The contribution of reduction in malaria as a cause of rapid decline of under-five mortality: evidence from the Rufiji Health and Demographic Surveillance System (HDSS) in rural Tanzania

The fourth Millennium Development Goal (MDG4) aims to achieve a two-thirds reduction in under-five mortality rate (U5MR) during 1990–2015 [1]. Between 1990 and 2010, the average U5MR in the world has fallen by only one-third from 89 to 60 deaths per 1,000 live births [2]. In 2011, 7.2 million children under-five died in the world with most of these deaths occurring in developing countries [3], of the regions of the world where achieving MDG4 has been most challenging, malaria has been a prominent cause of death. Yet, noteworthy progress has been registered in recent years. According to the Demographic and Health Surveys (DHS), the U5MR is declining in sub-Saharan Africa since 2000 after a decade of stagnation. Successive Tanzania DHS results have shown a decline of 40% of U5MR between 1992–1996 and 2006–2010 (from 137 to 81 per 1,000), with concomitant reduction in inequalities between age group, residence, maternal education and wealth quintile [4‐7]. At the same time, U5MR declined faster in rural Tanzania than urban areas, from 85/1,000 to 60/1,000 against 73/1,000 to 63/1,000, respectively. Inequality in progress is nonetheless evident by age at death: post-neonatal mortality declined from 41/1,000 in 2001–2005 to 25/1,000 in 2006–2010 while neonatal mortality declined by only 30/1,000 to 26/1,000 in the corresponding period.

While malaria-related mortality is known to represent a major component of childhood causes of death in Tanzania, and malaria remains a major focus of health interventions, the possible contribution to malaria reduction in the Tanzania’s childhood mortality decline has yet to be investigated. This paper addresses this evidence gap. Improving scientific knowledge of childhood mortality trends is essential to guide efforts to prioritizing interventions and gauging the effectiveness of disease specific interventions [8‐11]. In most developing countries however, there is a lack of precise data on demographic levels and trends because civil registration is incomplete. DHS in sub-Saharan Africa partially compensate for these gaps, but do not provide specific causes of death for children under-five years of age. Health and Demographic Surveillance systems (HDSS), a continuous monitoring of population in a delimited area, provide accurate information on mortality, fertility and migration trends [12]. The Rufiji HDSS (RHDSS), established in 1998 in a rural district of Pwani (Coast) Region, has collected data on causes of death by using the verbal autopsy (VA) study [13, 14]. This paper analyses the contribution of malaria relative to other causes of death as components of the rapid U5MR decline in a rural district of coastal Tanzania.

Although important insights emerge from this analysis, results must be interpreted with caution owing to data limitations. Findings on cause of deaths trends in the RHDSS could be validated only (1) by supposing that the data collection protocol and diagnostic method of cause of death remained unchanged over the reference period and (2) by accepting the proportional redistribution method of unknown causes. The VA technique is widely used to ascribe childhood causes of death. This method is however known to be reliable for diagnosing only certain causes of deaths, such as measles, whooping cough, cholera, injuries and violence, while less so for delineating causes of associated with different febrile illnesses [27‐30]. Caution must, therefore, be directed to ascribing to VA methods an ability to distinguish malaria from other causes of febrile illnesses that are common in the African context [31]. Although most of HDSS in sub-Saharan Africa are using VA methods with forms, procedures, and definitions that have not changed [32], VA data are compiled by field interviewers and medical VA readers. During the course of the project, different people rotated into these positions. Such subtle staffing changes may have introduced variation in the classification of causes of death over time. However, observers posit that such changes have a little impact on cause of death as diagnosed by physicians [33]. VA data also can be adjusted to allow for some of the misclassification biases [29] and for sensitivity and specificity in malaria mortality but this requires information on parasite prevalence also [32].

The proportion of unknown causes may also have an impact in the cause of deaths distribution. In 2002–2007, this proportion was however low in the RHDSS (19%) compared to Ifakara HDSS (31%), another HDSS in Tanzania [34]. The proportional redistribution method keeps unchanged cause of deaths trends over time. It allocates however a significant number to the most representative categories of causes of deaths. Thus, the proportionate contribution of presumptive malaria on U5MR decline would be at least 50% in RHDSS.

More recently, researchers have developed and applied standardized computer-based algorithm to assign causes of deaths using VA data. These probabilistic models have the advantage of being more internally consistent than physician generated causation estimates but mechanization of inference may lack essential subtlety and nuance [19, 35‐38]. These methods do not allocate specific cause for each death. They also produce undetermined causes, which should be included in the analysis [39]. This paper did not use such methods to compare with causes of deaths diagnosed by physicians. However, standardized computer-based algorithm methods can be useful to assess neonatal cause of deaths which remain more complicated to diagnose by physician [40].

The research legacy associated with the Rufiji study area may also have had an impact on results. The RHDSS, among other studies has been used as a platform of research on malaria. This focus on malaria and related pharmaceutical trials could influence the climate of diagnosis that physicians apply to causes of deaths. Due to their specialization, developed in the course of other studies, participating physicians may weight causes more heavily toward malaria than would be the case among physicians who lack this specialization [28, 37].

The combining of malaria and fever (presumptive malaria) is related to the changes in the percent of children deaths attributed to fever or malaria between 1999–2002 and 2003–2007. That changes in the diagnose VA could be related, in part, to the concomitant introduction in most primary health care facilities of malaria rapid diagnostic tests (mRDTs) [41], the absence of any other new modality that would reduce mortality related to febrile illness [42] and to a large proportion of malaria deaths being properly recoded by the RHDSS.

These limitations, while important to note, are less likely to apply to inference related to long-term trend data than to simple cross-sectional studies. The RHDSS data, though flawed, are standardized over time. Changing compositions of estimated cause of death are more likely to arise from actual shifts in mortality risk than from shifts in any particular source of bias.

Over the past two decades, successive DHS in sub-Saharan Africa have documented a rapid decline of U5MR particularly in rural areas. However, these surveys do not provide the specific causes of deaths. Thus, HDSS in rural settings remain the main source providing accurate information on trends and causes of deaths in delimited area [31, 43]. Results revealed that under-five mortality trends in the RHDSS were similar to the overall rural area in Tanzania [7]. If this trend is sustained, RHDSS and Tanzania as a whole [44] will be a locality that achieves the MDG4 target.

Second, three-quarters of the children under-five who died experienced one of only six causes (presumptive malaria, pneumonia, diarrheal diseases, malnutrition, PLBW and congenital abnormalities) as described for populations elsewhere in Africa [8‐11]. Though all of these causes are in the process of decline, they remain the major killers of children in RHDSS. Nonetheless, their declining importance has dominated the process of U5MR decline in the RHDSS during the last decade.

Despite its high contribution to the U5MR decline, these results corroborate with the previous findings that malaria remains the leading cause of death among children in sub-Saharan Africa [45]. However, improvements in malaria prevention and care may explain why the decline in malaria-related mortality decline predominates as the cause contributing most to this progress. The successful introduction of preventive measures, particularly the insecticide-treated mosquito net [46‐51] and more recent expansion of the availability of the mRDT [52, 53] have had their intended impact. The introduction of mRDT has allowed immediate treatment of patients through efficient and low-cost drugs (Artemisinin-based combination therapy, ACT) [51, 54]. Since March 2007, the ACT was introduced as first-line treatment for uncomplicated malaria and was provided free of charge through all fixed health facilities in Rufiji district [55]. Results from this study were also comparables to the Niakhar HDSS in Senegal [56] indicating that malaria-attributable mortality in under-5 children decreased from 13.5/1,000 during 1992–1999 to 2.2/1,000 in 2010 accelerating the achievement of MDG 4 in that rural area. That suggests that effective control of malaria can be sufficient to allow poor rural areas of Africa to meet MDG4.

Findings show that the contribution of anaemia to increase mortality goes against expected results regarding the contribution of malaria in that area of very high malarial transmission [57, 58]. In rural Kenya, malaria and/or anaemia accounted for the greatest increases in U5MR in 2008–2009 due to stock-outs of essential anti-malarial drugs [59]. In the RHDSS, the slight increased of anaemia mortality rates between 2003 and 2007 could be related to other causes of deaths. Malaria alone is not the cause of anaemia and other possibilities include nutritional anaemia and hookworm infection [7].

Despite its contribution to mortality decrease, pneumonia remains one of most serious killers of young Tanzanians [7] as well as sub-Saharan African children elsewhere [9, 60]. Mortality attributable to pneumonia has been reduced by third in the RHDSS over time probably due to reduction of the prevalence, the early diagnosis and the availability of treatment (antibiotics) with the successful introduction of the program known as the Integrated Management of Childhood Illness (IMCI) [61]. Most of the leading risk factors contributing to pneumonia incidence notably, lack of exclusive breastfeeding, malnutrition, indoor air pollution, low birth weight, crowding and lack of measles immunization [60] have been improved in the RHDSS over time.

Neonatal mortality remained high with only slight decline in the RHDSS and Tanzania in general [7, 62, 63]. PLBW and BIA remain the leading neonatal causes of deaths in the RHDSS as noted elsewhere in Africa [9, 11].

The results of this study suggest that several causes have contributed to rapid decline in U5MR in RHDSS. Presumptive malaria-related mortality was however the predominant contributor to the U5MR decline over the study period. Yet, despite this progress, presumptive malaria remains as a tragically high source of risk of mortality to young children. Thus, findings attest to the need (1) for renewed malaria control efforts and delivery of IMCI and (2) for improvement in the clinical management of complications of childbirth and preterm birth in commitment to strengthening maternal and newborn health to promote improved case of neonates. But the remaining challenge represented by malaria demonstrates that achieving MDG 4 requires continued attention to expanding efforts to reduce malaria related mortality.