Although often considered a predominantly rural disease in Africa, malaria represents a leading cause of morbidity and mortality among many urban African populations 1. With an estimated 200 million urban residents currently at risk for malaria 2 and a projected doubling of the African urban population by 2030 3, an urgent need to identify risk factors for malaria in urban settings has been identified 14.

In Africa, malaria transmission is generally lower in urban compared with rural areas 56. This is largely attributed to such factors as lower vector density, higher human density, better quality housing and improved drainage in urban areas 7. Urban populations are also better able to access healthcare facilities and consequently suffer lower malaria morbidity and mortality 5. Malaria risk is often unevenly distributed across urban environments, with heterogeneities in malariometric indices measured between and within many cities 89. Evidence is also accumulating that in urban areas, the poor may be most vulnerable to the impacts of malaria 8. While urban malaria should be amenable to targeted prevention and control programmes, a better understanding of the individual, environmental, and community-based determinants within specific communities is needed 12.

A baseline survey of malaria in children aged 6 months to 5 years, conducted in 2002–3 in Kumasi, Ghana, found marked prevalence differences between communities 8. Proximity to urban agriculture was not significantly associated with malaria prevalence, suggesting the importance of other risk factors. This study reports findings from a follow-up cross-sectional survey comparing two neighbouring communities (Moshie Zongo and Manhyia), which in the previous survey were found to have Plasmodium falciparum parasite rates of 32.7% and 3.8%, respectively. The primary objective of this study was to determine whether the previously observed marked difference in parasite rates between the two communities was maintained. The secondary objectives were to compare prevalences of clinical malaria, anaemia, intestinal helminths, and malnutrition between these communities, and to identify risk factors for P. falciparum infection and anaemia.

Kumasi is the second-largest city in Ghana, with a population of 1.2 million 10. Located in the Ashanti region of central Ghana, the climate is semi-humid tropical, with peak rainfall between April and June.

Table 1 summarizes characteristics of the sampled households (97 in Manhyia, 87 in Moshie Zongo). Household characteristics were markedly different between the communities, with the exception of reported use of bednets, housescreens, and mosquito repellents/coils/sprays (p > 0.05).

A total of 296 children were tested, 148 in each community (Table 2). Sample composition (sex ratio and median age) did not significantly differ between the communities (p > 0.05). Prevalences of P. falciparum (X² = 24.45, p < 0.001), clinical malaria (X² = 14.84, p < 0.001), and anaemia (X² = 29.85, p < 0.001) were higher, and mean Hb levels lower (p < 0.001), in Moshie Zongo compared to Manhyia. There were few cases of moderate to severe anaemia (< 3% prevalence in each community). More children from Moshie Zongo were stunted (X² = 11.26, p = 0.001), had a reported history of diarrhoea (X² = 8.64, p = 0.003), and had reportedly taken malaria medication in the previous two weeks (X² = 3.95, p = 0.047). Fewer than 5% in either community had travelled recently to a rural area. Stools were examined for the first 197 children surveyed. With the exception of 4 children from neighbouring households in Manhyia who had Dicrocoelium dendriticum infections, no helminth eggs were found. As a consequence, stool samples were not collected from the remaining 99 children.

Significantly higher prevalence of P. falciparum was associated with low carer education, Northerner/Immigrant ethnicity, low SES, using coils/sprays/repellents, greater distance from a health facility, older age, and recent rural travel (Table 3). The same variables were significantly associated with higher anaemia prevalence, with the exception of younger age, and the addition of male sex, stunting, P. falciparum, and clinical malaria. No significant differences in P. falciparum or anaemia prevalence were observed with reported use of bednets or housescreens, recent antimalarial medication, wasting, underweight, or low MUAC-for-height. The proportion of fever cases with P. falciparum infection were 17.9% in Manhyia (5/28 fever cases) and 50.0% in Moshie Zongo (25/50). The PAR% of anaemia attributable to P. falciparum infection was 16.5% (95%CI: 9.5–23.3), and to malnutrition was 7.6% (95%CI: 1.4–13.9).

In adjusted models (Table 4), risk factors for P. falciparum infection included being older (OR = 1.28 per 1-year increase), recent rural travel (OR = 23.45), and lower SES (OR = 3.95 per 1-unit decrease). Significant risk factors for anaemia were P. falciparum infection (OR = 3.92), Moshie Zongo residence (OR = 3.76), male sex (OR = 2.16), and younger age (OR = 0.79 per 1-year increase).

This study demonstrated that the higher prevalence of P. falciparum among children in Moshie Zongo compared to neighbouring communities is temporally stable and not an artefact of a single study period 8. The P. falciparum rate in Moshie Zongo was higher than recently reported rates in other urban areas; Kampala 14, Coutonou 15 and Dar es Salaam 16, though lower than rates reported in Ouagadougou 17. Parasite rates from these latter surveys were obtained through primary school parasitaemia surveys and represent aggregated rates for communities in each school's catchment area. Further breakdown of rates by residence in Coutonou indicated a large parasitaemia rate difference (27.3% and 9.2%) between two intermediate zone areas 15. Heterogeneities between neighbouring communities have also been reported in Brazzaville, where rates ranged from 5.6% to 78.9% among children aged 5–9 years in 4 urban communities 9. This study adds to the growing body of evidence that malaria transmission can vary widely across cities and highlights the importance of targeting interventions towards specific communities 1.

The finding that recent rural travel was associated with increased P. falciparum prevalence has been similarly reported among other urban populations, including Dar es Salaam 16, Abidjan 18, Lusaka 19, and previously in Kumasi 8. However, few children from either study community had travelled recently, which suggests that most infections were obtained through local transmission. High P. falciparum prevalence has been recorded in rural areas of the surrounding Ashanti region 20 and Northern Ghana 21. Urban residents who travel to these areas could be at high risk for infection, as well as at higher risk for more severe outcomes following infection if functional immunity is inadequately developed, and may become sources of infection when back home 22. Promoting increased vigilance in the use of mosquito avoidance methods and/or prophylactic treatment among urban residents who travel to endemic rural areas could be an effective target for reducing some subsequent transmission within urban areas.

Two-thirds of the Moshie Zongo children sampled in our study were anaemic (an approximately two-fold higher prevalence than found among children in Manhyia). Prevalence of severe anaemia was low in both communities. When restricted to children under 5 years, the anaemia prevalence in the present study was comparable to the previous Kumasi survey 8, and to the 2003 Ghana Demographic and Health Survey for urban residents (67.8%) 10. Demonstration of heterogeneity in anaemia rates within cities underscores the importance of targeting anaemia prevention programs to both rural and urban populations, and of targeting interventions to specific urban communities.

Anaemia is often caused by a combination of malnutrition, helminth infections, infectious diseases, and haemoglobinopathies 2324. Significant reductions in anaemia prevalence have been demonstrated following malaria control programs in stable endemic areas, and usefulness of anaemia as an indicator of malaria burden and malaria control progress will depend on its specificity 24. In lower transmission urban areas, malaria may make a less significant contribution to anaemia than other factors. The PAR% estimates indicated that approximately one-sixth of anaemia cases were attributable to P. falciparum infections, and P. falciparum infection was strongly associated with anaemia after adjustment for other factors. This suggests that P. falciparum infection is an important determinant of anaemia among these Kumasi children and that anaemia may represent a useful indicator for malaria control progress in these communities.

In comparison with these Ghanaian study populations, a higher prevalence of stunting but similar prevalence of wasting was found among urban Zairian children 25. In these children from Kumasi, malnutrition (stunting, in particular) was significantly associated with anaemia in univariate tests. Stunting has been previously associated with more severe occurrence of malarial anaemia 26, and it is hypothesized that nutritional inadequacies causing stunting and underweight also impair host immunity, further exacerbating the effects of malaria 2627. In contrast, the low prevalence of intestinal helminths was surprising, compared to higher rates reported in other urban areas 25. Parasitology laboratory records (1999–2005) from the Kumasi Komfo Anokye Teaching Hospital revealed similar low rates of intestinal helminths: of the 1,181 children aged 1–9 years tested for intestinal parasites over this period, only 164 (13.9%) were positive for one or more parasites (including 10 with hookworm, 8 A. lumbricoides, and 1 T. trichiura). Kumasi is exposed to problems of crowding and sewage disposal, which are often associated with increased risk of helminth infections 28. One explanation for the low prevalence of helminth infections in Kumasi could be that many parents may be regularly de-worming their children (as reported anecdotally), however, further investigation is needed to confirm these results.

The Paracheck-Pf RDTs used in this survey detect the presence ofHistidine Rich Protein 2 (HRP-2), a water soluble protein produced by asexual stages and young gametocytes of P. falciparum 29. The benefits of using Paracheck Pf RDTs in community-based surveys are their high sensitivity at low levels of parasitaemia (> 150 parasites/μl) and high specificity to P. falciparum, and that they provide results quickly (within 15 minutes) 30. One limitation of HRP-2-based RDTs is that HRP-2 has been shown to persist and be detectable after the clinical symptoms of malaria have disappeared and the parasites have been cleared 31. Surveys using RDTs detecting HRP-2 may, therefore, overestimate prevalence of current infection, representing instead a combined prevalence of current and recent infections.

Twenty-five percent (25.3%) of children with P. falciparum infection and 36.7% of children with clinical malaria had reported taking an antimalarial within the previous 2 weeks. The proportion of P. falciparum cases with reported recent antimalarial use was higher in Moshie Zongo (28.6%) than Manhyia (15.8%). It is possible that some false positives were detected due to residual HRP-2, while recall bias may also have resulted in overestimation of reported antimalarial use. However, this evidence also suggests that treatments being used by some children may not be effective in clearing parasitaemia. Home treatment or reliance on informal drug shops as a primary treatment method for child malaria is reported among many urban populations 3233. In eastern Uganda, treatments and dosages recommended by drug shops and government health units were often incorrect, and poor caregiver compliance with prescribed dosages was common 34. Poor quality drugs with insufficient active ingredients have also been found among informal drug vendors in Cameroon 35. In addition, only 50% of fever cases in Moshie Zongo and 18% in Manhyia had a P. falciparum infection. As many urban caregivers in Ghana report treating all child fevers with antimalarials 10, this suggests some over-utilization of antimalarials by children without P. falciparum infection in our study communities. Further investigation is needed to determine the quality of drugs being sold in these areas and to assess patterns of drug utilization by caregivers.

In addition, this study highlights the vulnerability of the urban poor to the impact of malaria, as a strong association was found between SES and disease prevalence. Disaggregating the mechanisms through which poverty acts is challenging due to the complex interactions of factors 36. The SES measure was a composite and could partly reflect greater exposure to infected mosquitoes through poor quality house construction, or greater infection risk due to crowding 36. Among sampled households, Moshie Zongo residents lived further on average to the nearest health facility, and more than 1 km further to the nearest government hospital. Studies in rural areas have found an association with distance to health clinics and treatment patterns for child malaria 3738, however, the influence of distance on treatment choice in urban areas is not well-documented. It is possible that distance to health facilities in our study reflected a proxy measure for other factors which differed between the two communities. Furthermore, sample sizes were calculated to detect a difference in P. falciparum prevalence, and as a consequence, this study may have more limited power to assess differences in other factors between the communities. No data were collected to differentiate clinics by ownership type, treatment costs, or the range of services provided by each clinic; however, all of the clinics operating in Moshie Zongo were private clinics, which could have different treatment fees than government facilities. Good access to health services and timely provision of effective drug treatment is a self-targeting intervention, yet the costs of treatment and prevention often preclude the poorest people from using them 39. Greater distance to affordable health services could result in delayed seeking of treatment. Further research is needed to determine if treatment-seeking behaviours in these communities differ and to what extent these behaviours are associated with availability of different types of health services.

Heterogeneities in malariometric indices between neighbouring urban communities of Kumasi are consistent over time. The low prevalence of intestinal helminths, and the twofold higher PAR% of anaemia attributable to P. falciparum infection compared to malnutrition, indicate the importance of malaria as a cause of anaemia in this urban population. Further investigation is needed to determine if treatment-seeking behaviours in these two communities differ and to what extent behaviours are associated with availability of different types of health services.

LAR helped to design the study, performed data collection, analysis and drafted the manuscript. SLK helped to design the study, performed data collection and analysis. EK helped to design the study and assisted with the analysis. AOA helped to design the study and supervised data collection. IB performed data collection. GB designed and supervised the study. MJD conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript.

The author(s) declare that they have no competing interests.