• View in gallery
    Figure 1.

    Malaria incidence in children by age and residential distance to breeding site. Incidence estimates were derived from the final child multivariate model using age and distance splines, with all other model covariates set to baseline values.

  • View in gallery
    Figure 2.

    A, Malaria incidence in children by distance from breeding site and age group. Adjusted incidence estimates were derived from the final child multivariate model using distance and age splines, with all other model covariates set to baseline values. B, Malaria incidence in adults by distance from breeding site and age group. Adjusted incidence estimates were derived from the final adult multivariate model using distance and age splines, with all other model covariates set to baseline values.

  • 1

    van der Hoek W, Konradsen F, Dijkstra DS, Amerasinghe PH, Amerasinghe FP, 1998. Risk factors for malaria: a microepidemiological study in a village in Sri Lanka. Trans R Soc Trop Med Hyg 92 :265–269.

    • Search Google Scholar
    • Export Citation
  • 2

    Guthmann JP, Hall AJ, Jaffar S, Palacios A, Lines J, Llanos-Cuentas A, 2001. Environmental risk factors for clinical malaria: a case-control study in the Grau region of Peru. Trans R Soc Trop Med Hyg 95 :577–583.

    • Search Google Scholar
    • Export Citation
  • 3

    Trape JF, Lefebvre-Zante E, Legros F, Ndiaye G, Bouganali H, Druilhe P, Salem G, 1992. Vector density gradients and the epidemiology of urban malaria in Dakar, Senegal. Am J Trop Med Hyg 47 :181–189.

    • Search Google Scholar
    • Export Citation
  • 4

    Carter R, Mendis KN, Roberts D, 2000. Spatial targeting of interventions against malaria. Bull World Health Organ 78 :1401–1411.

  • 5

    Creasey A, Giha H, Hamad AA, El Hassan IM, Theander TG, Arnot DE, 2004. Eleven years of malaria surveillance in a Sudanese village highlights unexpected variation in individual disease susceptibility and outbreak severity. Parasitology 129 :263–271.

    • Search Google Scholar
    • Export Citation
  • 6

    Ghebreyesus TA, Haile M, Witten KH, Getachew A, Yohannes M, Lindsay SW, Byass P, 2000. Household risk factors for malaria among children in the Ethiopian highlands. Trans R Soc Trop Med Hyg 94 :17–21.

    • Search Google Scholar
    • Export Citation
  • 7

    Gunawardena DM, Wickremasinghe AR, Muthuwatta L, Weerasingha S, Rajakaruna J, Senanayaka T, Kotta PK, Attanayake N, Carter R, Mendis KN, 1998. Malaria risk factors in an endemic region of Sri Lanka, and the impact and cost implications of risk factor-based interventions. Am J Trop Med Hyg 58 :533–542.

    • Search Google Scholar
    • Export Citation
  • 8

    Snow RW, Peshu N, Forster D, Bomu G, Mitsanze E, Ngumbao E, Chisengwa R, Schellenberg JR, Hayes RJ, Newbold CI, Marsh K, 1998. Environmental and entomological risk factors for the development of clinical malaria among children on the Kenyan coast. Trans R Soc Trop Med Hyg 92 :381–385.

    • Search Google Scholar
    • Export Citation
  • 9

    Adiamah JH, Koram KA, Thomson MC, Lindsay SW, Todd J, Greenwood BM, 1993. Entomological risk factors for severe malaria in a peri-urban area of The Gambia. Ann Trop Med Parasitol 87 :491–500.

    • Search Google Scholar
    • Export Citation
  • 10

    Thomas CJ, Lindsay SW, 2000. Local-scale variation in malaria infection amongst rural Gambian children estimated by satellite remote sensing. Trans R Soc Trop Med Hyg 94 :159–163.

    • Search Google Scholar
    • Export Citation
  • 11

    Wolff CG, Schroeder DG, Young MW, 2001. Effect of improved housing on illness in children under 5 years old in northern Malawi: cross sectional study. BMJ 322 :1209–1212.

    • Search Google Scholar
    • Export Citation
  • 12

    Gamage-Mendis AC, Carter R, Mendis C, De Zoysa AP, Herath PR, Mendis KN, 1991. Clustering of malaria infections within an endemic population: risk of malaria associated with the type of housing construction. Am J Trop Med Hyg 45 :77–85.

    • Search Google Scholar
    • Export Citation
  • 13

    Lindsay SW, Emerson PM, Charlwood JD, 2002. Reducing malaria by mosquito-proofing houses. Trends Parasitol 18 :510–514.

  • 14

    Lindsay SW, Snow RW, 1988. The trouble with eaves; house entry by vectors of malaria. Trans R Soc Trop Med Hyg 82 :645–646.

  • 15

    Guthmann JP, Llanos-Cuentas A, Palacios A, Hall AJ, 2002. Environmental factors as determinants of malaria risk. A descriptive study on the northern coast of Peru. Trop Med Int Health 7 :518–525.

    • Search Google Scholar
    • Export Citation
  • 16

    Ghebreyesus TA, Haile M, Witten KH, Getachew A, Yohannes AM, Yohannes M, Teklehaimanot HD, Lindsay SW, Byass P, 1999. Incidence of malaria among children living near dams in northern Ethiopia: community based incidence survey. BMJ 319 :663–666.

    • Search Google Scholar
    • Export Citation
  • 17

    Clarke SE, Bøgh C, Brown RC, Walraven GE, Thomas CJ, Lindsay SW, 2002. Risk of malaria attacks in Gambian children is greater away from malaria vector breeding sites. Trans R Soc Trop Med Hyg 96 :499–506.

    • Search Google Scholar
    • Export Citation
  • 18

    Thomson MC, D’Alessandro U, Bennett S, Connor SJ, Langerock P, Jawara M, Todd J, Greenwood BM, 1994. Malaria prevalence is inversely related to vector density in The Gambia, West Africa. Trans R Soc Trop Med Hyg 88 :638–643.

    • Search Google Scholar
    • Export Citation
  • 19

    Lenz R, 1988. Jakarta kampung morbidity variations: some policy implications. Soc Sci Med 26 :641–649.

  • 20

    Mendez F, Carrasquilla G, Munoz A, 2000. Risk factors associated with malaria infection in an urban setting. Trans R Soc Trop Med Hyg 94 :367–371.

    • Search Google Scholar
    • Export Citation
  • 21

    Bashford G, Richens J, 1992. Travel to the coast by highlanders and its implications for malaria control. P N G Med J 35 :306–307.

  • 22

    Sevilla-Casas E, 1993. Human mobility and malaria risk in the Naya river basin of Colombia. Soc Sci Med 37 :1155–1167.

  • 23

    Burkot TR, Dye C, Graves PM, 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am J Trop Med Hyg 40 :229–234.

    • Search Google Scholar
    • Export Citation
  • 24

    Binka FN, Indome F, Smith T, 1998. Impact of spatial distribution of permethrin-impregnated bed nets on child mortality in rural northern Ghana. Am J Trop Med Hyg 59 :80–85.

    • Search Google Scholar
    • Export Citation
  • 25

    Mauny F, Viel JF, Handschumacher P, Sellin B, 2004. Multilevel modelling and malaria: a new method for an old disease. Int J Epidemiol 33 :1337–1344.

    • Search Google Scholar
    • Export Citation
  • 26

    Deressa W, Olana D, Chibsa S, 2004. Magnitude of malaria admissions and deaths at hospitals and health centers in Oromia, Ethiopia. Ethiop Med J 42 :237–246.

    • Search Google Scholar
    • Export Citation
  • 27

    Yohannes M, Petros B, 1996. Urban malaria in Nazareth, Ethiopia: parasitological studies. Ethiop Med J 34 :83–91.

  • 28

    Cameron AC, Trivedi PK, 1998. Basic count regression. Regression Analysis of Count Data. Cambridge University Press: Cambridge, 59–95.

  • 29

    Willis S, Akhwale JKL, Kaneko A, Eto H, Obonyo C, Bjorkman A, Kobayakawa T, 2004. Anemia and malaria at different altitudes in the western highlands of Kenya. Acta Trop 91 :167–175.

    • Search Google Scholar
    • Export Citation
  • 30

    Sintasath D, Lynch M, Kleinhau E, Bretas G, Shilulu J, Brantly E, Graves P, Beier J, 2005. Malaria prevalence and associated risk factors in Eritrea. Am J Trop Med Hyg 72 :682–687.

    • Search Google Scholar
    • Export Citation
  • 31

    Lindsay SW, Martens WJ, 1998. Malaria in the African highlands: past, present and future. Bull World Health Organ 76 :33–45.

  • 32

    Robert V, Macintyre K, Keating J, Trape JF, Duchemin JB, Warren M, Beier JC, 2003. Malaria transmission in urban sub-Saharan Africa. Am J Trop Med Hyg 68 :169–176.

    • Search Google Scholar
    • Export Citation
  • 33

    Mahande A, Mosha F, Mahande J, Kweka E, 2007. Feeding and resting behaviour of malaria vector, Anopheles arabiensis with reference to zooprophylaxis. Malar J 6 :100.

    • Search Google Scholar
    • Export Citation
  • 34

    Koram KA, Bennett S, Adiamah JH, Greenwood BM, 1995. Socioeconomic risk factors for malaria in a peri-urban area of The Gambia. Trans R Soc Trop Med Hyg 89 :146–150.

    • Search Google Scholar
    • Export Citation
  • 35

    Gu W, Novak RJ, 2005. Habitat-based modeling of impacts of mosquito larval interventions on entomological inoculation rates, incidence, and prevalence of malaria. Am J Trop Med Hyg 73 :546–552.

    • Search Google Scholar
    • Export Citation
  • 36

    Ribeiro JM, Seulu F, Abose T, Kidane G, Teklehaimanot A, 1996. Temporal and spatial distribution of anopheline mosquitos in an Ethiopian village: implications for malaria control strategies. Bull World Health Organ 74 :299–305.

    • Search Google Scholar
    • Export Citation
Past two years Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 328 148 19
PDF Downloads 97 43 9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

 

Individual and Household Level Factors Associated with Malaria Incidence in a Highland Region of Ethiopia: A Multilevel Analysis

Ingrid PetersonMedical Research Council Laboratories, Fajara, The Gambia; Department of Health Sciences, Lehman College, City University of New York, New York, New York; Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York; International Center for AIDS Care and Treatment Programs (ICAP), Mailman School of Public Health, Columbia University, New York, New York

Search for other papers by Ingrid Peterson in
Current site
Google Scholar
PubMed
Close
,
Luisa N. BorrellMedical Research Council Laboratories, Fajara, The Gambia; Department of Health Sciences, Lehman College, City University of New York, New York, New York; Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York; International Center for AIDS Care and Treatment Programs (ICAP), Mailman School of Public Health, Columbia University, New York, New York

Search for other papers by Luisa N. Borrell in
Current site
Google Scholar
PubMed
Close
,
Wafaa El-SadrMedical Research Council Laboratories, Fajara, The Gambia; Department of Health Sciences, Lehman College, City University of New York, New York, New York; Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York; International Center for AIDS Care and Treatment Programs (ICAP), Mailman School of Public Health, Columbia University, New York, New York

Search for other papers by Wafaa El-Sadr in
Current site
Google Scholar
PubMed
Close
, and
Awash TeklehaimanotMedical Research Council Laboratories, Fajara, The Gambia; Department of Health Sciences, Lehman College, City University of New York, New York, New York; Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York; International Center for AIDS Care and Treatment Programs (ICAP), Mailman School of Public Health, Columbia University, New York, New York

Search for other papers by Awash Teklehaimanot in
Current site
Google Scholar
PubMed
Close

Multilevel analysis was used to identify individual and household level factors associated with malaria risk in 1,367 individuals in a peri-urban area of highland Ethiopia. Living within 450 m of a major vector-breeding site accounted for 38.78% and 78.49% of between-household variance in malaria incidence in adults and children, respectively. In adults, other individual level factors associated with malaria risk were regular or recent travel to rural areas (incidence rate ratio [IRR] = 12.96; 95% confidence interval [CI] = 4.05, 41.48) and having an indoor job (IRR = 0.37; 95% CI = 0.16, 0.87). Household level factors associated with adult malaria risk were low vegetation level in compound (IRR = 0.27; 95% CI = 0.10, 0.78), tidy compound (IRR = 0.29; 95% CI = 0.12, 0.71), household use of preventive measures (IRR = 0.31; 95% CI = 0.13, 0.74), and the number of 5- to 9-year-old children in the household (IRR = 1.66; 95% CI = 1.08, 2.53). Aside from distance to the vector-breeding site, few other malaria risk factors were identified in children. Malaria interventions in highland African communities should address household level factors associated with malaria clustering.

INTRODUCTION

There is ample epidemiologic evidence that malaria is often highly clustered within communities,13 particularly in areas of low transmission.4 Estimates of the proportion of households carrying 50% of the malaria burden in a community were 8% and 18% in studies from Belize and Tigray in Ethiopia, respectively.4 Such clustering has been noted to occur over small distances; in Senegal, for example, there was a 3-fold difference in childhood malaria incidence in households separated by only 100 m.3 Susceptibility to malaria infection may also vary greatly between individuals. In a low-transmission region of Sudan, 32% of village residents reported no malaria symptoms, whereas others experienced up to eight clinical episodes over 11 years of follow-up.5

Given this observed clustering of malarial illness, community-based malaria interventions should efficiently target individuals and households at highest risk of malaria. A large number studies have attempted to identify household and individual level factors associated with malaria. Factors examined include housing type, 1,2,615 proximity to vector-breeding site and vector abundance,13,7,1618 socioeconomic status, 1,2,6,19 age,13,6,20 occupation,2,19,20 sex,1,2,6,19 residential mobility and travel, 2022 knowledge of malaria, 2,20 household size, 19 sleeping room density, 2,6,8 presence of domestic animals near home, 2,8,23 use of preventive methods (e.g., coils, house spraying), 1,2,8 bed net use, 1,8,20,24 and local area population density.8 However, malaria risk factor studies have often failed to examine a wide range of both individual and household level factors; their results may therefore be confounded by factors omitted from analysis. Furthermore, only one study has used multilevel modeling to quantify the proportion of variance in malaria incidence attributable to specific risk factors; however, individual level factors were not examined. 25 Finally, few studies examine malaria risk factors in urban settings located in epidemic prone, highland areas of Africa. In such settings, the targeting of intervention activities toward individual and household level factors may be particularly effective as epidemics are initially confined to small geographic areas. This paper attempts to fill the gaps in the malaria risk factor literature by using multilevel modeling to identify individual and household factors simultaneously, associated with malaria risk in an urban community located in Ethiopia’s epidemic-prone highland fringe.

MATERIALS AND METHODS

Study area and population

The study was conducted in an area of the city of Adama (formerly Nazareth), a city of ~160,000 residents located 120 km southeast of the capital city of Addis Ababa. The study area was a 1.8-km2 section of kebele 11, one of 20 administrative units that make up the city. In August 2003, the study area had 8,697 inhabitants of various ethnic origins (Oromo, Amhara, Tigraway, Guragie, and others). At an altitude of 1,600 m, malaria epidemics are frequent, and clinical immunity to malaria is low in the population. 26 During peak transmission months (September to November), population prevalence of parasitemia reaches 2–3%, of which ~40% are caused by P. falciparum and ~60% are caused by P. vivax.27 Because of seasonal flooding, malaria incidence is higher in kebele 11 compared with other kebeles; from 1995 to 2002, rates ranged from 15.84 to 82.03 episodes per 1,000 population (Oromiya Regional Health Ministry, unpublished data).

Subjects

Inclusion criteria of children and adults in the study were as follows: age > 1 year, continuous residence in study area household since July 1, 2003, and intention to remain in the study area for the duration of study follow-up (through November 30, 2003). Study households were identified through systematic random sampling of every fourth house to obtain a sample of 310 households (1,460 individuals). After excluding individuals reporting non-continuous residence and those with missing data, the final sample contained 294 household (1,367 individuals). Oral consent was sought from the head of household before collection of household survey data collection. The study protocol was approved by the ethical review committees of the Ethiopian Ministry of Health and Columbia University’s Mailman School of Public Health.

Data collection

Malaria incidence

Data on incident malaria infections occurring from August 1 to November 30, 2003 were obtained by assigning a unique study identification number to study households, which was used to link malaria infections to that household. The head of household was given a study identification card with instructions that household members should present the card on all visits to the Adama Malaria Laboratory, which is 2 km from the study site. This is the only facility in Adama where free microscopic diagnosis and treatment of malaria are offered.

House distance to the vector-breeding site

From August 1–3, 2003, two experienced vector control technicians surveyed the study area to identify seasonally permanent Anopheles vector breeding sites. A single site was identified; this was a 0.50-km2 flood plain bordering the study area, which was used as a watering hole for cattle during the heavy rains (July to October) and for crop cultivation from January to October. Over study follow-up, water temperature and larva samples were taken every 2 weeks from the perimeter of the site closest to the study area (~0.10 km long). The median larva/pupae count was 124 per 100 dips (Interquartile Range = 150.5); median water temperature was 22.3°C. Using a hand-held Global Positioning System (GPS), the geographic coordinates of all households were measured to a horizontal accuracy of < 10 m. Continuous variables were created to define the distance of each household to the vector-breeding site.

Household survey data

Over the month of August 2003, four trained interviewers visited study households. Using a standardized form, a census was taken of all household residents. Data were collected on demographic, travel, employment, and bed net use factors, as well as house characteristics, livestock ownership, and use of malaria preventive measures, including insecticide-treated net (ITN) use. Study interviewers used exemplary photographs to rate housing quality, vegetation cover near homes, and compound tidiness on a three-point scale. To control for potential confounding by ITN use that was initiated after the household survey was completed, study households were revisited in December 2003 and asked about the number and date of acquisition of ITNs in the household. (ITNs were promoted and sold by local government in the study area starting in September 2003 in an attempt to halt the ongoing malaria epidemic.)

Outcome

The primary outcome of the study was the development of incident P. falcipiarum and P. vivax clinical malaria cases in children and adults occurring from August 1 to November 30, 2003. The outcome factor in all analyses was malaria incidence, calculated as the cumulative number of malaria infections (P. vivax, P. falciparum, mixed infections) in individuals over 4 months of study follow-up, divided by the person-time in the study that was taken as 4 months for all study subjects.

Independent variables

Explanatory factors were obtained from the household survey. Several factors were constructed from the primary data for use in analysis. Linear splines were constructed for age and residential distance from the breeding site. Splines were constructed by identifying breakpoints (knots) where the slope of the regression line of malaria incidence on age and distance, respectively, was thought to change. Within categories defined by the knots, spline values were continuous. For age splines, knots were placed at 9 and 17 years for children and at 25, 59, and 95 years for adults. For distance splines, knots were placed at 450, 750, and 1,250 m for children and 350 and 1,250 m for adults. Knot locations for linear splines and cut-off points for dichotomized factors were determined by comparing Aikake information criteria (AIC) values in univariate models, which regressed malaria counts on the factor of interest using different knots or cutoff points. Distance to the unique public tap in the study area was defined dichotomously as < 200 and ≥ 200 m. A dichotomous factor for livestock keeping was set to one for households with more than seven livestock animals within 10 m of house and was otherwise set to zero. In adults, the factor married was set to one for currently married subjects and set to zero for all others. The factor works indoors was set to one for adults working for pay (in the past 2 months) at a job classified as indoor and was otherwise set to zero. Indoor jobs included factory worker, office worker, clerk, teacher, nurse, nurse’s aid, shop owner, and shop clerk. The factor works outdoors was set to one for adults working for pay (in the past 2 months) at a job classified as outdoor and was otherwise set to zero. Outdoor jobs included agricultural worker, herder, construction worker, day or night watchman, trucker, and casual day laborer. The dichotomous factors rural travel and urban travel were set to one for subjects who had traveled in the past month or traveled regularly (at last once every 6 months) to rural and urban areas, respectively. The dichotomous factor tidy compound was set to one for households with compounds scored by study interviewer as “very tidy” and set to zero for those scored as “somewhat tidy” or “not tidy.” Compounds with areas covered with vegetative or household litter (such as discarded cans, buckets, or dishes), open pits (used to make mud bricks), or piles of construction materials were labeled as “somewhat tidy” or “not tidy.” The dichotomous factor household uses preventive measures was set to one for households using any of the following measures: cutting grass near house, filling in ditches near house, using store bought insecticides, and burning coils; this factor was set to zero for households using none of these measures. Preventive measures were aggregated because several were infrequent in the population.

Statistical analysis

Study data were analyzed by first examining univariate associations between malaria incidence and explanatory factors by regressing a single factor against individual malaria counts. In multivariable modeling, factors were selected based on hypothesized relationships, the statistical performance of factors in univariate analysis, and correlations among the factors. The selected factors were entered into the model in the order of their presumed importance; those that were not statistically significant were removed unless deemed important based on evidence from the literature. Within- and cross-level interactions were assessed by adding one interaction term at a time to the final model. All analyses were performed separately for children (< 18 years) and adults (≥ 18 years) using random effects Poisson regression (REPR) with a household-specific random intercept term and common slope coefficients for explanatory factors. The REPR model was chosen after fixed effect Poisson regression (FEPR) showed significant over-dispersion in the study data, indicating the temporal or spatial clustering of malaria counts. 28 Over-dispersed count data may be modeled using an individual or a household random intercept. A household random intercept term was chosen because malaria clusters to a greater extent on the household rather than the individual level.

The REPR model is fit by assuming that the number of malaria infections for each household (di) has a greater variance than that predicted by the Poisson distribution. The model is generated by allowing malaria counts to follow the Poisson distribution, conditional on random household effects (ui). In the model, each household has a unique over-dispersion term (ƒi), which follows a gamma distribution with a mean of 1 and an unknown variance.

Huber/White/sandwich estimators were used to calculate coefficient SEs. Incidence rate ratios (IRRs) and 95% confidence intervals (CIs) were calculated from univariate and multivariable analyses.

To estimate the percent of between-household variance in malaria incidence caused by explanatory factors, changes in random effect terms were assessed in various models. First, an empty model, containing only a random effect term, was developed. An individual factors only model, containing only a random effect term and individual-level factor, was developed. Household level explanatory factors were added one by one to the empty model and to the individual factors only model to examine reductions in the random effect term associated with each factor. The random effect term in each model represents between-household variance in malaria incidence not accounted for by factors in the model. The percent reduction in this variance associated with explanatory factors was calculated by comparing random effect terms between models.

The percent of this variance explained by all individual level factors was obtained as follows:

RESULTS

Malaria incidence

Malaria incidence over study follow-up was 90.5 per 1,000 population, a rate higher than previously recorded in Adama during the period from 1994 to 2001 (Oromiya Regional Health Ministry, unpublished data). Malaria was highly clustered in households in the study population. Sixty-five percent of malaria cases occurred in only 5% of households; only 16% of households had any malaria cases. Overall, 5.8% of individuals had a malaria episode; 2% of individuals had two or more malaria episodes to a maximum of four (analyses not shown in tables). Malaria incidence was significantly higher in children (127.5 per 1,000 population) than adults (64.9 per 1,000 population; Table 1).

Individual factors associated with malaria incidence

Few individual level factors were associated with malaria incidence in univariate analysis ( Table 1 ). In children, only “age 5–9 years” (referent, 15–17 years) was statistically significant; in adults, only age and work status were associated with malaria risk. In both adults and children, ITN use was rare at baseline; however, no malaria case occurred in adults using ITNs at baseline.

Household factors associated with malaria incidence

The mean household size was 4.7 individuals (Table 2). On average, there were 0.4 malaria episodes per household. Although located in an urban area, many study households possessed characteristics prevalent in rural settings. Seventy-one percent of houses were of mud-brick construction, 39% kept livestock animals, 81% had no running water in their home, and 11.2% had no toilet or outhouse. Small scale agriculture was common in study households, a fact reflected by the low occurrence (31%) of household compounds with a low level of vegetation. Many households used household-level prevention against malaria (65%; primarily burning coils or cutting vegetation around the home). However, few (3.1%) had any window screens or owned an ITN at baseline (3.4%). A number of household level factors were associated with malaria in univariate analysis including household use of preventive measures, absence of toilet or outhouse on compound, greater than seven livestock animals kept within 10 m of home, distance in meters vector breeding site, < 350 m from vector breeding site, household has child(ren) age 5–9 years, and household has member(s) with outdoor job(s), or indoor job(s) (Table 2).

Factors associated with malaria infection in multivariable analysis

In multivariable modeling, age (see Table 3 for IRRs of linear splines), distance from the major vector breeding site (see Table 3 for IRRs of linear splines), and number of adults with indoor jobs (IRR = 0.54; 95% CI = 0.34, 0.88) were the only factors significantly associated with malaria incidence in children (Table 3). No interaction terms were found to be significant. To show the influence of age and distance on malaria incidence, adjusted and unadjusted malaria incidence was graphed by age and distance. In children, residing at all distances from the breeding site malaria incidence rose to a peak at age 9 and gradually declined to age 17 years (Figure 1). In children of all ages, house distance to the breeding site exerted a profound effect on malaria incidence, particularly within 350 m from the breeding site (Figure 2A). At 150 m from the breeding site, the mean adjusted malaria incidence in children 0–17 years of age was 1,374 per 1,000 population compared with 373 per 1,000 population at 350 m (analysis not shown).

In the adult multivariable model, two individual level factors were significantly associated with malaria incidence; no interaction terms were significant in the model. Recent or regular travel to rural areas was associated with a substantial increase in adult malaria incidence (IRR = 12.96; 95% CI = 4.05, 41.48). Conversely, working indoors was associated with decreased malaria incidence (IRR = 0.37; 95% CI = 0.16, 0.87). Having a tidy compound (IRR = 0.29; 95% CI = 0.12, 0.71), low vegetation cover in compound (IRR = 0.27; 95% CI = 0.10, 0.78), and household use of preventive measures (IRR = 0.31; 95% CI = 0.13, 0.74) were household level factors associated with lower malaria incidence. Keeping more than seven livestock animals < 10 m from the house (IRR = 3.49; 95% CI = 0.98, 12.45) was associated with higher incidence (Table 3). Adult malaria incidence was also strongly associated with house distance to the vector breeding site, particularly for adults ≤ 60 years of age. For example, in adults 18–25 years of age, malaria incidence was 910.5 per 1,000 populations at 150 m compared with 50.7 per 1,000 populations at 350 m; the incidence curve is nearly flat at distances beyond 350 m (Figure 2B).

One household composition factor, number of children in household 5–9 years of age, was associated with increased malaria risk in adults (IRR = 1.66; 95% CI = 1.08, 2.53). To examine whether this effect may have been caused by transmission from children to adults, monthly adult malaria incidence was compared before and after the first 5- to 9-year-old malaria case in the household. In households with at least one 5- to 9-year-old malaria case (26 cases in 22 households), cumulative adult malaria incidence was higher before (0.10) compared with after (0.06) the first case. Thus, there is no evidence that the observed association is attributable to malaria transmission from children to adults within households.

Between-household variance in malaria incidence caused by explanatory factors

Table 4 shows reductions in the household random effect term attributable to explanatory factors. The household random effect from the empty models shows that, in the absence of any explanatory factors, between-household variance in malaria incidence was highly clustered at the household level. It was 4.88 (between-household variance is calculated as the random effect plus one, i.e., 1 + 3.88) and 5.03 (i.e., 1 + 4.03) times the Poisson variance in children and adults, respectively. Percent reductions in the random effect term comparing the empty model to the model with individual level factors showed that individual level factors explained only 4.12% of the between-household variance in malaria incidence in children and 10.42% in adults. Thus, household level clustering of malaria is not caused by groupings of high-risk individuals in certain households.

In children, house distance to vector-breeding site and number of adults in household with an indoor job accounted for 79% and 10.8% of the between-household variance in child malaria incidence, respectively, compared with the model containing only individual level factors. Despite containing few explanatory factors, the final model reduced the random effect term by 82% compared with the empty model, indicating that the model explained a large proportion of between household differences in child malaria incidence.

In adults, several household level factors were important in explaining between-household differences in malaria incidence. House distance to vector breeding site, tidy compound, and household uses preventive measures accounted for reductions in the between-household variance in malaria incidence by 38.8%, 15.2%, and 16.6%, respectively, compared with the model containing only individual level factors. Thus, each of these single household factors alone accounted for a moderate portion of the between-household variance in adult malaria incidence. The final adult model explained 67.7% of the between-household variance in adult malaria incidence.

DISCUSSION

This study identified individual and household level factors associated with malaria in an urban setting in an epidemic-prone highland area of Ethiopia. Malaria was highly clustered in households. House proximity to the major vector-breeding site was the most important factor determining household differences in malaria incidence, accounting for 79.38% and 38.78% of between-household variance in malaria incidence in children and adults, respectively. Overall, only age, distance to vector breeding site, and number of adults in household with an indoor job were identified as malaria risk factors for children; nevertheless, the final model explained 82% of between-household differences in child malaria. In adults, two individual level factors (work indoors and regular or recent travel to rural areas) and six household level factors (distance to vector breeding site, low vegetation in compound, tidy compound, household uses preventive measures, more than seven livestock animals, and number of children 5–9 years of age in household) were associated with malaria risk.

In malaria-epidemic prone areas, such as Adama where this study was conducted, acquired immunity to malaria is assumed to be low in all age groups. However, in our study, malaria incidence was significantly higher in children than adults and peaked in children at age 9. This finding is consistent with other studies of malaria conducted in highland Africa, which have found that malaria infection occurs in all age groups and is not consistently highest in the < 5 age group. 6,26,29,30 It is unclear why, in high altitude settings, older children often have higher malaria incidence than younger children. However, it is possible that, at high altitudes, Anopheles activity is quite spatially clustered and that human activity within these clusters plays a pronounced role in malaria exposure. Older children may be more likely to work and play where the Anopheles vector is present, especially at dusk when Anopheles becomes active.

In Africa, lowland31 and rural settings 32 generally have higher malaria transmission compared with highland and urban settings. This study concurred with this in its finding that rural travel was strongly associated with adult malaria incidence. Thus, rural travel should be recognized as an important malaria risk factor in urban African settings. Urban travel was not found to be associated with malaria risk. Perhaps this is because most urban travel in the study area was to Addis Ababa, which is largely malaria free. Future studies should examine the association between the altitude of travel destinations and malaria risk.

The finding of an association between the number of adults in the household with indoor job and decreased malaria risk in children and of the association between the number of children age 5–9 years and increased malaria risk in adults provides evidence that household composition influences malaria risk in individuals, a topic that has not yet received attention in malaria risk factor studies. Household composition factors may correlate with the number of household members infected with malaria, who generate transmission within the household. In this study, however, adult incidence was higher before rather than after child malaria cases (age 5–9 years), which suggests that the number of children in the household may be an indicator of behavioral characteristics associated with malaria risk.

The finding that keeping seven or more livestock animals was associated with increased malaria risk in adults is highly relevant in peri-urban African settings, where keeping cattle close to houses is common. The Anopheles arabiensis vector is a mildly zoophilic feeder, and it has been posited that the presence of cattle near homes diverts biting mosquitoes away from humans. 33 However, cattle may also attract mosquitoes, either by serving as bait or by creating mosquito breeding and resting sites near livestock pens. Our study found no association between malaria incidence and housing quality, which concurs with some studies 1,8,11,19 but contradicts others. 6,7,34 This null finding may be explained by the low prevalence in the study area of housing characteristics typically associated with malaria risk (open eaves, open ceilings, and window screens) 19 or because malaria transmission occurred primarily outdoors near but not in the house.

There are a couple of possible explanations for the study’s finding that distance to breeding site was a very important malaria risk factor in children but only one of several malaria risk factors for adults. First, it is possible that there is less variability among children in risk-modifying activities compared with adults. For example, living in a tidy compound may have a greater impact on malaria risk in adults than in children if children tend to move more between compounds (to play or work) than adults. Second, in general, as the incidence of an infectious diseases increases, there opportunity for behavioral and other factors to modify risk of infection decreases. This study was carried out under epidemic conditions, where malaria incidence over study follow-up for subjects residing within 350 m of the vector breeding site was very high in children (721 per 1,000) and only moderately high in adults (248 per 1,000). Thus, whereas several risk factors of moderate strength were associated with malaria risk in adults, few were identified in children.

This study possesses a number of strengths including its use of a large, population-based cohort, its assessment of a wide range of both individual and household factors with regard to malaria risk, and the use of multilevel modeling. In addition, the study identified important malaria risk factors in a highland urban setting in Africa under epidemic conditions. Relatively few studies have been conducted in such settings, yet malaria epidemics in the densely populated highlands of Africa comprise a large and growing proportion of Africa’s malaria burden. Despite these obvious strengths, the study also had some limitations. The study did not assess malaria episodes treated at home or in private clinics; study results may therefore reflect the combined effect of malaria risk factors and health care utilization patterns. Another limitation is that travel was determined retrospectively based on travel in the past month or regular travel (at least every 6 months). The observed association between travel and malaria risk may therefore reflect unmeasured factors associated with being a frequent traveler rather than to the travel itself.

Findings from our study have important policy implications. The importance of house distance to the major vector-breeding site in determining malaria risk suggests that vector control strategies targeted at such sites could greatly reduce the malaria burden in urban communities in Africa. Computer modeling and observational studies have shown that larvicide application can be highly effective in reducing malaria incidence in low and moderate transmission areas, if implemented at most major breeding sites, with a minimum lag time. 35,36 In addition, it is clear that urban malaria control programs in Africa must encourage community participation to reduce vector-breeding and resting sites around residential compounds. Interventions such as maintaining a tidy household, reducing residential vegetation cover, and decreasing the presence of livestock near homes are all likely to be fruitful in lowering malaria incidence.

Table 1

Individual attributes of children and adults and their crude association with malaria risk

Table 1
Table 2

Household attributes of 294 study households and their crude association with malaria risk

Table 2
Table 3

Full model of individual and household level malaria risk factors in children and adults

Table 3
Table 4

Percent reduction in household level variance in malaria incidence attributable to risk factors in children and adults

Table 4
Figure 1.
Figure 1.

Malaria incidence in children by age and residential distance to breeding site. Incidence estimates were derived from the final child multivariate model using age and distance splines, with all other model covariates set to baseline values.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 80, 1; 10.4269/ajtmh.2009.80.103

Figure 2.
Figure 2.

A, Malaria incidence in children by distance from breeding site and age group. Adjusted incidence estimates were derived from the final child multivariate model using distance and age splines, with all other model covariates set to baseline values. B, Malaria incidence in adults by distance from breeding site and age group. Adjusted incidence estimates were derived from the final adult multivariate model using distance and age splines, with all other model covariates set to baseline values.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 80, 1; 10.4269/ajtmh.2009.80.103

*

Address correspondence to Ingrid Peterson, MRC Laboratories, Fajara, Atlantic Rd., P.O. Box 273, Banjul, The Gambia. E-mail: idpet2@gmail.com

Authors’ addresses: Ingrid Peterson, MRC Laboratories, Fajara, Atlantic Rd., P.O. Box 273, Banjul, The Gambia. Luisa N. Borrell, Department of Health Sciences, Lehman College, City University of New York, 250 Bedford Park Blvd. West, Gillet 336, Bronx, NY 10468. Wafaa El-Sadr, International Center for AIDS Care & Treatment Programs, Columbia University, Mailman School of Public Heath, 722 West 168th St., Room #709, New York, NY 10032. Awash Teklehaimanot, Earth Institute Center for Global Health & Econ Dev, Columbia University, Hogan Hall, 100 Level, Mail Code: 3277, New York, NY 10027.

Acknowledgments: The authors thank all the community members and patients who participated in the study, as well as the laboratory staff, vector control technicians and data collectors whose diligence made this study possible.

REFERENCES

  • 1

    van der Hoek W, Konradsen F, Dijkstra DS, Amerasinghe PH, Amerasinghe FP, 1998. Risk factors for malaria: a microepidemiological study in a village in Sri Lanka. Trans R Soc Trop Med Hyg 92 :265–269.

    • Search Google Scholar
    • Export Citation
  • 2

    Guthmann JP, Hall AJ, Jaffar S, Palacios A, Lines J, Llanos-Cuentas A, 2001. Environmental risk factors for clinical malaria: a case-control study in the Grau region of Peru. Trans R Soc Trop Med Hyg 95 :577–583.

    • Search Google Scholar
    • Export Citation
  • 3

    Trape JF, Lefebvre-Zante E, Legros F, Ndiaye G, Bouganali H, Druilhe P, Salem G, 1992. Vector density gradients and the epidemiology of urban malaria in Dakar, Senegal. Am J Trop Med Hyg 47 :181–189.

    • Search Google Scholar
    • Export Citation
  • 4

    Carter R, Mendis KN, Roberts D, 2000. Spatial targeting of interventions against malaria. Bull World Health Organ 78 :1401–1411.

  • 5

    Creasey A, Giha H, Hamad AA, El Hassan IM, Theander TG, Arnot DE, 2004. Eleven years of malaria surveillance in a Sudanese village highlights unexpected variation in individual disease susceptibility and outbreak severity. Parasitology 129 :263–271.

    • Search Google Scholar
    • Export Citation
  • 6

    Ghebreyesus TA, Haile M, Witten KH, Getachew A, Yohannes M, Lindsay SW, Byass P, 2000. Household risk factors for malaria among children in the Ethiopian highlands. Trans R Soc Trop Med Hyg 94 :17–21.

    • Search Google Scholar
    • Export Citation
  • 7

    Gunawardena DM, Wickremasinghe AR, Muthuwatta L, Weerasingha S, Rajakaruna J, Senanayaka T, Kotta PK, Attanayake N, Carter R, Mendis KN, 1998. Malaria risk factors in an endemic region of Sri Lanka, and the impact and cost implications of risk factor-based interventions. Am J Trop Med Hyg 58 :533–542.

    • Search Google Scholar
    • Export Citation
  • 8

    Snow RW, Peshu N, Forster D, Bomu G, Mitsanze E, Ngumbao E, Chisengwa R, Schellenberg JR, Hayes RJ, Newbold CI, Marsh K, 1998. Environmental and entomological risk factors for the development of clinical malaria among children on the Kenyan coast. Trans R Soc Trop Med Hyg 92 :381–385.

    • Search Google Scholar
    • Export Citation
  • 9

    Adiamah JH, Koram KA, Thomson MC, Lindsay SW, Todd J, Greenwood BM, 1993. Entomological risk factors for severe malaria in a peri-urban area of The Gambia. Ann Trop Med Parasitol 87 :491–500.

    • Search Google Scholar
    • Export Citation
  • 10

    Thomas CJ, Lindsay SW, 2000. Local-scale variation in malaria infection amongst rural Gambian children estimated by satellite remote sensing. Trans R Soc Trop Med Hyg 94 :159–163.

    • Search Google Scholar
    • Export Citation
  • 11

    Wolff CG, Schroeder DG, Young MW, 2001. Effect of improved housing on illness in children under 5 years old in northern Malawi: cross sectional study. BMJ 322 :1209–1212.

    • Search Google Scholar
    • Export Citation
  • 12

    Gamage-Mendis AC, Carter R, Mendis C, De Zoysa AP, Herath PR, Mendis KN, 1991. Clustering of malaria infections within an endemic population: risk of malaria associated with the type of housing construction. Am J Trop Med Hyg 45 :77–85.

    • Search Google Scholar
    • Export Citation
  • 13

    Lindsay SW, Emerson PM, Charlwood JD, 2002. Reducing malaria by mosquito-proofing houses. Trends Parasitol 18 :510–514.

  • 14

    Lindsay SW, Snow RW, 1988. The trouble with eaves; house entry by vectors of malaria. Trans R Soc Trop Med Hyg 82 :645–646.

  • 15

    Guthmann JP, Llanos-Cuentas A, Palacios A, Hall AJ, 2002. Environmental factors as determinants of malaria risk. A descriptive study on the northern coast of Peru. Trop Med Int Health 7 :518–525.

    • Search Google Scholar
    • Export Citation
  • 16

    Ghebreyesus TA, Haile M, Witten KH, Getachew A, Yohannes AM, Yohannes M, Teklehaimanot HD, Lindsay SW, Byass P, 1999. Incidence of malaria among children living near dams in northern Ethiopia: community based incidence survey. BMJ 319 :663–666.

    • Search Google Scholar
    • Export Citation
  • 17

    Clarke SE, Bøgh C, Brown RC, Walraven GE, Thomas CJ, Lindsay SW, 2002. Risk of malaria attacks in Gambian children is greater away from malaria vector breeding sites. Trans R Soc Trop Med Hyg 96 :499–506.

    • Search Google Scholar
    • Export Citation
  • 18

    Thomson MC, D’Alessandro U, Bennett S, Connor SJ, Langerock P, Jawara M, Todd J, Greenwood BM, 1994. Malaria prevalence is inversely related to vector density in The Gambia, West Africa. Trans R Soc Trop Med Hyg 88 :638–643.

    • Search Google Scholar
    • Export Citation
  • 19

    Lenz R, 1988. Jakarta kampung morbidity variations: some policy implications. Soc Sci Med 26 :641–649.

  • 20

    Mendez F, Carrasquilla G, Munoz A, 2000. Risk factors associated with malaria infection in an urban setting. Trans R Soc Trop Med Hyg 94 :367–371.

    • Search Google Scholar
    • Export Citation
  • 21

    Bashford G, Richens J, 1992. Travel to the coast by highlanders and its implications for malaria control. P N G Med J 35 :306–307.

  • 22

    Sevilla-Casas E, 1993. Human mobility and malaria risk in the Naya river basin of Colombia. Soc Sci Med 37 :1155–1167.

  • 23

    Burkot TR, Dye C, Graves PM, 1989. An analysis of some factors determining the sporozoite rates, human blood indexes, and biting rates of members of the Anopheles punctulatus complex in Papua New Guinea. Am J Trop Med Hyg 40 :229–234.

    • Search Google Scholar
    • Export Citation
  • 24

    Binka FN, Indome F, Smith T, 1998. Impact of spatial distribution of permethrin-impregnated bed nets on child mortality in rural northern Ghana. Am J Trop Med Hyg 59 :80–85.

    • Search Google Scholar
    • Export Citation
  • 25

    Mauny F, Viel JF, Handschumacher P, Sellin B, 2004. Multilevel modelling and malaria: a new method for an old disease. Int J Epidemiol 33 :1337–1344.

    • Search Google Scholar
    • Export Citation
  • 26

    Deressa W, Olana D, Chibsa S, 2004. Magnitude of malaria admissions and deaths at hospitals and health centers in Oromia, Ethiopia. Ethiop Med J 42 :237–246.

    • Search Google Scholar
    • Export Citation
  • 27

    Yohannes M, Petros B, 1996. Urban malaria in Nazareth, Ethiopia: parasitological studies. Ethiop Med J 34 :83–91.

  • 28

    Cameron AC, Trivedi PK, 1998. Basic count regression. Regression Analysis of Count Data. Cambridge University Press: Cambridge, 59–95.

  • 29

    Willis S, Akhwale JKL, Kaneko A, Eto H, Obonyo C, Bjorkman A, Kobayakawa T, 2004. Anemia and malaria at different altitudes in the western highlands of Kenya. Acta Trop 91 :167–175.

    • Search Google Scholar
    • Export Citation
  • 30

    Sintasath D, Lynch M, Kleinhau E, Bretas G, Shilulu J, Brantly E, Graves P, Beier J, 2005. Malaria prevalence and associated risk factors in Eritrea. Am J Trop Med Hyg 72 :682–687.

    • Search Google Scholar
    • Export Citation
  • 31

    Lindsay SW, Martens WJ, 1998. Malaria in the African highlands: past, present and future. Bull World Health Organ 76 :33–45.

  • 32

    Robert V, Macintyre K, Keating J, Trape JF, Duchemin JB, Warren M, Beier JC, 2003. Malaria transmission in urban sub-Saharan Africa. Am J Trop Med Hyg 68 :169–176.

    • Search Google Scholar
    • Export Citation
  • 33

    Mahande A, Mosha F, Mahande J, Kweka E, 2007. Feeding and resting behaviour of malaria vector, Anopheles arabiensis with reference to zooprophylaxis. Malar J 6 :100.

    • Search Google Scholar
    • Export Citation
  • 34

    Koram KA, Bennett S, Adiamah JH, Greenwood BM, 1995. Socioeconomic risk factors for malaria in a peri-urban area of The Gambia. Trans R Soc Trop Med Hyg 89 :146–150.

    • Search Google Scholar
    • Export Citation
  • 35

    Gu W, Novak RJ, 2005. Habitat-based modeling of impacts of mosquito larval interventions on entomological inoculation rates, incidence, and prevalence of malaria. Am J Trop Med Hyg 73 :546–552.

    • Search Google Scholar
    • Export Citation
  • 36

    Ribeiro JM, Seulu F, Abose T, Kidane G, Teklehaimanot A, 1996. Temporal and spatial distribution of anopheline mosquitos in an Ethiopian village: implications for malaria control strategies. Bull World Health Organ 74 :299–305.

    • Search Google Scholar
    • Export Citation
Save