The burden of disease from neonatal mortality: a review of South Asia and Sub-Saharan Africa


* Dr A. A. Hyder, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Suite E-8132, 615 North Wolfe Street, Baltimore, Maryland 21205, USA.


Objective To assess the burden of neonatal mortality in two developing regions.

Design Review of secondary data collated through literature review.

Setting Community and facility based studies and national health surveys.

Population Neonatal (<28 days) population of South Asia and Sub-Saharan Africa.

Methods Data on neonatal mortality were gathered from peer-reviewed literature, reports of the Demographic and Health Surveys and websites of country-based organisations. The base year for this study is 1995. For each country, a weighted mean neonatal mortality rate was calculated and the total number of neonatal deaths estimated. Country data were summarised over each region to estimate annual regional neonatal deaths and rates. The burden of disease from neonatal mortality was determined using a summary measure of health—healthy life years.

Main outcome measures Numbers of deaths and healthy life years (HeaLYs).

Results Neonatal mortality rate for South Asia ranged from 41.9 to 56.9 per 1000 live births for 1995. Sri Lanka was an exception with a neonatal mortality rate between 16.3 and 18.6 per 1000 live births. The estimated regional neonatal mortality rate for South Asia was 46.27 per 1000 live births for 1995. There was a significant lack of data from Sub-Saharan Africa, resulting in highly variable neonatal mortality rates, ranging from 13 per 1000 live births in Kenya to 108 per 1000 live births in Senegal. The mean regional neonatal mortality for Sub-Saharan Africa for 1995 was estimated at 38.8 per 1000 live births.

Conclusion The burden of neonatal mortality in only these two regions of the developing world represents more than 2 million annual deaths. A call for greater investments in neonatal research and health programs is a logical extension to this review of evidence.


Neonatal mortality, defined as death of an infant during the first 28 days of life, is a major contributor towards infant mortality, accounting for two-thirds of deaths in children under the age of one year1. Nearly 5 million neonatal deaths have been estimated to occur around the world annually, 98% of these being in developing countries2. Neonatal mortality contributes substantially towards under-five and infant mortality and while infant mortality rates have declined during the past decades, neonatal mortality rates have remained relatively unchanged, especially in developing countries3. Besides mortality, neonatal conditions, such as prematurity and low birthweight, have implications from complicating conditions (such as pulmonary disease) and affect quality of life. Conditions affecting neonates also result in retarded growth and disability through neurological and cognitive impairment.

South Asia and Sub-Saharan Africa are the poorest and the most underdeveloped parts of the developing world, with the worst absolute and relative indicators of health and poverty in the world. These two regions have some of the highest child and infant mortality rates in the developing world and will be the focus of this burden of disease study. Previous attempts to define the burden of disease, globally or regionally, have not defined the neonatal age group separately4. The inclusion of neonatal deaths within the 0–4 age category results in a loss of focus on the first 28 days of life and undercounting of neonatal deaths. This paper is a first step in such an overall consideration of neonatal health. It assesses country- and region-specific data and uses standardised methods to estimate the burden of disease from neonatal mortality in Africa and Asia. The overall purpose of this paper is to define the impact of neonatal mortality on the developing world. The specific objectives are to define the number of neonatal deaths in Sub-Saharan Africa and South Asia, to estimate the burden of disease from neonatal mortality and to explore the distribution of that burden by country. This work will highlight the implications of neonatal health issues for health research and health care policy in the developing world.


This study is focussed on estimating the burden of disease from neonatal mortality in only two developing regions of the world. The estimates do not account for perinatal mortality, stillbirths or abortions. The study traces the consequences of premature mortality after a live birth, using the following definition of neonatal mortality: death occurring after a live birth and not later than the end of the 28th day of life. For the purposes of this study, 1995 was chosen as the baseline year.

An extensive search of the PubMed database was conducted to retrieve literature on neonatal mortality in Sub-Saharan Africa and South Asia. Combinations of following keywords were used: neonatal mortality, neonatal morbidity, neonatal death, neonatal infections, neonatal tetanus, birth injury, birth trauma, birth asphyxia and congenital malformations. The name of each country was added to these terms to get databases for respective countries. The search was extended by using the ‘related articles’ link on PubMed. During review of the articles, additional papers were identified from the respective reference lists. Articles published between 1980 and 2001, in English, were included in the study.

Studies considered for inclusion in the review were restricted to original reports of facility- or community-based studies giving estimates of neonatal mortality, studies giving proportional mortality data, critical reviews that provided reports of original research results and quantitative data and reports of country-based or nationally representative samples in the non-published literature. Studies that were excluded from the review included reports that did not give the size of study population or denominators (such as hospital case reports), studies in which the criteria for recruiting the subjects and age groups were not defined, studies that only described the presence of neonatal diseases and injuries in the study population, but did not give estimates for incidence or mortality rates, studies that gave estimates for refugee populations instead of the native population of any country and studies that exclusively evaluated twin births rather than singleton births.

Estimates for neonatal and infant mortality rates were also collected from the reports of Demographic and Health Surveys (DHS; DHS data for both regions acted as the primary source for overall neonatal mortality rates. For certain countries within each region, there were alternative national data resources that were used to compare the data reported by DHS. Contact was made with a range of individuals and organisations for data and unpublished information. In addition, three data sources were explored to get information on neonatal and infant mortality in South Asia, which included Sample Registration System of India (, data from the Matlab field area in Bangladesh and the Civil Vital Registration System of Sri Lanka ( The databases of papers, for each country, were maintained in Reference Manager8.

Due to the variability in data and broad range of rates reported by community-based studies, the imperative to estimate a single national neonatal mortality rate was challenging. Comparisons of neonatal mortality rate with infant mortality rate and time trends were used for exploring the data and to define and exclude outliers. Outliers were a single data point more than five years before or after the majority of data points for a country, data points that reflected neonatal mortality at a vastly different level of infant mortality and studies with small sample sizes (less than 50). Neonatal mortality rates reported by various community-based studies, nationally representative surveys and DHS were used to determine simple mean and weighted mean national neonatal mortality rate for 1995. For countries with very few studies reporting neonatal mortality rate, such as Nepal and Pakistan in South Asia and Benin, Botswana, Cameroon, Liberia, etc. in Sub-Saharan Africa, the relationship between neonatal and infant mortality was used to estimate a 1995 neonatal mortality rate. These extrapolated rates were used to estimate the burden of premature mortality.

UN projections for live births averaged for 1995 were used to estimate total neonatal deaths for 1995 in each country. Regional estimates of neonatal mortality rate were generated by using a weighted mean of national neonatal mortality rates for 1995.

The burden of neonatal deaths was calculated using summary measures of health or ‘health gap’ measures that estimate the loss of healthy life from premature mortality. The healthy life year (HeaLY) methodology has been used previously in burden of disease assessment and is used in this study9,10. A detailed description of the HeaLY method is described elsewhere11–13.


The search for articles in indexed journals yielded 106 articles from South Asia and 148 articles for Sub-Saharan Africa; these comprised two major categories of studies—community- and facility-based studies. Although both sources were evaluated, the neonatal rates presented in this paper are based on community-based studies. Overall, 25 papers (14 community-based studies) were included in evaluating neonatal mortality for South Asia and 35 (23 community-based studies) for Sub-Saharan Africa. Data for five countries in South Asia were available from DHS and for 30 countries in Sub-Saharan Africa. (A detailed bibliography of data sources is available from the authors.)

Neonatal mortality rates reported by community-based studies including DHS for South Asia (except for Sri Lanka) ranged between 41.9 and 56.9 per 1000 live births. The range for Sri Lanka was much lower, between 16.3 and 18.6 per 1000 live births. Few studies provided neonatal and infant mortality rates together, and Fig. 1a shows the plot of neonatal and infant mortality rates for these studies. The time trend for neonatal mortality rates is shown in Fig. 1b. Weighted means of neonatal mortality rates based on non-DHS data sources compared with the DHS rate for each country are shown in Table 1. The estimated national neonatal mortality rates for 1995 for all countries were within 42 and 47 per 1000 live births, except Sri Lanka. The regional neonatal mortality rate for South Asia was estimated to be 46.27 per 1000 live births.

Figure 1.

(a) Plot of neonatal mortality rate (NNMR) against infant mortality rate (IMR) for South Asia: ○ DHS reports; ♦ community-based studies. (b) Trend in neonatal mortality (South Asia). Data points represent community-based studies as well as DHS data.

Table 1.  Neonatal mortality rates (NNMR): South Asia and Sub-Saharan Africa.
CountryNon-DHSDHS (year)National 1995 NNMR
Number of studiesCumulative sample sizeWeighted mean
  1. * * One estimate for neonatal mortality rate included in the study was taken from non-DHS nationally representative databases6,7.

  2. * Neonatal mortality rates extrapolated from the trend line of relationship between infant and neonatal mortality rates for each country.

Bangladesh793,89347.2942 (1999–2000)47.2
India**515,98850.9643.4 (1999)47.5
Nepal00049.9 (1996)46.48*
Pakistan1147656.948.9 (1991)42.18*
Sri Lanka**1 18.616.3 (1987)10.44
Regional NNMR    46.27
Benin00038.2 (1996)38.72*
Botswana00022.5 (1988)27.84*
Burkina Faso121073040.8 (1999)39.5
Burundi00035.2 (1987)47.6*
Cameroon00037.2 (1991)34.04*
Central African Republic00042.1 (1994)42.14*
Chad00043.9 (1997)45.92*
Comoros00038.2 (1996)34.96*
Cote d'Ivoire00042 (1994)38.28*
Eritrea00024.8 (1995)40.04*
Ghana00029.74 (1998)35.2
Guinea00048.4 (1999)48.36*
Kenya415,37615.328.4 (1998)27
Liberia1530915.967.9 (1986)46.4*
Madagascar00040.4 (1997)36.9*
Malawi514,78147.241.2 (1992)38
Mali00060.4 (1996)47.2*
Mozambique00053.9 (1997)45.6*
Namibia00031.5 (1992)30.55*
Niger00044.2 (1998)46*
Nigeria2422424.842.2 (1990)35.23*
Ondo State00026.3 (1986) 
Rwanda00038.6 (1992)48.36*
Senegal2667047.837.4 (1997)37
South Africa1784,30418.219.8 (1998)28.32*
Sudan224452043.8 (1990)33*
Tanzania00040.4 (1999)36.5
Togo00041.3 (1998)37.38*
Uganda00027 (1995)44.94
Zambia1050.635.4 (1996)37*
Zimbabwe00028.9 (1999)27
Regional NNMR    38.76

Using United Nations projections for live births by country and the regional neonatal mortality, it is estimated that each year neonatal mortality claims 1.6 million lives in South Asia (Table 2). This represents deaths in five countries ranging from 3400 deaths in Sri Lanka to 1.18 million deaths in India. Neonatal deaths resulted in the loss of 48 million HeaLYs in 1995, representing a loss of 39.80 HeaLYs per 1000 population in South Asia (Table 2). When compared with all the countries within South Asia, Nepal had the greatest number of HeaLYs per population lost as a result of neonatal mortality. Sri Lanka remains an outlier for the low impact of neonatal mortality among its neighbors, at least for the mid-nineties.

Table 2.  Burden of neonatal mortality: South Asia and Sub-Saharan Africa.
CountryNumber of neonatal deaths*Healthy life years lost (HeaLYs) due to neonatal deaths**Healthy life years lost (HeaLYs) per 1000 population
  1. * Based on UN Projections 1990–1995 and 1995–2000 averaged for 1995.

  2. * * Based on HeaLYs9.

Sri Lanka3401103,8265.8
Burkina Faso20,971619,85159.5
Central African Republic5473167,24950.9
Cote d'Ivoire22,176617,06745.6
South Africa17,408912,16324.3

There was a significant lack of data, especially population-based, from Sub-Saharan Africa. Most of the information on neonatal mortality in the region was provided by DHS data. The reported neonatal mortality rate by community-based studies was highly variable and ranged from 13 per 1000 live births in Kenya to 108.9 per 1000 live births in the Senegal. The trend of neonatal mortality rates with infant mortality and time is shown in Figs 2a and b. The weighted means of non-DHS neonatal rates and DHS rates by country are shown in Table 1. As can be seen, non-DHS data sources were uncommon, and for several countries, neonatal mortality for 1995 had to be estimated based on reported infant mortality rates. The overall regional neonatal mortality rate for Sub-Saharan Africa was estimated at 38.8 per 1000 live births (Table 1).

Figure 2.

(a) Plot of neonatal mortality rate (NNMR) against infant mortality rate (IMR) for Sub-Saharan Africa: ○ DHS reports. (b) Trend in neonatal mortality (Sub-Saharan Africa). Data points represent community-based studies as well as DHS data.

Nearly 0.8 million neonatal deaths (Table 2) occurred in Sub-Saharan Africa in 1995, based on UN projections for live births. Country-specific mortality ranges from only 917 deaths in the Comoros to 172,000 neonatal deaths in Nigeria. These deaths result in more than 24 million HeaLYs being lost in Sub-Saharan Africa, equivalent to a rate of 157 HeaLYs per 1000 population. Although Nigeria, Ethiopia and Tanzania lead in terms of the absolute loss of healthy life, the per capita losses are highest in Somalia, Rwanda and Liberia.

The estimates for neonatal mortality in this study were compared with those for the same time period as reported by the World Health Organization (WHO) in the World Health Report 1996. The neonatal mortality rates reported by WHO for the year 1996 range from 20 per 1000 live births in Sri Lanka to 65 per 1000 live births in Bangladesh. Absolute and proportional differences between the rates reported by WHO and estimated by this study were examined and the comparison reflects that overall this study is reporting lower estimates when compared with the WHO for South Asia. This is especially true for Sri Lanka, where estimates for 1995 are 50% of what WHO reported. For Sub-Saharan Africa, the lowest neonatal mortality rate reported by WHO is 20 per 1000 live births in Botswana and the highest in the region being 70 per 1000 live births in Guinea and Somalia. In comparing the estimates for Sub-Saharan Africa with the rates reported by WHO, it is clear that there is a mixed picture; while the majority of the rates estimated by current study are lower than WHO's figures, 8 out of 38 are greater, with three of them by more than 25%.


There is a paucity of information on neonatal health, especially population-based studies, and the most consistently available source of neonatal mortality was the nationally representative DHS. Community-based studies were found in South Asia, but were severely lacking in Sub-Saharan Africa. Such studies usually monitor geographically defined populations and follow outcomes of live births, thus providing mortality rates. On the other hand, hospital-based studies in these two regions are not plentiful either and often do not describe the methods used or sample selection in great detail. The inaccessibility of raw data from such studies further complicates the ability to review information. In general, the majority of studies published in the peer-reviewed literature on neonatal mortality were conducted in hospital settings with relatively small sample sizes, thus with limited generalisability of results. Therefore, they were excluded from neonatal mortality rate estimation for this study.

In South Asia, the majority of published studies were conducted in Bangladesh and India; no data were available for Bhutan and Maldives, and our only source of information for Nepal was the DHS report. For most countries in Sub-Saharan Africa, DHS reports were the only available information regarding neonatal mortality. No country-specific reports or annual government data sources (usually reported by countries to WHO) were included in this review given that their validity is hard to determine and the use of primary means of verifying the reliability of their statistical methods were beyond the scope of this study.

Neonatal mortality accounts for up to two-thirds of deaths during the first year of life3. It is an indicator for maternal and fetal health during pregnancy, delivery and the immediate postnatal period. Although there has been a decline in infant mortality rates in the developing world, no substantial decrease in neonatal mortality rates has been noted (see Figs 1b and 2b)14. South Asia and Sub-Saharan Africa have one of the highest neonatal mortality rates in the world. This study estimates that neonatal mortality claimed 1.6 million deaths in 1995 in South Asia and these deaths were distributed across only five countries. The neonatal mortality rate of 46.2 per 1000 live births for the region is a reflection of the high impact of mortality in this age group. The overall rate is an aggregate with pockets of much higher neonatal mortality (∼60 per 1000 live births) within the region. Except for Sri Lanka, the other four countries have similar levels of neonatal mortality, while recent reports indicate that Sri Lanka may have experienced a recent rise in its mortality rates15. The explicit definition of neonatal issues is therefore critical to attract more attention and resources to this public health issue in South Asia.

The impact of neonatal mortality on Sub-Saharan Africa is lower in absolute (0.8 million deaths) and relative terms than South Asia. The neonatal mortality rate for Sub-Saharan Africa (38.7 per 1000 live births) is actually lower than South Asia (46.2 per 1000 live births), but reflects greater intra-regional variation, from 15 in Kenya to 68 in Liberia. This reported variation is a result of both the quality and availability of data and the national variations in determinants of neonatal and child health, as well as differences in health care systems.

An exploration of the relationship between neonatal and infant mortality rate reflects the lack of information on infant mortality from those studies reporting neonatal mortality rate. The fact that research studies report only one rate is surprising in view of the well-known association between neonatal and infant mortality rate. When available, scatter plots have shown that studies that report a higher infant mortality rate also report a higher rate for neonatal mortality in both regions. In general, national time trends demonstrate that neonatal mortality does not decline at the same rate as infant mortality, which makes further improvements in child health dependent on the reduction of neonatal mortality rate.

It is difficult to interpret trends in neonatal mortality over time from diverse data sources. In general, the neonatal mortality rate in South Asia seems to have declined since 1980, especially in countries such as Bangladesh where regular data are available. In Sub-Saharan Africa, data over time for most countries show that the rate has remained either largely unchanged or has shown minimal decrease. Exceptions include countries such as Kenya and Liberia where the neonatal mortality rate seems to have increased over time and present a major concern. It may warrant some focussed research on such nations to better define such negative trends and determine their causative factors.

The only other easily available review of neonatal mortality rates was reported in the World Health Report 199616. The methodology for the WHO review is not clear, except that it was based on a ‘review of information’ and data reported by countries to WHO. WHO estimates from that report, in general, were higher than those reported by this study. Countries such as Sri Lanka, Botswana, Cote d'Ivoire, Guinea, and Mozambique demonstrate a high variation between the two estimates. It is important to note that even small differences may mean thousands of lives and so there needs to be an appreciation of the impact of small differences. On the other hand, the quality and coverage of data, on which such estimates are based, are a cause for caution in reading too much in minor differences. It may be time for WHO to revise the estimates of neonatal mortality in a systematic and transparent manner.

Composite indicators of health status, such as HeaLYs and disability-adjusted life years, have become important tools that help researchers and policymakers estimate the impact of morbidity and mortality. This study has not evaluated the impact of neonatal morbidity or disability. Neonatal mortality (1995) results in a loss of 48 million HeaLYs (39 HeaLYs per 1000 population) in South Asia and 25 million HeaLYs (157 HeaLYs per 1000 population) in Sub-Saharan Africa. Although the absolute loss of healthy life from neonatal mortality is twice as high in South Asia as it is in Sub-Saharan Africa, the rate (per population) for Sub-Saharan Africa is about four times higher—reflecting the very high burden in Sub-Saharan Africa relative to the population. These figures demonstrate the loss of potential life as a result of these neonatal deaths, which have not been specifically evaluated in previous burden of disease estimates4.

Comparison of the regional burden of disease estimates for neonatal mortality in this study to the global burden of disease estimates of WHO is not possible due to differences in time frames, methods and regional groupings. The 1998 estimates by WHO are the closest to 1995 and do not have South Asia or Sub-Saharan Africa as regions. They include regions called ‘Africa’ and ‘South East Asia’, do not give neonatal deaths and do not give national estimates. Neonatal deaths in Sub-Saharan Africa and South Asia based on this study, when combined, represent 4.5% of all deaths estimated by WHO in the world. Neonatal deaths in Sub-Saharan Africa represent 1.5%, while neonatal deaths in South Asia represent 3% of all global deaths. Similarly, the burden of disease from neonatal mortality in Sub-Saharan Africa and South Asia represent 1.9% and 3.8% of the global losses, respectively. These percentages are meant to reflect the fact that greater efforts to measure the impact of neonatal mortality will lead to a better appreciation of this important public health issue. The WHO global burden of disease estimates for 1998 or other years do not provide figures for either neonatal deaths or perinatal deaths; however, they do estimate the number of deaths from ‘perinatal causes’ in the world and for each region. If estimates like those in this study were to be included in the global WHO analysis, then the ranking of the leading causes of deaths by region would see a jump in the rank order of the perinatal category. This is again a reflection of the undercounting of neonatal mortality in previous global estimates.

The lack of data for neonatal mortality was the major limitation of the present study, especially in Sub-Saharan Africa. In addition, studies published in French only or not on PubMed did not qualify the inclusion criteria for this study. Very few studies provided data on early and late neonatal mortality and that has limited such analysis in this study. The studies that were used had variations in the study population, study settings and methods. This study uses 1995 as the base year for estimates since the availability of data from both pre- and post-1995 time periods allowed for such an analysis. Changes in neonatal mortality rate may have taken place since 1995 and therefore current (2002) estimates may be different, especially for countries such as Bangladesh and Sri Lanka in South Asia or Zimbabwe and South Africa in Sub-Saharan Africa. As more recent data becomes available, this report recommends a re-estimation of regional neonatal mortality. The estimations provided in this study also do not include disability and morbidity and, therefore, have not accounted for the loss of healthy life from non-fatal health outcomes in neonates. Additional research studies are needed to explore available data on neonatal morbidity and long term disability.

The main goal of this paper is to emphasise the lack of valid data on neonatal mortality and as a consequence, the potential to undercount the impact of this public health problem. Neonatal mortality claimed more than 2 million lives for 1995 in only two regions of the world. This figure could easily double if other parts of the developing world were included in the analysis. This is a huge public health disaster, which is potentially preventable and therefore deserves urgent and immediate action.


This project was funded by Saving Newborn Lives—Save the Children, USA. The authors would like to thank Gary Darmstadt and Kenneth Hill for their special encouragement and assistance. They also thank the following colleagues for their helpful comments at various steps: Abdullah Baqui, Robert Black, Richard Morrow, Stan Becker, and Khurram Nasir.