This paper looks at changes in causes of death among under-fives in a population experiencing major decline in fertility and mortality rates over thirty-three years period. Stagnant childhood mortality rates, however, call for a careful review of current strategies of disease specific or selective primary health care interventions largely focusing on better management of diseases implemented through private sector health services.
Logically, differences in causes of death between different time periods and populations can be due to three main reasons: difference in incidence, difference in case fatality rates and misclassification. Misclassification in cause of death has been reported to be low for congenital anomalies and injuries and high for neonatal sepsis, diarrhea, pneumonias and malnutrition [
16,
31]. Incidence of a disease would be affected by differences in its determinants (for example, source and quality of water for diarrhea) or presence of preventive measures (vaccination for tetanus). The case fatality rate is determined by a combination of three factors: health seeking behavior, availability and access to health services and quality of care provided in the health services. We attempt to explain the differences noted in this study between time periods and sex using this framework.
Temporal trends in causes of death
The decrease in neonatal mortality was mainly because of decrease in prematurity and infections related deaths including tetanus. Despite an increase in institutional deliveries to a quarter of all births, deaths due to birth asphyxia did not come down. Less than half of these deliveries are in government health facilities. Probably, these institutional deliveries are occurring in low risk mothers in higher socio-economic groups. Also, it is possible that the facilities where the deliveries are now occurring are not fully equipped in terms of trained human resource and equipment resulting in higher cases of birth asphyxia. As neonatal asphyxia has major developmental consequences to the surviving child it would result in serious burden on the society to care for the affected neonates with various forms of disability. Misclassification of neonatal deaths as still births or prematurity as sepsis in the initial period which could have gradually decreased over a period of time due to better training and monitoring is a possibility. The increase in coverage of pregnant mothers with tetanus toxoid from about 24% in 1972 to almost universal coverage in early nineties and a strategy of supply of disposable delivery kits (a sterilized kit having a gauze piece, razor blade and thread) resulted in the decline in neonatal tetanus mortality [
21].
The initial decline followed by a rise in prematurity related deaths is difficult to explain. There is no information on incidence of low birth weight for the period before 1990 in the study area. Since then, different sources have reported the incidence of low birth weight ranging from 9% to 26%, including data from electronic database (19% for 2010) [
29,
30]. The decrease in deaths due to low birth weight is thus, unlikely to be due to decreased incidence but perhaps due to decreased case fatality due to improved access. Despite an apparent and not well-documented increase in health facilities where deliveries can be conducted, health facilities at Ballabgarh still lack capacity for managing babies below 1.5 Kg. These babies would be referred to centres in Faridabad and Delhi which are now more accessible due to better roads and affordable due to improvement in socio-economic status. However, seeking costly care also has the potential to accentuate gender differences.
The main causes of death in 1-59 months age group were ARI, diarrhea and malnutrition. It has been reported from the study area that measles vaccination introduced in 1985 resulted in 57% decline in all cause mortality and that case management approach resulted in 26% decline in ARI mortality [
17,
21]. These relate well with the results reported. A case fatality rate of 1.3% was reported for pneumonia from Ballabgarh in 1987, a figure which is on the lower side of interquartile range (1.3-2.6%) of case fatality rate globally [
32]. This was because mothers were able to recognize danger signs for pneumonia and take them for treatment with private practitioners [
25,
26]. It is likely that the case fatality rate has further declined in Ballabgarh since then. The reported ARI incidence from the study area has remained around 3 per child per year from 1987 to 2005 [
33]. For the same period, the pneumonia incidence was reported as 0.3 per child per year [
26]. There is a wide difference in the estimated incidence of pneumonia between developing and developed countries (0.29 Vs. 0.05 per child per year) [
33]. The large potential in reduction of pneumonia incidence and an already low case fatality rate point to the need for a change of strategy to address reduction in incidence of respiratory infections. Among the established risk factors for pneumonia , the important one in the study area are those of indoor air pollution due to cooking fuel and tobacco smoking, poor housing and child care practices.
The promotion of Oral Rehydration Therapy (ORT) has been the major intervention for reduction of diarrhea-related deaths. High use of fluid therapy was reported in late eighties [
22]. The same study reported a case fatality rate of 0.6%, an estimate confirmed by another study done in the area around these villages [
34]. As diarrheal deaths have further reduced, case fatality rates must have come down further. Though exact information is not available, it is unlikely that the incidence rate of 2.6 diarrheal episodes per child per year reported in this population in late eighties has come down significantly [
22]. Three fifths of the population did not have latrines at home and 81% had access to drinking water at or close to home as per our data in 2002-3 [
27]. The stagnation in diarrheal incidence rates and a decline in diarrhea mortality rates have been noted globally as well [
35]. This means that the focus of the efforts to control diarrhea has to shift to water supply, sanitation and hygiene (WASH) interventions and possibly vaccination against common agents of diarrhea. These have already been identified as potentially cost-effective interventions [
35].
Data from studies done in this area over the study period have reported more or less similar rates of grade II or more malnutrition in late sixties and eighties to a prevalence of 24.9% in 2004 [
36,
37]. Anecdotally, it appears that there is a decline in rates of severe malnutrition, with grade I and II malnutrition remaining at about the same levels. One would therefore expect that the contribution of malnutrition to deaths would decrease. Malnutrition as a cause of death also has high potential for misclassification [
16,
31]. Many of those who died due to ARI and diarrhea would also have severe malnutrition and could have been differently classified by different medical officers.
The Million Death Study (MDS), done in 2001-03 reported that three leading causes (prematurity and low birthweight, neonatal infections and birth asphyxia) accounted for 78% of all neonatal deaths as compared to 62% in this study in 2002-04 [
12]. Two causes accounted for 50% of all deaths at 1–59 months: pneumonia and diarrhoeal diseases as compared to 42% in Ballabgarh during roughly the same time period. The mortality rate due to prematurity, infections and asphyxia as estimated by MDS were however, in absolute terms, much higher than in Ballabgarh. The overall mortality rate was also higher with u5MR being 85.8 per 1000 lb and NMR being 36.9 per 1000 lb as compared to 71 and 22, respectively in our study. However, deaths due to congenital abnormalities in the neonatal age group (1.2 vs. 2.1) and injuries in the 1-59 months age group (2.9 vs. 2.3) were reasonably similar to our study. Tetanus deaths were still higher at 1.2 per 1000 lb in MDS. This similarity in major causes of death supports external validity of our study in Ballabgarh.
Trends in childhood mortality reported from other parts of the world
The global estimates between 2000 to 2010 show a decrease in all causes of under-five mortality during this decade, mainly due to the decline in diarrhea and pneumonia deaths, with annual decline rates of 4% and 3.1%, respectively [
1] There are not many studies which have reported on changing patterns of causes of childhood deaths over time. In Matlab HDSS, using routinely collected demographic surveillance data from a rural area of Bangladesh between 1975 and 2002, the overall reduction in early and late neonatal mortality comparing the same period was 39% and 73%, respectively. The dramatic decline in neonatal mortality was mainly due to a fall in deaths from neonatal tetanus [
38]. Data from three Bangladesh Demographic and Health Surveys done during 1993 to 2004 show an increase in death rates due to birth asphyxia (from 4 to 9 per 1000 lb) and prematurity (from 4 to 7 per 1000 lb). The decline in mortality in the 1-59 months age group were also similar to those reported in this study [
39]. A population-based survey conducted through a nationwide multi-level surveillance network in China showed that the greatest rate reduction in under-five mortality in rural China between 1996 and 2006 was seen in deaths due to pneumonia and diarrhoea; 69.4% and 69.1%, respectively. Deaths due to premature birth/ low birth weight and birth asphyxia decreased by 46.8% and 46.0%, respectively, while deaths due to congenital heart disease increased by 24.5% [
40]. Deaths due to injuries have been reported to have increased in the last decade in both Africas and Americas and probably reflect increased mechanization of agricultural as well as introduction of motorised transport in villages [
1].
Gender differences in childhood causes of death
In societies in which care is equal for boys and girls, baby girls have a lower mortality rate than baby boys: the ratio of neonatal mortality for boys to girls is usually at least 1.2 [
41]. In the current study however, this was reversed. Data from different sources in India show that the major causes of death in the first week of life are due to asphyxia and prematurity whereas most of deaths in the 7-28 days are due to sepsis [
42,
43]. A review of all child deaths in the study area between 1991-95 showed that for early neonatal deaths (<7 days), there was a slight preponderance of boys (55: 45) whereas for late neonatal deaths, the ratio was reversed to 40:60, which was more or less maintained till 5 years of age [
44]. The gender difference in late neonatal mortality has also been reported from Pakistan which shares a similar cultural practice of female neglect [
45]. Gender difference in care seeking for neonatal illnesses was confirmed in a recent study in Ballabgarh which showed that sick male neonates were more likely to be taken to a health facility especially to a private one, which is perceived by community to be better than public ones [
46]. Higher unclassified deaths among girls could be due to their care givers not providing sufficient information on their deaths, either due to disinterest or because of hesitation that this may reflect on their neglect.
A community based study done in eighties in our study area, reported similar incidence rates for boys and girls for diarrhea and pneumonia but a higher case fatality rate among girls [
22,
26]. This has been confirmed by large scale national level household surveys in India [
47,
48]. Data from our primary health centre (PHCs) for the year 1983 showed that emergency consultation among under-fives, which were mainly for ALRI and diarrhoea, were higher for boys as compared to girls [
49]. A review of all admissions due to ALRI to the secondary level hospital at Ballabgarh for the year 1990 showed that 74% of admissions were of boys [
50]. Other studies have also shown that girls are often brought to a health facility in more advanced stage of illness than boys, are taken to less qualified doctors when they are ill, and less money is spent on medicines for them than for boys [
10,
11]. Data from a HDSS site in Bangladesh also showed higher mortality rates among girls for malnutrition (2.5 times) and diarrhea (2.1) as in this study area [
51].
The strengths of this study are that the information on causes of death is available over a period of time along with information on coverage with many childhood interventions for the same population. The main weakness is that the method of ascertainment of causes of death underwent changes especially for the last time period. However, a comparison done in 1995 showed acceptable agreement between causes of death arrived by VA tool and the previous system [
15]. Inter-observer variation between physicians who coded the cause of death is possible. However, the fact that injuries and birth asphyxia remained stable during this period with no major interventions targeting these illnesses having been implemented in the study area and decline in tetanus deaths in conformity with interventions reflects the robustness of the data. Lack of data on quality and outcome of care of private health care facilities makes it difficult to comment on their role in reduction of childhood mortality in the study area.