Why the increase in under five mortality in Uganda from 1995 to 2000? A retrospective analysis

The chance of dying before the fifth birth day (during childhood) otherwise known as the under-five mortality rate (U5MR) or childhood mortality rate, is a composite indicator measuring human as well as economic progress [1]. The U5MR measures an end result of the development process from multiple inputs such as income, income distribution, food availability in the family, the availability of clean water and safe sanitation, use of maternal and child health services (including prenatal care), level of immunization, and the overall safety of the child's environment [2]. Moreover, the U5MR is less susceptible than, say, per capita Gross Domestic Product (GDP) to the fallacy of the average [3]. This is because the natural scale does not allow the children of the rich to be one thousand times as likely to survive, even if the man-made scale does permit them to have one thousand times as much income. In other words, it is much more difficult for a wealthy minority to affect a nation's U5MR, and the indicator therefore presents a more accurate, if far from perfect, picture of the health status of the majority of children (and of society as a whole) [1, 3].

The United Nations and its agencies use U5MR as measure of wellbeing and consequently decreasing the U5MR is included in the millennium development goals (MDGs) [4]. According to the MDG number four, Uganda is expected to reduce U5MR by two thirds from 195 deaths per 1000 (in 1990) to 65 deaths per 1000 in 2015. In order to achieve this goal, the annual average reduction rate (AARR) of U5MR should be more than 4.4% [5]. Recent analyses showed that the AARR of 4.4% has never been reached in Uganda [6]. Besides, according to the Uganda Demographic Health Surveys (UDHS) of 1995 and 2000, the U5MR increased from 147.3 deaths per 1000 live births to 151.5 deaths per 1000 live births [7, 8]. The reasons for this increase in U5MR are not clearly known and there has been a lot of speculation both internationally and locally regarding possible causes [9]. Moreover, the increase appeared paradoxical since the major determinants of U5MR such as GDP per capita, prevalence of HIV among pregnant women, and the quality and quantity health services and other social services were favourable during the same time [10]. There is thus need to the understand factors that could have been responsible for the increase in U5MR during the period 1995-2000 in the country. It has been stressed that understanding of factors influencing changes in U5MR at country level rather than in geopolitical areas is a pre-requisite for designing effective intervention methods aimed at decreasing childhood mortalities [11]. This article reports on factors that most likely influenced the increase in the under five mortality rate in Uganda for the period 1995-2000.

Table 1 shows the childhood mortality rates in Uganda disaggregated by age of the child and by region for the period 1995-2000. It is interesting to note that only the western region showed an increase in U5MR whereas other regions of the country showed a decrease (figure 2). Furthermore, only in western Uganda did the change in U5MR reach statistical significance (i.e. the 95% confidence interval for the years 1995 and 2000 do not overlap). In western region most of the observed increase in U5MR is contributed by an increase in neonatal mortality and by increase in child mortality. Figure 2 shows that the AARR was most remarkable in terms of magnitude and direction (being negative) in the western region compared to other regions of the country.

	1995			2000
	Rate (Per 1000)	95% CI	Weighed Sample	Rate (Per 1000)	95% CI	Weighed Sample
Uganda
NNMR	27.0	22.1-31.9	7681	33.2	27.8-38.5	7834
PNMR	54.3	47.0-61.6	7717	55.2	47.9-62.6	7853
IMR	81.3	71.1-90.4	7719	88.4	78.8-98.0	7854
CMR	71.9	63.6-80.2	7834	69.2	60.0-78.4	8007
U5MR	147.3	135.1-159.5	7875	151.5	138.2-164.8	8028
Central
NNMR	29.6	21.9-37.2	3579	29.8	23.4-36.1	4039
PNMR	47.0	37.6-58.4	3585	42.2	33.2-51.2	4042
IMR	76.6	63.9-89.3	3586	71.9	60.0-84.0	4044
CMR	70.1	57.8-82.4	3612	68.1	52.8-83.4	4082
U5MR	141.3	122.2-160.5	3621	135.1	113.6-156.6	4088
Eastern
NNMR	38.4	31.1-45.7	3615	29.5	21.6-37.4	4201
PNMR	59.7	50.0-69.5	3623	59.8	48.2-71.4	4204
IMR	98.1	84.9-111.3	3623	89.3	74.7-104.0	4204
CMR	86.0	71.6-100.4	3665	63.7	52.3-75.1	4231
U5MR	175.7	156.5-194.6	3672	147.3	131.5-163.1	4234
Northern
NNMR	33.6	23.8-43.4	2653	42.2	31.6-52.7	2413
PNMR	65.8	49.2-82.3	2655	63.7	49.3-78.1	2421
IMR	99.3	78.7-120.0	2655	105.9	89.7-122.0	2421
CMR	100.6	73.1-128.1	2693	80.6	63.3-97.9	2439
U5MR	190.0	162.6-217.3	2965	178.0	155.4-200.5	2447
Western
NNMR	26.8	20.7-32.9	4029	41.5	33.8-49.2	3586
PNMR	48.3	37.2-59.5	4030	56.3	44.3-68.3	7853
IMR	75.1	61.3-89.0	4032	97.8	82.3-113.4	3597
CMR	60.1	47.4-72.8	4057	87.0	69.5-104.5	3638
U5MR	130.7	110-150.1	4061	176.3	152.3-200.3	3638

As shown in table 2 correlation coefficients for all possible determinants of U5MR for the period 1995 and for 2000 were not statistically significant (P > 0.05). It is also interesting to note that most of the determinants were favourable for the year 2000 compared to 1995 and the expected effect on U5MR would have been a reduction not an increase!

These data show that the period 1995-2000 was characterised by an increase in U5MR in Uganda from 147.3 to 151.5 deaths per 1000 live births. Disaggregation of the data shows the increase of U5MR in the country was only contributed by the western region with other regions of the country showing a decline in U5MR. Furthermore, this analysis demonstrated that the increase in U5MR in the western region could not be explained by the changes in the major determinants of childhood mortalities such as changes in poverty levels, quality and quantity of health and other social services.

In order to explain the observations related to changes in U5MR in Uganda for the period 1995-2000, there is need to find something that happened only in western Uganda during the period 1995-2000 and has the potential to drastically change the U5MR by a big margin. It is hypothesized that this increase in U5MR in western Uganda could be attributable to the severe malaria epidemic that happened in western Uganda in 1997/1998. This was the worst malaria epidemic happening in Uganda and it was characterised by el-Niño weather phenomenon with almost twice the level of rainfall [16–18].

Several attributes makes the malaria epidemic in western Uganda a likely explanation for the observed trends in childhood mortality. First, there is temporal sequence [24]. The increase in childhood mortality was preceded by the malaria epidemic which followed the el-Niño weather phenomenon [16–18]. Second the malaria epidemic happened in western Uganda only and this is the same region that experienced the increase in childhood mortalities. Third the assertion linking malaria as a major determinant of U5MR is consistent with other studies that demonstrated an increase of childhood mortality with occurrence of malaria epidemics in various parts of Africa [25, 26] or that show relative low childhood mortalities in malaria free areas [27]. For example, according to the 2002 census in Tanzania the U5MR for relatively malaria free areas of Arusha and Kilimanjaro regions were 58 and 67 deaths per 1000 live births respectively compared to the U5MR for Tanzania (154 deaths per 1000 live births) and the U5MR for Dar-es-salaam (123 deaths per 1000 live births) [27]. Furthermore, data from intense malaria control strategies also show that control of malaria reduces the overall U5MR by up to 60% [28].

The explanation that malaria epidemics increase U5MR is biologically and theoretically plausible from several perspectives. First, malaria is more likely to affect children aged 0 to 4 years hence the corresponding negative consequences on childhood mortalities [29–31]. Although, malaria causes significant childhood mortality in areas of stable high transmission, the effect of malaria on childhood mortality under such conditions of transmission is stable and therefore not detectable by time trends. Moreover, under conditions of stable transmission children in the first 2 years of life are more susceptible to malaria whereas in unstable transmission areas the effect of malaria on mortality is felt throughout all the years of childhood [29]. Thus, the negative effect of malaria on childhood mortality for unstable malaria is more likely to be more than that of stable malaria. Second, in stable transmission areas malaria in pregnancy predominantly affects prime-gravida (that forms only about 15% of total mothers who become pregnant) whereas in unstable transmission areas malaria affects both prime and multi-para (100% of pregnant women) [32, 33]. Therefore, the effect of malaria epidemics on childhood mortality by this factor is that many more children in unstable malaria areas are affected compared to children in stable malaria areas. Third, unstable malaria is more likely to lead to prematurity in low transmission areas whereas in high transmission areas malaria infection in pregnancy is more likely to lead to underweight [29, 32, 33]. Although, both underweight and prematurity increases the risk of children death, the risk of premature deaths is about twice that of underweight children [34]. Fourth, the effects of anti-malarial drug resistance on mortality are more likely to be felt among the non-immune as found in low transmission areas compared to the semi-immune that are more likely to be found in high transmission areas [35]. Indeed, at the time of the 1997/98 epidemic, the malaria treatment policy was still chloroquine whose level of resistance in Uganda particularly in the western region was very high [36]. Finally, another factor that leads credence to these assertions is that the western region of Uganda according to the 2000 UDHS [8] was the least prepared to prevent malaria or to mitigate its effects. For example the percent of household using bed nets was 5% in western Uganda compared to 13% for the rest of the country; the percent of pregnant women receiving malaria prophylaxis was 21% for western Uganda compared to 40% for the rest of the country and the percent of women delivering under medical supervision was 22% in the west compared to 37% in Uganda.

A limitation of this study was that information used in the analyses was collected retrospectively and therefore susceptible to recall bias especially on the age of death, where respondents could have forgotten events that happened in the past. We think that this bias was spread through all the regions of Uganda and therefore less likely to affect trends. Another limitation is that since most of the women in Uganda deliver at home and their children are not weighed at birth; it was thus not possible to analyze the trends of prematurity and low birth weight for adducing more evidence. However, data from neighbouring Tanzania suggest that during the 1997/1998 El Niño weather phenomenon there was an epidemic of low birth weights especially in low malaria transmission areas [37]. Besides, the 2000 UDHS demonstrated that the western region had a more unfavourable peri-natal mortality of 52.6 per 1000 pregnancies compared to the national average of 42.6 per 1000 pregnancies [8]. The peri-natal mortality is a measure of still births and deaths within the first week of life. Prematurity, low birth weight, malaria and home deliveries are major determinants of perinatal deaths [38].

Nevertheless our methods indicate the utility of disaggregated data using ecological evidence to understand events that would otherwise be unethical to study using randomization. Besides, these data are almost as good as a natural experiment where events were analyzed before and after the possible determinant occurred and in places where the possible determinant occurred and did not occur [39].

Thus it is highly plausible that malaria was a major determinant of childhood mortality in Uganda during the period 1995-2000. The mechanisms through which malaria causes childhood deaths are both direct and indirect. Directly malaria increases childhood deaths through causing infection and severe disease among affected children. Indirectly malaria causes deaths by leading to low-birth weight and prematurity. Thus in order to speed the reduction in childhood mortality in Uganda, there is need to aggressively control malaria as through universal coverage with intermittent presumptive treatment in pregnancy (IPT), use of long lasting insecticide treated nets, residual household spraying and increasing access to prompt treatment [40]. Additionally, there is need to increase the proportion of children born under qualified medical supervision through improvement in peri-natal health care and in increasing the number of health personnel that can supervise deliveries [41].