Do improvements in outreach, clinical, and family and community-based services predict improvements in child survival? An analysis of serial cross-sectional national surveys

A series of recent metanalyses have identified over 60 interventions that are effective in reducing mortality in children under five years of age [1]. Broadly speaking, they can be divided into three groups: i) interventions for direct treatment of illnesses ii) preventive interventions such as immunization and iii) health promotion/behavioral change interventions such as breastfeeding. Coverage remains low for many of the interventions, and unless it can be accelerated, it is unlikely that the Millennium Development Goal (MDG) 4 of a two-thirds reduction in under-five mortality from 1990 by 2015 will be met at the global level [2]. Efforts have thus been intensified by UNICEF and others to assist countries in identifying the most effective packages of interventions to address their leading causes of death within the context of the country's health care system and identify the bottlenecks that would need to be overcome to achieve high coverage.

Although in fragile countries or those with minimal infrastructure there may be a role for programs delivering a limited number of interventions [3], there has been increased emphasis in recent years on the provision of broader packages that share a common delivery mechanism and/or target population (e.g., the treatment of diarrhea, pneumonia, and malaria at community or facility level or the bundling of interventions that can be delivered through periodic outreach, such as vaccination, vitamin A, and bed net distribution) [4, 5]. Whilst there is a large and growing evidence base regarding the composition and packaging of essential interventions, the evidence for which types of interventions should receive priority when considering marginal investments by governments and donors is unclear. A persistent concern has been that merely increasing expenditure on priority sets of interventions will not be sustainable without commensurate investments in health systems. The first step in combining a focus on critical interventions with investments in health systems is to identify the different ways in which interventions are delivered.

There are essentially three domains for the delivery of such packages, as outlined in the 2004 World Development Report [6]. The first domain is clinically-based services delivered at the individual level, which include not only treatment of such major childhood killers as diarrhea, pneumonia and malaria but also the availability of skilled attendants at delivery and emergency obstetrical care. The second consists of services that can be delivered primarily through episodic contact or outreach and include immunization, distribution of insecticide-treated bed nets for the prevention of malaria, distribution of vitamin A, and antenatal care. The final domain is community and family-oriented services that support self-care and includes activities such as information and social support for breastfeeding promotion and appropriate child nutrition. The three domains operate against a complex backdrop of the social, physical, economic, and cultural environment, which may influence the extent to which the interventions achieve their potential for change.

A recent study has demonstrated that countries which are making progress towards MDG4 have in common the scaling up of services such as family planning and immunization, which are delivered through episodic contact, while the countries making the most progress have also successfully scaled up their clinic-based services [7]. Among those working in maternal and child health, however, there has been a great deal of controversy regarding which domain represents the most appropriate and cost-effective means to reduce mortality in countries which have made poor progress. Proponents of community-based services have argued that strengthening this approach presents a much quicker and pragmatic approach than the longer and more expensive building of clinic-based services in settings with existing low levels of coverage and health system infrastructure [8].

Many countries have undergone rapid changes in the three service domains and also in contextual factors such as female literacy and wealth. However the extent to which changes in these domains have had an impact on under-five mortality rates is not known. In this paper, we use data from sequential Demographic and Health Surveys (DHS) from 27 countries to examine the relationship between progress in the three service domains and average annual change (AAC) in under-five mortality rates.

Our study demonstrated considerable variability in the AAC in under-five mortality among the 27 countries and that the strongest predictor of changes in under-five mortality rates was improvements in access to and coverage of clinical services. In addition, improvements in selected outreach interventions were also correlated with changes in mortality, but the effect was not as strong as for clinical services. For community and family services, we found that an improvement in breastfeeding practices/hygiene was significantly correlated with under-five mortality but that the effect was weaker and not statistically significant among the African countries. By contrast, childhood nutritional status was strongly correlated to a positive impact in under-five mortality rates only in the 19 African countries.

The importance of access to clinical services is not completely surprising as the major causes of child deaths such as pneumonia, diarrhea, malaria and various conditions of the neonatal period historically have required access to clinical services provided through health facilities. Our findings are also in keeping with those of Rohde et al [7], who demonstrated that the countries that had made the most progress as measured by average annual reduction of mortality were those which had higher coverage of comprehensive primary health care, which in their study was measured by a high prevalence of skilled attendants at birth. They, too, suggest that countries can achieve progress in child mortality even in the absence of improvements in contextual factors such as per capita income and HIV prevalence.

The observed correlation between the changes in antenatal care and family planning outreach services and mortality, though weaker than that observed for access to clinical services, suggests that improvements in access to these services can also have an important impact on mortality. With respect to immunization, some studies have reported a positive effect of immunization on child mortality [14‐16], while others failed to demonstrate its effect on mortality [17‐19]. Indeed, in our study, annual increases in immunization coverage were not correlated with changes in under-five mortality rates, although clearly it is essential to maintain or improve such coverage. Furthermore, the anticipated addition of pneumococcal and rotavirus vaccines to many national immunization programs may well boost the impact of improved immunization coverage on mortality.

With respect to community and family services, improved breastfeeding coverage and an increase in households with improved sanitary facilities also had an impact on under-5 mortality, although the impact was less in Africa than in the 27 countries as a whole. Indeed, exclusive breastfeeding for 6 months has been identified as having the greatest potential impact of any single intervention on under-5 mortality [20]. Regarding nutritional status, malnourished children are known to be at an increased risk of death [21‐23], and it has been estimated that malnutrition contributes to about one-third of the child deaths that occur each year [24]. Our study demonstrated that an improvement in childhood nutritional status is associated with a decline in under-5 mortality rates, although the difference was significant only in the subgroup of African countries. These findings are compatible with previous studies [23, 25, 26]. It has been estimated that implementing interventions that improve child nutrition as well as provide clean water and sanitation to all children younger than 5 years would result in an estimated 31% reduction in child deaths in sub-Saharan Africa [25].

While these findings support further scaling up services in the outreach and community and family domains, they also would appear to underline the importance of additional investment in access to services. Prior efforts to scale up of clinical services have demonstrated major challenges in identifying and training adequate skilled staff to provide these services and in providing the facilities and logistics to ensure quality and continuity. In addition, the high costs of delivering such services have been a major inhibitory factor. Indeed, the overall costs of improving and sustaining a system capable of delivering rapid gains in health improvement are large, estimated at $36-45 billion annually above and beyond current expenditures, and the difficulty of raising such funds given current worldwide economic conditions, is substantial [27]. At the same time, however, our findings suggest that the benefits of such improvements may also be large.

One of the reasons behind the high costs of clinical services is the difficulty of providing care to those most in need, many of whom live in remote areas where access may be difficult. An encouraging development that may improve clinical coverage and also offers some excellent opportunities for synergy with outreach and family and community efforts is integrated community case management (iCCM) [28]. In iCCM, treatment for conditions such as pneumonia diarrhea, and malaria, which account for approximately 40% of under-5 mortality worldwide [29], is delivered at community level by specially trained community health workers. Such efforts also have the advantage of emphasizing family and community-based prevention and promotion and provide opportunities to promote better coverage of outreach services, resulting an excellent opportunity for synergy of those interventions which appear to have the greatest impact on reductions in mortality.

While other analyses have focused on the relationship between prevalence or coverage of risk factors and current mortality levels in study populations or at country level [[7]30], to our knowledge, this is the first analysis that examined the correlation between changes in individual and composite variables and changes in under-5 mortality at country level. We believe this approach provides a more dynamic way of examining how the scaling up of different types of interventions effect changes in under-5 mortality rates that is not provided by the simple correlation of existing coverage and mortality levels.

At the same time, however, these findings need to be interpreted cautiously. First, the results may not be generalizable because of the limited number of countries included in the analysis. We believe, however, that the countries included represent a wide range of experiences, both in terms of geography and initial levels of development. A second limitation is that consistent data on potentially important variables such as vitamin A administration and use of insecticide treated bed nets, which are highly effective in preventing malaria, were not available from the penultimate surveys. Furthermore, it is clear that many of the variables we included were markers rather than directly associated with child mortality as an outcome. For example, changes in Cesarean section, which as an individual variable produced one of the highest correlations with under-5 mortality change, was less strongly associated with reductions in the infant mortality, as might have been expected with a more direct effect; which strongly suggests that it is a marker of access to quality clinical services. A fourth limitation was that some of the variables that might be more plausibly linked to mortality such as treatment for acute respiratory infection could not be included because the two-week reference frame in the questionnaire results in prevalence values of < 5%, making calculation of average annual change problematic. A fifth limitation concerns the assumption that the effects of individual interventions and their interactions will be similar for our countries, which may not be true as suggested by the differences observed between the overall set of countries and the African countries. Finally, mortality data collected in surveys reflect events occurring in the population an average of 2.5 years previously. The mortality figures we used thus reflect mortality that occurred in the early 2000s, before many of the interventions had been scaled up, which may underestimate the influence that changes in coverage have had on mortality.

Our study looked at change rather than absolute coverage levels. It is important to keep in mind that coverage was low for many of the indicators. Except for the immunization variables, where coverage tended to be high, it was much lower for many other potentially life-saving interventions. For example, the percentage of women reporting the use of a skilled birth attendant ranged from 6% to 83%, with a median value of 47% for the 27 countries; for appropriate treatment of diarrhea, values ranged from a low of 18% to a high of 75%, with a median of 37%. These data suggest that there is great room for improvement, even in countries that have already made substantial progress.

Although this study is limited by its ecologic nature and the limited number of countries and information that were available for inclusion, we nonetheless feel that the method of examining average annual change rather than looking at absolute levels of outcomes should be considered in future studies examining the impact of change at a population level. More importantly, however, the findings on the apparent role of clinical services in further reducing under-five mortality may provide support to current efforts to improve the availability of care at community level and to make needed improvements in the health infrastructure, manpower, and logistic and supply systems.

The authors wish to thank David Brown and Attila Hancioglu for their helpful comments and suggestions, to Mary Gray and David Olsen for their invaluable advice on data analysis, and Danielle Burke and Carlos Carrera for their assistance in identification of data bases used in these analyses. No specific funding was provided for this project, which was considered a routine UNICEF activity.