T R A N S A C T I O N S OF T H E ROYAL S O C I E T Y OF T R O P I C A L M E D I C I N E A N D H Y G I E N E (2003) 97, 655-660
Malaria in southern Mozambique: incidence of clinical malaria in children living in a rural community in Manhi~a district Francisco Safitte 1'2, John Aponte 1'2, Jesus A l m e d a 1, Carlos Ascaso 2, Neide Vaz 1, Martinho D g e d g e 1 and Pedro A l o n s o 1'2 ZCentro de InvestigafSo em Safde da Manhifa, Instituto Nacional de Saf~de, Ministdrio da Safzde, C.P. 1929 Maputo, Mozambique; 2Centro de Salud Intemacional, Hospital Clinic, Institut d'Investigacions Biomediques August Pi I Sunyer, Universidad de Barcelona, Villarroel 170, E-08036, Barcelona, Spain Abstract We estimated the incidence of clinical malaria episodes by weekly home-based active case detection between December 1996 and July 1999 in 2 groups of children in Manhi~a district in southern Mozambique. Cohort 1 comprised a random sample of children aged < 1 0 years at recruitment and cohort 2 comprised newborns. A blood slide was taken if the axillary temperature was ~>37.5 °C or if the child was reported to have been febrile over the previous 24 h. A total of 1966 children were followed-up. Malaria occurred all year round. T h e number of clinical malaria episodes ranged from 0 to 6 per child. No clinical malaria episoses were detected in 71% of children or in children aged < 2 months. Those aged 6 months to < 4 years showed the highest incidence ranging from 0.65 to 0.74 episodes per 100 person-weeks at risk, indicating that infants and young children were at highest risk. Malaria transmission is perennial in this district of Mozambique with some seasonality. Significant differences were observed in the spatial incidence of malaria episodes in regions just a few miles apart with a higher incidence in children living near the river or in swampy areas. These findings strengthen the need for improved control measures targeted at infancy and early childhood.
Keywords: malaria, clinical episodes, incidence, children, Mozambique Introduction Estimates of the burden and distribution of malaria in a given country are usually based on national statistics often exclusively produced from data from health facilities. As such, the accuracy of national malaria statistics is heavily dependent on access and patterns of health seeking behaviour. In many countries the silent burden of malaria is greatest in the rural areas where people often live far from health facilities and most preventive methods are absent. Mozambique is no exception. Approximately 60% of the population has no access to curative facilities from the National Health Service. Data on the incidence of clinical malaria episodes and which age groups and populations are at greatest risk are critical for the planning and conduct of control strategies, as well as for the evaluation of new interventions in a research setting. Young infants are relatively protected against clinical episodes of malaria in endemic countries (McGregor, 1984). This is believed to be due to a variety of factors including immunity acquired in utero (Sehgal et al., 1989) and through breastfeeding (Maegraith, 1952), as well as the persistence of fetal haemoglobin in the first 3 months of life which can retard the growth of Plasmodium falciparum parasites (Pasvol et al., 1976, 1977). However, the magnitude and duration of this protection as well as the mechanisms mediating it remain a matter of discussion (Riley et al., 2001). T h e duration of this protection has been found to vary from child to child within the same epidemiological setting (Akanmori et al., 1995), and from place to place according to the endemicity level (Kitua et al., 1996; McGuiness et al., 1998; Snow et al., 1998). Therefore, an improved description of the local burden of malaria in infants would be useful both in guiding control strategies and clinical management. Epidemiological studies relating cases of malaria to environmental conditions are of most interest in supporting the need for a renewed emphasis on environmental management for malaria control in endemic countries. Indeed, though this method has been successfully and cost-effectively used for malaria control in Address for correspondence: Dr Francisco Saflte, Centro de Salud Internacional, Hospital Clinic, Villarroel 170, E-08036, Barcelona, Spain; phone +34 93 2275706, fax +34 93 4515272, e-mail
[email protected] or
[email protected]
many places (Utzinger et al., 2001), it is not a common practice in many endemic countries. Th e community incidence of malaria in Mozambique has been poorly described. There are a limited number of published community-based malariometric studies in the country (Hogh et al., 1995; Thompson et al., 1997; Mendis et al., 2000), all conducted in urban or semi-urban settings. Th e present study on clinical malaria incidence in infants and children was part of a larger study of the epidemiology of malaria in a rural area in southern Mozambique.
Methods Study area and population Th e study took place in Manhiqa district in rural southern Mozambique, where the Centro de Investigaqao em Safide da Manhi~a (CISM) has been running a continuous demographic surveillance system since 1996. A full description of the Manhi~a demographic surveillance system area and methods has been reported elsewhere (Alonso et al., 2001). In brief, 35 000 people live in 28 neighbourhoods within a 10 km radius of Manhi~a Health Centre (Manhiqa District Hospital). Th e area is a flat savannah with moderate vegetation. There are 2 seasons, a hot and wet season (October-May) and a dry and cold season (JuneAugust); average annual rainfall was 1100 m m 3 in 1998 (Fig. 1). Most villagers are farmers growing sugar cane, bananas and rice, and some are workers in 2 large sugar processing factories. Houses are typically made of reeds with thatched or corrugated roofs. Two reference district hospitals (Manhi~a with 80 beds and Xinavane with 59 beds) and 8 peripheral health facilities constitute the government health network in the district. Plasmodium falciparum accounts for over 90% of all malaria cases and the estimated average entomological inoculation rate in the area was 15 infective bites per person per year in 1998 (C. Aranda, personal communication).
Study design and subjects In order to estimate the community-based incidence of clinical malaria the study area, we designed a study consisting of 2 cohorts. Cohort 1 consisted of children aged < 10 years at recruitment and followed-up for 31 months. Given that, as time progressed, this cohort would have few infants being followed, a second cohort was recruited. Cohort 2 complemented cohort 1 as it
656 aimed to produce robust estimates of the incidence of malaria during the first year of life by recruiting newborns who were followed-up for 1 year. For cohort 1 an age-stratified simple random sample was generated from the updated census of the Manhi~a demographic surveillance system area. There were 6 age strata ( 0 15 years and at home at the time of the visit. Axillary temperature was recorded using electronic digital thermometers (Maximum Thermometer CE 0120). If there was reported fever over the past 24 h, or if the measured axillary temperature was i>37.5 °C, a fingerprick blood sample was collected and 2 blood slides were prepared for malaria parasitaemia detection. Participants received the laboratory results within 24 h of sample collection. Subjects with parasitaemia and those reporting fever and/or axillary temperature ~>37.5°C were advised to seek medical care at Manhiqa Health Centre. Data collection was supervised by a trained field supervisor on a motorbike. Cohort 1 was run from December 1996 to July 1999 and cohort 2 was run from December 1997 to July 1999. This study falls within the national ethical clearance granted to the malaria epidemiology studies of the C I S M by the Ministry of Health/National Institute of Health in 1996.
Laboratory methods Thick and thin blood films were Giemsa-stained and read for malaria parasites following standard, quality control procedures (Alonso et al., 1994). For rapid feedback of results, a semi-quantitative system of reporting parasitaemia was used (Bruce-Chwatt, 1986).
Study definitions, data handling and statistics Fever was defined as axillary temperature i>37.5 °C and/or patient-reported fever within the previous 24 h. Clinical malaria was defined as fever with a positive blood slide (with 1 or more asexual P. falciparum parasites). Only slides with agreement according to the previously described standard criteria were included. T h e lag period for clinical malaria, the maximum duration for a given malaria episode, was standardized and set at 28 d. Th e sample size was calculated using Epi Info, version 6 (C DC , Atlanta, GA, USA). For this purpose, we assumed an expected malaria prevalence of 50 + 6.4%, 50 -4- 8.8%, and 50 4- 6.3% for infants 7 years) experienced relatively few malaria episodes, with incidence rates ranging between 0.2 and 0.4 episodes per 100 PWAR. In the univariate analysis, age, mother's education, and the presence of a latrine in the household were significantly associated with occurrence of clinical malaria episodes (Table). Indeed, using the age group of 8-11 months as the baseline group, the age groups showing significantly fewer malaria episodes were 2 4 years (though there was a borderline significance in the age group of 6 - < 7 years). Using uneducated mothers as the reference group, we found that children of mothers with a secondary or higher level of education had significantly fewer malaria episodes. Finally, children from households with a latrine
CLINICAL MALARIA EPISODES IN CHILDREN
657
iRate per 100 P W A R
r
............................ Rainfall Imm3) i I
1.5 <
1.0 600
O
500
0.5
4OO 0.0
300 200 100
\...........i ................. \,.,/" ................. 1
i
]
I
]
I
I
F
\ ..................... 1
I
i
Month Fig. 1. Monthly incidence of clinical malaria episodes (CME) per 100 person-weeks at risk (PWAR) in relation to rainfall, Manhi~a district, Mozambique. X2 test for unequal rates of C M E in both univariate and multivariate analysis = 197, d.f. = 32; P < 0.0001.
~'~ T a r = = ~
,"'
,/
Zambia~~I
/ ?
!kM°zambaqu~ / ''/
Zimbabwe
C
~/,,''"
~
SoutU-~ica/'"~.T sw4
-__S" ",,
Malarial 000 PwARcasesper
",,.].
[~
0-1.35 1.35-2.89 2.89-8.58 5.01-8.58 [] 8.58-50 >>>> Railway Swampy area . . . . Main road ~..., Incomati River
,.
4
"'" "':" "". '".,, ",,
Fig. 2. Geographic distribution of clinical malaria episodes (CME) in Manhiqa district, Mozambique. 1000 PWAR, 1000 personweeks at risk. X2 test for unequal rates of C M E between zones = 20 668, d.f. = 118; P < 0.001.
F. SAOTEETAL.
658
1.0-
~
0.80.6-
0.4 0.20.0Age (years) Fig. 3. Crude age-specific rate of clinical malaria episodes per 100 person-weeks at risk (PWAR), Manhi~a district, Mozambique. Error bars indicate 95% confidence intervals. Error bars for infants aged 0 - < 2 months were not included because there were no cases of malaria in this age group.
had significantly fewer malaria episodes than those from households without one. Likewise, in the multivariate analysis (Table), the age groups showing significantly fewer malaria episodes than the baseline group (8-11 months) were 2 - < 6 months and />4 years, after adjusting for presence of latrine, region, and month. Children with latrines in their homes had significantly lower rates of clinical episodes compared with those not owning one after adjusting for age, month, and region.
Discussion This study confirms that current malaria transmission continues to be stable (Bruce-Chwatt, 1986) in the study area, although with some seasonality. Over 70% of the children did not have a single malaria
episode during the study period. Moreover, among those who did, half of them had only one episode. Similar findings from Senegal have been reported previously (Rogier et al., 1999). Together these studies argue in favour of a clustering effect of clinical episodes whereby relatively few people account for a large proportion of all malaria cases in a given area. The significant differences observed regarding spatial incidence of clinical episodes in the area support these data. The highest incidences correlate with areas closer to the Incomati river flood plain and/or to swampy areas (Fig. 2). Those children having socio-demographic and/or geographic risk factors are at regular and higher risk of clinical episodes than their age-matched counterparts living just a few miles away, despite minor ecological differences. In other words, relatively few children are at risk of clinical episodes within a given area. There is a considerable similarity among the high incidence areas detected by active case detection in the present study and those detected by an ongoing hospital disease surveillance system (J. Aponte, unpublished information), and those implicated by an entomological survey conducted in 1998 (C. Aranda, personal communication). Repeated community surveys in this area also confirm these high-risk zones (Saute et al., 2003). The youngest infants, particularly those aged
