child morbidity. other illnesses as well . It is also possible that subjects Additional files over- or underreported fever. If such incorrect reporting is correlated with programme rollout, estimated results Additional file 1: Figure S1. Location of TKMI programme area. The may be biased. In terms of the main results presented in main map shows Angola (dark gray) and Namibia (light gray) and depicts this paper, potential social desirability biases are more provincial boundaries within each country. The numerical labels indicate likely to apply to the overall programme impact esti- the administrative areas in which the TKMI program was implemented: (1) = Namibe; (2) = Cunene; (3) = Kunene; (4) = Omusati; (5) = Ohang- mates since the provision of (highly appreciated) nets may wena. The black rectangle within the main map shows the region induce grateful households to over-report positive health being demonstrated in the inset map. The inset map shows the TKMI outcomes. In terms of the spillover estimates, all house- programme area and the crosses show the 64 villages selected for the evaluation of the malaria control programme. The horizontal line in the holds received the same interventions and should thus be middle of the inset map represents the Angola–Namibia border. equally prone to such biases unless they directly take the Additional file 2: Figure S2. Programme roll-out map. Maps illustrating relative treatment status of neighboring villages into con- the coverage of the programme at each survey round. Panel (a) highlights sideration in their responses; however, this seems unlikely. the boundaries of the intervention corridor and shows how none of the areas had received any treatment at baseline. Panel (b) highlights areas A second limitation of the analysis is that the rollout of in Angola and Namibia that received treatment during Phase I. Panel the treatment was not randomized at the village or clus- (c) shows that all programme areas had received the intervention by ter level, but rather followed a two-phase blocked rollout endline. The horizontal line across the middle of the map represents the Angola–Namibia border. The labels Cunene, Omusati, and Ohangwena schedule. Exploring this rollout schedule will yield unbiased represent the Angolan province and Namibian regions encompassed by impact results as long as treated and control villages would the intervention corridor. The double lines represent major road networks have experienced the same trajectories in the outcomes in in the programme area. the absence of treatment. While this assumption cannot be Additional file 3: Figure S3. Programme and evaluation implementation schedule. Figure depicting the rollout of the TKMI evaluation surveys and tested directly, it is likely to hold since no other major health the interventions. programmes were run in the intervention corridor during Additional file 4: Figure S4. Gantt Chart describing timing of pro - the study period. Furthermore, given the large effects found gramme rollout and evaluation. Gantt Chart describing timing of pro- and the somewhat arbitrary blocking of villages, major con- gramme rollout and evaluation. founding biases seem somewhat unlikely overall. Additional file 5: Table S1. Baseline child fever, LLITN ownership, usage, Given the data, it is also not possible to directly distin- and malaria knowledge levels. Baseline statistics on a per village basis disaggregated by country. guish the impact of LLITNs from the impact of behavior Additional file 6: Table S2. Disaggregating TKMI impact into immediate change oriented home visits on morbidity. Similarly, the and follow-up effect. Programme impact disaggregated by immediate data do not allow for evaluating the relative effectiveness and 1-year follow-up period. of guidance on using and maintaining LLITNs versus Additional file 7: Figure S5. Forest plots illustrating unadjusted mean provision of malaria prevention information on changes difference in fever prevalence between baseline and midline among in morbidity. The combination of the two interventions Namibian villages. a presents mean differences in villages exposed to coordinated cross-border efforts while b presents mean differences for may at least partially explain the relatively large impacts villages unexposed to coordinated efforts. seen in this study compared to the previous literature Khadka et al. Malar J (2018) 17:224 Page 9 of 13 Authors’ contributions Availability of data and materials AK curated the data, conducted the formal analysis, wrote the original draft, The anonymized datasets used and/or analysed during the current study are and corresponded with all authors to finalize the manuscript. GF conceptual- available from the senior author on reasonable request. ized and supervised the entire study. NH, ADG, BP were involved in curating the data and revising the manuscript. NAP, DJW, and SV made significant Consent for publication contributions to editing and finalizing the manuscript as well. All authors read Not applicable. and approved the final manuscript. Ethics approval and consent to participate Author details This study was approved by the Institutional Review Board at the Harvard Department of Global Health and Population, Harvard T.H. Chan School T.H. Chan School of Public Health, the Ministry of Health and Social Services, of Public Health, 655 Huntington Avenue, Boston, MA 02115, USA. University Republic of Namibia, and the Comité de Ética do Ministerio da Saude, Ministé- of California Berkeley, Berkeley, CA 94720, USA. Carolina Population Center, rio da Saúde, República de Angola (Ethics Committee of the Ministry of Health, University of North Carolina, Chapel Hill, NC 27516, USA. Swiss Tropical Ministry of Health, Republic of Angola). All surveyed participants provided and Public Health Institute and University of Basel, Basel, Switzerland. prior consent to be surveyed. Acknowledgements Funding The authors would like to thank the J.C. Flowers Foundation for the financial This work was supported by the J.C. Flowers Foundation. Study sponsors were and logistical support received for this project. Special thanks to Susan Lassen not involved in any aspect of the implementation, analysis, and manuscript and Rebecca Vander Muelen for their assistance during the manuscript revi- writing. sion phase of the study. Competing interests The authors declare that they have no competing interests. Appendix 1 Questionnaire Quesonnaire # BASELINE SURVEY QUESTIONNAIRE Country: Municipality: Province / Region: Study Area / Community/ Village: NAME OF HOUSEHOLD Interviewer’s Name: HEAD: Status: Reason incomplete Checked by: Complete In-complete First visit date:Second visit date:Third visit date: [Interviewer instructions: Please make sure to arrange time to talk to head of household/most senior member of household. Please circle the answer provided by the respondent or write the answer in the spaces provided where applicable] Khadka et al. Malar J (2018) 17:224 Page 10 of 13 No QUESTIONS CODING CATEGORY CODE (For office use) SECTION A : SOCIODEMOGRAPHIC DATA OF RESPONDANT (CIRCLE THE ANSWER) A1 What is your gender? MALE / FEMALE A2 Can you tell us your age, in ┌──┬──┐ years? │░░│░░│ └──┴──┘ A3 What was the highest grade you None 1 2 3 4 5 6 7 8 9 10 11 12 reached in school? Higher A4 What is your employment?1 -WORKING THE FIELDS 2 - TEACHER 3 - NURSE 4 - SHOP ASSISTANT 5 - SELLING LOCALLY PRODUCED GOODS 6 - NDF 7 - UNEMPLOYED 8 -OTHER: (Pls. Specify) ______________________ A5 Does your job require you to YES / NO cross the country border? A6 Can you read the following Yes (can read) No (can’t read phrase below) phrase below: Ovakulunhu nava file oshisho ounona ile navatekule ounona nawa [Read to respondent]: }”Now we would like some information about the people who usually live in your household, that is, usually live more than 5 days per week with you “ A6aHow many people (including yourself) are part of this ____ members household right now: A6bHow many children under the age of 5 currently live in the ____ chlildren household? under 5 A6cHow many sleeping spaces are there in this household ____ spaces Can you please list all children under the age of five living in this household, starting with the youngest [Instructions: Make sure all children just mentioned are listed. If there are more than five, just list the youngest five children] Khadka et al. Malar J (2018) 17:224 Page 11 of 13 A10 A11 A12 A7 A8 A9 Sex A10 Did (NAME) Did (NAME) Did (NAME) get LINE First name Male or How old is sleep under a have a fever in medication for NO. Female? (NAME)? net last the last 2 malaria in last 2 night? weeks? weeks? 0 1 2 3 4 YES / NO YES / NO YES / NO M F 01 DK 0 1 2 3 4 YES / NO YES / NO YES / NO M F 02 DK 0 1 2 3 4 YES / NO YES / NO YES / NO 03 M F DK 0 1 2 3 4 YES / NO YES / NO YES / NO 04 M F DK 0 1 2 3 4 YES / NO YES / NO YES / NO 05 M F DK [Now ask respondent about him/herself] A13Have you had a fever in the last 2 weeks? YES NO DK A14How many months ago did you have your last fever? 1. Less than 4 weeks 2. ______ months 3. DK A15Last time you had a fever, did you get treatment at a health facility? YES NO DK A16Last time you had a fever, did you get a blood test for malaria? YES NO DK A.17 Last time you had a fever, what medication did you get (list all) [Read to respondent]: “I would like to hear about your beliefs related to fevers and malaria. I will read a few statements to you - for each of them Iwill kindly ask you tell me if you if you think they are true or false”. TRUE FALSE DON’T B.1One can get malaria by taking in dirty water or food KNOW TRUE FALSE DON’T B.2One can get malaria through bites from infected mosquitoes KNOW TRUE FALSE DON’T B.3One can get malaria through witchcraft KNOW TRUE FALSE DON’T B.4One can get malaria from eating unripe mangoes or fruits KNOW TRUE FALSE DON’T B.5One can get malaria by getting beaten by the rain or sun KNOW TRUE FALSE DON’T B.6Malaria can kill – people can die from malaria KNOW Khadka et al. Malar J (2018) 17:224 Page 12 of 13 [Read to respondent]: Now I would like to ask you a few questions about malaria. I will read a few statements to you - some of them are true, some are false. Once again, for each of them I will kindly ask you tell me if you if you think they are true or false. TRUE FALSE DON’T B.8Malaria can be cured if one uses appropriate treatment KNOW TRUE FALSE DON’T B.9Malaria can be prevented through witchcraft KNOW TRUE FALSE DON’T B.10 Malaria can be prevented by sleeping under a bed net KNOW TRUE FALSE DON’T B.11 The risk of malaria is highest in the rainy season KNOW TRUE FALSE DON’T B.12 Paracetamol/Panadol is an effective treatment for malaria KNOW TRUE FALSE DON’T B.13 Coartem (ACT) is an effective treatment for malaria KNOW TRUE FALSE DON’T B.14 Most people who have fevers have malaria KNOW TRUE FALSE DON’T B.15 Fever is a symptom of malaria KNOW TRUE FALSE DON’T B.16 Blindness is a symptom of malaria KNOW TRUE FALSE DON’T B.17 Diarrhoeais a symptom of malaria KNOW TRUE FALSE DON’T B.18 Itching is a symptom of malaria KNOW TRUE FALSE DON’T B.19 Headache is a symptom of malaria KNOW TRUE FALSE DON’T B.20 Vomiting is a symptom of malaria KNOW SECTION C : PROGRAMME EVALUATION C1 For the past six months, do you remember receiving or hearing YES or seeing some education on malaria prevention/treatment? NO DON’T KNOW C2 If yes to C1, what was your source of information (where did you hear, see or receive the information)? [Read to respondent]: Now I would like to ask you just a few more questions about the household: How many nets does the household currently own? D1 D2 How many nets are hanging right now? D3` How many nets were used last night? D4 1 - ONE HOUR OR LESS How long does it take you to reach your nearest health 2 - HALF A DAY clinic/hospital (walking)? 3 - ONE DAY 4 -TWO DAYS OR MORE D5 Have you had any visitors stay the night at your YES / NO / DON’T KNOW household that came from across the border in the last month? D6 Have you or any other household member slept YES / NO / DON’T KNOW overnight across the border in the last month? Thank you for your cooperation! [Interviewer: please complete interview information on front page] Khadka et al. Malar J (2018) 17:224 Page 13 of 13 10. Wang R-B, Dong J-Q, Xia Z-G, Cai T, Zhang Q-F, Zhang Y, et al. Lessons Publisher’s Note on malaria control in the ethnic minority regions in Northern Myanmar Springer Nature remains neutral with regard to jurisdictional claims in pub- along the China border, 2007–2014. Infect Dis Poverty. 2016;5:95. https :// lished maps and institutional affiliations. doi.org/10.1186/s4024 9-016-0191-0. 11. Cosep Consultoria, Consaude, ICF International. Angola malaria indicator Received: 17 November 2017 Accepted: 25 May 2018 survey 2011. 2011. http://dhspr ogram .com/pubs/pdf/MIS11 /MIS11 .pdf. Accessed 27 Jun 2017. 12. Instituto Nacional de Estatística (INE), Ministério da Saúde (MINSA), Ministério da Planeamento e do, Desenvolvimento Territorial (MPDT ), ICF. Key Findings of the 2015–2016 Angola IIMS. Luanda, Angola. Rockville, References Maryland, USA.; 2017. https ://dhspr ogram .com/pubs/pdf/SR238 /SR238 1. WHO. World Malaria Report 2017. http://apps.who.int/iris/bitst ream/ .pdf. Accessed 15 May 2018. handl e/10665 /25949 2/97892 41565 523-eng.pdf;jsess ionid =9BDF7 2122B 13. Chanda E, Ameneshewa B, Angula HA, Iitula I, Uusiku P, Trune D, et al. 5E54E 92C63 7EFBF 88827 3D?seque nce=1. Accessed 15 May 2018. Strengthening tactical planning and operational frameworks for vec- 2. Gething PW, Casey DC, Weiss DJ, Bisanzio D, Bhatt S, Cameron E, et al. tor control: the roadmap for malaria elimination in Namibia. Malar J. Mapping Plasmodium falciparum Mortality in Africa between 1990 and 2015;14:302. https ://doi.org/10.1186/s1293 6-015-0785-1. 2015. N Engl J Med. 2016;375:2435–45. https ://doi.org/10.1056/NEJMo 14. Smith Gueye C, Gerigk M, Newby G, Lourenco C, Uusiku P, Liu J. Namibia’s a1606 701. path toward malaria elimination: a case study of malaria strategies and 3. Carter R, Mendis KN. Evolutionary and historical aspects of the burden of costs along the northern border. BMC Public Health. 2014;14:1190. https malaria. Clin Microbiol Rev. 2002;15:564–94. ://doi.org/10.1186/1471-2458-14-1190. 4. Hemingway J, Shretta R, Wells TNC, Bell D, Djimdé AA, Achee N, et al. Tools 15. Norris LC, Norris DE. Efficacy of long-lasting insecticidal nets in use in and strategies for malaria control and elimination: what do we need to Macha, Zambia, against the local Anopheles arabiensis population. Malar J. achieve a grand convergence in malaria? PLoS Biol. 2016;14:e1002380. 2011;10:254. https ://doi.org/10.1186/1475-2875-10-254. 5. Pindolia DK, Garcia AJ, Huang Z, Fik T, Smith DL, Tatem AJ. Quantifying 16. Lengeler C. Insecticide-treated bed nets and curtains for preventing cross-border movements and migrations for guiding the strategic plan- malaria. In: Lengeler C, editor. Cochrane database of systematic reviews. ning of malaria control and elimination. Malar J. 2014;13:169. https ://doi. Chichester: Wiley; 2004. p. CD000363. https ://doi.org/10.1002/14651 858. org/10.1186/1475-2875-13-169. CD000 363.pub2. 6. Wesolowski A, Eagle N, Tatem AJ, Smith DL, Noor AM, Snow RW, 17. Huber PJ. The behavior of maximum likelihood estimates under non- et al. Quantifying the impact of human mobility on malaria. Science standard conditions. In: Proceedings of the fifth Berkeley symposium on 2012;80:338. mathematical statistics and probability. Berkeley: University of California 7. Smith JL, Auala J, Haindongo E, Uusiku P, Gosling R, Kleinschmidt I, et al. Press; 1967. p. 221–33. https ://proje cteuc lid.org/eucli d.bsmsp /12005 Malaria risk in young male travellers but local transmission persists: a 12988 . Accessed 14 Aug 2017. case-control study in low transmission Namibia. Malar J. 2017;16:70. https 18. StataCorp. Stata Statistical Software: Release 15. 2017. ://doi.org/10.1186/s1293 6-017-1719-x. 19. ESRI. ArcGIS Desktop: Release 10. 2011. 8. Moonasar D, Maharaj R, Kunene S, Candrinho B, Saute F, Ntshalintshali 20. WHO. Global technical strategy for malaria 2016–2030. 2015. http://apps. N, et al. Towards malaria elimination in the MOSASWA (Mozambique, who.int/iris/bitst ream/10665 /17671 2/1/97892 41564 991_eng.pdf?ua=1. South Africa and Swaziland) region. Malar J. 2016;15:419. https ://doi. org/10.1186/s1293 6-016-1470-8. 9. Krisher LK, Krisher J, Ambuludi M, Arichabala A, Beltrán-Ayala E, Navarrete P, et al. Successful malaria elimination in the Ecuador-Peru border region: epidemiology and lessons learned. Malar J. 2016;15:573. https ://doi. org/10.1186/s1293 6-016-1630-x. Ready to submit your research ? Choose BMC and benefit from: fast, convenient online submission thorough peer review by experienced researchers in your ﬁeld rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions
Malaria Journal – Springer Journals
Published: Jun 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera