NATIONAL UNIVERSITY OF SCIENCE AND

TECHNOLOGY

FACULTY OF APPLIED SCIENCES

DEPARTMENT OF OPERATIONS RESEARCH AND STATISTICS

Determinants of adverse neonatal outcomes in a low resource

setting.

by

Cindy Ndou

(N0183353E)

SUPERVISOR : Mr D. Mwembe

This dissertation was submitted to the Department of Operations Research and Statistics of
the National University of Science and Technology in partial fulfillment of the requirements
for the Degree of BSc. in Operations Research and Statistics, Bulawayo, Zimbabwe

JUNE 2022
Declaration

I, Cindy Ndou, declare that the project which is hereby submitted for the qualification of
Bachelor of Science(Hons) in Operations Research and Statistics at the National University
of Science and Technology, is my own independent work and has not been handed in before
for a qualification at/in another University/Faculty/School. I further declare that all sources
cited or quoted are indicated and acknowledged by means of a comprehensive list of refer-
ences. I further cede copyright of the dissertation to the National University of Science and
Technology.

Signature..................................................................

Date: June 2022

Copyright c 2022 National University of Science and Technology

All rights reserved

i
Abstract
This study examines crucial topics relating to neonatal mortality in low resource setting.Ensuring
that policies and practice in neonatal care are informed by evidence from high-quality re-
search is fundamental to improving outcomes for newborn infants and their families. Effec-
tive interventions in the neonatal period can have a life-long impact disproportionate to their
costs. Many of the major advances in care that have transformed outcomes for preterm and
sick newborn infants have been informed by empirical and applied health research.In order
to model the survival time of neonates at risk, the study’s goal is to identify the variables
that affect neonatal mortality rates. Mpilo Central Hospital provided the secondary data set
that was used in this investigation. Cox regression was used to estimate the survival time
of individuals at risk, and binary regression was used to identify the components that are
relevant in neonatal mortality rates. The study’s hypotheses were all tested for significance
at a level of 95%, and Stata was used to analyze the data.Empirical and applied health re-
search has been a primary source of inspiration for many of the significant improvements in
care that have improved outcomes for preterm and ill newborn newborns. The challenge in
the 21st century is to maintain the trajectory of improvements in care and outcomes. The
findings indicated that factors such as ANC visits, mother’s age, education level, booking sta-
tus, patient’s diagnosis at birth, mode of delivery, amniotic fluid odor, and amniotic fluid color
play a role in unfavorable outcomes. then, the variables were further. Gestational age and
ANC visits were discovered to be important when they were evaluated to determine their
effect on neonatal survival time and the mother’s booking status. If all other factors are held
constant, mothers who do not plan their pregnancy have (2.4) more chances of having a baby
who develops poorly than mothers who do.Neonatal infants with normal and low apgar scores
had reduced survival rates, according to Kaplain Meir survival curves. Mortality rate is 2.4
times more likely to occur in low-apgar-score infants than in those with a normal apgar score.
Kaplain Meir survival curves were also used to model survival patterns of delivery modes,
which were divided into three groups: NVD, C-Section, and other modes. It was found that
NVD births had higher survival rates than C-Section births, as well as births by forceps deliv-
ery, vacuum extraction, and other methods. To reduce infant mortality, the government must
enhance its maternity and neonatal care programs, particularly those at community-based
health care facilities.

ii
Dedication

To my loving mum, sister, brothers and uncle. Love you totally.

iii
Acknowledgments

First and foremost, I would like to express my gratitude to the Almighty God for His unfail-
ing love, tender mercies, and guiding hand as I approached my final year at the National
University of Science and Technology. Only through Him is everything made possible. I also
want to express my gratitude to Mr. D. Mwembe my academic supervisor, for his advice,
ongoing support, and help in making this research project a success. In addition, I want to
express my gratitude for the direction, support, and advice I received from the department
of statistics and operations research employees. Special thanks to my loved ones, my family
and classmates for the love through it all, your positive outlook on life gave me the strength
to drive through all.

—God Bless you all

iv
Contents

Declaration i

Abstract ii

Dedication iii

Acknowledgments iii

Table of Contents v

List of Figures ix

List of Tables ix

List of Abbreviations and Acronyms ix

1 Introduction 1
1.1 DEFINATION OF TERMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 BACKGROUND OF THE STUDY . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Mpilo Central Hospital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Significance of the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Aim of the study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.1 Objectives of the study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Research questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5.1 What factors contribute to poor newborn outcomes in low resource set-
tings? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6.1 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

v
 1.6.2 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Literature review 8
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Situation of neonatal mortality in Developing and Wealthy Nations . . . . . . . 8
2.3 Africa’s neonatal mortality situation . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Causes of neonatal mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Ways of reducing neonatal mortality . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Models Used In Previous Researches . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3 Methodology 16
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 RESEARCH DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 STUDY AREA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 STUDY Population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3.1 Population to be targeted . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3.2 DATA SETTING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 LOGISTIC REGRESSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4.1 PARAMETER ESTIMATION . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.2 MODEL ASSUMPTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.3 DEPENDENT VARIABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.4 Explanatory Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.5 STEPWISE REGRESSION . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.6 BACKWARD ELIMINATION . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4.7 Backward Elimination has several steps. . . . . . . . . . . . . . . . . . . 21
3.5 COX REGRESSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5.1 ASSUMPTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5.2 Specifications for the Model . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5.3 PARAMETER ESTIMATION . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5.4 HAZARD FUNCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5.5 THE HAZARD RATIO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5.6 KAPLAIN MEIER ESTIMATES . . . . . . . . . . . . . . . . . . . . . . . 23

vi
 3.5.7 LOG RANK TEST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5.8 CONFIDENCE INTERVAL . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 CENSORING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.7 RATIOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7.1 ODDS RATIO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7.2 Relative Risk Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.7.3 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4 Data Analysis 28
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 SOCIO-DEMOGRAPHIC AND ECONOMIC CHARACTERISTICS . . . . . . . 28
4.2.1 PERCENTAGE FREQUENCIES OF THE VARIABLES . . . . . . . . . 29
4.3 LOGISTIC REGRESSION ESTIMATION . . . . . . . . . . . . . . . . . . . . . . 32
4.3.1 Testing for Multicollinearity . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.2 Model Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.3 Likelihood Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3.4 Model Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3.5 Variables In The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.6 Variables Not In The Model . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 Goodness Of Fit Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4.1 Hosmer Lemeshow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.5 Cox Proportional Hazards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.5.1 Model Significance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.5.2 Significant Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.5.3 Hazard Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.6 Logrank Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6.1 Amniotic Fluid Odor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6.2 Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6.3 ANC Visits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6.4 Survivor Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.6.5 Logrank Test For Equality Of Survivor Functions . . . . . . . . . . . . . 45
4.7 Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.7.1 Relative Risk Ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.8 Odds Ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

vii
 4.8.1 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Results 50
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Appendix 58

viii
List of Tables

4.1 Maternal Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.2 Percentage Of Frequencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3 Sampling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.4 VIF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 Classification Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.6 Classification Percentage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.7 Parameter estimates for the fitted model . . . . . . . . . . . . . . . . . . . . . . 36
4.8 Parameter estimates for the fitted model . . . . . . . . . . . . . . . . . . . . . . 39
4.9 Hosmer Lemeshow Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.10 Cox Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.11 Probability among males and females. . . . . . . . . . . . . . . . . . . . . . . . . 46
4.12 Odds for death among males and females, π denotes the probability of death. . 46

ix
List of Abbreviations and Acronyms

APH Antepartum Haemorrhage

NVD Normal Virginal Delivery

C-Section Caesarean section

LBW Low Birth Weight

AMA Advanced Martenal Age

NMR Neonatal Mortality Rate

LRS Low Resource Setting

WHO Wealth Health Organisation

VIF Variance Inflated Factor

ANC Antenental Clinic

ENND Early Neonatal Death

NDR Neonatal Death Rate

SDG Sustainable Development Goals

RR Relative Risk Ratio

OR Odds Ratio

x
 HR Hazard Ratio

HIE Hypoxic-ischemic encephalopathy

PROM Premature Rapture Of Membranes

LNND Late Neonatal Death

MCH Mpilo Central Hospital

SDG Sustainable Development Goals

MDG Millennium Development Goals

xi
Chapter 1

Introduction

Neonatal is also called a new born baby. The health status of a woman before and during
pregnancy is a key determinant of pregnancy outcomes, and poor nutritional status and ex-
posure of infectious diseases during pregnancy contribute to maternal and as well as infant
mortality. A low resource setting (LRS) is a setting where health care systems do not meet
the minimum standards set by World Health Organization (WHO) or any other quas govern-
mental organization.Neonatal is the most vulnerable time for a child’s survival. The neonatal
outcome is the consideration of the first four weeks of a child’s life. It is a time when changes
are very rapid, neonatal deaths may be subdivided into early neonatal deaths, occurring dur-
ing the first seven days of life, and late neonatal deaths, occurring after the seventh day but
before the 28 completed days of life. Early neonatal death (ENND) can be caused by low
birth weight (LBW), mode of delivery, premature birth. Late neonatal death major causes
are congenital disorders, sepsis and pneumonia as to (Lawn et al). The incidence of neonatal
death rate (NDR) is calculated as the number of infant deaths that occur between (0-28) days
of life (the first month of life) divided by number of live births, multiplied by 1000. The WHO
definition specifies that the number of live births or total number of births can be used as a
denominator, and calculations can be made (Johnston et al, 2011).

1.1 DEFINATION OF TERMS

• Neonate:a new born baby within 28 days of life.

• Adverse neonatal outcome: it is the presence of birth asphyxia, respiratory distress,

1
 birth trauma , hypothermia , meconium apiration syndrome , neonatal intensive care
admission and neonatal death.

• Neonatal Mortality Rate (NMR): number of infant deaths for every 1000 live births.

• Neonatal death:Neonatal death occurs when a baby dies within 28 days following birth.

• Early neonatal death (ENND) is defined as death of a newborn within the first week of
life, while late neonatal death (LNND) is defined as death of an infant between the ages
of 8 and 28 days.

• Low resource setting: settings where health care systems do not meet the minimum
standards set by world health organization(WHO).

• NVD : is a vaginal delivery, whether or not assisted or induced, usually used in statistics
or studies to contrast with a delivery by cesarean section.

• CS : is the use of surgery to deliver babies, often necessary when a vaginal delivery
would put the baby or mother at risk.

• Gravida or gravidity: describes the total number of confirmed pregnancies that a woman
has had, regardless of the outcome.

• Parity: is defined as the number of births that a woman has had after weeks gestation.

• LBW(low birth weight): A low birth-weight infant is an infant weighing less than 2500
g or 2.5 kilograms (kg) at birth, regardless of gestational age.

• Apgar Score:is a way to evaluate the health of all newborns at 1 and 5 minutes after
birth and in response to resuscitation.

• Prenatal Care (ANC): If the moms visited a health facility for antenatal care at least
once for this index birth (current newborn), we counted them as having ANC visits
(Yes/No), regardless of the number of visits.

• Premature rupture of membranes (PROM): Pre-labor rupture of membranes, is a con-

dition that can occur in pregnancy. It is defined as rupture of membranes (breakage of
the amniotic sac), commonly called breaking of the mother’s water(s), more than 1 hour
before the onset of labor.

• Preterm: one who is born before the 37th week of pregnancy.

2
 • Birth Asphyxia: defined as failure to initiate and sustain normal breathing at the first
and fifth minute of birth.

• Antepartum Haemorrhage(APH):is the bleeding from or in the genital tract, occurring

from 24+ weeks of pregnancy and prior to the birth of the baby.

• Hypoxic-ischemic encephalopathy (HIE): is a kind of brain injury in newborns caused

by a lack of oxygen and restricted blood flow. HIE is a form of birth injury, which is a
wide term that refers to any harm a newborn suffers during or shortly after birth.

• Congenatal Abnormality:Refers to structural or functional anomalies that occur during

intauterine life.

• Neonatal Sepsis:Is a bloodstream illness that affects newborn infants under the age of
28 days.

• Neonatal Mortality:Refers to the number of death during the first -28 completed days of
life.

• Neonatal Morbidity:Refers to the death during the newborn period -the first 28 com-
pleted days of life.

• Haemoglobin:Its a red protein responsible for transporting oxygen in the blood of verte-
brates.

• Neonatal sepsis is a type of neonatal infection and specifically refers to the presence in
a newborn baby of a bacterial blood stream infection (BSI) (such as meningitis, pneu-
monia, pyelonephritis, or gastroenteritis) in the setting of fever.

1.2 BACKGROUND OF THE STUDY

1.2.1 Mpilo Central Hospital

Mpilo Central Hospital is Zimbabwe’s third largest hospital and the largest in the Southern
Region. The institute provides all main health services that necessitate experts; these are
the highest and most specialized levels of care, with speciality services for both maternal and
child health. It is the hospital’s obligation to look into the best ways to improve the well-being
of patients.

3
Adverse neonatal statistics

Neonatal mortality rate is a benchmark for infant care and the health of a society as a whole.
Global efforts to reduce these mortality rates have been led by the World Health Organization
(WHO) and the United Nations. A further 3.8 million babies die in their first month of life
up to half of these on the first day. Twenty million low birth weight (LBW) babies and others
with neonatal complications live but may not reach their full potential. This enormous bur-
den is wholly related to adverse maternal, yet remains under-appreciated and has attracted
less investment than other conditions with a lower burden. Many of these stillbirths and
early neonatal deaths occur at home, unseen and uncounted in official statistics by (Phibbs
et al, 2019). For both neonatal and maternal deaths, over two-thirds occur in Sub-Saharan
Africa and South Asia. The regional variations in neonatal mortality rate (NMR) and mater-
nal mortality ratio (MMR) are wide. However some low-income countries, such as Thailand
and Sri Lanka, have managed to reduce their NMR to under ten deaths per 1000 live births.
Low income countries that have achieved major reductions in maternal and neonatal mortal-
ity have also reached high coverage of skilled attendance during childbirth. In India alone
over million babies die every year. Poverty and ill health of newborns are closely linked.
Mothers and newborns in poor families are at increased risk of illness and face more chal-
lenges in accessing timely, high quality care compared with wealthier families. The newborn
health gap between rich and poor countries is unacceptably high. Maternal mortality has
decreased by approximately 45% from the Millennium development goals by 1990 according
to (Chopra et al, 2009). One of the four babies worldwide are delivered without the presence
of a skilled birth attended. On a daily basis hundreds of preventable maternal deaths oc-
cur due to pregnancy and childbirth related complications. Neonatal deaths now account for
44% of the childhood mortality which is the highest percent according to (Liu et al, 2015).
Approximately 2.8 million babies worldwide die each year during the first month after birth,
with a high rate in developing countries. Scaling up preventions proven to be effective could
substantially reduce the regrettable loss of young lives. The article “temporal and spatial evo-
lution of maternal and neonatal mortality rates in Brazil 1997-2012” by reports the trends in
maternal mortality ratio and neonatal mortality rate in large geographical area with signif-
icant disparities in socioeconomic status. The study highlights that the maternal mortality
rate remained relatively constant during the study period in spite of decreases in the neona-
tal mortality rate. Efforts related to the Millennium Development Goals have been associated
with a reduction of approximately 45% in maternal mortality and over 50% in neonatal and

4
child mortality. Globally 2.4 million children died in the first month of life in 2019 approx-
imately 6700 neonatal deaths everyday with about a third of all neonatal deaths occurring
within the first day after birth, and close to three-quarters occurring within the first week of
life. Among the countries that have been not making progress in reducing maternal death
is Zimbabwe. In 2003 Zimbabwe together with Afghanistan, Kenya, Mozambique, Tanza-
nia and Malawi were among the countries with the highest MMRs of 1000 or greater in the
world (Abouzahr et al, 2003) as well as high rate of NMRs. Maternal and neonatal mortality
remains disturbingly high in Zimbabwe. Since 1990 there has been a large and significant
increase in maternal and neonatal mortality in Zimbabwe and this erases any potential gain
in maternal survival achieved by safe motherhood programs during the preceding decade.
Greater progress is, therefore, needed in significantly reducing maternal and neonatal death
in Zimbabwe. This study seeks to investigate the causes of maternal and neonatal deaths
and contribute to the accurate assessment of maternal and neonatal mortality.

1.2.2 Problem statement

The high neonatal mortality rate is a worldwide concern, and it has been agreed that mater-
nal and child health is an essential indication of national health and socioeconomic progress.
The explanation of the high infant mortality rate at Mpilo Central Hospital is unknown,
necessitating this study. One of the triggering factors in negative outcomes is the lack of pre-
diction models. Because of the inability to forecast bad neonatal outcomes, pregnant women
and their babies are missing out on potentially life-saving therapy and care options. By iden-
tifying the moms and babies who are most at risk of difficulties, interventions can be tailored
to those who will benefit the most. Current risk prediction models, on the other hand, cover
a wide range of scenarios sensitivity (42–81%) and specificity (87–92%) indicating that there
are a lot of challenges there for improvements are needed as noted by (Ngwenya et al, 2019).
Furthermore, no predictive models have been developed hence this research will be of help.

1.3 Significance of the study

The findings will help us understand the factors that contribute to poor neonatal outcomes in
low-resource settings, as well as effective strategies for lowering neonatal mortality rates.

5
1.4 Aim of the study.
To identify factors that contribute to poor newborn outcomes in low-resource settings.

1.4.1 Objectives of the study.

1. To determine the factors that influence neonatal survival.

2. To determine the survival function of neonates.

3. To model survival patterns of neonates.

1.5 Research questions.

The research seeks to address several questions which are as follows:

1.5.1 What factors contribute to poor newborn outcomes in low re-

source settings?

• What is the average neonatal survival time?

• What are the current neonatal survival rates in low resource setting?

• What factors play a role in low resource setting’s high NMR?

1.6 Assumptions
We will utilize non parametric models to analyze neonate survival patterns if all of the data
used in the research does not follow any distribution.

1.6.1 Limitations

The research will concentrate on patients who have been admitted, as there may be some
missing or erroneous information in the hospital statistics.

6
1.6.2 Delimitations

It will assess the effectiveness of the current clinical procedures used to nurture neonates
within the Hospital. This will also identify the procedures needed to improve survival chances
of all neonates.

1.7 Summary
The chapter highlighted the problem statement and gave a detailed background about the
study. It also stated the significance of the research how will it prove clinical procedures in
improving maternal and neonatal health care.

7
Chapter 2

Literature review

2.1 Introduction
This chapter focuses on evaluations of literature from various researchers who have studied
the factors that contribute to a high neonatal death rate. This chapter begins with a review
of survival analysis, followed by theories on hospital neonatal care.

2.2 Situation of neonatal mortality in Developing and

Wealthy Nations
The neonatal mortality rate (NMR) is a risk indicator for newborn death per live birth that is
comparable across time and nations. The number of newborn fatalities is a valuable metric of
burden as compared to other conditions because it is dependent on the NMR and livebirths.
The newborn mortality rate (NMR) and the absolute number of neonatal fatalities are both
important metrics for determining the breadth of neonatal deaths in a country as well as the
causes of morbidity and mortality.Sustainable Development Goal 3 (SDG 3 or Global Goal
3) is one of the 17 Sustainable Development Goals created by the United Nations in 2015.
It focuses on ”Good Health and Well-Being”. To ensure healthy lifestyles and promote well-
being for everybody at all ages,” says the official statement. SDG 3 includes and focuses on

8
Literature review

a variety of components of a healthy living and lifestyle. Twenty-one indicators are used to
track progress toward the goals. There is very little evidence of a link between GDP an d new-
born mortality. Contextual factors (poverty, female literacy, sanitation, and access to clean
water), according to a larger study using principal component analysis, explain less of the
difference in country mortality rates for newborn mortality than for post neonatal and early
child mortality (WHO, The World Health Report). There is also some indication that income
may be more important at the home level for older children than for neonates. Kinney et
al(2010) discovered that mortality disparities between the richest and poorest quintiles were
generally higher for post neonatal deaths than for neonatal deaths.The WHO (Domine et al,
2011) created user-friendly clinical standards for maternal and neonatal care. Starting at
the initial level of care, the guidelines contain the key recommendations for the delivery of
maternal and neonatal care in health facilities. The criteria were created in response to high
maternal and newborn mortality caused by insufficient and poor-quality health services, as
well as lower health-care utilization (WHO 2007b:1). Neonatal mortality is a major public
health issue, accounting for almost one-third of all fatalities in children under the age of five.
Akombi et al (2017) noticed that poor mother health, insufficient prenatal care, incorrect han-
dling of pregnancy and delivery difficulties, poor hygiene during delivery and the first critical
hours after birth, and a lack of resources all contribute to neonatal deaths and stillbirths
and lack of neonatal care.Several factors, such as women’s social status, nutritional status
at conception, early childbearing, too many closely spaced pregnancies, and harmful prac-
tices, such as inadequate cord care, letting the baby stay wet and cold, discarding colostrum,
and feeding other food, are deeply rooted in the cultural fabric of societies and interact in
ways that are not always well understood. Over the last several decades, the globe has made
significant progress in lowering child mortality. Poor mothers are more likely to have a new-
born death than their richer counterparts (Lawn et al,2005), and these disparities persist
even as national neonatal mortality rates decline: indeed, even in industrialized countries, a
socio-economic gradient is visible (UK Office of National Statistics 1999). In developing coun-
tries that have made significant progress in reducing neonatal mortality, reducing inequities
experienced by poor and disadvantaged communities (e.g. Vietnam) remains a challenge.
Nonetheless, mortality rates in developing countries remain high, and progress in reduc-
ing neonatal mortality is slower than progress in reducing mortality in children aged 1–59
months. Hospital acquired infection(HAI) is a major cause of neonatal mortality and morbid-
ity with prevalence ratios in low and middle-income countries 3-20 times higher than high
income countries (Zaidi et al, 2005). It is well understood that there is a wealth gradient in

9
Literature review

child mortality.The link between GDP and national levels of child mortality has been well
documented (Filmer et al, 1997), but data has never been disaggregated to look at different
age groups. Recent research employing family/individual data indicates a wealth gradient in
neonatal mortality, with the poorest 20 percent of households reporting consistently higher
NMRs than the wealthiest.Further evidence from UK historical data collected in England
and Wales between 1911 and 1932 shows that while a class gradient existed in neonatal
mortality, it was much less steep than in newborn mortality (Neal et al, 2014). The correct
assessment of newborn mortality in underdeveloped countries has a number of obstacles,
and a lack of data has likely contributed to the lack of attention paid to this area in the
past (Lawn et al, 2005). In most underdeveloped countries, vital registration is either par-
tial or non-existent, and many newborn fatalities occur at home without any contact with
medical services, therefore health information systems do not register them.Even in places
where institutional delivery is widespread, different policies for classifying newborn deaths
and stillbirths might cause measurement differences (Aleshina et al, 2005). There is also
evidence that in some health systems, employees are rewarded for misreporting newborn fa-
talities as stillbirths in order to escape audits or boost hospital ratings when NMR is used
as a quality indicator (ibid.). Many countries are unlikely to implement effective and com-
prehensive vital registration systems in the foreseeable future. Almost no nations now have
both child mortality rates of over 25 per 1000 live births and comprehensive vital registra-
tion coverage (defined as 95 percent of all deaths documented) (Morris et al, 2003).Another
possibility is to construct continuing retrospective surveys or sample registration systems,
similar to those pioneered in China and India. The Indian Sample Registration Survey (SRS)
collects data in two ways: births and deaths are regularly counted in a sample of locations
by a part-time worker, and six monthly retrospective studies are conducted. However, it ap-
pears that even the SRS’s multiple techniques result in underestimating of mortality (Chan
et al, 2002). Historical data also suggests that, in addition to dropping more slowly, neona-
tal mortality in many affluent nations did not begin to diminish significantly until several
years after post-neonatal and childhood mortality had decreased. Neonates are increasingly
a key focus of child health in high-income countries (HIC), both for mortality and morbidity
reduction. However, compared to maternal deaths or deaths among older children under the
age of 5, neonatal mortality rates (NMRs), trends, and causes have received relatively little
attention in lower-income countries, and neonatal deaths still do not receive attention com-
mensurate with their burden in international public health policy and programs (Shiffman
et al, 2010). Adam et al (2005) researchers investigating the impact of ANC interventions on

10
Literature review

maternal and neonatal health outcomes have come up with mixed results. There is insuffi-
cient evidence of the effect of prenatal care on the reduction of neonatal mortality, according
to two recent systematic reviews undertaken in both high and low-income countries.

2.3 Africa’s neonatal mortality situation

Every year, around 14 million mothers aged 15 to 19 give birth around the world. In Sub-
Saharan Africa (SSA), the frequency of teenage births is exceptionally high, with an esti-
mated 50% of mothers under the age of 20. In comparison to adults, adolescent moms had
a much increased risk of newborn mortality. Neonatal mortality has been dropping in Sub-
Saharan Africa, especially Ethiopia, during the past two decades. The leading causes of death
can be avoided or treated. Recognizing the variables, however, could be a critical step in re-
ducing infant death.However, research on newborn survival and/or determinants of neonatal
death was scarce in Ethiopia, particularly in the study area. Within SSA, neonatal mortality
rates are unevenly distributed, and countries in the great lake region (Burundi, DRC, Kenya,
Rwanda, Tanzania, and Rwanda) contribute to this burden due to decades of political insta-
bility, conflicts, and poor quaternary care (Akombi et al, 2019). The majority of children that
die in their first four weeks of life (99 percent) do so in poorer parts of the world, particularly
in Sub-Saharan Africa and South Asia. In 2018, the newborn mortality rate in Sub-Saharan
Africa was 28 deaths per 1000 live births. While child mortality has decreased significantly
in poor nations over the last few decades, newborn mortality has remained relatively stable.
Previous research (Hill et al, 1997) has shown that neonatal mortality has declined more
slowly in many countries than post-neonatal infant mortality (PNMR) or early childhood
mortality (ECMR).Although this broad pattern has been well documented in recent years,
more cross-national research is needed to gain a better understanding of how death rates
and progress differ throughout nations and regions, as well as over time.No study has deter-
mined if the trend of slower NMR reductions is universal or whether trends vary by country.
Furthermore, little is known about how recent patterns in neonatal mortality have evolved
in light of standstill and reversal in child health improvements, notably in Sub-Saharan
Africa. Many nations in Sub-Saharan Africa are still not on track to meet the Sustainable
Development Goal of 12 newborn deaths per 1000 live births by 2030, despite decades of low-
ering neonatal mortality rates. Betran et al, (2016) believe that maternal and newborn care
services given in health care facilities are crucial to accelerating improvement on neonatal
health outcomes. Improving the number and quality of maternity and newborn care services

11
Literature review

(Khanna et al, 2021) says it has been a major focus of performance-based financing (PBF)
programs in Africa during the last two decades, thanks in part to funding from the World
Bank’s Health Results Innovation Trust Fund (HRITF).

2.4 Causes of neonatal mortality

Prematurity and low birth weight (LBW) (29 percent), newborn infections (25 percent), birth
asphyxia and birth trauma (23 percent), congenital abnormalities (8 percent), neonatal tetanus
(2 percent), and diarrhoeal disease (2 percent) are the top five causes of neonatal mortality
globally (2 percent )Newborn Health Epidemiology. 2011, WHO.The risk of infant mortal-
ity is linked to maternal age, with young women having a higher risk. Every year, over 14
million teenagers aged 15 to 19 give birth around the world. In SSA, 16% of young adults
live, with a fertility rate of 112.84/1000, compared to 40.50/1000 for older women (Fatusi
and Blum (2008)). In SSA, 50% of births are to moms under the age of 20, in comparison to
moms aged 20 years, mothers aged 15 years, 16–17 years, and 18–19 years have a 55 per-
cent, 19 percent, and 6% higher risk of newborn mortality(Ramaiya et al, 2014). Various risk
factors related with newborn mortality have been discovered in studies conducted in devel-
oping nations. Multiple births, a lack of antenatal care, maternal illnesses during pregnancy,
preterm, birth asphyxia, and neonatal sepsis are among them (Lee et al, 2013). Furthermore,
various reasons of newborn fatalities have been identified in Ethiopian investigations. Birth
asphyxia, newborn infections, and preterm were each responsible for 67 percent of neonatal
mortality (Kokeb et al, 2016). Another cause in neonatal deaths in hospitals is the misuse of
intravenous medicines to speed up labor. In India, a recent study indicated that giving oxyto-
cics to help with labor was linked to a three-fold increase in birth asphyxia-related mortality
[odds ratio (OR)=2.6; 95 percent confidence interval (CI) 1.9-3.6] (Bang et al, 2005).The trial
of labor was linked to a higher risk of newborn death in health institutions in a Bangladeshi
rural district (Nasreen et al, 2012). According to the findings of a recent study conducted in
South Africa, insufficient foetal monitoring by health staff is a key preventable factor linked
to birth asphyxia-related mortality (Velaphi et al, 2007).The high rate of newborn mortality
could be explained by a lack of competent health services, which result in insufficient prena-
tal care and late presentation at NRH, as well as the low socioeconomic status of a significant
portion of the population in Mauritania. Many of the newborns who are moved to this hospi-
tal have significant problems and a high risk of death, as we have previously demonstrated.
Hypothermia and transfer status, both of which might be considered markers of late presen-

12
Literature review

tation at health care facilities, were identified as risk factors. Other emerging countries have
reported similar outcomes(UNICEF/WHO/The World Bank/UN Pop Div).The most common
causes of death in the first week were prenatal anomalies and postnatal infections. Prematu-
rity was also found as the major cause of death in the first week of life in other studies(Liuet
al, 1999). The education level of the father is linked to neonatal mortality, but not the ed-
ucation level of the mother. Fathers are the heads of families in many societies, and their
education level, in contrast to the mother’s education level, may thus play a more important
role, as their socioeconomic status and education largely define the general family situation,
as well as health care seeking behavior during critical conditions (Yaya et al, 2014).Adult
pregnancy, AMA(Advanced Maternal Age) is a substantial risk and a key contributor to a
variety of unfavorable perinatal outcomes(Silvermanet al, 2006). Still birth, intrauterine fe-
tal growth restriction (IUGR), and neonatal death have all been linked to AMA. It was also
linked to perinatal morbidities such as low birth weight (LBW), premature birth, and a low
Apgar score(Ghezzi et al, 2005). AMA also increases the risk of congenital malformations and
chromosomal abnormalities like trisomy (Ghezzi et al, 2005) during pregnancy. According to
a Danish study, poor neonatal outcomes were 10.8% among AMA women compared to 5.4
percent among adult women(Buccellato et al, 2000).

2.5 Ways of reducing neonatal mortality

The majority of neonatal deaths occur in low- and middle-income nations. By achieving high
coverage of excellent antenatal care, competent birth care, postnatal care for mother and
baby, and treatment of small and sick infants, it is possible to increase newborn survival and
health and reduce unnecessary stillbirths. Midwife-led continuity of care (MLCC) can prevent
premature births by up to 24 percent in settings with well-functioning midwife programs.
MLCC is a model of care in which the same woman is cared for by a midwife or a team of
midwives throughout her pregnancy, labor, and postpartum period, with medical assistance
when needed.According to a recent UNICEF research, the majority of newborn fatalities can
be avoided with simple, cost-effective interventions given during pregnancy and childbirth,
but there is a lot of variation in the utilization and quality of maternal health care services
provided to pregnant women and their newborns. Many pregnant women and babies lose out
on life-saving interventions [UNICEF, Committing to Child Survival].With the rise in facility
deliveries (almost 80% worldwide), there is an excellent chance to provide vital infant care
as well as identify and manage high-risk newborns. However, few mothers and newborns

13
Literature review

stay at the facility for the full 24 hours following birth, which is when difficulties are most
likely to occur. Furthermore, premature discharge from the hospital, access restrictions, and
delays in obtaining care cause far too many neonates to die at home. It is recommended that
postnatal care contacts be offered at a health facility or through home visits to reach these
newborns and their families. According to NICHQ (National Institute for Children’s Health
Quality). Preterm babies often have a low birth weight and may have underdeveloped organs;
they are at risk for infection, as well as growth and development disabilities; and they may
have difficulty feeding because they are too small or sick to feed by bottle or breast, according
to the National Institute for Children’s Health Quality. Optimizing these babies’ diet can
aid their growth and provide life-changing and life-saving advantages.Premature neonates
are usually fed intravenously until they are ready to be transitioned to tube feeding through
the gastrointestinal tract (known as enteral nutrition). Early enteral nutrition, which begins
feeding babies through the gastrointestinal tract sooner than more typical procedures, is one
strategy Campbell’s team tried to help them reach a healthy weight.

2.6 Models Used In Previous Researches

The most recent global estimate comes from (Lawn et al, 2006), who used multinomial re-
gression techniques to model cause of death based on vital registry data (where available)
and both published and unpublished studies. According to the report, infections are respon-
sible for 36 percent of all deaths worldwide (26 percent sepsis/pneumonia, 7% tetanus, and
3% diarrhoea). Preterm birth is responsible for another 28%, hypoxia is responsible for 23%,
and congenital abnormalities are responsible for 7%. The study does, however, reveal that
there are large differences in cause of death distribution between countries and regions, and
it is also limited in that it cannot capture the multi-causal nature of many of these deaths.

Premature newborns, for example, are more susceptible to infection, but the methods used
cannot investigate these connected aspects.Low birth weight, low 5th APGAR score, preterm
birth, sex, ANC visit, hypothermia, initiation EBF, time of rupture of membranes, and man-
ner of delivery were all determined to be significant predictors of death with a p-value of 0.25
by (Berhanu et al. (2021)).

According to (Toussia et al, 2021), 391 newborn medical records were used, with a data com-
plete rate of 97.02 percent.A multivariable logistic regression showed Lack of antenatal care

14
Literature review

(ANC) follow up [AOR = 4.69: 95 percent CI (1.77, 12.47)], giving birth via cesarean section
[AOR 3.59, 95 percent CI (1.22, 10.55)], admission temperature less than 36.5◦ C [AOR 10.75,
95 percent CI (3.75, 30.80)], birth asphyxia [AOR 7.16, 95 percent CI (2.22, 23.10)], and
having a length of stay greater than the associated neonatal mortality.This study revealed
that the rate of neonatal mortality is still high compared to the national data. Antenatal
treatment, cesarean section delivery, length of stay in the hospital, low temperature upon
admission and birth hypoxia were significantly related with newborn mortality. As a result,
health facilities should focus on improving antenatal care, intrapartum care, and standard-
ized neonatal care. In addition, prospective investigations are suggested.

Reggente et al, (2018) applied multivariate multiple regression using natural log of gross na-
tional income (GNI) per capital as independent variable (with dummy variables for region)
and natural log of neonatal, post neonatal, and early childhood mortality rates (NMR, PNMR,
and ECMR) as dependent variable. The coefficient for change in GNI was significant for all
component mortalities, but the results showed higher elasticities for post neonatal and early
childhood mortality than for neonatal mortality: a 10% increase in GNI over time is associ-
ated with decreases of 2.6%, 4.8%, and 5.3% in NMR, PNMR, and ECMR, respectively. It can
be noted that changes in GNI over time appear to have a greater impact on mortality than
differences in GNI between countries at the same point in time for all component mortality
rates. While confidence intervals are quite wide, postestimation testing of the hypothesis
that the coefficients for log of change in GNI are equal in the equations for log of change in
NMR and log of change in PNMR, and log of change in NMR and log of change in ECMR
demonstrate the coefficients to be significantly different at the 5% level.

2.7 Summary
Survival analysis is a branch of statistics that studies the time it takes for an event to occur,
such as death in biological organisms or mechanical system failure. In engineering, this is
known as reliability theory or reliability analysis, as well as duration analysis or duration
modeling.

15
Chapter 3

Methodology

3.1 Introduction
This chapter sets out our project’s methodological framework. It describes the methods used
to collect the necessary information and the procedure for analysing and interpreting the
collected information. It will provide a detailed algorithm to how the techniques will be used
to analyze and describe patterns of the neonatal survival. All the variables will be explained
dependent and independent.

3.1.1 RESEARCH DESIGN

The organization of settings for data collecting and analysis is referred to as a study design.
In a way that tries to combine relevance to the research goal with procedural economy
(Kothari, 2004). The study design adopted by the researcher was a case control study. This
form of research compares a group of neonates who have had the desired outcome, referred
to as cases (in this example).
It contrasts neonates who died as a result of a neonatal cause) with newborns who did not.
Controls (neonates of reproductive gestational age) are the outcome of interest (neonatal
death).
who born and survived at Mpilo Central Hospital during the research period). This is a ret-
rospective study that compares how often each group is exposed to a risk factor to see if there
Methodology 17

is a link between the risk factor and neonatal death. The study used secondary data from the
facility-based review data from 2019-2021 to examine the causes and various risk factors of
neonatal fatalities at Mpilo Central Hospital Methodology . Through a survey of published
literature from both international and national sources, the study synthesized the available
operational and academic research findings.

3.2 STUDY AREA

This study was conducted at Mpilo Central Hospital.Mpilo Central Hospital is the biggest
hospital in the Southern Region and the third largest in Zimbabwe after Harare Central
and Pararenyatwa Hospitals. As a major quaternary referral centre, Mpilo Central Hospital
serves the population of Bulawayo Metropolitan Province, Matebeleland North and South,
Midlands and Masvingo Provinces. This area was chosen because as a referral centre of
neonates admitted at neonatal intensive care unit (NICU)from different places with different
cultural backgrounds.

3.3 STUDY Population

The population for this study was made up of all neonates admitted at Mpilo Central Hospital
during the period of 2018-2020. This is due to the fact that, there are those who experience
problems during delivery and extra care maybe needed. It also involved health workers
because under normal circumstances they are the ones who provide maternal services during
pregnancy, delivery and after delivery. For this reason it was believed that health workers
have information on the whole issue of neonatal deaths.

3.3.1 Population to be targeted

The study’s target demographics were all neonates from 0-28 days who died between the
years of 2018 and 2020. Neonatal fatalities discovered during the audit were treated as
cases. Neonate’s parents socioeconomic variables including education and marital status, as
well as biological parameters like age, gravida, and obstetric history, were identified as risk
factors for the case-control study. The data used for the cases and controls for time-dependent
Methodology 18

characteristics such gestational age , birth attended mode of delivery , booking status, and
neonatal-related factors were those reported at the time of admissions at NICU.

3.3.2 DATA SETTING

3.4 LOGISTIC REGRESSION

Logistic regression is a statistical model that uses a logistic function to represent a binary
dependent variable in its most basic form, though there are many more advanced variants.
Logistic regression (or logit regression) is a technique for estimating the parameters of a
logistic model in regression analysis (a form of binary regression). Logistic regression is a
semi-parametric modeling, it estimates the conditional probability of a given event with re-
spect to some independent variable Xi.It allows you to claim that the existence of a risk factor
enhances the likelihood of a specific result by a certain amount. In statistics, odds represent
the probability that an event will occur, whereas odds against indicate the probability that it
will not. Assume Pi is a linear function of xi in the following way:

P (Y = 1|X = x) (3.1)

We’ll call log(Pi ) a linear function of x because Pi must be between 0 and 1, and linear func-
tions are unbounded:

Pi = β0 + β1 Xi +  (3.2)

Because logarithms are only bounded in one direction, we apply the logistic or logit transfor-
mation to modify log(Pi ) in an unbounded direction:

Pi
log( ) = β0 + β1 Xi + 
1 − Pi

log(Pi ) = β0 + β1 Xi +  (3.3)

Solving Pi results
Methodology 19

eβ0 +βi Xi
Pi = (3.4)
1 + eβ0 +βi Xi

where:
Pi is the probability that a death will occur for the ith individual.
βθ is the intercept.
βi is the regression coefficient for the explanatory variable Xi .

3.4.1 PARAMETER ESTIMATION

The binomial distribution is used to estimate in the univariate situation. For a single subject
with covariate values x1i ; x2i ; · · · xni . The likelihood function is:

n
i
Y
π(Xi )y i(1 − π(Xi ))1−y (3.5)
i=1

The number of trails is represented by yi . As in the uni-variate situation, we must maximize

this likelihood function to obtain estimates of the unknown parameters. To build a set of
(p + 1) equations in (p + 1) unknowns, we perform the standard processes of calculating the
logarithm of the likelihood function, taking (p + 1) partial derivatives with respect to each
parameter, and setting these (p + 1)equations equal to zero. The greatest likelihood equations
are obtained by solving this system of equations.

3.4.2 MODEL ASSUMPTIONS

Because the model implies that P (Y = 1) is the likelihood of an event occurring, the de-
pendent variable must be coded accordingly. Only the relevant variables should be included
in the model (neither overfitting nor underfitting should occur). The error phrases should
be self-contained. Each observation must be independent in order to use logistic regres-
sion. That is, no data points should come from a dependent samples design. Multicollinearity
should be minimal or absent in the model. This implies that the independent variables should
be unrelated to one another. The linearity of independent variables and log chances are both
assumed in logistic regression. While it is not necessary for the dependent and independent
variables to be linearly related, it is necessary for the independent variables to be linearly
Methodology 20

related to each other.

3.4.3 DEPENDENT VARIABLE

Otherwise, the test undervalues the strength of the link and rejects it too quickly, resulting
in the association being deemed insignificant (not rejecting the null hypothesis). It should
be significant in some way. The categorization of the independents is one answer to this
challenge variables. This entails converting metric variables to ordinal levels and then in-
corporating them into a model the prototype. The dependent variable has only 2 outcomes ,
therefore the logistic regression technique was implemented to model the study. Y i repre-
sents maternal death as follows:




1 if Y es


Yi = (3.6)



0
 if N o

3.4.4 Explanatory Variables

These are independent variables that influence a dependent variable’s result. The researcher
will collect all significant neonatal mortality indicators in this study and utilise them to fore-
cast the probability of survival for newborns. There will be explanatory variables which are
gestational age, birth weight,booking status,mode of delivery,diagnosis at birth,color of am-
niotic fluid,amniotic fluid odor,maternal age,maternal educational level,ANC visits and Birth
Attended. The variables can be modeled using the vector that follows:

Zi = [x1 , x2 ..., x11 ] (3.7)

3.4.5 STEPWISE REGRESSION

Stepwise regression is a type of regression model fitting in which the selection of predictive
variables is done automatically. Each phase considers a variable for inclusion or deletion from
the set of explanatory variables based on a predetermined criterion. It’s commonly done with
a series of F-tests or t-tests, although it can also be done with alternative methods. Backward
elimination, forward selection, and bi-directional elimination are the three different methods
Methodology 21

of stepwise regression. Backward elimination will be employed by the researcher.

3.4.6 BACKWARD ELIMINATION

It entails starting with all candidate explanatory variables, testing their absence using a
chosen model fit criterion, deleting the variable whose failure results in the least statistically
significant deterioration of the model fit, and repeating the process until no more variables
can be deleted without causing a statistically significant loss of fit.

3.4.7 Backward Elimination has several steps.

• We begin by selecting all variables as potential predictors and including them in the
regression equation.

• For each of the variables left in the regression equation, we compute partial F statistics.

• If this F is less than 5%, we eliminate it from the model and return to step 2.

• We’ll keep going until no partial F is located that’s low enough.

3.5 COX REGRESSION

The Cox proportional-hazards model (Cox, 1972) is a regression model that is often used
in medical research to look into the relationship between patient survival time and one or
more predictor factors. The hazard function, indicated by h, expresses the Cox model (t).
In a nutshell, the hazard function represents the danger of dying at time t. The dependent
variable was the amount of time spent in the hospital, and it was compared to the effects of
many factors that influence neonatal death.

3.5.1 ASSUMPTIONS

Individuals should have independent survival times, a stable hazard ratio throughout time,
and a multiplicative relationship between independent variables and the hazard. The base-
line hazard function might be positive or negative, but it cannot be negative. To ensure that
the hazard is positive, an exponential function of the covariates is utilized. In the Cox Model,
there is no intercept (any intercept would be incorporated into the baseline hazard).
Methodology 22

3.5.2 Specifications for the Model

The key determinants impacting the time of survival of preterm newborns will be modeled
using proportional hazard regression. A row vector Zp = [X1 ; X2 ; · · · ; Xp ] will be used to
display the factors. The regressors will be discovered through the use of a logistic regression
model to determine the elements that are important in neonate survival. The hazard function
will be as follows:

h(t; xi ) = h0 (x)exp(β1 x1 + β2 x2 + · · · + βp xp ) (3.8)

where:

• The number of regressors (x1 , x2 , ..., xp ) is given by xi .

• h0 is the default function, with no regressor effects (exp(0) = 1).

• βi represents the proportional effect of the various elements.

• The relative risk ratio is exp(βixi ).

the linearized equation will give:

log (h(txi ))
= β 1 x1 + β 2 x2 + · · · + β p x p (3.9)
log (h0 (t))

log (h(txi ))
HR = (3.10)
log (h0 (t))

The hazard ratio (HR) is the function. A positive Bi indicates that the component is a predic-
tor of neonatal survival time. The hypothesis test will be used to examine the survival times
of preterm newborns from two different groups. Against the alternative hypotheses of signif-
icance between the two independent groups at the 95% significant level, the null hypothesis
will reveal that there is no significant difference between the two groups.

3.5.3 PARAMETER ESTIMATION

Without the base line hazard function (h0(t)), partial likelihood will be employed to derive
the values of Bi at a particular time period of t1 t2 t3 · · · tk. The partial likelihood is calculated
as follows:
Methodology 23

3.5.4 HAZARD FUNCTION

This graph depicts the probability of surviving to the time interval (t; t + δt), or the likelihood
that if you survive to t, you will succumb to the event in the following moment:

P (t ≤ T ≤ ∇t/T ≥ t
h(t) = lim (3.11)
x→0 ∇t

f (t)
h(t) = (3.12)
S(t)
S(t) = 1 − F (t) (3.13)
S 0 (t)
h(t) = (3.14)
S(t)

The cumulative function will be:

R
h(t)dt = −ln(S(t))

3.5.5 THE HAZARD RATIO

The hazard ratio (HR) is a measure of the relative survival experience of two groups that was
established expressly for survival data. It calculates the overall difference between the two
survival curves. The HR will be used to compare different groups of neonates who have ex-
perienced neonatal death. The following hypotheses will be tested in a hypothesis test: H 0:
There is no statistically significant difference in the risk of neonatal death across the various
exposed groups. H 1:There is a considerable variation in terms of the risk of neonatal death
amongst the various exposed groups. The following formula will be used to compute the ratio:

h0 exp( pi=1 βi x∗ )
P
HR = (3.15)
h0 exp( pi=1 βxi
P

This will result to:

p
X
HR = exp( β i x∗i ) (3.16)
i=1

3.5.6 KAPLAIN MEIER ESTIMATES

Kaplan-Meier the survival function s(t), which is the likelihood of surviving beyond time t,
is estimated using a non parametric method called survival analysis. It is used to calculate
Methodology 24

the survival time of newborns and to create a survival curve to compare neonates’s survival
experiences based on categorical characteristics. The survivorship function’s Kaplan-Meier
estimator is defined as:

(3.17)

where t ≥ 0
Survival function

n1 − d1 n2 − d2 nn3 − d3
S(t) = ( )( )( ) (3.18)
n1 n2 n3

Y di
Ŝ = (1 − ) (3.19)
ni
ti ≤(t)

where:

• the set of m is ti , ..tm represents death times observed in the sample.

• di is the number of deaths at(ti ).

• ni is the number of individuals at risk before ith death time.

Survival curves can be generated using the survivor-ship function. Curves are a type of curve.
The survival time related to any proportion of the population is plotted and can be used to
compute the survival time. the instance The number of moms who survive each period is
used to compute the survival probability. The number of people who survive is divided by the
number of people who are at risk. Women who have died are dropped from the program or
are never contacted. Women who are lost are regarded ”censored” and are not counted as ”at
risk” at this time. are not included in the numerator The likelihood of surviving to any point
is calculated. calculated as the cumulative probability of surviving each of the preceding time
intervals the sum of the possibilities that came before it). Despite the fact that the probability
computed at any particular time Because of the short size of the interval, it is not particularly
accurate.
Methodology 25

3.5.7 LOG RANK TEST

. To compare the survival times of two groups, the Kaplan-Meier survival curves for each
are produced first, and then the Log rank test is used to compare them formally. The null
hypothesis is that there isn’t any data.
The disparity between groups can be stated as the median survival times of the two groups.
There are two equal groupings. The test is given the term Log rank since it is related to
another test.
A statistical test that use the logarithms of the data’s ranks. A test statistic is provided.
by:
(OA − EA )2 (OB − EB )2
χ2logrank = +
EA EB

3.5.8 CONFIDENCE INTERVAL

In our research, we’ll utilize a confidence interval with a 95% level of significance. A 95%
confidence interval (CI) is a set of numbers within which we may be 95% confident that the
true population mean is contained. For the CI variances, we’ll utilize Greenwood’s formula.

X di
V ar[S(t)] = S(t) (3.20)
ti ≤t
ni (ni − di )
s
di
Lowerlimit = S(t) − 1.96S(t) (3.21)
ni (ni − di )
s
di
U pperlimit = S(t) + 1.96S(t) (3.22)
ni (ni − di )

3.6 CENSORING
Censoring occurs when we have some knowledge regarding a person’s survival time but not
the exact time. The subject has been suppressed in the sense that nothing has been observed
or learned about it since the censoring. After the observation period ends, a censored subject
may or may not experience an occurrence. this is happens when:

• a patient drops out of the study.

• The study has a fixed time-line and the event occurs after the cut off time.

• a candidate withdraws from a study.

Methodology 26

The only observable information we have is that, at the end of 28 days, a subject ’neonate’ did
not experience the event of death. Hence , the survival time of the neonate is censored. This
means that until the censoring event occurred a child did not experience the event (which in
case is death).

3.7 RATIOS

3.7.1 ODDS RATIO

An odds ratio (OR) is a statistic that quantifies the strength of the association between two
events.The odds ratio is the ratio of two sets of odds: the chances of an event happening in
an exposed group versus the chances of an event happening in a non-exposed group. Case-
control studies are frequently reported using odds ratios. The odds ratio is a tool that may
be used to determine how probable an exposure is to result in a given occurrence.Odds are
the ratio of the probability of an event occurring in a group, divided by the probability of that
event not occurring, it is given by the following formula:

π
Odds = (3.23)
1−π

3.7.2 Relative Risk Ratio

Relative risk is a ratio of two probabilities: the likelihood of an event in one group versus the
probability of an event in the other group after exposure to a risk variable. For comparison of
risks between groups, the ratio of risks, or the relative risk, is a statistic of choice. Formally,
if π1 is the probability of the event in group 1, and π2 is the probability of the event in group
2, then the relative risk is:

π1
RR = (3.24)
π2
Methodology 27

the Relative Difference (RD) is given by:

RD = π1 − π2 (3.25)

A relative risk ratio if its≥1 indicates an exposure to be harmful, while a less than 1 indicates
a protective effect.

3.7.3 CONCLUSION

The methodology employed in the study to attain the study’s goals was detailed in detail in
this chapter. To create a clear image of the study’s objectives, the concerns linked to research
methodology were explained.
Chapter 4

Data Analysis

4.1 Introduction
All data analysis was used using Stata. The results obtained from the analysis will be ex-
plained in relation to the study objectives. The purpose of the study is to explore the contrib-
utory factors in neonatal deaths at selected district hospital which is Mpilo Central Hospital
with the aim of bringing down high neonatal death rate.

4.2 SOCIO-DEMOGRAPHIC AND ECONOMIC CHARAC-

TERISTICS
The table below gives an informative information on socio-demographic and economic char-
acteristics for patients who were admitted at MCH for 28days. Maternal age had 4 categories
and 15-25 had 42.5%, 25-29 contributed 23%, 30-39 had 32% and 2.5% were of mothers who
were 40 years and above. Maternal education had 5 categories since there were missing val-
ues for maternal education level which contributed 5.75, those who were not educated were
21%, those who ended up on primary level had highest percentage of 36.75, whereas 15%
ended up on secondary level, tertiary educated mothers were 21.5% of mothers admitted at
NICU. From the observed data 25.75% were married, 11.75% were single, those who were
cohabiting contributed 53% 0f the observed data however 9.5% that were recorded could be
Data Analysis 29

missing data, or widowed. In terms of husband employment status 41.85% are employed and
58.15% are unemployed. The analysis showed that 25.25% were referred from other places
or from clinics and 74.75% were not referral cases.

Table 4.1: Maternal Background

Variables Percentage%

MATERNAL AGE

15-25 Years 42.5

25-29 Years 23.00

30-39 Years 32.00

40 and above 2.50

MATERNAL EDUCATION

Unknown 5.75

None 21.00

Primary 36.75

Secondary 15.00

Tertiary 21.50

Marital Status

Married 25.75

single 11.75

Cohabiting 53.00

other 9.50

Husband Employment Status

Employed 41.85

Not Employed 58.15

Referral

Referred 25.25

Not Referred 74.75

4.2.1 PERCENTAGE FREQUENCIES OF THE VARIABLES

The percentage frequencies of the variables utilized in this study are listed below, with regard
to whether or not the patient experienced neonatal death. There is a 50/50 split between the
two binary cases, i.e. 50% neonatal death cases and 50% censored observations, based on the
data used. The percentages for each of the 2 gestational age groups are as follows: We had
65.6% neonatal deaths of those who were preterm and 34.4% censored observations for those
Data Analysis 30

who were ≥37weeks. Regarding birth attended those who were helped Doctors, Nurses or
midwifes and survived neonatal death were 59.98% and 46.08% respectively, and those who
were attended by traditional healer or other(who didn’t deliver at the hospital) and survived
76.19% neonatal death were 43.02% , of the collected data. Neonates who were diagnosed
as preterm 56.19% survived, those who were born of low birth weight i.e ≥2500g, 68.80%
did not survive neonatal period were as 54.8% survived with birth weight ≥2500g. 59.59%
babies via NVD did not survive and the remaining 40.41% survived, 63.95% did not survive
C-section the remaining 36.05% survived, and 3.66% born via other methods such as breech,
vacuum, and forceps did not survive neonatal period whereas 96.34% managed to survive.
According to the data, 63.9 percent of patients had scheduled their pregnancies, while the re-
maining 36.1 percent had not scheduled their pregnancies with any hospital or clinic. 31.35%
who booked did not survive and a high rate of observed data of 81.76% who did not book did
not survive. In terms of maternal educational level 39.13% of the unknown maternal educa-
tional level, 54.76% of those who are not educated, 58.50% who had educational level, 80% of
those who attained secondary level and 12.79% did not survive. The variable diagnosis had
8 categories, LBW 31.2%, 34.4% PROM/Preterm,54.22% Intrapartum asphyxia,53.48% An-
tepartum haemorrhage,17.5% Injection (sepsis, pneumonia and meningitis), 17.5% HIE,0%
Intrauterine growth retardation and 14.29%Congenital abnormality survived the listed di-
agnosis. Neonates who were born in a foul smelling amniotic fluid odor and did not survive
were 50.86% whereas the remaining 49.14% survived, however 48.81% born in a environment
without foul smelling did not survive, the remaining 51.19 survived being born without any
foul smelling amniotic fluid odor.
Data Analysis 31

Table 4.2: Percentage Of Frequencies.

Variable Died %Died Alive %Alive Total

Gestational Age

<32 Weeks 82 65.6 43 34.4 125

≥ 32W eeks 118 42.91 157 57.09 275

Birth Attended

Doctor 37 43.02 49 56.98 86

Nurse/Midwife/Equivalent 158 53.92 135 46.08 293

Traditional Birth Attended or Other 5 23.81 16 76.19 21

Mode Of Delivery

NVD 87 59.59 59 40.41 146

C-Section 110 63.95 62 36.05 172

Other 3 3.66 79 96.34 82

Booking Status

Booked 79 31.35 173 68.65 252

Not Booked 121 81.76 27 18.24 148

Maternal Educational Level

Unknown 9 39.13 14 60.87 23

None 46 54.76 38 45.24 84

Primary 86 58.50 61 41.50 147

Secondary 48 80 12 20 60

Tertiary 11 12.79 75 87.21 86

Diagnosis

LBW 183 68.80 83 31.20 266

PROM/PRETERM 82 65.6 43 34.4 125

Intrapartum asphyxia 168 45.78 199 54.22 367

Antepartum haemorrhage 174 46.52 200 53.48 374

Injection (sepsis, pneumonia and meningitis) 160 44.44 200 55.56 360

HIE 198 82.85 141 17.15 239

Intrauterine growth 17 100 0 0 17

Congenital abnormality 24 85.71 4 14.29 28

Amniotic fluid odor

Foul smelling 118 50.86 114 49.14 232

No foul smelling 82 48.81 86 51.19 168

Data Analysis 32

4.3 LOGISTIC REGRESSION ESTIMATION

Variables
The following variables are used throughout the analysis and construction of the logistic re-
gression model with Y being dependent variable and the X j being independent variables for
j = 1;2;...;11 as indicated below:

• Y represents the probability of default which is a dichotomous variable (survival status).

• X1 represents neonate’s gestational age

• X2 represents booking status

• X3 represents mode of delivery

• X4 represents diagnosis at birth

• X5 Represents ANC visits

• X6 represents amniotic fluid odor

• X7 Maternal educational level

• X8 represents color of amniotic fluid

The function for the likelihood of default is given by the following equation, the resulting
model using the variables specified above will be:

eβ0 +β1 X1 +β2 X2 +...+β8 X8

Pi = (4.1)
1 + eβ0 +β1 X1 +β2 X2 +···+β8 X8

As a result, the regressor co-efficients β0 ; β1 ; β2 ;...; β8 must be estimated using maximum

likelihood estimation. Stata, a statistical package, was used to accomplish this, as explained
below.
Before beginning any study, it was necessary to verify for sample bias and data imbalance.
The table below presents a summary of the data. It can be seen that half of the observations
experienced the event of interest, death, while the other half was suppressed.
Data Analysis 33

Table 4.3: Sampling.

Neonatal Survival Status Frequency % Cumulative%

Yes 200 50 50

No 200 50 100

Total 400 100

4.3.1 Testing for Multicollinearity

VIF
The researcher used the statistical software Stata to test for multicollinearity and used Vari-
ance Inflated Factor to do so. The VIFs of all of the possible explanatory variables were cal-
culated. Multicollinearity values larger than 3 were found for birth weight and LBW, which
were 7.55 and 7.58, respectively. It’s possible that this is due to the fact that birth weight and
neonates born with low birth weight are linearly associated. The researcher excluded birth
weight and computed VIFs which were less than 3 showing that there is no multicollinearity:
results are shown in the table below:

Table 4.4: VIF

Variable VIF

Booking 2.02

ANC Visits 1.74

Mode 1.36

LBW 1.12

Gestational age 1.04

Previous pregnancy loss 1.03

Amniotic fluid odor 1.01

Mean 1.33

4.3.2 Model Classification

The table shows that there were 200 neonates who survived and 200 who died due to neonatal
death. According to the findings, the model correctly predicted 157 of the 200 people who
died. The accuracy rate for forecasting neonatal death was discovered to be 82.71 percent,
indicating that the model is good. It was discovered that 173 of the 200 babies who survived
were accurately predicted, yielding a 86.50 percent accuracy rate, indicating a good model.
Data Analysis 34

The overall rate of proper classification, which is an indicative of a good model, is predicted
to represent model 82.71 percent of the variation in bad neonatal outcome.

Table 4.5: Classification Results

-True-

Classified D -D Total

+ 157 27 184

- 43 173 216

Total 200 200 400

Table 4.6: Classification Percentage

Sensitivity P r(+kD) 78.50%

Specificity Pr (−k D) 86.50%

Positive predictive value P r(Dk+) 85.79%

Negative predictive value P r( Dk−) 64.86%

False +rate for true D Pr (+k D) 80.09%

False - rate for true D Pr (−kD) 13.07%

False + rate for classified Pr ( Dk+) 21.50%

False – rate for classified Pr (Dk−) 14.21%

Correctly classified 82.50%

4.3.3 Likelihood Test

Model evaluation also entails comparing a created model to the intercept alone model to see
if it provides a better fit to the data (the null model). The likelihood ratio, score test, and
Wald test are the three tests that can be used in general, however the likelihood ratio test
was chosen because it was recommended by (Menard (2000)). Because the p value is less
than 0.05, the difference is statistically significant. This indicates that, as compared to the
null model, the model with many parameters accurately predicts the output, and hence the
estimated model provides a better fit for the data.

4.3.4 Model Estimation

The co-efficients were computed using Stata statistical software, and the fitted logistic re-
gression model was summarized in the table below. The table has six columns, one of which
is called co-efficient and contains the estimated value of the coefficient of an independent
Data Analysis 35

variable (β1;β2;..;β12). These are the outcomes of the likelihood estimation method, which
calculates co-efficients to maximize the amount of data accessible (maximally likely esti-
mates). The logistic regression coefficients show how a one-unit increase in the predictor
variable affects the result’s log probability. The table also includes the standard errors of the
co-efficients, as well as their p values and the z statistic (sometimes known as the Wald z
statistic).In the last column, the odds ratio is determined using exp(i), where I is the regres-
sion co-efficient estimate for the variable Xi . The intercept shows the average log of neonatal
death probability when all explanatory factors are set to zero. As a result, the odds ratio of
0.0286898 in the Table depicts the probability of newborn mortality in any neonate, regard-
less of explanatory variables. The score is slightly higher than zero and positive, indicating
that everyone has a natural possibility of dying as a newborn. To convert odds ratios to prob-
abilities and create a probability forecast, using the formula below:

odds
P robability = (4.2)
1 + odds

Substituing odds values to the formular above

0.0286898
P robabilty = (4.3)
1 + 0.0286898

Without any explanatory variables, offers a probability of 0.0286898, indicating that any pa-
tient will experience neonatal death 2.87 percent of the time. Each explanatory variable in
Table is addressed in detail below, with three aspects determining the meaning of each coeffi-
cient: the sign of the coefficient, which indicates the direction of the relationship, the degree
of the correlation, and finally the statistical significance of the estimate. A consideration of
statistically unimportant estimation abilities will also be discussed:
Data Analysis 36

Table 4.7: Parameter estimates for the fitted model

Co-efficient Standard Error Z=Value P-Value Odds ratio

Gestational age

<37weeks 3.93784 0.5690169 4.15 0.001 2.537249

Booking

Not Booked 12.59539 1.991516 7.54 0.00 15.0847

Mode of delivery

C-Section 3.3968801 7.492876 3.89 0.005 11.79661

Other 1.30873 1.925332 -2.30 0.021 8.311248

Diagnosis

LBW 1.617851 0.2551148 5.85 0.00 2.061856

Preterm 3.972753 0.5690169 4.15 0.013 2.537249

Intrapartum asphyxia 7.44879 14.38462 13.15 0.000 14.38462

Antepartum haemorrhage 1.371321 0.679362 2.55 0.011 5.646673

Injection 1.731067 0.679362 3.00 0.003 5.646673

HIE 4.586238 1.222669 -4.71 0.00 5.557377

Visits 0.9693893 0.1948491 5.71 0.00 2.447413

Amniotic fluid odor

Foul Smelling 4.348837 3.330466 1.92 0.005 19.50959

Maternal Education

Tertiary -2.90335 0.860544 -3.41 0.001 0.1607143

Color of Amniotic fluid

Meconium stained 2.78812 0.1913415 2.48 0.013 3.166072

4.3.5 Variables In The Model

• Gestational Age:With a p value of 0:001, the variable was determined to be significant.

The co-efficient has a positive value of 3.93784, as shown in Table, indicating that there
is a positive association between the number of gestational age and the log of the odds of
neonatal mortality. ANC visits have an odd ratio of 2.537249 and a confidence interval
of (1.774056 ; 2.708406), which suggests that the risk of neonatal mortality increases
by a factor of 2.537249 for every unit increase in the number of gestational age, with all
other factors held constant.

• Booking: In our model, the booking status was identified as a categorical variable. The
positive co-efficient of 12.59539 indicates that booking status and newborn death have
a positive association. Using unbooked moms as a control group, results revealed a
Data Analysis 37

15.0847 odds ratio for those who booked a pregnancy, those who did not and those who
were referred to MCH , with a 95% confidence interval of ( 3.142814 ; 13.0847). Booking
a pregnancy appointment is essential for getting frequent doctor visits.this shows that
mothers who did not book are 15.08 times at risk.

• Mode of delivery: It had three categories that were dummy coded. Other modes of
delivery, such as suction, breech, and others, have a favorable connection, according to
the estimate 3.3968801. The confidence intervals were calculated as follows: (0.4872773
; 9.438204). It demonstrates that a mother who gives birth via alternative ways of
delivery has an 8-fold chance of dying, whilst those who give birth via C-Section have
an 11.8-fold chance of the neonate not surviving the neonatal period.

• Diagnosis: With eight categories, this variable was dummy coded. The study found that
a neonate’s chance of neonatal death increases if they are diagnosed with any of the
eight conditions.For LBW, the co-efficient of 1.617851 and CI of (0.5742934 ; 2.558935)
indicated that neonates born underweight are at risk, and the odd ratio of 2.061856
indicates that they are two times more at risk.Preterm neonates had a co-efficient of
3.972753 and a confidence interval of (1.648264; 3.972753), indicating that they were
born before 37 weeks. The odd ratio of 2.537249 indicates that they are 2.5 times more
vulnerable. The co-efficient for intrapartum asphyxia was 7.44879, with a 95% confi-
dence interval of (9.666268;21.40611), indicating that neonates born with asphyxia are
at risk. They are fourteen times more at risk, according to the odd ratio of 14.38462.
The co-efficient for antepartum haemorrhage was 1.371321, with a 95% confidence in-
terval of (2.991356 6.45931), indicating that newborns with bleeding are at risk. They
are 5.6 times more at risk, according to the chances ratio of 5.646673.The co-efficient for
neonates who had an injection for sepsis, pneumonia, or meningitis was 1.731067, with
a 95% confidence interval of (4.730617; 9.848578), indicating that newborn infections
constitute a risk. They are 5 times more at risk, according to the odd ratio of 5.646673.
For neonates with HIE, the co-efficient was 4.586238, with a 95% confidence interval of
(4.231405; 7.298862). They are five times more at risk, as evidenced by the odd ratio of
5.557377.

• Visits:With a p value of 0:00, the variable was judged to be significant. The co-efficient
has a positive value of 0.9355478, indicating that there is a positive link between the
number of antenatal clinic visits and the log of odds of newborn death, as shown in
Data Analysis 38

Table. A confidence interval of 2.447413 and an odd ratio of 2.447413 with confidence
interval of (1.6460905 3.0661562), odds ratio therefore confirm that neonates are 2 times
at risk.

• Amniotic fluid odor : The estimate is positive co-efficient of 4.348837 which shows that a
neonate born in a foul smelling environments increases the log odds of neonatal death.
It was shown from the analysis that the confidence interval was (7.298862;20.80848 )
with odds ratio valuing 19.50959 meaning that a foul smelling environments are 19
times at risk of adverse outcomes.

• Education level: The estimate is negative (-2.90335), indicating that better educated
moms have a lower log of likelihood of newborn death. The study revealed that the con-
fidence interval was ( -3.593838 ; -0.6733061). When all other things are maintained
constant, the probability of dying lowers by a factor of 0.1607143 for moms with ad-
vanced education.

• Color of amniotic fluid: The study found that a neonate’s chance of neonatal death
increases if they are given birth in a meconium stained environment.For meconium
stained fluid, the co-efficient of 2.78812 and CI of (1.073115;1.832135) and the odds
ratio of 3.166072 indicates that they are 3 times at risk of adverse outcome.
Data Analysis 39

4.3.6 Variables Not In The Model

Some of the variables in our data were found to be non-significant in our model, thus they
were left out at the 95 percent level of significance. The p-values obtained are listed in the
table below.

Table 4.8: Parameter estimates for the fitted model

Variable P-Value

Parity 0.093

Education(Primary) 0.188

Education(Secondary) 0.087

Gestational Age(≥ 37) 0.67

Mode of delivery(NVD) 0.506

Birth Attended(Doctor) 0.076

Birth Attended(Nurse/Midwife/Equivalent) 0.113

Birth Attended(Traditional Birth Attended or Other) 0.197

Color of amniotic fluid odor(Clear) 1

Amniotic fluid odor(No foul smelling) 1

4.4 Goodness Of Fit Model

4.4.1 Hosmer Lemeshow

Before examining the system’s outputs, we should first examine the model’s significance. To
determine the model’s goodness of fit, the Hosmer and Lemeshow test is utilized. The good-
ness of fit is a metric that indicates how closely the data matches the model. The test is
calculated using the following formula:

The Hosmer and Lemshow test was used to determine the goodness of fit under the following
hypothesis:
Data Analysis 40

Table 4.9: Hosmer Lemeshow Results

Number of Observations 400

Number of groups 12

Hosmer - Lemeshow Chi(10) 13.71

2
Pro  χ 0.2904

H0 :There exists no significant difference in the model predicted values and observed.
H1 :There is a significant difference in the model predicted values and observed.

The Hosmer and Lemshow tests review the test and explain the model’s goodness of fit based
on the P values. The test yielded a p=0.2904, which is greater than the 0.05 significance level,
demonstrating insignificance and implying a good fit. This means that the values predicted
by the model and the values observed were identical.

4.5 Cox Proportional Hazards.

The results of logistic regression were further analyzed using Cox regression analysis. The
variables that were found to be significant in mortality will be investigated further to see if
they have any bearing on survival.

4.5.1 Model Significance.

Before we can assess our findings, we must first determine the model’s significance. The null
hypothesis indicates that the model fits the data; nevertheless, if the model’s p-value is less
than 0.050, the null hypothesis that the data fits the model is not rejected. Our model’s p-
value was 0.000, indicating that it fits the data and may be used to predict survival.
Data Analysis 41

4.5.2 Significant Factors

Table 4.10: Cox Regression

Variable β Hazard Ratio Standard error Z-Value P-Value

Color of amniotic fluid odor

Clear 1 1 1 1

Meconium Stained 1.257936 4.032255 0.1041368 0.753 0.000

Diagnosis

LBW 1.1067 2.001317 0.3265241 4.25 0.000

Preterm 0.7181 1.210371 0.2155868 1.07 0.026

Intrapartum asphyxia 1.1274 0.5509994 0.1405756 -5.15 0.000

Antepartum haemorrhage 0.8805 0.299432 0.0701162 -1.98 0.048

Injection (sepsis, pneumonia and meningitis) 1.0913 6.0400158 0.1444206 2.86 0.004

HIE 0.8618 7.861338 5.663276 2.86 0.004

Intrauterine growth retardation 1.2530 0.6034178 0.1746622 -1.75 0.003

Congenital abnormality 0.4465 0.6603861 0.1726005 -1.59 0.112

Apgar score 1.85747 2.535006 0.4022206 5.86 0.000

ANC Visits 0.8534 0.9149255 0.0424408 -1.92 0.05

Parity 3.60563 1.121872 0.0424764 3.04 0.002

Mother’s educational level 4.38536 1.162815 0.0660404 2.66 0.008

The system’s output is shown in the table above. The variables used in the analysis were
gestational age, booking status, mode of delivery, diagnosis, number of ANC visits, and amni-
otic fluid odor. We do not reject the null hypothesis that the variable has an association with
survival time if the p-value is less than 0.05. Age, mode of delivery, and educational level all
had p values greater than 0.05, so we conclude that they have no relationship with patient
survival time. Since all the variables p values are less than 0.05 we conclude that all the
stated variables have a therefore we conclude that it has no relationship with the neonate’s
survival time. The hazard function is as follows:

h(t) = h0 exp(1.26x1 + 1.11x2 + 0.72x3 + 1.13x4 + 0.88x5 + 1.09x6 + 0.86x6 + 1.25x7 + 0.45x8 )
(4.4)

• x1 represents LBW

• x2 represents Intrapartum asphyxia

Data Analysis 42

• x3 represents Antepartum haemorrhage

• x4 represents injection

• x5 represents HIE

• x6 represents Intrauterine growth retardation

• x7 represents congenital abnormality

• x8 represents ANC visits

4.5.3 Hazard Ratio

Color of amniotic fluid odor: The color of amniotic fluid was divided into two categories: clear
and meconium stained. For meconium stained, the algorithm calculated a hazard ratio of
4.032255. When all other variables are held constant, neonates born in a meconium-stained
environment had a 4.03 higher likelihood of experiencing an incident than those born in a
clear environment. The graph below demonstrates that neonates born in a meconium-stained
environment have a poorer survival rate than those born in a clear environment.

ANC Visits: For the number of ANC visits that a patient had, the algorithm calculated a
hazard ratio of 1.638377. Mothers who do not attend ANC visits have a 1.6 times higher
likelihood of placing their unborn baby at risk of neonatal death than those who go to the
clinic for prenatal checkups. ANC visits are beneficial to a patient’s and the unborn baby’s
health because doctors perform checkups to ensure the mother and the unborn baby are in
good health.

Apgar score:When compared to neonates with low apgar score which is <7 with those with
apgar score ≥ 7, those with low apgar score had a 2.55 higher likelihood of experiencing
neonatal death. Apgar score was found to be significant in predicting the survival of neonatal
death, with a p value of 0.000.In general, neonates who experience problems during their
newborn period have a higher risk of dying than those who do not. This discovery can be
explained in the medical area by medical facts, which show that mothers from a low socioe-
conomic background contribute to an increase in infant death.For both the one minute and
five minute apgar tests, a score of 7 to 10 is considered normal. A score in this range often
Data Analysis 43

indicates that the baby is healthy and will only require routine post-delivery care.

Diagnosis: Neonates admitted with LBW have 2.0 greater chance to event compared to those
born with normal weight. The analysis showed that preterm neonates were significant in
determining survival of newborns with a p value of 0.026.Premature birth raises the risk
of infant death by 1.21 times.With a p value of 0.004, HIE was revealed to be important
in determining survival analysis. It demonstrates that born with HIE raises the chance of
neonatal death by 7.9.In general, neonates who have problems during birth have a higher
chance of neonatal death than those who do not. This observation in the medical profes-
sion can be explained by medical data, which show that newborns who experience problems
during birth have a high mortality rate. Neonates suffering from HIE and those induced
for injection (sepsis, pneumonia and meningitis) is very poor, followed by those with LBW,
whereas the survival of newborns who are prematures is better.

Parity: Parity has a 1.12 effect on adverse neonatal outcome, mothers with more children are
at high risk of having a baby not surviving the firstt 28 days of life.parity was found to be
significant in determining the survival analysis with p value of 0.002.
Data Analysis 44

4.6 Logrank Tests

4.6.1 Amniotic Fluid Odor

Using Logrank test we test the following hypothesis: Null Hypothesis: There is no significant
difference in survival functions of survival curves of Amniotic fluid odor.
Alternative Hypothesis: There is significant difference in survival functions of survival curves
of Amniotic fluid odor.
The system Log rank computations found the χ2 of 60.60 with one degree of freedom and a
p-value of 0.06, which is greater than 0.05 therefore we reject the null hypothesis that they
is no difference in survival functions of survival curves of amniotic fluid odor. In general this
shows that the survival curves foul smelling and with no foul smelling are not proportional.

4.6.2 Diagnosis

The Logrank test was used to test the following hypothesis:

Null Hypothesis: There is no significant difference in survival functions of survival curves of
diagnosis at birth.
Alternative Hypothesis: There is significant difference in survival functions of survival curves
of diagnosis at birth.
We do reject the null hypothesis, we therefore conclude that there is a significant difference
in survival functions of survival curves of diagnosis at birth since the system Log rank calcu-
lations yielded a value of 17.70 with seven degrees of freedom and a p-value of 0.047, which
is smaller than 0.05.

4.6.3 ANC Visits

The following hypothesis was evaluated using the Logrank test:

Null Hypothesis: The survival functions of the survival curves of the number of ANC visits
do not differ significantly.

Alternative Hypothesis: The survival functions of survivors differ significantly curves of num-
ber of ANC visits.

We reject the null hypothesis that there is no difference in survival functions of survival
Data Analysis 45

curves of number of ANC visits since the system Log rank calculations found a true value of
54.67 with one degree of freedom and a p-value of 0.00, which is less than 0.05. In general,
the survival curves for neonate’s mother who received antenatal care versus those who did
not receive antenatal care are not proportional.

4.6.4 Survivor Functions

Null Hypothesis: There is no significant difference in survival functions of gestational age

(survival functions of gestational age are equal). Alternative Hypothesis: There is a signifi-
cant difference in survival functions of gestational age (survival functions of gestational age
are not equal).

Gestational Age Events observed Events expected

<37Weeks 43 57.02

≥ 37W eeks 157 142.98

Total 200 200

4.6.5 Logrank Test For Equality Of Survivor Functions

Chi-square value for gestational age is 5.72 and p-value is 0.0167, since the p-value is less
than 0.05 we reject the null hypothesis (that the survival functions are equal). In general,
PROM/preterm life expectancy is not equal to full term neonates, therefore one can con-
clude that PROM/Preterm are more likely to experience neonatal death compared to full
term neonates.
Data Analysis 46

4.7 Ratios

4.7.1 Relative Risk Ratio.

Table 4.11: Probability among males and females.

Sex Died Survived Risk

Males 118 113 0.51

Females 82 87 0.49

The relative risk of death of men compared to women is:

π1 0.51
RR = = = 1.05 (4.5)
π2 0.485

Since RR ≥1 this means that gender also plays a role on adverse neonatal with 0.05%.

4.8 Odds Ratio.

Table 4.12: Odds for death among males and females, π denotes the probability of death.
Death Survival

Sex π 1-π Odds

Male 0.51 0.49 1.04

Female 0.49 0.51 0.96

the odds ratio value is the obtained from the following :

1.04
OR = = 1.08 (4.6)
0.96

Using the odds ratio on gender, gender plays a 0.08% on adverse neonatal outcome which is
greater than 0.05 therefore there is a signifance difference in gender.

4.8.1 Discussion of results

Meyer et al. (2021)Investigated maternal and intrapartum factors linked to a poor newborn
outcome in meconium-stained amniotic fluid births. A retrospective cohort study of all women
Data Analysis 47

with singleton pregnancies who had MSAF labor trials between 2011 and 2020. The out-
comes of deliveries with and without a negative neonatal outcome were examined. The fol-
lowing factors were shown to be independently linked with composite unfavorable neonatal
outcome in multivariable regression analysis. The findings of this study show that deliveries
complicated by MSAF and any of the following characteristics should be given extra atten-
tion: polyhydramnios, intrapartum fever, amnioinfusion, and pregestational diabetes. MSAF
causes pregestational diabetes, polyhydramniosis, fever, and amnioinfusion, according to this
study, and hence plays a role in poor birth outcomes. Elias et al, (2022) researched about
determinants of adverse neonatal outcomes, only liquor status and birth weight remained
statistically significant predictors of early neonatal outcomes in the multi-variable logistic re-
gression study. During vaginal examination, mothers with meconium-stained amniotic fluid
were nearly 6-times more likely to have unfavorable early newborn outcomes than mothers
with clear amniotic fluid. Low birth weight, on the other hand, raised the risk of unfavor-
able early neonatal outcomes by 14 times. This researcher concluded that children born in
a meconium-stained environment are more likely not to survive neonatal period. The stated
researches found that color of amniotic fluid odor plays a role in adverse neonatal outcome.
The results obtained from this research using logistic regression indicates that neonates are
3 times at risk of adverse outcomes.

Su et al,(2022) data was subjected to a retrospective study, at a 1:1 ratio, a propensity score
matching analysis was performed. Before matching, the male group had significantly higher
rates of neonatal respiratory distress syndrome, bronchopulmonary dysplasia (BPD), severe
intraventricular hemorrhage, periventricular leukomalacia, necrotizing enterocolitis, and patent
ductus arteriosus (P value 0.05), whereas after matching, the male group only had a signifi-
cantly higher rate of BPD than the female group (p value 0.05). Before and after matching,
there was no significant difference in the rate of survival at discharge between the two groups
(P≥ 0.05). However this analysis contradicts with the one that the researcher did using Ra-
tios, RR showed a value which ≥1 this means that gender also plays a role on adverse neona-
tal with 0.05%, using odds ratio as well gender plays a 0.08% on adverse neonatal outcome
therefore the researcher concluded that gender of the neonate plays a role. To account for
confounding factors in unfavorable neonatal outcomes, Tolossa et al, (2020) used bivariate
and multivariable logistic regression analysis. When the p-value was less than 0.05, statis-
tical significance was declared. Adolescent girls and women were less likely to receive ANC.
Babies born to adolescent mothers are more likely to have negative neonatal outcomes such
Data Analysis 48

as LBW and preterm birth than babies born to adult mothers.

Workineh et al, (2022) did study based on the multivariate logistic regression analysis,low
birth weight, preterm birth, and low Apgar score at fifth minutes were the major identified
adverse neonatal outcomes.The adverse neonatal outcomes and the risk factors identified in
this research have the potential to harm the health of the neonates. Thus, it needs emphasis
to tackle the problems and save the life of the newborn through better and strengthened ANC
follow-up, accesses to health care.

Abadiga et al,(2022) did a research about Infants with one or more unfavorable birth out-
comes are more likely to die and develop a variety of health and developmental issues. Age,
antepartum hemorrhage, abortion history, gestational age, anemia, and maternal malnutri-
tion were all factors that predisposed the woman to a poor delivery outcome.Advanced odds
ratios were used to conclude that poor birth outcomes were caused by factors such as low
antenatal care visits, anemia, preterm membrane rupture, pregnancy-induced hypertension,
a lack of food supplements, and physical maltreatment. Clinicians should play an important
role in antenatal care follow-up, counseling, and supplementing prescribed foods, as well as
reducing violence and abuse throughout pregnancy.Using hazard ratio the researcher found
that for the number of ANC visits that a patient had, the algorithm calculated a hazard ratio
of 1.638377. Mothers who do not attend ANC visits have a 1.6 times higher likelihood of plac-
ing their unborn baby at risk of neonatal death than those who go to the clinic for prenatal
checkups,therefore we agree that ANC visits play a role in adverse neonatal outcome. Turi
et al,(2020) advised that community- and health-care-based intervention programs be used
to prevent adolescent pregnancy and improve unfavorable neonatal outcomes in adolescent
girls. At a P-value of 0.05, multivariable logistic regression was utilized to find determinants
of adverse birth outcomes). Low antenatal care visits, anemia, preterm membrane rupture,
pregnancy-induced hypertension, lack of food supplementation, and physical maltreatment
were all factors that contributed to poor delivery outcomes. Clinicians should play a key role
in improving antenatal care follow-up, counseling, and supplementing suggested foods, as
well as reducing violence and abuse throughout pregnancy. To account for confounding fac-
tors in unfavorable neonatal outcomes, Tolossa et al, (2020) used bivariate and multivariable
logistic regression analysis. When the p-value was less than 0.05, statistical significance was
declared. Adolescent girls and women were less likely to receive ANC. Babies born to adoles-
cent mothers are more likely to have negative neonatal outcomes such as LBW and preterm
Data Analysis 49

birth than babies born to adult mothers.

Chapter 5

Results

5.1 Introduction
The overall aim of the study was to propose strategies to reduce the high neonatal mortality
rate in a low resource setting.The purpose of the study was to explore the factors contributing
to neonatal deaths at the hospital under study so as to reduce the neonatal mortality rate.
The research objectives that guided the study were to determine underlying contributory fac-
tors to neonatal deaths in an obstetric unit and propose strategies for midwifery practice in
order to prevent neonatal deaths.The study would be of significance to health care practition-
ers regarding the contributory factors which caused neonatal deaths and it would impact on
the training implications for personnel working in a maternity unit, resulting in improved
maternal and neonatal care and facilities.

5.2 Conclusions
Neonatal mortality is an essential health indicator that is widely recognized as a general indi-
cator of a population’s overall health, the status of infants in a society, and the healthcare sys-
tem’s functioning. Neonatal mortality is influenced by a variety of socio-demographic factors.
The first goal was to develop a logistic model that predicted neonatal death in neonates, with
all variables taken into account.The model was first assessed for multicollinearity between
Results 51

variables, and those that indicated collinearity were removed. The Hosmer and Lemeshow
test was used to determine the model’s goodness of fit, or how well the data fits the model.
According to Hosmer and Lemeshow, the logistic model fit well indicating that the model
predicted values and the observed.When the logistic regression model was fitted, several
covariates, including marital status and husband employment status, were shown to be sta-
tistically insignificant. However, in terms of their influence to neonatal mortality, gestational
age, mode of delivery, booking status, number of ANC visits, amniotic fluid odor, and diagno-
sis at birth were determined to be statistically significant. The Researchers discovered that
preterm babies have a greater mortality rate.The survival periods of neonates were modeled
using Cox regression analysis. The survival times of pregnant women in the hospital were
modeled using regression analysis. The researcher performed a model significance test, and
the results revealed that our model’s p-value was 0.000, indicating that it matches the data
and can be used to predict survival. ANC visits, amniotic fluid odor color, apgar score, and
intrapartum hypoxia were revealed to be significant predictors of neonate survival time in
the NICU.The Kaplain Meir survival curves were also utilized in neonatal survival analysis.
The analysis also revealed that women who schedule their pregnancies had a greater benefit
in terms of infant mortality. According to the findings, babies born with a low apgar score
are more likely to die. The study also used log ranks tests, which revealed that the survival
curves for mothers who attended ANC Visits are not proportional.

5.3 Recommendations
The purpose of this study was to determine, by record auditing, the underlying contributory
reasons in neonatal mortality in an obstetric unit and to provide methods for NICU prac-
tice to prevent neonatal deaths.The researcher provided the following recommendations for
improving neonatal treatment and future research based on the findings and conclusions of
this study.The conclusions drawn from this study supported the assumptions that there are
factors that contribute to neonatal deaths.The government should build NICU with proper
medical services wards in clinics in remote areas.Foe neonates with birth asphyxia Invasive
ventilation should be available. In other cases, due to terrible roads and a lack of transporta-
tion, women give birth on their route to larger hospitals, which causes issues because they
give birth without the presence of medical personnel. Neonatal mortality owing to delays in
assessing a health institution will be decreased if the government constructs roads in rural
areas and provides additional ambulances for referral centers to ensure they are available at
Results 52

all times. Because some women in rural regions do not book their pregnancies due to a lack of
cash or because they are unaware that maternity care are now free, the government should
conduct campaigns in rural areas to educate them on the subject.To the hospital Record au-
dits, perinatal morbidity and mortality review sessions are used to ensure quality. Motivating
health workers who perform deliveries and care for newborn newborns to receive training in
neonatal resuscitation and newborn care. The majority of the population is unaware of the
importance of scheduling their pregnancies. Booking lowers the danger of their infants by
keeping their mothers’ health in check by scheduling doctor’s appointments and taking ade-
quate medicines as needed. All women of reproductive age, especially those in remote regions,
must be educated on the necessity of scheduling pregnancies. Some people in rural areas be-
come pregnant and give birth at home instead of going to the hospital, which might be deadly
if difficulties emerge.To the hospital staff in maternity and neonatal units are required to be
able to resuscitate newborn babies. In caring for expectant moms and newborn babies, all
health workers must follow defined standards. After a patient’s examination, interpretation,
and management, all employees attending to them must keep adequate records.The risk of
infant death from intrapartum asphyxia can be reduced if hospitals have adequate intensive
care facilities. HIE effect on infants should be nursed on a radiant heat table, and LBW can
be minimized by consuming correct and healthful foods throughout pregnancy. To prevent
death, infants with sepsis, pneumonia, or meningitis must be given an injection and infection
control techniques must be closely followed. To reduce infant mortality, the government must
expand its maternity and child health care activities, particularly community-based health
care centers.
References

A BADIGA , M., M OSISA , G., T SEGAYE , R., O LUMA , A., A BDISA , E., AND B EKELE , T. (2022).
Determinants of adverse birth outcomes among women delivered in public hospitals of
ethiopia, 2020. Archives of Public Health, 80 (1), 1–17.

A BOU Z AHR , C. (2003). Global burden of maternal death and disability. British medical
bulletin, 67 (1), 1–11.

A DAM , T., L IM , S. S., M EHTA , S., B HUTTA , Z. A., F OGSTAD, H., M ATHAI , M., Z UPAN, J.,
AND D ARMSTADT, G. L. (2005). Achieving the millennium development goals for health:
Cost effectiveness analysis of strategies for maternal and neonatal health in developing
countries. BMJ-British Medical Journal-International Edition, 331 (7525), 1107–1109.

A KOMBI , B. J., A GHO, K. E., H ALL , J. J., WALI , N., R ENZAHO, A., AND M EROM , D. (2017).
Stunting, wasting and underweight in sub-saharan africa: a systematic review. Interna-
tional journal of environmental research and public health, 14 (8), 863.

A KOMBI , B. J., G HIMIRE , P. R., AND R ENZAHO, A. M. (2019). Factors associated with
neonatal mortality in the african great lakes region: A pooled analysis of national surveys.
Journal of Global Health Reports, 3, e2019043.

A LESHINA , N. AND R EDMOND, G. (2005). How high is infant mortality in central and eastern
europe and the commonwealth of independent states? Population studies, 59 (1), 39–54.

B ANG, A. T., R EDDY, H. M., D ESHMUKH , M. D., B AITULE , S. B., AND B ANG, R. A. (2005).
Neonatal and infant mortality in the ten years (1993 to 2003) of the gadchiroli field trial:
effect of home-based neonatal care. Journal of perinatology, 25 (1), S92–S107.

B ERHANU, B., O LJIRA , L., D EMANA , M., N EGASH , B., AYANA , G. M., R ARU, T. B., AND

H AILE , D. (2021). Survival and predictors of mortality among neonates admitted to neona-
tal intensive care unit at bombe primary hospital, southern ethiopia: Institution-based
retrospective cohort study. Pediatric Health, Medicine and Therapeutics, 12, 239.
REFERENCES 54

B UCCELLATO, C. A., S TIKA , C. S., AND F REDERIKSEN, M. C. (2000). A randomized trial of

misoprostol versus extra-amniotic sodium chloride infusion with oxytocin for induction of
labor. American journal of obstetrics and gynecology, 182 (5), 1039–1044.

C HAN, A. W., C HEW, D. P., B HATT, D. L., M OLITERNO, D. J., T OPOL , E. J., AND E LLIS,
S. G. (2002). Long-term mortality benefit with the combination of stents and abciximab
for cardiogenic shock complicating acute myocardial infarction. The American journal of
cardiology, 89 (2), 132–136.

C HOPRA , M., L AWN, J. E., S ANDERS, D., B ARRON, P., K ARIM , S. S. A., B RADSHAW, D.,
J EWKES, R., K ARIM , Q. A., F LISHER , A. J., M AYOSI , B. M., ET AL . (2009). Achieving
the health millennium development goals for south africa: challenges and priorities. The
Lancet, 374 (9694), 1023–1031.

D ESTA , M., A KIBU, M., T ADESE , M., AND T ESFAYE , M. (2019). Dietary diversity and asso-
ciated factors among pregnant women attending antenatal clinic in shashemane, oromia,
central ethiopia: a cross-sectional study. Journal of nutrition and metabolism, 2019.

D OMINE , V. (2011). The coming of age of media literacy. Journal of Media Literacy Education,
3 (1), 5.

E LIAS, S., W OLDE , Z., T ANTU, T., G UNTA , M., AND Z EWUDU, D. (2022). Determinants of
early neonatal outcomes after emergency cesarean delivery at hawassa university compre-
hensive specialised hospital, hawassa, ethiopia. Plos one, 17 (3), e0263837.

FATUSI , A. O. AND B LUM , R. W. (2008). Predictors of early sexual initiation among a na-
tionally representative sample of nigerian adolescents. BMC public health, 8 (1), 1–14.

F ILMER , D. AND P RITCHETT, L. (1997). Child mortality and public spending on health: how
much does money matter?, volume 1864. World Bank Publications.

G HEZZI , F., R AIO, L., G ÜNTER D UWE , D., C ROMI , A., K AROUSOU, E., AND D ÜRIG, P.
(2005). Sonographic umbilical vessel morphometry and perinatal outcome of fetuses with
a lean umbilical cord. Journal of Clinical Ultrasound, 33 (1), 18–23.

G OLDING, N., B URSTEIN, R., L ONGBOTTOM , J., B ROWNE , A. J., F ULLMAN, N., O SGOOD -
Z IMMERMAN, A., E ARL , L., B HATT, S., C AMERON, E., C ASEY, D. C., ET AL . (2017). Map-
ping under-5 and neonatal mortality in africa, 2000–15: a baseline analysis for the sustain-
able development goals. The Lancet, 390 (10108), 2171–2182.
REFERENCES 55

H ILL , K. AND PANDE , R. (1997). The recent evolution of child mortality in the developing
world. Arlington, VA, Partnership for Child Health Care, Basic Support for Institutionaliz-
ing Child Survival [BASICS]. Current Issues in Child Survival Series.

J OHNSTON, C. C., F ERNANDES, A. M., AND C AMPBELL -Y EO, M. (2011). Pain in neonates is
different. Pain, 152 (3), S65–S73.

K HANNA , M., L OEVINSOHN, B., P RADHAN, E., FADEYIBI , O., M C G EE , K., O DUTOLU, O.,
F RITSCHE , G. B., M ERIBOLE , E., V ERMEERSCH , C. M., AND K ANDPAL , E. (2021). De-
centralized facility financing versus performance-based payments in primary health care:
a large-scale randomized controlled trial in nigeria. BMC medicine, 19 (1), 1–12.

K INNEY, M. V., K ERBER , K. J., B LACK , R. E., C OHEN, B., N KRUMAH , F., C OOVA -
DIA , H., N AMPALA , P. M., L AWN, J. E., S CIENCE IN A CTION : S AVING THE L IVES OF

A FRICA’ S M OTHERS, N., AND G ROUP, C. W. (2010). Sub-saharan africa’s mothers, new-
borns, and children: where and why do they die? PLoS medicine, 7 (6), e1000294.

K OKEB, M. AND D ESTA , T. (2016). Institution based prospective cross-sectional study on

patterns of neonatal morbidity at gondar university hospital neonatal unit, north-west
ethiopia. Ethiopian journal of health sciences, 26 (1), 73–79.

L AWN, J., S HIBUYA , K., AND S TEIN, C. (2005a). No cry at birth: global estimates of intra-
partum stillbirths and intrapartum-related neonatal deaths. Bulletin of the World Health
Organization, 83, 409–417.

L AWN, J. E., C OUSENS, S., Z UPAN, J., T EAM , L. N. S. S., ET AL . (2005b). 4 million neonatal
deaths: when? where? why? The lancet, 365 (9462), 891–900.

L AWN, J. E., W ILCZYNSKA -K ETENDE , K., AND C OUSENS, S. N. (2006). Estimating the
causes of 4 million neonatal deaths in the year 2000. International journal of epidemiology,
35 (3), 706–718.

L EE , A. C., K ATZ , J., B LENCOWE , H., C OUSENS, S., K OZUKI , N., V OGEL , J. P., A DAIR , L.,
B AQUI , A. H., B HUTTA , Z. A., C AULFIELD, L. E., ET AL . (2013). National and regional
estimates of term and preterm babies born small for gestational age in 138 low-income and
middle-income countries in 2010. The Lancet Global Health, 1 (1), e26–e36.

L IU, L., O ZA , S., H OGAN, D., P ERIN, J., R UDAN, I., L AWN, J. E., C OUSENS, S., M ATHERS,
C., AND B LACK , R. E. (2015). Global, regional, and national causes of child mortality in
REFERENCES 56

2000–13, with projections to inform post-2015 priorities: an updated systematic analysis.

The Lancet, 385 (9966), 430–440.

L IU, R. Y., PARELIUS, J. M., AND S INGH , K. (1999). Multivariate analysis by data depth:
descriptive statistics, graphics and inference,(with discussion and a rejoinder by liu and
singh). The annals of statistics, 27 (3), 783–858.

M AC D ONALD, T., G ENTLES, A., L IU, R., S TEVENS -U NINSKY, M., A NNI , N. S., R EHMAN,
N., D ARLING, E. K., G REENE , S., M OLL , S., AND M BUAGBAW, L. (2022). Community
level youth-led interventions to improve maternal-neonatal outcomes in low-and middle-
income countries: Protocol for a systematic review. International Journal of Reproductive
Medicine, 2022.

M ENARD, S. (2000). Coefficients of determination for multiple logistic regression analysis.

The American Statistician, 54 (1), 17–24.

M EYER , R., B ART, Y., T SUR , A., Y INON, Y., F RIEDRICH , L., M AIXNER , N., AND L EVIN,
G. (2021). A marked decrease in preterm deliveries during the coronavirus disease 2019
pandemic. American Journal of Obstetrics & Gynecology, 224 (2), 234–237.

M ORRIS, B. L., L AWRENCE , A. R., C HILTON, P., A DAMS, B., C ALOW, R. C., AND K LINCK ,
B. A. (2003). Groundwater and its susceptibility to degradation: a global assessment of the
problem and options for management.

N ASREEN, H. E., L EPPARD, M., A L M AMUN, M., B ILLAH , M., M ISTRY, S. K., R AHMAN,
M., AND N ICHOLLS, P. (2012). Men’s knowledge and awareness of maternal, neonatal and
child health care in rural bangladesh: a comparative cross sectional study. Reproductive
health, 9 (1), 1–9.

N EAL , S. AND FALKINGHAM , J. (2014). Neonatal death and national income in developing
countries: will economic growth reduce deaths in the first month of life? International
Journal of Population Research, 2014.

N GWENYA , S., J ONES, B., H EAZELL , A. E. P., AND M WEMBE , D. (2019). Statistical risk
prediction models for adverse maternal and neonatal outcomes in severe preeclampsia in
a low-resource setting: proposal for a single-centre cross-sectional study at mpilo central
hospital, bulawayo, zimbabwe. BMC Research Notes, 12 (1), 1–11.
REFERENCES 57

P HIBBS, C. S., S CHMITT, S. K., C OOPER , M., G OULD, J. B., L EE , H. C., P ROFIT, J., AND

L ORCH , S. A. (2019). Birth hospitalization costs and days of care for mothers and neonates
in california, 2009-2011. The Journal of pediatrics, 204, 118–125.

R AMAIYA , A., K ISS, L., B ARAITSER , P., M BARUKU, G., AND H ILDON, Z. (2014). A systematic
review of risk factors for neonatal mortality in adolescent mother’s in sub saharan africa.
BMC research notes, 7 (1), 1–6.

R EGGENTE , N., M OODY, T. D., M ORFINI , F., S HEEN, C., R ISSMAN, J., O’N EILL , J., AND

F EUSNER , J. D. (2018). Multivariate resting-state functional connectivity predicts re-

sponse to cognitive behavioral therapy in obsessive–compulsive disorder. Proceedings of
the National Academy of Sciences, 115 (9), 2222–2227.

S HIFFMAN, J. (2010). Issue attention in global health: the case of newborn survival. The
Lancet, 375 (9730), 2045–2049.

S ILVERMAN, J. G., D ECKER , M. R., R EED, E., AND R AJ, A. (2006). Intimate partner violence
victimization prior to and during pregnancy among women residing in 26 us states: associ-
ations with maternal and neonatal health. American journal of obstetrics and gynecology,
195 (1), 140–148.

S U, Z.-W., L IN, L.-L., S HI , B.-J., H UANG, X.-X., W EI , J.-W., J IA , C.-H., W U, F., AND C UI ,
Q.-L. (2022). Sex differences in clinical outcomes of extremely preterm infants/extremely
low birth weight infants: a propensity score matching study. Zhongguo dang dai er ke za
zhi= Chinese journal of contemporary pediatrics, 24 (5), 514–520.

T OLOSSA , T., M ULISA , D., F ETENSA , G., AND F EKADU, G. (2020). Magnitude and factors
associated with lost to follow-up among women under option b+ pmtct program at east
wollega public health facilities, western ethiopia. International Journal of Africa Nursing
Sciences, 13, 100212.

T OUSSIA -C OHEN, S., M OHR -S ASSON, A., T SUR , A., L EVIN, G., O RVIETO, R.,
M ACHTINGER , R., AND M EYER , R. (2021). Pregnancy and neonatal outcomes of twin
pregnancies–the role of maternal age. Journal of perinatal medicine, 49 (5), 559–565.

T URI , E., F EKADU, G., T AYE , B., K EJELA , G., D ESALEGN, M., M OSISA , G., E TAFA , W.,
T SEGAYE , R., S IMEGNEW, D., AND T ILAHUN, T. (2020). The impact of antenatal care on
maternal near-miss events in ethiopia: a systematic review and meta-analysis. Interna-
tional Journal of Africa Nursing Sciences, 13, 100246.
REFERENCES 58

V ELAPHI , S. AND PATTINSON, R. (2007). Avoidable factors and causes of neonatal deaths
from perinatal asphyxia-hypoxia in south africa: national perinatal survey. Annals of trop-
ical paediatrics, 27 (2), 99–106.

W ORKINEH , Y. A. AND W ORKIE , H. M. (2022). Adverse neonatal outcomes and associated

risk factors: A case-control study. Global Pediatric Health, 9, 2333794X221084070.

YAYA , Y., E IDE , K. T., N ORHEIM , O. F., AND L INDTJØRN, B. (2014). Maternal and neonatal
mortality in south-west ethiopia: estimates and socio-economic inequality. PloS one, 9 (4),
e96294.

Z AIDI , A. K., H USKINS, W. C., T HAVER , D., B HUTTA , Z. A., A BBAS, Z., AND G OLDMANN,
D. A. (2005). Hospital-acquired neonatal infections in developing countries. The Lancet,
365 (9465), 1175–1188.
Appendix
Appendices 60
Appendices 61