Analyzing spatial clustering and the spatiotemporal nature and trends of HIV/AIDS prevalence using GIS: the case of Malawi, 1994-2010
© Zulu et al.; licensee BioMed Central Ltd. 2014
Received: 4 July 2013
Accepted: 2 May 2014
Published: 23 May 2014
Although local spatiotemporal analysis can improve understanding of geographic variation of the HIV epidemic, its drivers, and the search for targeted interventions, it is limited in sub-Saharan Africa. Despite recent declines, Malawi’s estimated 10.0% HIV prevalence (2011) remained among the highest globally. Using data on pregnant women in Malawi, this study 1) examines spatiotemporal trends in HIV prevalence 1994-2010, and 2) for 2010, identifies and maps the spatial variation/clustering of factors associated with HIV prevalence at district level.
Inverse distance weighting was used within ArcGIS Geographic Information Systems (GIS) software to generate continuous surfaces of HIV prevalence from point data (1994, 1996, 1999, 2001, 2003, 2005, 2007, and 2010) obtained from surveillance antenatal clinics. From the surfaces prevalence estimates were extracted at district level and the results mapped nationally. Spatial dependency (autocorrelation) and clustering of HIV prevalence were also analyzed. Correlation and multiple regression analyses were used to identify factors associated with HIV prevalence for 2010 and their spatial variation/clustering mapped and compared to HIV clustering.
Analysis revealed wide spatial variation in HIV prevalence at regional, urban/rural, district and sub-district levels. However, prevalence was spatially leveling out within and across ‘sub-epidemics’ while declining significantly after 1999. Prevalence exhibited statistically significant spatial dependence nationally following initial (1995-1999) localized, patchy low/high patterns as the epidemic spread rapidly. Locally, HIV “hotspots” clustered among eleven southern districts/cities while a “coldspot” captured configurations of six central region districts. Preliminary multiple regression of 2010 HIV prevalence produced a model with four significant explanatory factors (adjusted R2 = 0.688): mean distance to main roads, mean travel time to nearest transport, percentage that had taken an HIV test ever, and percentage attaining a senior primary education. Spatial clustering linked some factors to particular subsets of high HIV-prevalence districts.
Spatial analysis enhanced understanding of local spatiotemporal variation in HIV prevalence, possible underlying factors, and potential for differentiated spatial targeting of interventions. Findings suggest that intervention strategies should also emphasize improved access to health/HIV services, basic education, and syphilis management, particularly in rural hotspot districts, as further research is done on drivers at finer scale.
KeywordsHIV/AIDS GIS Spatiotemporal analysis Clustering Hotspots Coldspots Drivers Malawi Africa
The 2012 global AIDS report placed Malawi among 25 countries with declines of 50% or more in new cases of adult (age 15-49) HIV infection globally. Malawi is also among countries reporting positive behavioral change or covering 60-79% of eligible people in antiretroviral therapy (2001- 2011), and is among 32 countries with a 25-49% decline in HIV deaths (2005-2011) . Yet at 10.0% in 2011 [12, 16], HIV prevalence remains a formidable challenge 26 years after the first diagnosis in1985. Among sex workers, prevalence was up to 70%, while 923,058 of Malawi’s 14 million people were living with HIV/AIDS in 2010 . Nearly 600,000 children were orphaned due to HIV/AIDS , and there are glaring gender and urban/rural disparities. HIV prevalence was 2.2 times higher among female than male youth aged 15-24, and 2.8 times higher among urban women than their rural counterparts (22.7% versus 10.5%) in 2010 . Further, decline in HIV prevalence in the general population has been modest – from 11.8% in 2004 to 10.6% in 2010. As with other epidemics, understanding such spatial variation in HIV/AIDS prevalence and its drivers within a social, spatial and temporal context is crucial for spatial targeting of interventions and resources [1, 7, 18].
This article is divided into five parts. First, we briefly review the literature on the use of GIS and spatial analysis in analyzing HIV prevalence, and on commonly found drivers of HIV risk and prevalence. A presentation of broader spatiotemporal patterns of HIV prevalence at national and regional scale, including the significant urban/rural divide, follows. Third, the overall spatial structure of the epidemic is examined based on spatial statistical analysis of the presence, nature, temporal trends and implications of spatial dependency in prevalence rates. This knowledge is used to decompose spatial patterns of HIV prevalence first into continuous surfaces using spatial interpolation techniques in order to show continuous spatiotemporal variation nationally, and second at district level after spatial aggregation (averages). The fourth section presents analysis of local clustering patterns, or “hotspots” and “coldspots,” of HIV prevalence at district level.a The fifth part presents models of indicative drivers of HIV prevalence identified from multiple regression analysis and mapping of the spatial variation and clustering of the drivers. Study findings and implications are then discussed, and limitations presented before offering a concluding section. The methodology used is presented at the end.
Use of GIS and geospatial analytical methods in understanding HIV prevalence
Among the limited but growing uses of GIS in HIV analyses in Africa, very few studies have analyzed spatiotemporal variation in HIV prevalence, including clustering patterns, at meso (e.g., district) or lower scales, where better understanding can enhance the effectiveness of interventions .b Occasional studies during early parts of the epidemic demonstrated the value of geographic analysis in understanding HIV prevalence, risk, and spread; including in mapping the distribution of at-risk populations of commercial sex workers and uncircumcised males [20, 21]. However, most of the few recent GIS-based analyses of HIV/AIDS in Africa have been mainly at coarse continental scale with limited national policy relevance (e.g., [5, 6]), or at fine local scales requiring heavy data collection/use but with a limited spatial scope, e.g., assessment of local HIV variability/clustering and risk in a rural sub-district in northern KwaZulu-Natal Province, South Africa [1, 22].c Very few exceptions have analyzed country-wide spatial variation and clustering of HIV prevalence, and their drivers. Messina et al.  conducted such a study involving sex differentiated spatial variation in HIV prevalence using 2007 demographic and health survey (DHS) data, GIS, and regression analysis of HIV drivers at “community” (neighborhood or village-clusters) scale in the Democratic Republic of the Congo (DRC). Moise and Kalipeni  recently used GIS and HIV sentinel data for pregnant women attending ANCs in Zambia to analyze spatiotemporal patterns of HIV prevalence from 1994 to 2004. They also used spatial statistical regression modeling to identify possible drivers of HIV prevalence for 2004 at the district scale, where health-related data are reported and planning and decision-making done. Despite limitations involving sample representativeness and paucity of data compared to population-based (e.g. DHS) HIV data, ANC data are more readily available, allow longitudinal trend analysis, and can be re-scaled to district or other scales using spatial interpolation techniques [2, 6]. Our study uniquely goes beyond HIV “hotspot” analysis to map the spatial variation of likely drivers allowing identification of configurations of explanatory variables for particular clusters of HIV hotspot districts for more effective intervention targeting.
Common drivers of HIV/AIDS prevalence
HIV epidemics are diverse and complex to deal with because their drivers are numerous, diverse and vary over space and time. Given that most HIV transmission in sub-Saharan Africa is through heterosexual intercourse, the major proximate driver is known - having unprotected sex with an infected person, and the higher the number of cumulative and concurrent sexual partners, the higher the risk of transmission [12, 14, 24]. Yet recent research shows that there are underlying HIV drivers that go deeper than biomedical or narrow epidemiological susceptibility factors. These include diverse socio-economic, demographic, cultural, historical, and geographic factors and their configurations that affect the vulnerability of particular groups of men and women who engage in such risky sexual behavior to HIV infection [25–29].
For Malawi, reported underlying HIV drivers have included high levels of poverty (and wealth), low literacy, high rates of unprotected casual and transactional sex, low male and female condom use, cultural (e.g., widow cleansing and/or ‘hyena customs’) and religious factors, gender inequity and low social and economic status of women, high-risk livelihoods, high migration/mobility levels, high incidences of sexually transmitted diseases and tuberculosis, and geographic factors mainly involving poor access to health services or increased exposure to risks. For instance, several curable sexually transmitted diseases can increase the risk of HIV transmission 2-20 times per sexual contact, and prevalence is higher among those with poor access to prompt treatment . Poverty has forced some women into commercial sex or other risky sexual behavior in order to survive, increasing their risk of contracting/spreading HIV/AIDS [30, 31]. Therefore, conditions of high unemployment, and low and insecure wages that lead to such behavior may help explain high HIV rates among the urban poor. Tough economic conditions have also driven international male labor migration, including into mining hubs in South Africa where men are separated from their spouses for extended time periods and vulnerable to casual sex and HIV infection, and spreading HIV in their areas on their return [32, 33]. Geographically, high HIV infection has been associated with close proximity to major transportation networks which have provided transmission arteries for HIV within and across countries, and to urban and trading centers and transport networks linking them [3, 6, 34].
HIV prevalence data and temporal trends at national, regional and urban/rural scales
HIV/AIDS data came from mean HIV prevalence rates among pregnant women attending a longitudinally rich network of 19 ante-natal clinics (ANCs) where HIV surveillance has been conducted from 1994 to 2010. In 2007, the network was nearly tripled to 54 ANCs while maintaining the original 19, allowing continuity and spatiotemporal analysis (Figure 1). Data included the location (latitude and longitude coordinates) of the ANCs, allowing their mapping. HIV data collection frequency was annual from 1994-1999 and subsequently largely biennial. Although limited demographic information is also collected, only HIV prevalence was available to the authors for all the years, sourced from U.S. based Centra Technology Inc. (1994-2003) and HIV/syphilis surveillance reports produced by Malawi’s National AIDS Commission, NAC [13, 35, 36]. Malawi uses standard sampling, HIV testing, and data analysis methods and models recommended by the Joint United Nations Program on HIV/AIDS (UNAIDS) and the World Health Organization (WHO) (e.g., [37, 38]).d Despite known limitations of such sentinel based HIV prevalence estimates [18, 39], they remain the major source of HIV data in Malawi and other African countries, and the only longitudinal record for analyzing the spatiotemporal trends targeted in this study without projecting to the general population or generating predictions.e
Spatial dependence in HIV prevalence, spatial interpolation, and spatiotemporal trends
First, we plotted HIV prevalence rates for pregnant women 15 to 49 years old [13, 17] from 1995 to 2010) to provide a broad, multi-scalar, spatiotemporal perspective of the HIV epidemics at national, regional, urban and rural scales. Then GIS tools were used to 1) empirically test for spatial dependency in HIV prevalence nationally, 2) produce a continuous surface of HIV prevalence at 1 × 1 km spatial resolution for visualization and generation of prevalence estimates at district level for cluster/hotspot and regression analysis. GIS analysis was conducted with AcrGIS desktop 10.0 (Redlands, CA: Environmental Systems Research Institute, Inc., 1999).
where N is the number of spatial units (sentinel ANCs), Xi is the measured value for feature I (up to N), Xj is the measured value for a neighboring point j (up to N-1), and Wij represents a weight measure of the influence of neighboring feature j on measured value at I derived from the row-standardized spatial weight matrix.
In order to produce smooth surfaces of HIV prevalence for visualization and data generation at district level, the Inverse Distance Weighted (IDW) spatial interpolation method was used for the selected years (1994, 1996, 1999, 2001, 2003, 2005, 2007 and 2010). These years were chosen for trend continuity while including years of positively autocorrelated HIV prevalence (partly justifying use of IDW) and early years (1996, 1999, 2001) of non-significant and/or negative autocorrelation which nevertheless illuminate spatiotemporal patterns. Spatial interpolation methods apply mathematical models to measured point values of a continuous variable at known locations to predict values at locations that do not have values, thereby creating a continuous surface [43, 44]. In predicting values, interpolation methods generally use distance-based weights that assign more influence to measured values nearest an unmeasured location than to measured values located farther away. Deterministic interpolators, including IDW, use weights based only on distance between measured and unmeasured points while geostatistical (or stochastic, e.g., kriging) use sophisticated weights combining distance with probabilistic statistical models of the spatial variation among measured points. IDW produced stable and reasonably reliable predictions for cross-year comparisons with the small sample size (from 19 ANCs). It has been used reliably with small-medium samples in HIV studies [2, 23], at times preferred over (potentially superior) kriging whose performance often suffers more with small samples because of probability distribution requirements [44, 45]. With IDW, we used a variable setting of 6-10 points to predict values at each unknown location based on iterative testing to minimize mean error and root mean square error (RMSE). We then used GIS tools to extract HIV estimates for the 31 ‘districts’ (27 of Malawi’s 28 districts and four main cities of Blantyre, Lilongwe, Zomba and Mzuzu, Figure 1) by averaging prevalence values in constituent 1 × 1 km spatial cells.
Local spatiotemporal variation in HIV prevalence and cluster/‘hotspot’ analysis
Regression analysis and indicative drivers of HIV prevalence for 2010
In order to identify and illustrate potential linking of indicative drivers of observed spatial variation of HIV prevalence to particular hotspot/coldspot clusters of districts, we conducted multiple regression analysis of HIV prevalence for 2010 only, and then mapped the spatial distribution and clustering of selected factors. HIV prevalence per district among pregnant women attending ANCs, the dependent variable, was estimated using GIS as explained earlier for cluster analysis, but using the latest sentinel HIV data (2010) from the 54 ANCs (instead of 19). The year 2010 also matched dates of available explanatory factors closely.
Summary of selected variables significantly correlated with 2010 HIV prevalence
Variable type, description
Estimated HIV prevalence, 2010 (%)
2010 HIV/syphilis report, NAC 2011
Persons per km2, 2008
1998 census, NSO 2008
Attended senior primary sch. (%)
WMS 2011, NSO 2012
Gross migration (in and out migration), 2008 (%)
1998 census, NSO 2008
Unemployed female, 2011 (%)
WMS 2011, NSO 2012a
Population, age 25_49, 2008 (%)
1998 census, NSO 2008
District share of total urban population, 2008 (%)
1998 census, NSO 2008
Per capita consumption, 2011 (MK)
IHS3, NSO 2012b
Unemployed male, 2011 (%)
WMS 2011, NSO 2012a
HIV awareness and behavior
Percentage that had taken an HIV test ever
WMS 2011, NSO 2012a
Percentage who know a confidential HIV test is possible
WMS 2011, NSO 2012a
Proportion of women positive for syphilis (%, 2010)
2010 HIV, syphilis report, NAC 2011
Geographic (proximity, access, exposure)
Mean dist. to HF
GIS derived, 2001 HF data, MoH
Mean dist. to HF, age 30_44
WMS 2011, NSO 2012a
Mean dist. to main roads (km)
GIS based, roads layer from Malawi Surveys Dept.
Mean dist. to major city (km)
GIS based, cities map, NSO
Mean time to all-weather road, age 30_44 (min)
WMS 2011, NSO 2012a
Mean time to transport, age 30_44 (min)
WMS 2011, NSO 2012a
Mean dist. to HF, age 45_59
WMS 2011, NSO 2012a
Correlation analysis against HIV prevalence was used to screen the initial variable pool, yielding the 18 statistically significant (p ≤ 0.10) ones listed in Table 1 (the full list is available on request). We used further correlation analysis among the 18 to narrow the significant variables to 13 (variable names marked with the superscript a in Table 1) by removing highly correlated variables (generally r > 0.7, p ≤ 0.05). For instance, mean distance to health facilities was dropped because it was highly correlated (r = 0.784, p = 0.000) with distance to main roads but slightly less correlated with HIV prevalence. We entered the 13 independent variables into SPSS 20.0 (IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.) for multiple regression using forward stepwise entry after standardizing them to Z values to stabilize variability and curb observed remnant multi-collinearity. Several collinearity diagnostics and partial significance statistics were used to further limitmulticollinearity problems and to pick a “best” model among the four produced.
Variables in the “best” model had to have a Variance Inflation Factor (VIF) below 2, Condition Index below 30, and tolerance values above 0.5 to signify non-significant collinearity. Additional diagnostics on the “best” model confirmed multi-collinearity and heteroskedasticity not to be significant problems. Cluster/hotspot analysis and mapping were performed on variables from the “best” model and ancillary correlation analysis used to explain observations.
The narrow focus of the second study objective on identifying and mapping spatial patterns of indicative explanatory factors of HIV prevalence, rather than producing predictive models, and their basis on the literature, should mitigate concerns over automated variable-selection methods . For the same reasons, we maintained a simple ordinary least squares (OLS) model. Additional spatial diagnostics on the “best” OLS model using the GEODA spatial statistical software (GeoDa Center for Geospatial Analysis and Computation, Arizona State University) confirmed expected breach of the OLS independence assumption given the significant spatial dependence. Although the corrective spatial lag model [51, 53] and significant of its autocorrelative coefficients and individual variable coefficients were generally an improvement on the OLS model, except a slight decline in the significance of one variable, it did not sufficiently change the essentials of the OLS model. Thus, we report only results of the OLS model for purposes of this study.
Temporal and spatial trends in HIV prevalence at national and regional scales
Spatial autocorrelation, GIS interpolation mapping and spatiotemporal trends
Local spatial variation and hotspot analyses of HIV prevalence
Potential drivers of HIV prevalence for 2010 and their local spatial patterns
Summary of regression coefficients for possible drivers of HIV prevalence
Attended senior primary sch. (%)
Percentage that had taken an HIV test ever
Mean time to transport, age 30_44 (min)
Mean dist. to main roads (km)
The proportion that had ever taken an HIV test was the only behavioral, and the second most influential, explanatory factor for HIV prevalence, and was (counterintuitively) positively associated with HIV prevalence (Table 2). Only Zomba city emerged as a core hotspot for this variable and Lilongwe City as a spatial outlier (Figure 12). The percentage of those who had taken at least an HIV test was highest in Lilongwe city (81.9% of adults, hence an HL outlier) in relation to the exceptionally low Lilongwe Rural (52.9%, hence an LH outlier). However, analysis with the Getis-OrdGi* statistic revealed a secondary intensity hotspot cluster (GiZScore = 1.96-2.58 SDs) perfectly spatially matching the eight core HIV hotspots for 2010.
Education, in particular the percentage of the population that had attended senior primary school (grades 6-8), was significantly negatively correlated with HIV prevalence, although it had the lowest Beta value among the four “best” model variables. This suggests that the more pregnant women who have a little education (senior primary) the less likely they are to have HIV/AIDS. Cluster analysis for this variable shows only three northern districts (Chitipa, Karonga and Mzimba) standing out as hotspots. Analysis with Getis-OrdGi* statistic adds two secondary hotspot districts (Nkhata Bay and Mzuzu City) and one coldspot district. None of the HIV hotspot districts were in education “hotspots.”
HIV clustering and spatiotemporal patterns of HIV prevalence in Malawi
This study has provided a visually powerful and empirically derived, multi-scalar analysis of spatiotemporal variation in HIV prevalence among pregnant women attending ANCs in Malawi from 1994 to 2010. It goes beyond current broad characterizations at national, regional and urban/rural variation in HIV prevalence (Figure 2) to local variation at district and lower (continuous) levels (Figures 4, 5, 6, 7, 8, 9, 10, 11, 12, and 13). GIS tools allowed the generation of HIV prevalence estimates at scales where HIV data are not traditionally collected (district and continuous) from point HIV sentinel data at 19 ANCs for spatiotemporal pattern analysis and from 54 ANCs for use in multivariate analysis of possible drivers and their spatial variation. Observed widespread spatial variation in HIV prevalence reveal the HIV epidemic as an aggregation of several spatially defined sub-epidemics - national, regional, urban, rural, and local (clusters). Findings broadly call into question the use/effectiveness of one-size-fits-all interventions and policies under such circumstances.
The most prominent temporal trend was the general decline in HIV prevalence after rapid spread/intensification up to 1999. For Malawi, the trend positively reflects on the concerted planning and considerable human and financial resources (most from donor aid) applied to prevention and treatment efforts by government and other agencies over the past decade . However, the impressive rates of national decline in HIV prevalence among ANC-attending pregnant women are 1) slowing down from 1.1% annually betweeen1999 and 2010, to 0.91% and 0.88% annually for the periods 2003-2010 and 2005-2010, respectively (Figures 4-11) tempered by a modest decline in the general population (from 12.0% in 2004 to 10.6% in 2010, 0.2% annually) based on MDHS data  - a cautionary tale against complacency.
Another significant spatiotemporal trend was a slow but emergent spatial evening in HIV prevalence as the epidemic stabilized and then declined after an initial period of localized spatial heterogeneity and explosive spatial spread and intensification to a peak in 1999, as illustrated empirically in temporal patterns of global Moran’s I (Figure 3) and in various ways in Figures 4-11. First, there was a general narrowing in the prevalence gap among the regional sub-epidemics (more so for the north and center, and between the urban/semi-urban and rural epidemics. Second, the negative autocorrelation evident in 1995 and 1996 indicated localized heterogeneity with high next to low prevalence pockets confirmed as spatial outliers in clustering patterns (Figures 10 and 11), which are generally associated with expansionary spatial processes and rapid spread . Subsequent increases in the level and significance of positive spatial autocorrelation (Figure 3) further indicate relative spatial evening of the epidemic with districts of similar HIV prevalence more clustered together. The (low) decrease in total statistically significant clustering (12 core hotspot and coldspot cluster districts 2001 to 2010 relative to 14 in 1994 and 1999) also indicates spatial evening as districts that stand out from the rest decrease. The fact that the spatial and population center of gravity was firmly located in rural areas where the majority (85%) of Malawians lived  and the HIV epidemic was more stable and lower in intensity (at least among pregnant women), may have provided the inertia that kept the national epidemic in relative check. The rural epidemic’s declining trend may thus be a key turning point nationally.
Identification of a hotspot cluster of 5-11 districts making up the HIV epicenter consistently located in the south was the most significant outcome of cluster analysis. Potential explanations include the long history of urbanization in the south forming a hub for HIV transmission. Cluster analysis revealed a hierarchy among the four biggest cities in the country in HIV prevalence relative to surrounding districts. Blantyre, the first city in Malawi and the commercial capital, anchored the HIV hotspot/epicenter every year of analysis, followed by Zomba (twice in a hotspot cluster), then Lilongwe (twice an HL outlier but part of a primary or secondary coldspot in most of the study years), and then Mzuzu city [Figures 10 and 11]. The Southern Region was also the most densely populated , and had the highest levels of rural poverty  and prevalence of syphilis . The 2010 MDHS also showed that women and men in the Southern Region had the highest percentage of those who had two or more sexual partners in the previous 12 months and mean number of sexual partners per lifetime, while men had the lowest percentage reporting using a condom during last sexual intercourse (22.9%) . Elevated labor migrancy, particularly returning Malawian mine workers from South Africa late1980s/early 1990s, may partly explain emergence of an HIV sub-epicenter from 1994 to 1999 (Figures 6-9) among the northern districts of Nkhata Bay and Rumphi, and Mzuzu city [32, 55].
Potential drivers of HIV prevalence for 2010 and their local spatial patterns
Geographic variables, particularly mean travel time to nearest road, had the most explanatory influence on HIV prevalence among the variables in the “best” regression model (Table 2), but the two geographic variables in the model were virtually uncorrelated and they appeared to capture subtly different aspects of “access”. The spatial variation of these two variables in relation to HIV prevalence (and hotspots) also differed. Main roads generally link places to main towns and cities and this variable mainly captured the urban/rural divide. Increasing distance from main roads reflected increasing distance from urban areas and the factors that elevated HIV prevalence there, while increasingly capturing more isolated and sparsely populated areas with limited mobility and sexual mixing and access to health services and conditions that are generally associated with lower risks of HIV infection. This explains the negative relationship with HIV prevalence. Cluster analysis of distance from main roads revealed only one statistically significant core cluster, a coldspot (districts exceptionally close to main roads and urban areas, hence high HIV risk areas) cluster of Blantyre city and neighboring Blantyre Rural and Chiradzulu districts. This suggests that distance to cities (or urban influences) is a major explaining factor for high HIV rates in this subset of the core HIV hotspot districts, although hotspot analysis with the Getis-OrdGi* statistic expanded the primary coldspot cluster to 8 districts which virtually matched the larger core HIV hotspot.
In contrast, core and secondary hotspots for mean travel time to nearest transport (30-44 age group), which varied positively with HIV prevalence, matched three rural HIV hotspot districts of Mulanje, Phalombe and Thyolo spatially. These districts (including Zomba Rural as a tertiary HH cluster) have hilly terrain, which is a physical obstacle to travel and access to health services. Further, high concentration of commercial farms (mainly tea) forced many people into smaller remaining areas, making for some of the highest population densities in relation to available arable land. The dense populations, including the predominantly male commercial farm workers (a high HIV risk group in Malawi ), facilitated increased sexual mixing and risky sexual encounters, and HIV spread. These areas need increased access to HIV information and other services, including through mobile services.
Our findings are generally consistent with recent studies elsewhere in Africa. Tanser et al.  also found an inverse significant relationship between HIV prevalence and distance to the main road in Kwazulu-Natal, South Africa. Travel time to nearest public transport captures a similar relationship in the inverse direction (longer/farther from transport, lower HIV rates). Studies reveal a mixed role of education, suggesting that it is mediated by complex factors. The 2004 and 2010 MDHS studies show the level of education generally positively associated with HIV prevalence . However, Moise and Kalipeni found a negative relationship for literacy rates above 20% in Zambia based on a spatial lag model. Given our more specific definition of education, and controlling for HIV testing, distance to main roads and time to main transport, the negative relation with the percentage of those who had attained this modest (senior primary education) suggests some level of education is good for AIDS prevention.
Further, we found that the percentage of people in the larger population (men and women) who had taken at least one HIV test ever was positively associated with HIV prevalence. The secondary hotspot clusters for HIV testing revealed by the Getis-OrdGi* statistic captured seven districts that closely matched the core HIV prevalence hotspot for the women’s ANC sample for 2010 (Figure 13, B4). This finding that levels of HIV testing were highest in districts that already had high HIV prevalence is plausible because the sense of risk and value of HIV testing would likely be heightened in such areas. Indeed, the 2011 Malawi Welfare Monitoring Survey found that the main reason given by 44% of non-tested respondents for not testing was that they did not feel at risk or in need of an HIV test . During Malawi’s 2010/11 fiscal year, 1.773 million people (28% of the sexually active population) were tested for HIV in Malawi . However, HIV testing was higher among urban residents who are generally wealthier and more educated, have better access to testing facilities, but also a higher risk of being infected than rural residents [3, 6, 14]. This relationship explains the emergence of Lilongwe City as a core HL outlier and Zomba City as a core hotspot (Figure 12, A4). However, Blantyre City does not appear as an outlier probably because the core HIV cluster in the south is more extensive and includes Blantyre and several (rural) districts (Figure 12, HIV, A1). Further nuanced analysis may be needed on HIV testing behavior and its impact on HIV prevention.
Although the four variables included in the “best” model explained a relatively high proportion of the variance in HIV prevalence (68.8%, see Table 2), closer examination suggests that these variables were essentially proxies for underlying factors related to geographic/physical and socio-economic access or exposure to health services and other amenities, HIV prevention knowledge, personal resources, sexual networks and risky sex and other factors that influence HIV-related behavior and risk, mainly mediated by urban/rural residence or proximity and poverty/wealth and education. Further, some variables not included in the final model were interesting. For instance, the core and secondary clustering patterns (hotspots) for 2010 syphilis prevalence, highly significant in binary correlations (Table 1), were a near-perfect match with HIV hotspot districts (Figures 12 and 13), suggesting that syphilis treatment/management also needs special attention. Therefore, more detailed analysis needs to be done, focusing on local HIV variation at finer spatial scales and drivers of HIV incidence to better inform policy, rather than on prevalence as was done in this study. Nevertheless, HIV incidence is often closely associated with prevalence, and as in Tanser et al. , we also assumed that time lags between cause and effect would not significantly model outcomes given our emphasis on preliminary associations.
Limitations of the study
While this study contributes significantly to advancing spatiotemporal analysis of HIV/AIDS in Malawi, Africa, and the developing world generally, it has limitations. The longitudinal depth of HIV sentinel data and caution with the interpolation process including consistency across years allowed adequate assessment of spatiotemporal patterns/trends among pregnant women, but the small number (19) of sentinel ANCs, used in the spatiotemporal analysis means that the interpolated products and especially the derived district estimates are and should be treated as indicative, although the HIV data for the regression analysis were from a larger sample (54 points). Local spatial-statistical pattern (hotspot) analysis also came with uncertainties, including the choice of method or search distance to define a neighborhood. In this study, we empirically determined it as the distance at which global Moran’s I was highest for most years, 80 km. Further research is needed to determine the spatial limit of the spatial dependence of HIV on neighboring values at different spatial scales to better guide interventions. Finally, although our choice of an OLS regression model in the presence of spatial dependence breached the assumption of independence of measurements, OLS was sufficient for the study purpose of identifying indicative explanatory factors for observed spatial variation (as opposed to producing predictive models). Moreover, a corrective spatial lag model of HIV prevalence using standardized variables in the “best” OLS model based on spatial diagnostics was performed within the GEODA spatial statistical software [53, 57]. Despite improvements in the explanatory power of the model and significance of coefficients for all but one of the four independent variables (the p-value for distance to roads fell slightly from 0.043 to 0.06), the spatial lag model did not change the directions of the variable relationships nor their relative importance.f In the interest of space, we left out the spatial lag model findings.
This study has 1) shown spatiotemporal trends in HIV prevalence at multiple scales (national, regional, district and sub-district/continuous) in Malawi from 1994 to 2010 using spatial analysis of data from pregnant women attending HIV sentinel surveillance centers and GIS tools and 2) identified five socio-demographic, behavioral, socio-biological and geographic variables found to be significantly associated with HIV prevalence in multiple ordinary least squares (OLS) regression analysis and mapped their spatial variation at district level in relation to the spatial distribution of HIV hotspots, coldspots and spatial outliers. A varying core hotspot of 6-11 districts was found in the south and a coldspot of 1-6 districts in the center, but the epidemic was slowly leveling out spatially in terms of prevalence. Findings illustrated the importance of spatially explicit geographic analysis to enhance understanding of the spatial and temporal variation and nature of the HIV pandemic along with configurations of factors that shape it in particular locations, with potential to enhance spatial targeting of HIV interventions and policies.
The results indicate that for Malawi there were several geographically differentiated HIV/AIDS epidemics rather than a single one. Further, the results of our analysis offer the most spatially explicit longitudinal analysis of HIV prevalence in the country that we are aware of. The study joins a small but growing number of studies with similar spatial specificity including mapping the spatial variation of proximate and underlying factors that influence HIV prevalence. Despite acknowledged shortcomings associated with limited sample size, this study demonstrates the broader importance of using explicit spatial analysis to understand the geographic nature of the epidemic, examine spatiotemporal trends, and use this knowledge for more effective spatial targeting to combat the HIV/AIDS pandemic.
aThe presence of spatial dependence implies that HIV prevalence values in one district are similar to and dependent on values in neighboring districts. It is also called spatial autocorrelation.
cVery localized analysis has included issues of access to HIV anti-retroviral treatment in part of Karonga district, northern Malawi .
dTarget sample sizes are 300, 500, and 800 women at rural, semi-urban, and urban ANCs, respectively.
eConcerns involve sample representativeness –use of purposive sampling favoring urban/high risk areas over rural ones, paucity of ANCs, and narrow population of pregnant women attending ANCs relative to the larger population. Limited access and historical depth reduce utility of emerging population-based HIV survey data, e.g. the MDHS.
fThe R square increased from 0.688 to (pseudo) 0. 837, the Log likelihood increased from -70.757 to -61.932, while the Akaike info criterion and Schwartz criterion decreased from 151.515 to 135.863 and 158.685 to 144.467, respectively. These changes are indicative that the spatial lag model is better than the OLS model, and confirm that space matters as a factor in HIV prevalence among pregnant women attending antenatal clinics in Malawi .
Demographic and health survey
The Democratic Republic of the Congo
Geographic information systems
Government of Malawi
High-high hotspot (clustering of districts with outstandingly high HIV or other values)
Hi-Low hotspot (an outlier high-value district surrounded by low value ones)
HIV testing and counseling
Inverse distance weighted
Integrated health survey
Local measures of spatial autocorrelation
Low-High outlier (a coldspot district surrounded by high value ones)
Low-Low coldspot (clustering of districts with very low values)
Malawi demographic and health survey
Malawi Kwacha (Malawi’s currency)
Ministry of health
National AIDS Commission
Malawi National Statistical Office
Ordinary least squares
Root mean square error
Sexually transmitted infection
The Joint United Nations Program on HIV/AIDS
The World Health Organization
Welfare monitoring survey.
Authors are grateful in acknowledging the help of Imelda Moise and Sarah Hession for commenting on earlier drafts of the manuscript.
- Tanser F, Bärnighausen T, Cooke GS, Newell ML: Localized spatial clustering of HIV infections in a widely disseminated rural South African epidemic. Int J Epidemiol. 2009, 38: 1008-1016. 10.1093/ije/dyp148.View ArticlePubMedPubMed CentralGoogle Scholar
- Kalipeni E, Zulu L: Using GIS to model and forecast HIV/AIDS rates in Africa, 1986–2010. Prof Geogr. 2008, 60: 33-53. 10.1080/00330120701724061.View ArticleGoogle Scholar
- Messina JP, Emch M, Muwonga J, Mwandagalirwa K, Edidi SB, Mama N, Okenge A, Meshnick SR: Spatial and socio-behavioral patterns of HIV prevalence in the Democratic Republic of Congo. Soc Sci Med. 2010, 71: 1428-1435. 10.1016/j.socscimed.2010.07.025.View ArticlePubMedPubMed CentralGoogle Scholar
- Mayer JD: The geographical understanding of HIV/AIDS in sub-Saharan Africa. NorskGeografiskTidsskrift - Norwegian J Geogr. 2005, 59: 6-13.Google Scholar
- Uthman O, Yahaya I, Ashfaq K, Uthman M: A trend analysis and sub-regional distribution in number of people living with HIV and dying with TB in Africa, 1991 to 2006. I J Health Geogr. 2009, 8: 65-10.1186/1476-072X-8-65.View ArticleGoogle Scholar
- Kalipeni E, Zulu L: HIV and AIDS in Africa: a geographic analysis at multiple spatial scales. GeoJournal. 2012, 77: 505-523. 10.1007/s10708-010-9358-6.View ArticleGoogle Scholar
- Cromley EK, McLafferty S: GIS and Public Health. 2002, New York: Guilford Press, 6td edGoogle Scholar
- Tanser F, le Sueur D: The application of geographical information systems to important public health problems in Africa. Int J Health Geogr. 2002, 1: 4-10.1186/1476-072X-1-4.View ArticlePubMedPubMed CentralGoogle Scholar
- Foody GM: GIS: Health applications. ProgPhysGeogr. 2006, 30: 691-695.Google Scholar
- Higgs G: The role of GIS for health utilization studies: literature review. Health Serv Outcomes Res Method. 2009, 9: 84-99. 10.1007/s10742-009-0046-2.View ArticleGoogle Scholar
- Nykiforuk CIJ, Flaman LM: Geographic information systems (GIS) for health promotion and public health: a review. Health Promot Pract. 2011, 12: 63-73. 10.1177/1524839909334624.View ArticlePubMedGoogle Scholar
- UNAIDS Global Report: UNAIDS Report on the Global Aids Epidemic; 2012. 2012, Geneva: Joint United Nations Program on HIV/AIDS (UNAIDS)Google Scholar
- NAC: HIV and Syphilis Sero–Survey and National HIV Prevalence and AIDS Estimates Report for 2010. 2011, Lilongwe, Malawi: National AIDS Commission (NAC)Google Scholar
- NSO, ICF Macro: Malawi Demographic and Health Survey 2010. 2011, Zomba, Malawi; Calverton, Maryland, USA: National Statistical Office (NSO) and ICF MacroGoogle Scholar
- Geubbels E, Bowie C: Epidemiology of HIV/AIDS in adults in Malawi. Malawi Med J. 2006, 18: 99-121.Google Scholar
- Malawi:: HIV and AIDS estimates 2011. 2012, Geneva: UNAIDSGoogle Scholar
- GoM: 2012: Global AIDS Response Progress Report: Malawi Country Report for 2010 and 2011. 2012, Lilongwe, Malawi: Government of Malawi (GoM)Google Scholar
- Peng ZH, Yue-Jia C, Reilly KH, Lu W, Qian-Qia Q, Zheng-Wei D, Guo-Wei D, Ding KQ, Yu RB, Chen F, Wang N: Spatial distribution of HIV/AIDS in Yunnan province, People’s Republic of China. Geospatial Health. 2011, 5: 177-182.View ArticlePubMedGoogle Scholar
- Getis A, Ord JK: The analysis of spatial association by use of distance statistics. Geogr Anal. 1992, 24: 189-206.View ArticleGoogle Scholar
- Obbo C: HIV transmission through social and geographical networks in Uganda. Soc Sci Med. 1993, 36: 949-955. 10.1016/0277-9536(93)90086-J.View ArticlePubMedGoogle Scholar
- Webb D: Mapping the AIDS pandemic: geographical progression of HIV in South Africa 1990-93. Nurs RSA. 1994, 9: 20-21.PubMedGoogle Scholar
- Houben R, van Boeckel T, Mwinuka V, Mzumara P, Branson K, Linard C, Chimbwandira F, French N, Glynn J, Crampin A: Monitoring the impact of decentralized chronic care services on patient travel time in rural Africa - methods and results in northern Malawi. Int J Health Geogr. 2012, 11: 49-10.1186/1476-072X-11-49.View ArticlePubMedPubMed CentralGoogle Scholar
- Moise I, Kalipeni E: Applications of geospatial analysis to surveillance data: a spatial examination of HIV/AIDS prevalence in Zambia. Geo J. 2012, 77: 525-540.Google Scholar
- Epstein H, Morris M: Concurrent partnerships and HIV: an inconvenient truth. J Int AIDS Soc. 2011, 14: 13-10.1186/1758-2652-14-13.View ArticlePubMedPubMed CentralGoogle Scholar
- Kalipeni E, Oppong J, Zerai A: HIV/AIDS, gender, agency and empowerment issues in Africa. Soc Sci Med. 2007, 64: 1015-1018. 10.1016/j.socscimed.2006.10.010.View ArticlePubMedGoogle Scholar
- Nyindo M: Complementary factors contributing to the rapid spread of HIV-1 in sub-Saharan Africa: a review. East Afr Med J. 2005, 82, 1: 40-46.Google Scholar
- Kalipeni E, Ghosh J, Mkandawire-Valhmu L: The multiple dimensions of vulnerability to HIV/AIDS in Africa: A social science perspective. Women, AIDS and Access to Health Care in Sub-Saharan Africa: Approaches from the Social Sciences. Edited by: Mundi M. 2007, Barcelona, Spain: Medicus Mundi Catalunya, 45-60.Google Scholar
- Yeboah IEA: HIV/AIDS and the construction of sub-Saharan Africa: heuristic lessons from the social sciences for policy. Soc Sci Med. 2007, 64: 1128-1150. 10.1016/j.socscimed.2006.10.003.View ArticlePubMedGoogle Scholar
- Oppong JR, Harold A: Disease, ecology and environment. A Companion of Medical Geography. Edited by: Brown T, McLafferty S, Graham M. 2009, Chichester, West Sussex, U.K: Wiley-BlackwellGoogle Scholar
- Kalipeni E, Ghosh J: Concern and practice among men about HIV/AIDS in low socioeconomic income areas of Lilongwe, Malawi. Soc Sci Med. 2007, 64: 1116-1127. 10.1016/j.socscimed.2006.10.013.View ArticlePubMedGoogle Scholar
- Dodoo FNA, Zulu EM, Ezeh AC: Urban–rural differences in the socioeconomic deprivation–sexual behavior link in Kenya. Soc Sci Med. 2007, 64: 1019-1031. 10.1016/j.socscimed.2006.10.007.View ArticlePubMedGoogle Scholar
- Chirwa WC: Migrant labor, sexual networking and multi-partnered sex in Malawi. Health Transit Rev. 1997, 7 (Supplement 3): 5-15.Google Scholar
- Coffee M, Lurie MN, Garnett GP: Modelling the impact of migration on the HIV epidemic in South Africa. AIDS. 2007, 21: 343-350. 10.1097/QAD.0b013e328011dac9. 310.1097/QAD.1090b1013e328011dac328019View ArticlePubMedGoogle Scholar
- Kumwenda NI, Taha TE, Hoover DR, Markakis D, Liomba GN, Chiphangwi JD, Celentano DD: HIV-1 incidence among male workers at a sugar estate in rural Malawi. JAIDS J Acquir Immune Defic Syndr. 2001, 27: 202-208. 10.1097/00126334-200106010-00016.View ArticlePubMedGoogle Scholar
- NAC: Hiv and Syphilis Sero–Survey and National Hiv Prevalence and Aids Estimates Report for 2005. 2005, Lilongwe, Malawi: National AIDS Commission, NACGoogle Scholar
- NAC: Hiv and Syphilis Sero–Survey and National Hiv Prevalence and Aids Estimates Report for 2007. 2008, Lilongwe, Malawi: National AIDS Commission, NACGoogle Scholar
- UNAIDS WHO: Guidelines for Effective Use of Data from Hiv Surveillance Systems. 2000, Geneva: WHOGoogle Scholar
- UNAIDS/WHO Working Group on Global HIV/AIDS and STI Surveillance: The Pre-Surveillance Assessment: Guidelines for Planning Serosurveillance of HIV, Prevalence of Sexually Transmitted Infections and the Behavioral Components of Second Generation Surveillance of HIV. 2005, Geneva, Switzerland: UNAIDS and WHOGoogle Scholar
- UNAIDS, WHO: Guidelines on Estimating the Size of Populations Most At Risk to HIV. 2010, Geneva, Switzerland: UNAIDS/WHO Working Group on Global HIV/AIDS and STI SurveillanceGoogle Scholar
- Moran PAP: Notes on continuous stochastic phenomena. Biometrika. 1950, 37: 17-33. 10.1093/biomet/37.1-2.17.View ArticlePubMedGoogle Scholar
- Golub A, Gorr WL, Gould PR: Spatial diffusion of the HIV/AIDS epidemic: modeling implications and case study of aids incidence in Ohio. Geogr Anal. 1993, 25: 85-100.View ArticleGoogle Scholar
- Tobler WR: A computer movie simulating urban growth in the Detroit region. Econ Geogr. 1970, 46: 234-240.View ArticleGoogle Scholar
- Mitasova H, Mitas L, Brown WM, Gerdes DP, Kosinovsky I, Baker T: Modeling spatially and temporally distributed phenomena: new methods and tools for GRASS GIS. Int J GeogrInfSyst. 1995, 9: 433-446.Google Scholar
- Krivoruchko K: Spatial Statistical Data Analysis for GIS Users. 2011, Redlands, CA: ESRI PressGoogle Scholar
- EPA: Developing Spatially Interpolated Surfaces and Estimating Uncertainty. 2004, Triangle Park, NC, USA: United States Environmental Protection AgencyGoogle Scholar
- Anselin L, Sridharan S, Gholston S: Using exploratory spatial data analysis to leverage social indicator databases: the discovery of interesting patterns. Soc Indic Res. 2007, 82: 287-309. 10.1007/s11205-006-9034-x.View ArticleGoogle Scholar
- Anselin L: Local indicators of spatial association—LISA. Geogr Anal. 1995, 27: 93-115.View ArticleGoogle Scholar
- Anselin L, Getis O: Spatial statistical analysis and geographic information systems. Ann Reg Sci. 1992, 26: 19-33. 10.1007/BF01581478.View ArticleGoogle Scholar
- NSO: The 2008 Population and Housing Census: Main Report. 2009, Zomba, Malawi: National Statistical Office (NSO)Google Scholar
- NSO: Integrated Household Survey 2010-2011: Household Socio-Economic Characteristics Report. 2012, Zomba, Malawi: National Statistical Office (NSO)Google Scholar
- NSO: Welfare Monitoring Survey 2011. 2012, Zomba, Malawi: National Statistical Office (NSO)Google Scholar
- Altman DG, Andersen PK: Bootstrap investigation of the stability of a Cox regression model. Stat Med. 1989, 8: 771-783. 10.1002/sim.4780080702.View ArticlePubMedGoogle Scholar
- Anselin L, Syabri I, Kho Y: Geoda: an introduction to spatial data analysis. Geogr Anal. 2006, 38: 5-22. 10.1111/j.0016-7363.2005.00671.x.View ArticleGoogle Scholar
- Williams BG, Gouws E: The epidemiology of human immunodeficiency virus in South Africa. Philos Trans R Soc Lond B Biol Sci. 2001, 356: 1077-1086. 10.1098/rstb.2001.0896.View ArticlePubMedPubMed CentralGoogle Scholar
- Chirwa WC: Aliens and AIDS in southern Africa: the Malawi-South Africa debate. Afr Aff. 1998, 97: 53-79. 10.1093/oxfordjournals.afraf.a007921.View ArticleGoogle Scholar
- NSO: Malawi Biological and Behavioral Surveillance Survey 2006 and Comparative Analysis of 2004 BSS and 2006 BBSS. 2007, Lilongwe, Malawi: National AIDS Commission, NACGoogle Scholar
- Anselin L: Exploring Spatial Data with GeoDaTM: A Workbook. 2005, Urbana-Champaigl, IL: Spatial Analysis Laboratory, Department of Geography, University of Illinois, Urbana-ChampaignGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2334/14/285/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.