Dwomoh BMC Public Health          (2021) 21:492 
https://doi.org/10.1186/s12889-021-10473-w
RESEARCH ARTICLE Open Access
Geospatial analysis of determinants of
neonatal mortality in Ghana
Duah Dwomoh
Abstract
Background: Ghana did not meet the Millennium Development Goal 4 of reducing child mortality by two-thirds
and may not meet SDG (2030). There is a need to direct scarce resources to mitigate the impact of the most
important risk factors influencing high neonatal deaths. This study applied both spatial and non-spatial regression
models to explore the differential impact of environmental, maternal, and child associated risk factors on neonatal
deaths in Ghana.
Methods: The study relied on data from the Ghana Demographic and Health Surveys (GDHS) and the Ghana
Maternal Health Survey (GMHS) conducted between 1998 and 2017 among 49,908 women of reproductive age and
31,367 children under five (GDHS-1998 = 3298, GDHS-2003 = 3844, GDHS-2008 = 2992, GDHS-2014 = 5884, GMHS-
2017 = 15,349). Spatial Autoregressive Models that account for spatial autocorrelation in the data at the cluster-level
and non-spatial statistical models with appropriate sampling weight adjustment were used to study factors
associated with neonatal deaths, and a p-value less than 0.05 was considered statistically significant.
Results: Population density, multiple births, smaller household sizes, high parity, and low birth weight significantly
increased the risk of neonatal deaths over the years. Among mothers who had multiple births, the risk of having
neonatal deaths was approximately four times as high as the risk of neonatal deaths among mothers who had only
single birth [aRR = 3.42, 95% CI: 1.63–7.17, p < 0.05]. Neonates who were perceived by their mothers to be small
were at a higher risk of neonatal death compared to very large neonates [aRR = 2.08, 95% CI: 1.19–3.63, p < 0.05]. A
unit increase in the number of children born to a woman of reproductive age was associated with a 49% increased
risk in neonatal deaths [aRR = 1.49, 95% CI: 1.30–1.69, p < 0.05].
Conclusion: Neonatal mortality in Ghana remains relatively high, and the factors that predisposed children to
neonatal death were birth size that were perceived to be small, low birth weight, higher parity, and multiple births.
Improving pregnant women’s nutritional patterns and providing special support to women who have multiple
deliveries will reduce neonatal mortality in Ghana.
Keywords: Neonatal mortality, Child health, Geospatial modeling, Ghana
Correspondence: duahdwomoh@yahoo.com
Department of Biostatistics, School of Public Health, College of Health
Sciences, University of Ghana, Legon, Accra, Greater Accra, Ghana
© The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Dwomoh BMC Public Health          (2021) 21:492 Page 2 of 18
Background spatially varying relationships between neonatal mortal-
Globally, neonatal deaths remains a major public health ity and the following indicators: maternal education,
concern as 5.3 million children under age five died in women’s empowerment, home births, mothers without
2018 [1]. The risk of a family having or experiencing an education, and mothers whose husbands decided on
neonatal deaths within the first 30 days of the child life contraceptive practices, rural residency.
is still highest in the World Health Organization (WHO) Full-scale interventions are costly, especially in
African Region (76 per 1000 live births), which is ap- poverty-driven economies in Sub-Saharan Africa, and
proximately eight times as high as the WHO European hence interventions must be proven to work, selective,
Region (9 per 1000 live births) [1]. For instance, in 2018 and targeted at only essential risk factors associated with
alone, the under-five mortality rate in low-income coun- neonatal deaths [10]. With the adoption of the Sustain-
tries was 68 deaths per 1000 live births, which is almost able Development Goals (SDGs) in 2015, which estab-
14 times the average rate in high-income countries (5 lished ambitious targets for improving child survival by
deaths per 1000 live births) [1]. Neonatal mortality re- 2030, optimal intervention planning and targeting will
mains a major contributor to under-five mortality [2]. require an understanding of factors influencing the high
Despite the huge investment made by the govern- incidence of neonatal mortality [5]. This study aimed to
ments in sub-Saharan Africa and foreign partners to- use a more rigorous geospatial and traditional statistical
wards reducing neonatal deaths in Sub-Saharan modeling approach to identify key risk factors associated
Africa, only 12 countries in the World Health Organi- with neonatal deaths in Ghana to target interventions.
zation’s African Region met the Millennium Develop-
ment Goal 4 (MDG4) to reduce U5MR by two-thirds Methods
by 2015. These countries are Eritrea, Ethiopia, Liberia, Data source
Madagascar, Malawi, Mozambique, Niger, Rwanda, This study pooled data from the Ghana Demographic
Senegal, Tanzania, Uganda, and Zambia [3]. Health Surveys (GDHS) conducted in 1998/1999, 2003,
One of the Sustainable Development Goals 3 (SDG 3) 2008, and 2014, and Ghana Maternal Health Survey
primary objectives is to reduce neonatal mortality to at (GMHS) conducted in 2017. The GDHS is a nationally
least as low as 12 per 1000 live births [4]. However, the representative household survey of 24,846 women age
rate of progress toward these goals significantly varied at 15–49 and an average response rate of 96.7%. The total
the national level, demonstrating an essential need for number of children from the GDHS was 16,018. The
tracking even more local trends in child mortality [5]. GMHS interviewed 25,062 eligible women aged 15–49
Global estimates from 2010 show that approximately in the selected households with an average response rate
40% of all deaths to children under age 5 occur during of 99.0%. With the GMHS, the study restricted the ana-
the neonatal period [6]. Many countries have made pro- lysis to only children who were born alive in the 5 y pre-
gress in reducing deaths among children under age 5, ceding the survey. The two distinct surveys have
but these gains have been predominantly among chil- variables on neonatal mortality and other relevant covar-
dren age 1–4 [7]. Little progress has been made in redu- iates of interest. The study relied on the 1998–2014
cing the mortality risk for children under age 12months, GDHS and 2017 GMHS, the latest survey conducted in
especially for neonates, in the first month of life [7]. Ghana that has data on neonatal mortality and relevant
Neonatal mortality rates in Ghana remain unaccept- predictors. The 2017 GMHS was analyzed independently
ably high despite the substantial financial investment in from the GDHS datasets. The analysis was restricted to
the health sector over the last decade. Ghana’s neonatal only women who had children in the last 5 y preceding
death rate fell gradually from 202.8 deaths per 1000 live the survey. The sample contains the total number of
births in 1968 to 49.3 deaths per 1000 live births in children under-five born in the 5 y preceding the survey,
2017. To meet the set target of sustainable development and that data contains the information on their respect-
goal 3 (SDG 3), there is the need to understand complex ive mothers and households in terms of education, place
causes of neonatal deaths [8]. To improve neonates’ of residence, household wealth, parity, household size,
health, it is necessary for decision-makers to effectively etc.
comprehend the correlation between maternal, child, A detailed description of the data sets used for the
and household characteristics and how they co-integrate study can be found in Table 1.
with geospatial factors to influence the survival probabil-
ity of neonates in Ghana. Policymakers’ ability to under- Sampling design
stand the variation in neonatal mortality and the The 1998–2017 GDHS and GMHS were based on a
underlying causes is important for targeted intervention multi-stage stratified cluster sampling. First, the country
[9]. Recently, Grady et al. [2] have used geographically was stratified by the 10 regions and urban and rural
weighted Poisson regression models to capture the areas, generating 20 sampling strata. Samples of
Dwomoh BMC Public Health          (2021) 21:492 Page 3 of 18
Table 1 Description of the data source used for the study
Year of survey Source of data Number of women aged Number of children born in the The average response rate
15–49 years interviewed five years preceding each survey of eligible women (%)
1998 GDHS 4843 3298 97.4
2003 GDHS 5691 3844 95.7
2008 GDHS 4916 2992 96.5
2014 GDHS 9396 5884 97.3
2017 GMHS 25,062 15,349 99.0
Total 49,908 31,367 Average response rate = 97.2%
Abbreviations: GDHS Ghana Demographic and Health Survey, GMHS Ghana Maternal Health Survey
enumeration areas (EAs) were selected independently in cooking fuel, the frequency at which household mem-
each stratum in two stages. In the first stage of obtaining bers smoke inside the house, place where household
the samples, EAs (residential households) were selected members wash their hands, place of cooking, access to
with probability proportional to the EA size (number of an improved water source and household access to an
residential households) and with independent selection improved sanitation facility, number of women aged 15–
in each of the 20-sampling stratum. The EA size was ob- 49 years in the household, whether the household has io-
tained from the 2010 Population and Housing Census. dized salt, whether any member of the household own
The sampling frame for the survey was obtained by con- any agricultural land, whether the household has any
ducting a household listing operation in all the selected livestock, herds, other farm, animals, or poultry, and
EAs, and the resulting lists of households served as a household wealth. The wealth index was used as a proxy
sampling frame for the selection of households in the to measure the socioeconomic status of the household.
second stage. In the second stage of selection, a fixed The composite wealth index constructed by the DHS is
number of households per cluster (EA) was selected with based on household-level data on assets, services, and
an equal probability systematic selection from the newly amenities and ranks households according to their
created household listing. The trained interviewers vis- wealth level. A household is defined as having access to
ited and interviewed only the selected households. The an improved water source if it has any of the following:
detailed sampling design and sample selection for the piped water into the dwelling, yard, or plot; public tap or
2014 GDHS can be found in [11]. standpipe, tubewell, or borehole; a protected dug well or
protected spring; rainwater; or bottled water. Studies
Outcome measure have found access to an improved water source to be as-
This study investigated neonatal deaths in Ghana. All sociated with infant survival [13]. A household is defined
the children who died within the 5 y but the death that as having improved sanitation if it has any of the follow-
occurred within the first 28 days after delivery were clas- ing types of facilities: a flush or pour-flush to a piped
sified as neonatal deaths. The neonatal mortality rate sewer system, septic tank, or pit latrine; a ventilated im-
was estimated within 5 y preceding the survey and esti- proved pit latrine; a pit latrine with a slab; or compost-
mated using the synthetic cohort probability method. ing sanitation. Household access to improved sanitation
is associated with lower levels of infant mortality [13].
Predictors of neonatal deaths Finally, this study investigated the impact of household
This study relied on the Mosley and Chen conceptual size on neonatal deaths.
framework for neonatal deaths in developing countries
[12]. Lifestyle factors
In the last seven days, the number of fruits and vegetable
Characteristics of the household consumption was used as a proxy to assess fruits and
The household characteristics that were studied include vegetable consumption during the antenatal period.
place of residence, region, zone (coastal, middle, north-
ern), age of the household head, sex of the household Characteristics of the child
head, type of oil the household mainly use for cooking, The characteristics of the child that were studied include
whether the household has access to the internet, num- sex of the child (male or female), the weight of the child
ber of household members covered with health insur- at birth, multiple births (single or multiple birth), birth
ance, number of sleeping rooms, whether the household order, birth spacing. The birth order was grouped into
has electricity, main floor material of the household, four categories: first births, second births, third births,
main roofing material, main wall material, type of and fourth or higher-order births. The birth spacing was
Dwomoh BMC Public Health          (2021) 21:492 Page 4 of 18
grouped into three categories: intervals of less than 24 in this study were used as a proxy to understand the
months, 24–35 months, and 36 or more months, similar characteristics of the respondents before pregnancy, dur-
to what has been reported by Pezzullo et al.,2016 [10]. ing pregnancy, and after delivery.
Maternal factors Geospatial covariates
The following maternal related factors were studied: the Studies have linked climatic and environmental factors
current age, age at first birth, parity, the current marital such as precipitation, vegetation density, and length of
status, number of unions, religion, ethnicity, highest growing season with the prevalence of vector-borne dis-
educational level of the mother, mothers body mass eases such as malaria because these factors are related to
index, whether the mother has ever terminated preg- favorable habitats for disease vectors [17]. The following
nancy, currently using a contraceptive, heard about fam- geospatial covariates: aridity, built population, drought
ily planning method, ever been diagnosed with episodes, enhanced vegetation index, global human foot-
hypertension, anemia, the ideal number of children, print, night light composite, malaria, insecticide-treated
owns of a house or land, level of autonomy, level of vio- net (ITN) coverage, proximity to national borders, prox-
lence against women and mothers stature. Short stature imity to a protected area, proximity to water, rainfall,
is an indication of the mother’s nutrition status from the growing season length, environmental temperature,
fetal period through adolescence and reflects the magni- travel time, slope and livestock density were obtained
tude of environmental exposures such as infection, ill- from the Spatial Data Repository under the DHS pro-
ness, and economic hardship during this period. Some gram (https://spatialdata.dhsprogram.com/covariates/).
studies have found a positive correlation between neo- The definition of the geospatial covariate can be found
natal mortality and a mother’s stature [14]. We also in Appendix 4 of supplementary material.
studied the partner/husband’s characteristics in the con-
text of their educational level and occupation. Statistical methods
The study employed two main statistical techniques
Other factors (spatial and non-spatial analyses) to explore the risk and
The following represent the characteristics of maternal protective factors of neonatal deaths in Ghana.
and delivery care and coverage of other interventions
that may have influenced neonatal deaths over the study The non-spatial statistical methods-traditional regression
period: the place of delivery (home or hospital), type of models
delivery (normal or cesarian section), whether the child In this study, the unit of analysis is all children born in
or the mother was covered by health insurance, early ini- the five years preceding each survey (1998, 2003, 2008,
tiation of breastfeeding, currently breastfeeding, health- 2014, and 2017).
seeking behavior (visited health facility in the last six The traditional regression models assume that both
months), the presence of a program helping women to geospatial and non-geospatial covariates are identically
access health services, dwelling sprayed in the last 12 and independently distributed over the geographical
month, household ownership of bednet, number of mal- space, although all statistical analyses adjusted for
aria messages heard, heard about TB, heard messages weighting, clustering, and stratification at the individual-
about antenatal care, mothers slept under-bednet the level analysis. The study employed a three-stage statis-
previous night, children under-five that slept under a tical analytic technique. First, since the study pooled
bednet, number of mosquito bednet in the household, data from different DHS at different survey periods in
whether mother received tetanus injection, iron and Ghana, the women’s standard weights were de-
Fansidar (sulfadoxine and pyrimethamine) for malaria normalized to obtained one pooled data set for all the
prevention during pregnancy, antenatal care attendance. GDHS (1998–2014). Second, bivariate analysis of the se-
Admittedly, household ownership of bednet was mea- lected predictors (household, maternal, child and inter-
sured at the time of the survey; however, we use this in- ventions) of neonatal deaths was conducted using the
formation as a proxy for mosquito net ownership during Rao-Scott designed based adjusted Chi-square test of in-
the pregnancy. During pregnancy, malaria can cause dependence due to the correlation among units within
intrauterine growth retardation and preterm delivery, the same cluster{Rao, 1981 #56}. Third, the study
leading to low birth weight and neonatal death [15]. Be- assessed each covariate’s effect by fitting a Poisson
sides, mothers who slept under-bednet the previous model that includes the covariate of interest and the year
night were used as a proxy for the mother’s net use dur- of the survey fixed effect. Finally, nine different Poisson
ing her pregnancy and during the neonatal period. Early regression models were fitted to assess the multivariable
initiation of breastfeeding could be associated with neo- effect of the predictors on neonatal deaths as follows.
natal survival [16]. In summary, most of the predictors Model 1 assessed the joint effect of all household factors
Dwomoh BMC Public Health          (2021) 21:492 Page 5 of 18
on neonatal deaths. Model 2 and 3 assessed the effect of continuous geospatial covariates were fitted via a four-
only maternal and child-related factors respectively. knot restricted cubic spline functions. The study
Model 4 evaluated the effect of household and child fac- assessed the covariates’ multicollinearity using the vari-
tors. Model 5 assessed the combined effect of household ance inflation factor (VIF) (Additional file 1: Appendix
and maternal factors. Model 6 has a combination of ma- B2). All statistical analyses were conducted using Stata
ternal and child factors. Model 7 assessed the joint effect 15 (StataCorp, College Station, Texas, USA), and p-
of household, maternal, and child-related characteristics. values less than 0.05 were considered statistically
Model 8 has only environmental factors (geospatial co- significant.
variates), and this includes vegetation index, population
density, malaria prevalence per cluster, environmental Results
temperature, rainfall pattern, the length of the growing Descriptive statistics
season, goat livestock, and proximity to a protected area. The average age of the head of the household was ap-
The 9th model integrates household, maternal, child, proximately 40 ±11.02 years (range = 16–99 years), and
and environmental factors. females headed 28% of the households with an average
The study assessed the performance of these nine household size of 5.7. The percentage of households in
models for predicting neonatal deaths by estimating the rural areas was approximately 64%. The percentage of
Schwarz’s Bayesian Information Criteria (BIC), Akaike male and female children born in the 7 y preceding each
Information Criterion (AIC), and a 10-fold cross- survey was almost equal (males = 51.1%, females = 49%).
validated area under the receiver operating characteristic The prevalence of multiple births was 4.5, and 3.7% of
curve (cvAUROC). The best model was selected based the children had a low birth weight. The percentage of 4
on the subject matter knowledge and the estimates from or more antenatal visits was 76.1, and 41% of the
three model performance metrics (AIC, BIC, mothers did not deliver in a recommended health facility
cvAUROC). The study excluded the 1998 GDHS data (home delivery). Seven percent of all the deliveries were
set from all the multivariable models since the survey conducted through a cesarian section, and 48.2% of the
did not measure most of the important predictors of women breastfed after 1 h of delivery. Children born
interest. For instance, household wealth, which was used within the optimal 24–35-month birth spacing was 26%.
as a proxy to measure socioeconomic status was not
measured in the 1998 GDHS, and including the data The trend of neonatal mortality in Ghana: 1998–2017
would have reduced the effective sample size There has been a steady decline in neonatal deaths be-
substantially. tween 1998 and 2017 over the past 15 years. Specifically,
neonatal mortality declined from 30 per 1000 live birth
The spatial statistical methods: assessing the effect of in 1998 to 25 per 1000 live births in 2017 for all births
geospatial covariates on neonatal deaths in the 5 y preceding the survey, although there were
This study performed a descriptive cluster-level analysis some fluctuations between 2003 and 2014. Figure 1 pro-
and spatial distribution of neonatal deaths by aggregat- vides a brief summary of mortality trends in Ghana be-
ing the variables from household survey data and geo- tween 2003 and 2014.
spatial covariates collected by the DHS program to the
cluster level for 2003–2014. Sampling weights were ap- Factors that independently influenced neonatal mortality
plied to calculate appropriate estimates for each indica- The only household factor that influences the risk of
tor at the cluster level. The units of analysis were neonatal deaths was the household size. Neonatal mor-
clusters of households between 2003 and 2014. In all, tality was higher among households with only a few
the study analyzed 1245 unique clusters between 2003 members (4 or less). The risk of neonatal mortality re-
and 2014. To determine whether there was a need for duced by 29 percentage points among households with
applying Autoregressive Spatial Models that account for eight or more members compared to a household with
spatial autocorrelation in the data, we first tested the as- four or less number of inhabitants [adjusted relative risk;
sumption of spatial autocorrelation using the Moran I aRR = 0.7, 95% CI:0.5–0.9; p < 0.05, Additional file 1: Ap-
index. None of the geospatial covariates showed any pendix 1A]. The risk of having neonatal deaths among
form of spatial autocorrelation (Additional file 1: Appen- mothers who had multiple births was approximately five
dix B1). Each geospatial covariate was independently times as high as the risk of neonatal deaths among
evaluated to assess how they contributed to the final mothers who had single births [aRR = 4.5, 95% CI: 3.25–
baseline model’s overall discriminating ability using 6.49, p < 0.01, Additional file 1: Appendix 1A]. The risk
cvAUROC. This was achieved by estimating the trad- of neonatal mortality among mothers who gave birth to
itional modified negative binomial regression model that children in less than 24months after the previous child-
adjusts for weighting, clustering, and stratification. All birth was approximately two times the risk of neonatal
Dwomoh BMC Public Health          (2021) 21:492 Page 6 of 18
Fig. 1 Mortality trends in Ghana between 2003 and 2014
deaths if the index baby is the first child of the Comparative analysis of nine models for neonatal
mothers [aRR = 1.7, 95% CI: 3.3–6.5, p < 0.001, Add- mortality 2003–2014
itional file 1: Appendix 1A]. Neonates who were per- Nine different models were built and evaluated to assess
ceived to be small were at a higher risk of neonatal how they explain neonatal mortality dynamics in Ghana.
mortality [aRR = 1.7, 95% CI: 1.2–2.5, p < 0.01, Add- The nine models’ construction was to enable us to
itional file 1: Appendix 1A]. The risk was reduced by evaluate the statistical significance of one set of inde-
32% among mothers who received two or more tet- pendent variables whiles simultaneously controlling for
anus injections than mothers who received no injec- another set of independent variables.
tion [aRR = 0.7, 95% CI: 0.5–0.9, p < 0.05, Additional Table 2 provides a summary of the nine models for
file 1: Appendix 1A]. Women who were delivered neonatal deaths. Although these models’ discrimination
using cesarean section were at a higher risk of having ability was not substantial, maternal and child factors
neonatal mortality compared to women who had vagi- virtually explain the bulk of the variations in neonatal
nal delivery [aRR = 1.7, 95% CI: 1.2–2.5, p < 0.01, Add- deaths (neonatal; cvAUROC = 73%; Table 2). Clearly, the
itional file 1: Appendix 1A]. Mothers who gave birth household factors and the geospatial covariates (environ-
after age 30 had a higher risk of experiencing neo- mental factors) alone did not explain much of child
natal mortality compared to those who are less than mortality variations. It presupposes that more emphasis
18 years [aRR = 1.9, 95% CI: 1.1–3.2, p < 0.01, Add- should be placed on addressing maternal and child fac-
itional file 1: Appendix A1]. Mothers who had five or tors. However, models that integrate household, mater-
more children were at a higher risk of neonatal mor- nal, child, and environmental factors had a minimum
tality compared to mothers with only one child bias in all cases (smallest BIC and AIC). We applied sub-
[aRR = 1.6, 95% CI: 1.1–2.2, p < 0.01, Additional file 1: ject matter knowledge in choosing the final model,
Appendix A1]. Children born into families where the coupled with the AIC, BIC, and cvAUROC estimates.
biological father has more than one wife were also at Model 9 was selected as the best model to predict the
a greater risk of neonatal deaths [aRR = 1.4, 95% CI: risk of neonatal mortality in Ghana.
1.1–1.8, p < 0.01, Additional file 1: Appendix A1]. The
use of contraceptives among women reduced the risk Multivariable regression analysis of factors associated
of neonatal mortality by 25 percentage points com- with neonatal mortality: evidence from the pooled Ghana
pared to women who do not use any form of contra- Demographic and health surveys (2003–2014)
ceptive [aRR = 0.8, 95% CI: 0.6–1.0, p < 0.01, The findings from the pooled data set from 2003 to
Additional file 1: Appendix A1]. The stature of a 2014 GDHS, including the environmental factors, re-
woman could either be genetic or reflective of long vealed that household size, multiple births, perceived
term exposure to poor diet. Tall women (within the birth weight, parity, and use of contraceptives influenced
75th percentile of height) were at a lower risk of hav- neonatal deaths. An increase in the household size re-
ing neonatal deaths compared to short women (within duced neonatal deaths by 18% [aRR = 0.82, 95% CI:
the 25th percentile of height distribution) [aRR = 0.7, 0.73–0.92, p < 0.05, Table 3]. Among mothers who had
95% CI: 0.5–0.9, p < 0.01, Additional file 1: Appendix multiple births, the risk of having neonatal deaths is ap-
A1]. proximately four times as high as the risk of neonatal
Dwomoh BMC Public Health          (2021) 21:492 Page 7 of 18
Table 2 Evaluating the predictive performance of the nine models for neonatal mortality: 2003–2014 Ghana Demographic and
Health Surveys
Models variables Neonatal Mortality
Adjusted for all models Persons year at risk, and year of survey AIC BIC cvAUROC
Model 1 = Household factors Age of the household head, sex of the household 8,390,012 8,390,131 58.0%
head, place of residence, household size, zone location,
access to an improved water source, access to improved
water facility, household ownership of bed-net
Model 2 = Child characteristics Sex of the child, multiple births, birth order, birth spacing, 7,709,512 7,709,616 66.0%
size of the baby at birth
Model 3 = Maternal factors Mothers age at birth, marital status, literacy, parity, type 5,761,778 5,761,889 61.0%
of delivery, place of delivery, early breastfeeding, contraceptive
use, history of terminating a pregnancy
Model 4 = Model 1 + Model 3 Household factors + Maternal factors 5,586,465 5,586,672 63.0%
Model 5== Model 1 + Model 2 Household factors + Child characteristics 7,528,049 7,528,250 67.0%
Model 6 = Model 2 + Model 3 Child +Maternal factors 5,335,693 5,335,886 73.0%
Model 7 = Model 1 + Model 2+ Model 3 Household factors + Maternal factors+ Child factors 5,156,220 5,156,509 73.0%
Model 8 = Geospatial covariates The length of the growing season in months, environmental 7,663,200 7,663,348 54%
temperature, population density, the prevalence of malaria,
proximity to a protected area, net vegetation index, livestock goat.
Model 9 = Model 7+ Model 8 Household factors + Maternal factors+ Child factors+ Geospatial covariates 4,589,203 4,589,614 74%
Abbreviation: AIC Akaike Information Criterion, BIC Bayes Information Criterion, cvAUROC Cross Validated Area Under the Receiver Operating
Characteristics Curve
deaths among mothers who had only single birth [aRR = babies [aRR = 4.1, 95% CI: 1.9–8.9, p < 0.05, Table 4].
3.42, 95% CI: 1.63–7.17, p < 0.05, Table 3]. Neonates The risk of neonatal deaths among children born in
who were perceived by their mothers to be small were at rural areas is approximately two times the risk of neo-
a higher risk of neonatal death compared to very large natal deaths among children born in urban areas [aRR =
[aRR = 2.08, 95% CI: 1.19–3.63, p < 0.05, Table 3]. A unit 1.8, 95% CI: 1.1–2.9, p < 0.05, Table 4]. The following
increase in the number of children born to a woman of variables (ANC attendance, Tetanus injection) were also
reproductive age is associated with a 49% increased risk associated with neonatal deaths. The results must be
in neonatal deaths [aRR = 1.49, 95% CI: 1.30–1.69, p < interpreted with caution because the data originate from
0.05, Table 3]. The use of contraceptive reduced the risk the most recent birth during the survey. That is, we did
of neonatal deaths by 46% [aRR = 0.56, 95% CI: 0.35– not adjust for other confounding factors due to the sam-
0.90, p < 0.05, Table 3]. On average, a one-unit increase ple size. Among mothers who did not attend ANC, the
in the log-population density is equivalent to a 24% re- risk of neonatal death was approximately five times as
duction in neonatal deaths [aRR = 0.76, 95%CI: 0.60– high as the risk of neonatal deaths for mothers who
0.96; p < 0.05, Table 3]. attended ANC [aRR = 4.7, 95% CI: 1.8–12.1, p < 0.05,
Table 4]. The mothers who did not receive any tetanus
Risk factors of neonatal deaths: Most recent evidence injection during pregnancy were also at high risk (ap-
from the 2017 Ghana maternal health survey proximately two times) of having neonatal deaths [aRR =
This study utilized data from 15,348 children born in 2.1, 95% CI: 1.1–4.0, p < 0.05, Table 4]. It is important
the five years preceding the Maternal Health Survey con- to emphasize National Health Insurance (NHI) did not
ducted in 2017. provide protection against neonatal death, as shown in
2017 GMHS data. The results showed that mothers who
Risk factors of neonatal mortality in 2017 had no valid NHI card were at a lower risk of neonatal
A unit increase in the number of children born to a deaths compared with mothers with valid NHI card
woman is associated with an approximately 26% increase [aRR = 0.6, 95% CI: 0.4–0.9, p < 0.05, Table 4]. This find-
in neonatal deaths [aRR = 1.26, 95% CI: 1.03–1.54, p < ing is quite surprising, but what could possibly explain
0.05, Table 4]. However, larger household size was asso- the observed phenomenon could be that women who
ciated with a reduced risk of neonatal deaths by 12% are fully aware of their high-risk status before delivery
[aRR = 0.88, 95% CI: 0.80–0.89, p < 0.05, Table 4]. The are more likely to secure a valid NHI card during the
mothers who gave birth to two or more babies had ap- antenatal period to mitigate the cost of seeking health-
proximately four times the risk of experiencing neonatal care in case of emergency compared to women who are
deaths compared to mothers who gave birth to single perceived to be in good health. The results should be
Dwomoh BMC Public Health          (2021) 21:492 Page 8 of 18
Table 3 Multivariable regression analysis of factors associated with neonatal mortality: Evidence from the pooled Ghana Demographic and Health Surveys (2003–2014)
Variables MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6 MODEL 7 MODEL 8 MODEL 9
aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI]
HOUSEHOLD FACTORS
Age of the household head 1.01 [1.01– 0.99 [0.97–1.00] 1.02 [1.01–1.02]** 1.00 [0.99–1.02] 1.00 [0.99–1.02]
1.02]**
Sex of household head
Male 1 1 1 1 1
Female 0.98 [0.74–1.29] 0.62 [0.40–0.97]* 0.90 [0.67–1.22] 0.64 [0.40–1.03] 0.76 [0.49–1.18]
Place of residence
Urban 1 1 1 1 1
Rural 1.44 [1.00–2.08] 1.24 [0.77–2.00] 1.35 [0.92–1.99] 1.25 [0.74–2.09] 1.39 [0.81–2.39]
Zone
Coastal 1 1 1 1 1
Southern 1.19 [0.89–1.6] 0.93 [0.67–1.29] 1.09 [0.81–1.48] 0.87 [0.61–1.23] 0.76 [0.44–1.32]
Northern 1.25 [0.85–1.84] 1.15 [0.73–1.83] 1.28 [0.85–1.94] 1.20 [0.71–2.02] 1.35 [0.49–3.72]
Household size 0.90[0.85– 0.85 [0.76–0.94]** 0.83 [0.76– 0.80 [0.71–0.9]*** 0.82 [0.73–0.92]**
0.97]** 0.92]***
Wealth index
Poorest 1 1 1 1 1
Poorer 0.83 [0.57–1.21] 0.77 [0.51–1.15] 0.86 [0.59–1.27] 0.80 [0.52–1.22] 0.92 [0.60–1.40]
Middle 1.17 [0.74–1.84] 0.97 [0.56–1.67] 1.09 [0.67–1.78] 0.97 [0.54–1.75] 1.1 [0.60–2.01]
Richer 0.99 [0.56–1.75] 0.71 [0.36–1.42] 0.93 [0.51–1.71] 0.72 [0.34–1.51] 0.81 [0.37–1.80]
Richest 1.15 [0.63–2.12] 0.96 [0.43–2.13] 0.84 [0.43–1.65] 0.77 [0.33–1.82] 0.98 [0.38–2.54]
Household access to an improved water source
Improved 1 1 1 1 1
Unimproved 1.07 [0.77–1.47] 0.98 [0.67–1.44] 0.95 [0.67–1.36] 0.87 [0.57–1.34] 0.88 [0.57–1.34]
Household access to an improved sanitation facility
Improved 1 1 1 1 1
Unimproved 0.78 [0.56–1.08] 0.85 [0.58–1.24] 0.73 [0.52–1.05] 0.73 [0.47–1.13] 0.65 [0.41–1.02]
Have bed net
No net 1 1 1 1 1
Have net 0.98 [0.72–1.33] 0.89 [0.62–1.27] 1.05 [0.75–1.47] 0.88 [0.60–1.29] 0.95 [0.63–1.43]
CHILD FACTORS
The sex of the child
Male 1 1 1 1 1
Female 0.82 [0.64–1.04] 0.83 [0.65–1.06] 0.75 [0.56–1]* 0.74 [0.55–1]* 0.72 [0.53–1.00]
Multiple births
Dwomoh BMC Public Health          (2021) 21:492 Page 9 of 18
Table 3 Multivariable regression analysis of factors associated with neonatal mortality: Evidence from the pooled Ghana Demographic and Health Surveys (2003–2014)
(Continued)
Variables MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6 MODEL 7 MODEL 8 MODEL 9
aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI]
Single birth 1 1 1 1 1
Multiple births 4.10 [2.59– 4.99 [3.01– 3.29 [1.73– 3.88 [1.96– 3.42 [1.63–7.17]**
6.49]*** 8.28]*** 6.26]*** 7.68]***
Birth order
1st child 1 1 1 1 1
2nd child 0.39 [0.08–2.00] 0.36 [0.07–1.88] 0.43 [0.08–2.38] 0.4 [0.07–2.31] 0.5 [0.09–2.91]
3rd child 0.47 [0.09–2.48] 0.46 [0.08–2.54] 0.39 [0.07–2.26] 0.38 [0.06–2.38] 0.46 [0.07–2.94]
4th + child 0.55 [0.1–2.94] 0.62 [0.11–3.41] 0.14 [0.02–0.84]* 0.14 [0.02–0.91]* 0.17 [0.03–1.09]
Birth spacing
First child 1 1 1 1 1
< 24 months 3.25 [0.62–16.97] 3.95 [0.74–21.15] 2.16 [0.35–13.25] 2.3 [0.36–14.59] 2.02 [0.32–12.75]
24–35 months 1.66 [0.32–8.70] 2.00 [0.37–10.85] 1.14 [0.19–6.89] 1.23 [0.19–7.75] 0.84 [0.14–5.02]
36+ months 1.28 [0.25–6.49] 1.37 [0.26–7.21] 0.66 [0.12–3.74] 0.62 [0.10–3.68] 0.48 [0.08–2.87]
Perceived birth weight
Very large 1 1 1 1 1
Large 0.75 [0.51–1.11] 0.78 [0.53–1.16] 0.8 [0.48–1.34] 0.8 [0.47–1.34] 0.88 [0.5–1.53]
Average 0.79 [0.53–1.18] 0.77 [0.51–1.16] 1.02 [0.6–1.73] 0.94 [0.54–1.63] 1.05 [0.58–1.90]
Small 1.53 [1.03–2.26]* 1.48 [0.99–2.2] 1.85 [1.09–3.16]* 1.74 [1.02–3.00]* 2.08 [1.19–3.63]*
MATERNAL FACTORS
Mothers age at first-child
< 18 1 1 1 1 1
18–29 1.03 [0.74–1.43] 1.11 [0.79–1.57] 0.93 [0.65–1.34] 0.98 [0.67–1.44] 1.15 [0.75–1.76]
30+ 1.31 [0.55–3.11] 1.57 [0.65–3.82] 1.05 [0.42–2.62] 1.14 [0.44–2.92] 1.03 [0.36–2.91]
The current marital status of the mother
Never married 1 1 1 1 1
Married 0.95 [0.59–1.54] 0.72 [0.42–1.26] 1.06 [0.64–1.77] 0.87 [0.48–1.55] 1.23 [0.67–2.25]
Not married but living together 0.87 [0.50–1.52] 0.68 [0.36–1.28] 0.89 [0.50–1.61] 0.71 [0.36–1.40] 0.89 [0.46–1.74]
Literacy
Cannot read 1 1 1 1 1
Able to read part of sentence 0.67 [0.36–1.22] 0.62 [0.32–1.18] 0.73 [0.39–1.39] 0.7 [0.36–1.37] 0.66 [0.31–1.37]
Able to read whole sentence 1.06 [0.72–1.57] 1.13 [0.71–1.81] 0.89 [0.59–1.36] 0.97 [0.59–1.60] 1.04 [0.601–1.77]
Total children ever born 1.18 [1.09– 1.29 [1.17– 1.39 [1.23– 1.51 [1.32– 1.49 [1.30–
1.26]*** 1.42]*** 1.57]*** 1.72]*** 1.69]***
Type of delivery
Dwomoh BMC Public Health          (2021) 21:492 Page 10 of 18
Table 3 Multivariable regression analysis of factors associated with neonatal mortality: Evidence from the pooled Ghana Demographic and Health Surveys (2003–2014)
(Continued)
Variables MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6 MODEL 7 MODEL 8 MODEL 9
aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI]
Normal 1 1 1 1 1
Caesarean 1.32 [0.76–2.32] 1.41 [0.80–2.50] 1.15 [0.61–2.18] 1.21 [0.63–2.31] 1.09 [0.51–2.32]
Place of delivery
Facility 1 1 1 1 1
Home 0.64 [0.45–0.91]* 0.63 [0.43–0.90]* 0.70 [0.47–1.04] 0.70 [0.46–1.07] 0.70 [0.45–1.09]
Early breastfeeding
Within one hour 1 1 1 1 1
After one hour or more 1.16 [0.87–1.53] 1.13 [0.84–1.53] 1.01 [0.74–1.38] 0.99 [0.71–1.39] 1.06 [0.74–1.5]
Currently using contraceptive
No 1 1 1 1 1
Yes 0.56 [0.39–0.83]** 0.55 [0.38–0.82]** 0.54 [0.35–0.84]** 0.54 [0.34–0.85]** 0.56 [0.35–0.90]*
Ever terminated pregnancy
Never 1 1 1 1 1
Yes 1.18 [0.82–1.70] 1.13 [0.79–1.62] 1.32 [0.90–1.93] 1.27 [0.87–1.86] 1.24 [0.81–1.90]
GEOSPATIAL FACTORS
The length of the growing season
< =8months 1 1
9-10 months 1.08 [0.64–1.8] 0.91 [0.49–1.72]
11–12 months 1.78 [0.85–3.71] 1.38 [0.46–4.17]
Environmental temperature via restricted cubic spline
Livestock_goats 1.00 [1.00–1.01] 1.00 [1.00–1.01]
Malaria prevalence 0.58 [0.1–3.25] 0.25 [0.03–2.17]
Log of the population density 0.92 [0.79–1.09] 0.76 [0.60–0.96]*
Rainfall via restricted cubic 1.00 [1.00–1.01] 1.00 [0.99–1.00]
splines
Vegetation index
Less than 25th percentile 1 1
25th–50th percentile 1.61 [0.97–2.65] 1.19 [0.67–2.12]
50th–75th percentile 1.20 [0.61–2.38] 0.68 [0.28–1.63]
> 75th percentile 1.19 [0.53–2.64] 0.52 [0.19–1.44]
Proximity to a protected area
Less than 25th percentile 1 1
25th–50th percentile 1.37 [0.95–1.97] 1.83 [1.15–2.92]*
50th–75th percentile 0.89 [0.62–1.27] 1.24 [0.76–2.02]
Dwomoh BMC Public Health          (2021) 21:492 Page 11 of 18
Table 3 Multivariable regression analysis of factors associated with neonatal mortality: Evidence from the pooled Ghana Demographic and Health Surveys (2003–2014)
(Continued)
Variables MODEL 1 MODEL 2 MODEL 3 MODEL 4 MODEL 5 MODEL 6 MODEL 7 MODEL 8 MODEL 9
aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI] aRR [95% CI]
> 75th percentile 0.89 [0.59–1.36] 1.08 [0.59–1.97]
Survey year
2003 1 1 1 1 1 1 1 1 1
2008 0.76 [0.55–1.06] 0.73 [0.53–1.01] 0.84 [0.6–1.18] 0.9 [0.61–1.34] 0.75 [0.52–1.07] 0.83 [0.58–1.2] 0.92 [0.59–1.42] 0.66 [0.49–0.9]** 0.87 [0.55–1.38]
2014 0.71 [0.53–0.96]* 0.65 [0.48–0.86]** 0.70 [0.5–0.97]* 0.76 [0.52–1.12] 0.68 [0.49–0.94]* 0.71 [0.5–1] 0.80 [0.53–1.22] 0.64 [0.49– 0.81 [0.54–1.23]
0.84]**
Abbreviations: aRR Adjusted Relative Risk. P-value notation: ***p < 0.001, **p < 0.01, *p < 0.05
Dwomoh BMC Public Health          (2021) 21:492 Page 12 of 18
Table 4 Risk factors of neonatal mortality in Ghana: Evidence from the 2017 Ghana maternal health survey
Variables Neonatal mortality
Poisson Negative Binomial
aRR[95% CI] aRR[95% CI]
Sex of the child
Male Ref
Female 0.84[0.54–1.33] 0.60[0.34–1.05]
Mothers education
None Ref
Primary/middle 1.03[0.53–2.02] 1.21 [0.51–2.85]
JHS/JSS 0.97[0.53–1.77] 1.74 [0.74–4.11]
SHS+ 2.43[1.09–5.40] * 7.25 [2.36–22.23]**
Current age 0.95[0.89–1.01] 1.01 [0.94–1.08]
Marital status
Yes, currently married Ref
Yes, living with a man 0.88[0.49–1.60] 1.00 [0.52–1.95]
No, not in union 1.06[0.48–2.32] 1.19 [0.45–3.09]
Parity 1.26[1.03–1.54] * 1.27 [1.01–1.59]*
Multiple birth
single Ref
Multiple 4.12 [1.91–8.85] *** 7.38 [3.15–17.3]***
Age at first sex 0.99[0.90–1.08] 0.98 [0.88–1.09]
Place of delivery
Hospital Ref
Home 1.24[0.67–2.29] 1.17 [0.51–2.67]
Household size 0.88[0.76–1.03] 0.88 [0.80–0.98]*
Water
Improved Ref
Not improved 0.97[0.54–1.73] 1.44 [0.63–3.29]
Sanitation
Improved Ref
Not improved 0.97[0.54–1.73] 0.51 [0.24–1.10]
Owns agriculture land
Yes Ref
interpreted with caution since we did not adjust for con- mortality among neonates using the findings from the
founding factors because the data were only available for multivariable regression models. The moderate contrib-
the most recent birth reducing the sample size uting factors are those that are biologically plausible, but
drastically. for lack of data over the period, they were not included
in the multivariable regression models to quantify the ef-
Synthesizing all the available evidence on the factors fect size. These factors are relevant, but the power of the
influencing neonatal mortality in Ghana-1998-2017 study will be affected if included in the multivariable
This section combines all the available evidence from model because of the sample size. For instance, antenatal
1998 to 2017 to reclassify neonatal mortality predictors and postnatal care attendance are good indicators of
into major contributing factors, moderate contributing neonatal deaths, but they were only measured for the
factors, and minor contributing factors. Major contribut- most recent birth reducing the effective sample size
ing factors are factors that are biologically plausible and drastically. The minor contributing factors are difficult
have been consistent over the years in increasing to explain why and how they contribute to neonatal
Dwomoh BMC Public Health          (2021) 21:492 Page 13 of 18
Table 4 Risk factors of neonatal mortality in Ghana: Evidence from the 2017 Ghana maternal health survey (Continued)
Variables Neonatal mortality
Poisson Negative Binomial
aRR[95% CI] aRR[95% CI]
No 0.83[0.49–1.41] 0.7 [0.35–1.38]
Owns Livestock
Yes Ref
No 1.20[0.69–2.06] 1.53 [0.78–3.00]
Ever had an abortion
Yes Ref
No 0.77[0.44–1.34] 1.14 [0.55–2.36]
Ever had miscarriage
Yes Ref
No 1.35[0.75–2.45] 2.16 [0.98–4.76]
Ever had stillbirth
Yes Ref
No 0.79[0.31–2.02] 0.58 [0.17–1.95]
Zone
Coastal Ref
Middle 1.08[0.68–1.72] 1.06 [0.55–2.06]
Northern 0.81[0.45–1.45] 1.58 [0.67–3.74]
Place of residence
Urban Ref
Rural 1.77 [1.07–2.90]* 3.77 [1.83–7.76]***
Antenatal care¥
Yes Ref
No 4.65[1.80–12.06] ** 5.16 [2.06–12.91]***
Size at birth¥
Very large Ref
Large 0.61[0.28–1.36] 0.57 [0.24–1.36]
Average 0.86[0.41–1.83] 0.77 [0.34–1.73]
Small 0.67[0.25–1.81] 0.59 [0.20–1.75]
Very small 1.23[0.46–3.26] 1.15 [0.41–3.24]
Tetanus during pregnancy¥
Yes Ref
No 2.10[1.11–3.97] * 2.06 [1.07–3.99]*
Iron
Yes Ref
No 2.02[0.89–4.62] 2.10 [0.88–5.01]
Fansidar (sulfadoxine and pyrimethamine) during pregnancy¥
Yes ref
No 2.10[0.95–4.64] 2.22 [0.98–5.06]
Valid NHI card¥
Yes Ref
No 0.58[0.37–0.91] * 0.64 [0.39–1.05]
P-value notation: ***p < 0.001, **p < 0.01, *p < 0.05, ¥ Unadjusted estimates from Poisson and Negative Binomial models since the variable was only measured for
most recent birth. Abbreviation: NHI National Health Insurance
Dwomoh BMC Public Health          (2021) 21:492 Page 14 of 18
deaths, and further studies are needed to understand the deaths in Ghana. Specifically, the discussion focused on
biological underpinnings. Table 5 summarizes these con- multiple births, household size, parity, contraceptive use,
tributing factors. birth weight, and population density.
Discussion Multiple births
This study investigated the impact of several risk factors Multiple births are generally classified as high-risk birth
of neonatal mortality in Ghana between 1998 and 2017. in the medical literature because they are associated with
It was observed that maternal and child characteristics fetal and neonatal complications that require special and
explain a larger proportion of neonatal deaths variations expensive medical care [18]. The biological complica-
compared to household and selected environmental tions associated with multiple births usually lead to a
(geospatial) factors. The following were identified to be greater risk of congenital disabilities and accounted for a
major contributing factors to neonatal mortality: mul- larger percentage of prenatal deaths [19]. Besides, twins,
tiple births, birth spacing, birth weight perceived to be triplets, etc., suffer lower birth weight, have intrauterine
small, smaller household size, an increasing number of growth restriction and congenital abnormalities, and
children born to a woman of reproductive age, not using preterm delivery, which are key determinants of neonatal
a contraceptive, living in the northern part of the coun- deaths in the early years of life. This could be explained
try, mothers age at childbirth, mothers who delivered via by the fact that there could be some underlying bio-
cesarean section, mothers who gave birth at the age of logical complications that extend beyond the neonatal
30 years or more, a household with no bed net, mothers and infant years, but this has not been studied exten-
with short stature, mothers who did not receive Fansidar sively in the literature. Our result showed that the risk
(sulfadoxine and pyrimethamine) during antenatal, and of neonatal deaths was higher among multiple births
place of residence. The next section discusses factors compared to singletons, which is consistent with a previ-
that were considered major determinants of neonatal ous study in sub-Saharan Africa that indicated that one-
Table 5 Synthesizing all the available evidence on the factors influencing neonatal mortality in Ghana-1998-2017
Contributing factors The extent of the High neonatal
contributing factors mortality
Multiple births (twin, triplet, etc.) Major Yes
Smaller household size Major Yes
Perceived smaller birth weight Major Yes
Women who are not using contraceptive Major Yes
A higher number of biological children (high parity) Major Yes
Children born in the northern sector Major† Yes†
Population density Major Yes
Birth spacing less than 24month Moderate Yes
Mothers age at birth (older mothers 30+) Moderate Yes
Mothers who had no tetanus injection during pregnancy Moderate Yes
Mothers who delivered via cesarean section Moderate Yes
Children who did not receive vitamin A two months after delivery Moderate No
Mothers who did not attend antenatal care Moderate Yes
No access to an improved water source Moderate No
Children who live in rural areas Moderate No
Not receiving Fansidar (sulfadoxine and pyrimethamine) against malaria Moderate No
Household ownership of bednet Moderate No
Children who are Muslims were at a higher risk of death between 2003 and 2014 Minor No
Violence against women Minor No
Mothers whose husband or partners have two or more wives Minor Yes
Shorter mothers Minor Yes
Mothers with SHS or higher education Minor Yes
†: Contributing factor before 2017 but the impact disappeared in the 2017 Ghana Maternal Health Survey Data Analysis
Dwomoh BMC Public Health          (2021) 21:492 Page 15 of 18
fifth of twins in the region die before age five years, three with shorter birth intervals. Women who used contra-
times the mortality rate among singletons [20]. There ceptives were likely to reduce the number of births to
should be a significant improvement in health-service three or fewer children alleviating the negative impact of
delivery during pregnancy, at delivery, and postpartum high parity-associated with child mortality. To design
[21, 22]. It was evident that the impact of multiple births and implement effective interventions aim at improving
has not reduced over the period which is indicative of the prevalence of contraceptive use, we need to under-
the fact that not much has been done about the stand the drivers of low contraceptive use. There should
phenomenon from the perspective of key stakeholders be religious and culturally sensitive intensive education
(mothers, Ghana Health Service, Ministry of Health, on contraceptive use among women, and these contra-
Midwives, Nurses, Doctors, and all other health practi- ceptives should be readily available and affordable to the
tioners). This study could not determine the proportion general population.
of multiple births attributable to assisted reproductive
technology and whether the increase in the number of Birth weight
multiple births resulted from multiple pregnancies Children that were perceived to be small were at a
caused by the practice of transferring more than one greater risk of child mortality, and the finding is consist-
embryo into the uterus during Vitro fertilization (IVF). ent with what was reported previously [26, 27]. Low
Looking at the trend of the effect of multiple births, this birth weight is associated with high preterm birth and
study recommends routine training of midwives, doc- congenital abnormalities leading to high neonatal and
tors, and nurses in general on how to manage multiple infant deaths. The consequences of low birth weight do
births during antenatal, delivery, and postnatal. not exclude long term effects as Watkins et al, 2018 [28]
showed that low birth weight is associated with in-
Household size creased death rates not only in infancy but also through
This study found that larger household size is associated to adolescence. Improving the mother’s nutritional sta-
with a reduced risk of child mortality. Admittedly larger tus and avoiding high-risk behavior such as smoking and
household size could be associated with people living alcohol intake during pregnancy is critical to improving
below the poverty line and competing for the same so- birth weight.
cioeconomic and health resources, increasing child mor-
tality risk. However, the observed protective effect of Cesarean section
larger household size could be explained by the Ghan- The study showed that increased cesarean section in-
aian culture of providing food security, financial and creases the risk of neonatal deaths in Ghana, which is
medical support to the needy in larger households. More consistent with the studies mentioned above. This could
research needs to be done to better understand house- be possible because obstetricians are more likely to ad-
hold size dynamics and its relative effect on neonatal vise women with birth complications to opt for a
mortality in Ghana. cesarean section, but on the contrary, the majority of
early-term elective cesarean sections can be postponed
Parity [29]. Elective cesarean deliveries are on the rise and be-
This study found that high parity births are associated come more popular among expectant mothers and
with a higher neonatal mortality than low parity births, health professionals because cesarean delivery globally
consistent with the findings from the literature [23, 24]. represents a potentially life-saving procedure for both
High parity correlates with lower coverage of maternal mother and neonate in labor complications and health
and child health services such as ANC and PNC attend- conditions of expectant mothers that require early or
ance and skilled birth delivery, vitamin A supplementa- immediate delivery. However, in the absence of serious
tion, vaccination and immunization, and iron medical complications and the highly negligible likeli-
supplementation that might increase mortality [25]. hood of having stillbirth, neonatal and maternal deaths,
There is a positive correlation between birth spacing and cesarean delivery can pose avoidable short-term and
high parity births. There should be intensive education near-future risks to both the mother and newborn, in-
on birth spacing and the risk associated with high parity cluding birth injury, seizures, jaundice and
births. Women should be encouraged to use contracep- hypoglycemia, sepsis, longer maternal recovery, neonatal
tive methods to increase birth spacing and control the respiratory problems, and potentially severe complica-
number of children born to women of reproductive age. tions in subsequent pregnancies [30–32]. Studies have
consistently shown an increased risk in prenatal deaths
Contraceptive use and infant mortality, though the causal pathway is not
The use of contraceptives increases birth spacing and always clear [33–36]. The increasing number of unwar-
unwanted pregnancies reducing complications associated ranted cesarean sections globally is a major concern to
Dwomoh BMC Public Health          (2021) 21:492 Page 16 of 18
both healthcare providers and policymakers as stipulated services such as ANC, PNC, tetanus injection, mode of
in agenda “Healthy People 2020” with the goal of redu- delivery, etc., is all subject to recall bias. Several key indi-
cing the cesarean delivery rate by 10 % among low-risk cators, such as PNC attendance, could not be studied
women giving birth for the first time and among low- because they were only available for the women’s most
risk women with a prior cesarean section [37]. There recent birth. We omitted the 1998 GDHS data set from
should be a conscious effort to reduce the cesarean sec- the final model because most of the interest indicators
tion rate among mothers who are at low risk of birth were not measured during the survey. For instance, no
complications. variable measured household wealth in the data set, and
therefore, we couldn’t have assessed the impact of socio-
Impact of environmental factors on neonatal mortality economic status on neonatal deaths if we had included
The population density was the only geospatial variable the 1998 GDHS in the multiple variable analysis.
associated with neonatal mortality. Higher population
density is usually associated with low health service Policy implications
utilization and quality healthcare. However, this study Through the Ministry of Health and the Ghana Health
found a protective effect of population density on neo- Service, the government of Ghana should organize rou-
natal deaths. This could be explained by the fact that tine practical simulation training for obstetricians, gen-
higher population density in Ghana may be associated eral medical practitioners, nurses, and midwives on how
with higher maternal health coverage [38], which could to handle complications associated with multiple births.
reduce the incidence of neonatal deaths. Mothers who There should be a clearly defined direct policy on mul-
live in regions and areas with highly dispersed popula- tiple births that incorporates routine training of health
tions and low socioeconomic status could face a higher professionals. There should be a policy for all health fa-
burden of assessing quality health healthcare. cilities to design a database that captures all multiple
births in the health facility or the community, including
Strength and limitations multiple births from In vitro fertilization (IVF) interven-
The co-integration of individual-level data and geospatial tion. This will facilitate the process of tracking these
covariates in cross-sectional ecological studies can help births during antenatal, delivery, postnatal through to
researchers and policymakers understand the causal age 5. This activity could be spearheaded by the commu-
pathway of the risk of neonatal deaths in Ghana [39]. nity health nurses and monitored by a specialist obstetri-
The study’s power to detect significant effect of covari- cian. Efforts should be made to monitor both twins,
ates improved by using four GDHS data and one GMHS triplet etc. during labor. There should be early preg-
data that were collected within the MDG and post MDG nancy ultrasound scans, and monochorionic twins
period. This study applied more rigorous geospatial and should be referred for specialist obstetrician care. Except
non-spatial statistical models to nationally representative for a few private health facilities, most pregnancies, in-
survey data to determine the impact of a wide-ranging cluding multiple births, are generally handled by mid-
set of possible correlates of neonatal deaths. wives at the health facility, but we recommend, in
Nonetheless, there are several limitations to the study addition to the midwives, there should be specialist ob-
that are worth noting. First, the DHS data used for the stetrician care for all multiple births during antenatal,
analysis is a cross-sectional survey, and we cannot infer delivery, and postnatal care. The Ministry of Health,
that the identified risk factors caused the death of the Ghana Health Service, Ghana Medical Association, and
children. The study only showed that these risk factors all fertility centers across the country may consider a
are just associated with neonatal mortality. The true policy of elective single embryo transfer (eSET) to re-
cause of death could only be determined by conducting duce the chances of a woman having multiple births and
an extensive death autopsy for all children under-five. to reduce to the barest minimum the complications as-
However, the survey data used do not contain these chil- sociated with multiple births among women and their
dren’s autopsy report, making it difficult to assess the respective children who are delivered via Assisted Repro-
cause, mode, and manner of death. ductive Technology (ART). To encourage eSET, blasto-
Some important predictors of neonatal deaths, such as cyst culture, preimplantation genetic screening, and
vaccination, fever, diarrhea, nutritional and respiratory time-lapse imaging could help identify the embryo with
infections, were only collected for surviving children and the greatest implantation potential. The MOH and the
were therefore excluded from this study. GHS should use the pregnancy school and the media
Third, there is the possibility of recall bias as the DHS (radio, television, and social media), together with com-
collects information from respondents about past events, munity health nurses and other healthcare professionals,
behaviors, and health outcomes. For instance, informa- to provide intensive education about the risk of multiple
tion concerning women’s receipt of maternal care births, high parity, birth spacing, antenatal care
Dwomoh BMC Public Health          (2021) 21:492 Page 17 of 18
attendance, facility delivery, postnatal care, Consent for publication
immunization, tetanus injection, contraceptive use by Not applicable.
mothers, efficient ways to improve the weight of the un-
born baby and nutritional requirements during Competing interests
pregnancy. The author declares that he has no competing interests.
Received: 26 May 2020 Accepted: 21 February 2021
Conclusion
This study identified several factors associated with neo-
natal deaths in Ghana. Some of the factors have been References
consistent over time, whiles the impact of other factors 1. Global Health Observatory data on Under-five mortality [https://www.who.
int/gho/child_health/mortality/mortality_under_five_text/en/].
fluctuates and diminishes between survey years. The en- 2. Grady SC, Frake AN, Zhang Q, Bene M, Jordan DR, Vertalka J, Dossantos TC,
vironmental factors did not have a significant impact on Kadhim A, Namanya J, Pierre L-MJGh: Neonatal mortality in East Africa and
child mortality. The household, maternal and child- West Africa: a geographic analysis of district-level demographic and health
survey data. 2017.
related factors such as household size, parity, multiple 3. AUC, UNECFA, ADB, UNDP: MDG Report 2014: Assessing Progress in Africa
births, birth weight, place of residence, antenatal care at- toward the Millennium Development Goals. In.; 2014.
tendance, access to improved water source among rural 4. Nations U: United Nations. 2015. The Millennium Development Goals
Report. United Nations. In.; 2015.
residence, immunization coverage, and tetanus injection 5. Golding N, Burstein R, Longbottom J, Browne AJ, Fullman N, Osgood-
were among the few factors that affected neonatal deaths Zimmerman A, Earl L, Bhatt S, Cameron E, DCJTL C, Hay SI. Mapping under-
in Ghana. This study recommends an integrated inter- 5 and neonatal mortality in Africa, 2000–15: a baseline analysis for the
Sustainable Development Goals. Lancet. 2017;390(10108):2171–82.
vention that simultaneously addressed these challenges. 6. Liu L, Johnson HL, Cousens S, Perin J, Scott S, Lawn JE, Rudan I, Campbell H,
Cibulskis R, Li MJTL, Child Health Epidemiology Reference Group of WHO
Abbreviations and UNICEF. Global, regional, and national causes of child mortality: an
AIC: Akaike Information Criterion; ANC: Antenatal Care; aRR: Adjusted Relative updated systematic analysis for 2010 with time trends since 2000. Lancet.
Risk; ART: Assisted Reproductive Technology; BIC: Bayesian Information 2012;379(9832):2151–61.
Criterion; CS: Cesarean Section; cvAUROC: Cross-validated Area Under the 7. Winter R, Pullum T, Langston A, Mivumbi NV, Rutayisire PC, Muhoza DNJCII:
Receiver Operating Characteristics Curve; EAs: Enumeration Areas; trends in neonatal mortality in Rwanda, 2000–2010. 2013.
ECD: Elective Cesarean Delivery; eSET: Elective Single Embryo Transfer; 8. Winter R, Pullum T, Langston A, Mivumbi NV, Rutayisire PC, Muhoza DN,
EVI: Enhanced Vegetation Index; GDHS: Ghana Demographic and Health Hakiba S. Trends in neonatal mortality in Rwanda, 2000-2010. DHS further
Surveys; GHS: Ghana Health Service; GIS: Geographic Information System; analysis reports No. 88. Calverton: ICF International.
GMHS: Ghana Maternal Health Survey; IPTp: Intermittent Preventive 9. Kayode GA, Grobbee DE, Amoakoh-Coleman M, Ansah E, Uthman OA,
Treatment; ITN: Insecticide Treated Net; MDG: Millennium Development Klipstein-Grobusch K. Variation in neonatal mortality and its relation to
Goals; MOH: Ministry of Health; NHI: National Health Insurance; country characteristics in sub-Saharan Africa: an ecological study. BMJ
PNC: Postnatal Care; SDG: Sustainable Development Goals; VIF: Variance global health. 2017:2(4):e000209.
Inflation Factor; WHO: World Health Organization 10. Pezzulo C, Bird T, Utazi EC, Sorichetta A, Tatem AJ, Yourkavitch J, Burgert-
Brucker CR: Geospatial modeling of child mortality across 27 countries in
Supplementary Information sub-Saharan Africa. 2016.11. GSS G, Demographic IGJR, MAryland, USA: Ghana Statistical Service , Ghana
The online version contains supplementary material available at https://doi. Health Service, International I: Health Survey 2014. 2015.
org/10.1186/s12889-021-10473-w. 12. Mosley WH, Chen LC. An analytical framework for the study of child survival
in developing countries. Popul Dev Rev. 1984:10:25–45.
Additional file 1. 13. Fuentes R, Pfütze T, Seck P: Does access to water and sanitation affect child
survival? A five country analysis. In.: Human Development Report Office
(HDRO), United Nations Development Programme …; 2006.
Acknowledgments 14. Christian PJJ. Maternal height and risk of child mortality and undernutrition.
I acknowledge all teaching assistants at the Department of Biostatistics, 2010;303(15):1539–40.
School of Public Health, University of Ghana who downloaded and cleaned 15. Takem E, D’Alessandro U: Malaria in pregnancy. Mediterr J Hematol Infect
the data set for the analysis. Dis 5: e2013010. In.; 2013.
16. Edmond KM, Zandoh C, Quigley MA, Amenga-Etego S, Owusu-Agyei S,
Author’s contributions Kirkwood BR. Delayed breastfeeding initiation increases risk of neonatal
DD conceived the idea, conducted a literature review, wrote statistical mortality. Pediatrics. 2006:117(3):e380–6.
methods and analysis, and writing of the full manuscript. The author read 17. Tottrup C, Tersbol BP, Lindeboom W, Meyrowitsch D. Putting child mortality
and approved the final manuscript. on a map: towards an understanding of inequity in health. Tropical Med Int
Health. 2009:14(6):653–62.
Funding 18. Ananth CV, Joseph KS, Demissie K, Vintzileos AM. Trends in twin preterm
None. This research did not receive any specific grant from funding agencies birth subtypes in the United States, 1989 through 2000: impact on perinatal
in the public, commercial, or not-for-profit sectors. mortality. Am J Obstet Gynecol. 2005:193(3):1076–e1.
19. Joseph KS, Marcoux S, Ohlsson A, Kramer MS, Allen AC, Liu S, Wen SW,
Availability of data and materials Demissie K, Sauve R, Liston RJP, Fetal Infant Health Study Group of the
The data used in the study are freely available upon request from the DHS Canadian Perinatal Surveillance System. Preterm birth, stillbirth and infant
website (https://dhsprogram.com/). mortality among triplet births in Canada, 1985–96. Paediatr Perinat
Epidemiol. 2002:16(2):141–8.
Declarations 20. Monden CW, Smits J. Mortality among twins and singletons in sub-Saharan
Africa between 1995 and 2014: a pooled analysis of data from 90
Ethics approval and consent to participate Demographic and Health Surveys in 30 countries. Lancet Glob Health. 2017:
The study used secondary data sets that are publicly available. 5(7):e673–9.
Dwomoh BMC Public Health          (2021) 21:492 Page 18 of 18
21. Bjerregaard-Andersen M, Lund N, Jepsen FS, Camala L, Gomes MA,
Christensen K, Christiansen L, Jensen DM, Aaby P, HJBp B-N, Sodemann M.
A prospective study of twinning and perinatal mortality in urban Guinea-
Bissau. BMC Pregnancy Childbirth. 2012:12(1):1–12.
22. Akaba GO, Agida TE, Onafowokan O, Offiong RA, Adewole ND. Review of
twin pregnancies in a tertiary hospital in Abuja, Nigeria. J Health Popul Nutr.
2013;31(2):272.
23. Shah PS, Knowledge Synthesis Group on Determinants of LBW/PT births.
Parity and low birth weight and preterm birth: a systematic review and
meta-analyses. Acta Obstet Gynecol Scand. 2010;89(7):862–75.
24. Gibbs CM, Wendt A, Peters S, Hogue CJ. The impact of early age at first
childbirth on maternal and infant health. Paediatr Perinat Epidemiol. 2012;
26:259–84.
25. Sonneveldt E, Plosky WD, Stover J. Linking high parity and maternal and
child mortality: what is the impact of lower health services coverage among
higher order births?. BMC Public Health. 2013;13(3):1–8.
26. Tshehla RM, Coetzee M, Becker PJ. Mortality and morbidity of very low-
birthweight and extremely low-birthweight infants in a tertiary hospital in
Tshwane. S Afr J Child Health. 2019;13(2):89–97.
27. Grisaru-Granovsky S, Boyko V, Lerner-Geva L, Hammerman C, Rottenstreich
M, Samueloff A, Schimmel MS, Reichman B, Maternal-Fetal INNJTJo,
Medicine N, Israel Neonatal Network. The mortality of very low birth weight
infants: the benefit and relative impact of changes in population and
therapeutic variables. J Matern Fetal Neonatal Med. 2019;32(15):2443–51.
28. Watkins WJ, Kotecha SJ, Kotecha S. All-cause mortality of low birthweight
infants in infancy, childhood, and adolescence: population study of England
and Wales. PLoS Med. 2016;13(5):e1002018.
29. Vigdis Rikhardsdottir J, Hardardottir H, Thorkelsson T. The majority of early-
term elective cesarean sections can be postponed. J Matern Fetal Neonatal
Med. 2019:1–6.
30. Zanardo V, Simbi AK, Franzoi M, Solda G, Salvadori A, Trevisanuto D.
Neonatal respiratory morbidity risk and mode of delivery at term: influence
of timing of elective caesarean delivery. Acta Pædiatrica. 2004;93(5):643–7.
31. Tita AT, Landon MB, Spong CY, Lai Y, Leveno KJ, Varner MW, Moawad AH,
Caritis SN, Meis PJ, RJJNEJoM W, Mercer BM. Timing of elective repeat
cesarean delivery at term and neonatal outcomes. N Engl J Med. 2009;
360(2):111–20.
32. Morrison JJ, Rennie JM, Milton PJ. Neonatal respiratory morbidity and mode
of delivery at term: influence of timing of elective caesarean section. J
Obstet Gynaecol. 1995;102(2)101–6.
33. Liu S, Heaman M, Joseph KS, Liston RM, Huang L, Sauve R, Kramer MS. Risk
of maternal postpartum readmission associated with mode of delivery.
Obstet Gynecol. 2005:105(4):836–42.
34. Liu S, Liston RM, Joseph KS, Heaman M, Sauve R, Kramer MS. Maternal
mortality and severe morbidity associated with low-risk planned cesarean
delivery versus planned vaginal delivery at term. Can Med Assoc J. 2007;
176(4):455–60.
35. Lumbiganon P, Laopaiboon M, Gülmezoglu AM, Souza JP, Taneepanichskul
S, Ruyan P, Attygalle DE, Shrestha N, Mori R, NDJTL H, Villar J. Method of
delivery and pregnancy outcomes in Asia: the WHO global survey on
maternal and perinatal health 2007–08. Lancet. 2010;375(9713):490–9.
36. Villar J, Carroli G, Zavaleta N, Donner A, Wojdyla D, Faundes A, Velazco A,
Bataglia V, Langer A, AJB N, Acosta A. Maternal and neonatal individual risks
and benefitsassociated with caesarean delivery: multicentre prospective
study. BMJ. 2007;335(7628):1025.
37. Healthy People 2020 Topics & Objectives: Maternal, Infant, and Child Health.
[https://mchb.hrsa.gov/chusa13/perinatal-health-status-indicators/p/cesarea
n-delivery.html].
38. Hanlon M, Burstein R, Masters SH, Zhang R. Exploring the relationship
between population density and maternal health coverage. BMC Health
Serv Res. 2012:12(1):1–7.
39. Huque MH, Anderson C, Walton R, Ryan L. Individual level covariate
adjusted conditional autoregressive (indiCAR) model for disease mapping.
Int J Health Geogr. 2016;15(1):1–13.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.