KNUST Thesis Proposal Format
ASSESSING THE RISK FACTORS ASSOCIATED WITH LOW BIRTH WEIGHT (LBW) AND MEAN ACTUAL BIRTH WEIGHT OF NEONATES: A CASE STUDY OF
ST. MARTIN‟S HOSPITAL, AGROYESUM.
By
ANTWI – BOASIAKO ISHMAEL (BSC. COMPUTER SCIENCE)
A Thesis submitted to the Department of Mathematics, Kwame Nkrumah University of
Science and Technology, Kumasi
in partial fulfilment of the requirement for the degree of
Master of Science
Industrial Mathematics
INSTITUTE OF DISTANCE LEARNING
October, 2011
1
DECLARATION
I, hereby declare that this submission is my own work towards the award of MSc and that,
to the best of my knowledge, it contains no materials previously published by another
person nor material which has been accepted for the award of any other degree of the
University, except where due acknowledgement has been made in the text.
AntwiBoasiako Ishmael (PG3009009) 
……………… 
…………… 
Student„s Name and ID 
Signature 
Date 
Certified by: 


Nana Kena Frempong 
……………… 
…………… 
Supervisor‟s Name 
Signature 
Date 
K. F. Darkwah 
…………….. 
……………… 
Head of Department‟s Name 
Signature 
Date 
Prof. I. K. Dontwi 
…………………. 
……………… 
(Dean of Institute of Distance Learning‟s Name) 
Signature 
Date 
2
DEDICATION
This work is dedicated to the Almighty God who is my source of wisdom, knowledge and
power.
3
ABSTRACT
A study on some selected sociobiological and demographic determinants of mean actual birth weight and Low Birth Weight (LBW) was conducted in St Martins Hospital, Agroyesum, Ghana. Records of 221 single live births over a period of one year (1st January to 31st December 2010) were analysed.
The study was aimed at gaining understanding on the determinants of LBW and mean Actual Birth weight in the study area. The data were analyzed using SAS package. Predictors of LBW were assessed through logistic regression analysis and probit regression analysis whilst that of mean actual birth weight was also assessed through multiple linear regression analysis. The factors which were significant for LBW were gestation age, age of the mother and weight gain during pregnancy and that of mean actual birth weight was gestation age whilst occupation, sex of the infant and hypertension were not significant for LBW.
The results from the three models indicate that gestation age was the predictor that has an influence on birth outcome. We also conclude from the probit and logit models that gestation age, mother‟s age and weight of the mother at last menstrual period(MWT) are risk factors of low birth weight.
4
ACKNOWLEDGEMENT
I am grateful to the Almighty God for He is my source of inspiration.
I wish to express my profound gratitude to Nana Kena Frempong, my supervisor, whose contributions towards the completion of this work are immeasurable.
Finally, I am grateful to my parents Mad. Adjei Rose and Mr. R. K. Otchere, my Uncle, Owusu Adjei, Clara AheneAffoh and Rebecca Manu for the encouragement and assistance given to me when this project was being carried out.
5
TABLE OF CONTENTS 

CONTENT 
PAGE 
Declaration …………………………………………………………………………………………………………… 
i 
Dedication …………………………………………………………………………………………………… 
…….ii 
Abstract ……………………………………………………………………………………………………………… 
iii 
Acknowledgement ……………………………………………………………………………………………. 
iv 
Table of contents………………………………………………………………………………………………….. 
v 
List of tables……………………………………………………………………………………………………… 
viii 
List of figures ……………………………………………………………………………………………………… 
ix 
List of acronyms/abbreviations………………………………………………………………………………. 
x 
CHAPTER ONE
1.0INTRODUCTION TO THE STUDY…………………………………………………1
1.1Background of the study………………………………………………………………2
1.2Statement of the Problem……………………………………………………………..3
1.3 Objectives ……………………………………………………………………………………………………. 
4 
1.4Methodology………………………………………………………………………….4
1.5Justification of the Study……………………………………………………………..5
1.6Limitations……………………………………………………………………………6
1.7.Organisation of the study…………………………………………………………….6
6
CHAPTER TWO:REVIEW OF LITERATURE 

2.0 Introduction……………………………………………………………………………8 

2.1 Definition of low birth weight ………………………………………………………………………….. 
8 
2.2.1Levels, Patterns and Determinants of LBW in developing countries……………. .12
2.2.2Behavioural risk factors ……………………………………………………………17
2.2.3Healthrelated risk factors……………………………………………………….. .19
2.2.4Sociodemographic risk factors…………………………………………………. 20
2.3Morbidity and mortality consequences of LBW in neonates …………………………..25
2.4Growth in LBW children…………………………………………………………….27
CHAPTER THREE: METHODOLOGY
3.0 Introduction………………………………………………………………………………………………….. 
31 
3.1Study Area……………………………………………………………………………31 3.1.2 Range of service of the Hospital………………………………………………… 32
3.2 Measurement and Data Analysis …………………………………………………………………….. 
32 
3.2.1 covariates………………………………………………………………………… 33 
3.3ChiSquare Test for independence………………………………………………………….34
3.4Multiple Linear Regression Model…………………………………………………..35
3.4.1Assumptions……………………………………………………………………..36
3.4.2Inferential Analysis………………………………………………………………36
3.4.3Analysis of Variance Table………………………………………………………37
3.4.4FTest of the overall Model…………………………………………………….38
3.4.5Multiple coefficient of determination………………………………………….38
3.4.6Confidence interval for estimated coefficients in multiple regression………….39
7
3.5Multiple Logistic Regression Model…………………………………………………40
3.5.1Fitting the Multiple Logistic regression model………………………………….42
3.5.2Testing the significance of the model……………………………………………44
3.5.3HosmerLemeshow statistic………………………………………………………45
3.5.4Confidence Interval estimation…………………………………………………..47
3.5.5Odds Ratio………………………….……………………………………………47
3.6Probit Regression Model…………………………………………………………….50
CHAPTER FOUR: RESULTS
4.0RESULTS AND DISCUSSION……………………………………………………52
4.1Descriptive Analysis…………………………………………………………………52
4.2Multiple Linear Regression Results…………………………………………………..55
4.3Multiple Logistic Regression Results…………………………………………………56
4.4Multiple Probit Regression Results…………………………………………………..57
CHAPTER FIVE: SUMMARY,CONCLUSION AND RECOMMENDATION
5.1Summary…………………………………………………………………………….59
5.2Conclusions ……………………………………………………………………………………….61
5.3Recommendation…………………………………………………………………….62
REFERENCES…………………………………………………………………………63 

APPENDIX A ……………………………………………………………………………………………………. 
73 
8
LIST OF TABLES 


Table 1.1: Incidence of low birth weight in selected African Countries…………………12 

Table 3.1: Contingency Table……………………………………………………………34 

Table 3.2: Analysis of Variance(based on k predictor variables) ………………..………38 

Table 3.3: Values of the Logistic Regression Model when the predictor Variable is 


Dichotomous……………………………………………………………………………..48 

Table 4.1 Summary statistics of selected variables……………………………………… 53 

Table 4.2: Test for independence(sex and low birth weight) 
…………………………………….. 
54 
Table 4.3: Parameter estimates and standard errors with associated pvalue for multiple 


regression model ………………………………………………………………………………………………… 
55 

Table 4.4: Parameter estimates and standard errors with associated pvalue and odd ratios 

for multiple logistic regression model ………………………………………………………. 
………..56 

Table 4.5: Parameter estimates and standard errors with associated pvalue for multiple 


probit regression model……………………………………………………………………………………….. 
57 

Table A.1 Test for independence(hypertension vs low birth weight)……………………73 

Table A.2 Test for independence(occupation vs low birth weight)………………………74 

Table A.3 Statistics of Test for independence(Sex vs low birth weight)………………… 
74 

Table A.4 Analysis of Variance…………………………………….……………………75 

Table A.5 model fit statistics for logistic regression……………….……………………75 

Table A.6 odd ratio estimate of logistic regression model………………………………76 

Table A.7 creiteria for assessing goodness of fit for logistic regression…………………76 

Table A.8 creiteria for assessing goodness of fit for probit regression…………………..76 

Table A.9 Summary of the dataset…………………………………………………………………..77 
9
LIST OF FIGURES 

Figure 1.1: Relationship between PTB and IUGR ………………………………………………….. 
11 
Figure 2.1:Incidence of LBW at term in selected Asian Countries …………………………… 
13 
Figure 2.2: Incidence of LBW at term in selected Afrian Countries ………………………….. 
13 
10
LIST OF ACRONYMS/ABBREVIATIONS
ART………..Artificial Reproductive Technology
ACC/SCN…Administrative Committee on Coordination/Subcommittee on Nutrition ALRI ………Acute Lower Respiratory Infections
ANOVA…..Analysis of Variance
BPD……….Brochopulmonary Dysplasis
BMI ………Body Mass Index
CHD………Coronary Heart Disease
CI…………Confidence Interval
ELBW…….Extremely Low Birth Weight
EUS……….Endoscopic ultrasound
EVS……….EUROPEAN VOLUNTARY SERVICE
IGT………..Impaired Glucose Tolerance
HIV/AIDS.. Human immunodeficiency virus /Acquired immune deficiency syndrome IMR……….Infant Mortality Rates
IUGR………IntraUterine Growth Restriction
IVR………..In Vitro Fertilization
LBW………Low Birth Weight
LMP……….Last Menstrual Period
MAR………Medical Assisted Reproduction
11
MDG………Millennium Development Goal
MLE……….Maximum Likelihood Estimation
MRC……….Medical Research Council
NICU………Neonatal Intensive Care Unit
NIDDM…….Noninsulin Dependent Diabetes Mellitus
OR…………Odd Ratio
PID……….. .Pelvic Inflammatory Disease
PTB…………Preterm Birth
SGA………. .Small for Gestational Age
SMA……. …..Society of African Missions
SAS…… ……Statistical Package for Social Sciences
UNICEF…...United Nations Children’s Fund
VLBW………Very Low Birth Weight
WHO………..World Health Organization
12
CHAPTER ONE
1.0 INTRODUCTION TO THE STUDY
The issue concerning post maternal outcomes is a concern to the World Health Organisation (WHO). One of the issues of birth outcomes is low birth weight (LBW). This is defined as weight at birth less than 2500g (5.5pounds). Low birth weight (LBW) is one of the major predictors of neonatal and perinatal morbidity and mortality both in developed and developing countries. It is believed that the better a population socioeconomic development, the better its health indicators, including LBW. That is LBW is a major factor contributing towards high infant mortality in developing countries. The proportion of infants with low birth weight reflects the socioeconomic development of any region or country (Murthy, 1991).
This is based on epidemiological observations that infants weighing less than 2,500 g are approximately 20 times more likely to die than heavier infants (UNICEF WHO, 2004). LBW is more common in developing than developed countries; a birth weight below 2,500 g contributes to a range of poor health outcomes. However, lower LBW rate of populations with lower socioeconomic levels has been observed compared to populations with better indicators. This is the socalled epidemiological paradox of LBW. LBW is mostly common in developing countries, where the burden of malnutrition and infectious diseases is heavy, and the incidence is estimated to be more than twice that of developed countries (Dreyfuss et al., 2001). In subSaharan Africa, estimated rate of LBW is 14 per 100 live birth based on statistics derived from health facilities, which constitute about 35% of all live births occurring in the region. Similarly, based on health facility statistics of 1999, which included only 45% of infants weighed at birth, incidence rate of LBW in Tanzania was 13 per 100 live births (UNICEF WHO, 2004).
13
1.1 BACKGROUND OF THE STUDY
The cohort of LBW (birthweight <2.5 kg) infants is likely to reflect two effects, namely a short gestational age (preterm births) and small for gestational age (SGA). Small for gestational age usually results from intrauterine growth restriction (IUGR). However, if the mother is small, it may be normal for her to have a small fetus. In the current study gestational age is included as an explanatory variable; therefore the study is focused on the identification of risk factors for the complementary effect which is SGA.
In a literature survey, de Onis et al (1998) found that IUGR infants are at increased risk of perinatal mortality and morbidity, i.e. sudden infant death syndrome, poor cognitive development and neurologic impairment, cardiovascular disease, high blood pressure, obstructive lung disease, diabetes, high cholesterol concentrations and renal damage in adulthood. Such infants remain a burden on government expense in developed countries and a permanent problem for their families in developing countries.
The incidence of LBW (<2.5 kg irrespective of gestational age) is estimated to be 16% worldwide, 19% in the least developed and developing countries and 7% in the developed countries. The incidence of LBW is 31% in South Asia followed by Middle East and North Africa (15%), SubSaharan Africa (14%) and East Asia and Pacific 7%. Of the total estimated IUGR infants (<2.5 kg and = >37 weeks), Asia accounts for 75%, and with 20% and 5% born in Africa and Latin America, respectively. UNICEF estimates show that in Ghana, the percentage of low birth weight infants is 9, adjusting for both mother‟s assessment and heaping on 2500g, yields higher estimates of the incidence of low birth weight (UNICEF WHO, 2004). The IUGR accounts for 11% of the total
14
infants in developing countries ranging from 2% to 21%, that is 6 times higher compared to developed countries.
The diversity in areastatus (Tribal/Settled), and differential in overall geodemographicsfactors suggest a need to investigate LBW and mean birth weight in the area. The current study derives an explanatory multivariate regression model for both LBW and actual birth weight that includes gestational age as an predictor covariate. The remaining selected covariates are thus interpreted as explanatory of incidence of LBW and actual birth weight under the category of SGA. Therefore, this prospective public hospitalbasedstudy in Amansie West focuses on LBW and Actual birth weight to investigate associated explanatory factors. These factors contribute towards the explanation of the observed births that are SGA.
1.2 STATEMENT OF THE PROBLEM
In Ghana, the issue of birthweight and factors influencing it has not received the much needed attention. This should not be the case because birthweight is a strong predictor of an individual baby‟s survival and a person‟s personality. The high rate of low birthweight (LBW), defined as weight at birth less than 2.5kg is a major problem in developing countries. From 1998 to 2004, Ghana recorded higher LBW cases of 16% compared to the average of 14% for subSaharan Africa (UNICEF,2004). Cases of low birthweight are more common in the developing countries than in the developed countries.
Prior to this study, Amansie West District was challenged with pregnancy outcomes as data at St Martin‟s Hospital over the years recorded a high number of infants born with LBW.
It is therefore imperative to identify risk factors for LBW in district in order to come up with feasible intervention strategies to minimize the problem.
15
1.3 OBJECTIVES
The main objectives of this study are:
1.To identify risk factors associated with probability of giving birth to a low birth weight using logistic and Probit regression models.
2.To model the effects on the mean actual birth weight.
1.4 METHODOLOGY:
The study was carried out from October1, 2010 to December 2010. The inclusion criteria were all infants with actual birth weight measured at the Neonatal Intensive Care Unit (NICU) of St Martin‟s Hospital from January 2010 to December 2010.
Data for the study were sourced from the birth registers available in the hospital. For this purpose, the cases of births over a year i.e. from 1st January 2010 to 31st December 2010 were taken into consideration. Since the data were recorded primarily for the use of the hospital, these data may be regarded as secondary data. The data consist of 221 observations. Incomplete records and multiple births were excluded.
The data included variables such as prenatal demographic details, maternal pregnancy history and antenatal care, details of the delivery, the infant‟s status at delivery and outcome at discharge. Maternal attributes included maternal age, maternal hypertension status and the maternal weight. Labor employment history, delivery, and newborn characteristics were: gestational age (determined according to at least 2 of the following parameters: last menstrual period) and birth weight.
The data was analysed using the Statistical Analysis System (SAS). Chisquare test was used to test independence of the categorical variables. Linear regression analysis model was used to test the effects of risk factors associated with mean Actual Birth weight and
16
also Multiple Logistic regression analysis and Probit regression analysis were used to determine the risk factors associated with Low Birth Weight. Pvalues of less than 0.05 and 0.1 were considered statistically significant because of the nature of the data.
Regression methods are essential to any data analysis which attempts to describe the relationship between a response variable and any number of predictor variables. Frequently, situations involving discrete variables arise. Logistic regression analysis and probit regression analysis extend the techniques of multiple linear regression analysis to research situations in which the outcome variable is categorical, that is, taking on two or more possible values.
1.5 JUSTIFICATION OF THE STUDY
Birth weight is a strong predictor of infant growth and survival. Infants born with low birth weights begin life immediately disadvantaged and face extremely poor survival rates. In most developing countries it was approximated that every ten seconds an infant dies from a disease or infection that can be attributed to low birth weight.
The reduction of the incidence of low birth weight also forms an important component of the Millennium Development Goals (MDGs) on child health. Activities towards the achievement of the MDGs will need to ensure a healthy start in life by making certain that women commence pregnancy healthy and well nourished, and go through pregnancy and childbirth safely. Low birth weight is, therefore, an important indicator for monitoring progress towards these internationally agreedupon goals.
Earlier works stated the birth weight of infants in Ghana ranged from 2.00 to 3.00 kg.
It is in this light that this research seeks to investigate the effect of some of the factors that affect birth weight in the Amansie West District, one of the poorest districts in Ghana.
17
It is hoped that the results of this study would help to inform the health authorities about the local risk factors for LBW and actual birth weight in order to introduce programmes to reduce the prevalence of low birth weight.
1.6 LIMITATIONS
As the research was completed in a relatively short period of time other factors and variables such as multiple birth, parity etc. were not considered. This might have an impact on the results of the study.
In spite of these limitations, the findings of the study constitute a strong basis for generalisation.
1.7 ORGANIZATION OF THE STUDY
The study has been organised into five chapters. Chapter one which is the introduction deals with the background to the study, statement of the problem, purpose of the study, justification of the study, limitations of the study and organisation of the study.
Chapter two discusses the literature related to the study. The review involves theoretical and empirical studies related to the problem under study.
The third chapter describes the methodology used in the study. Specifically, the research design, the research instrument, the procedure for data collection and the data analysis are discussed.
In chapter four, the main focus is the analysis of the data and the presentation of the results.
18
Finally, Discussion of the results, conclusions and recommendations are presented in
chapter five.
19
CHAPTER TWO
REVIEW OF LITERATURE
2.0 INTRODUCTION
The review, which involves both theoretical and empirical approaches, presents a broad overview about the literature related to the study.
2.1 DEFINITION OF LOW BIRTH WEIGHT
Birth weight is the first weight of the foetus or newborn obtained after birth. For live births, birth weight should preferably be measured within the first hour of life before significant postnatal weight loss has occurred.
Low birth weight (LBW) is defined as the weight of liveborn infants less than 2,500 g, for a given time period. LBW is closely associated with fetal and neonatal mortality and morbidity, inhibited growth and cognitive development, and chronic diseases later in life. At the population level, the proportion of infants with a LBW is an indicator of a multifaceted publichealth problem that includes longterm maternal malnutrition, ill health, hard work and poor health care in pregnancy. On an individual basis, LBW is an important predictor of newborn health and survival and is associated with higher risk of infant and childhood mortality (StevensSimon and Orleans, 1999).
The WHO 2,500 grams criterion for the definition of low birth weight mentioned at the beginning of the paper is quite arbitrary. There exists a significant relationship between the mean height of mothers in a population and the mean birth weight of children in that
20
same population. Therefore, if the standard WHO criterion is applied, populations with shorter women will almost by definition have a higher proportion of low birth weight. In those populations, the ‘optimal‟ birth weight, defined as the birth weight at which mortality is lowest, is different from the European one. This is noticeable when one compares for instance countries like Portugal, Italy, Spain and Great Britain, where the average maternal height is between 1.61 and 1.63 meters, with countries like Denmark, the Netherlands and Germany, where the average maternal height is between 1.67 and
1.69meters (Lack et al. 2003). The relationship is also noticeable when comparing Europe as a whole with nonEuropean populations. The study of Vangen et al. (2002), which compared the birth weight and perinatal mortality of Vietnamese, Pakistani, Norwegian and North African infants, found that mean birth weights were lower for Vietnamese and Pakistani children and higher for Norwegian and North African mothers. Another comment relates to the fact that, for quite a long time, low birth weight and premature birth (defined as a gestational age of less than 37 weeks – Kramer, 1987) were used as synonyms. It has now been recognised, however, that not all premature births result in low birth weight and that not all low birth weights result from premature births. In the latter case, the terms intrauterine growth retardation (IUGR) or smallforgestational age (SGA) are used – defined as “birth weight less than 10th (or 5th) percentile for gestational age; birth weight less than 2,500g and gestational age greater than or equal 37 weeks; and birth weight less than 2 standard deviations below the mean value for gestational age” (Kramer, 1987). The fact that a distinction is made between SGA and other low birth weight infants already points to differences in health consequences for the two groups.
The reason why it has become possible to make this distinction more precisely is that the measurement of gestational age, while still not perfect, has greatly improved recently,
21
with an increased use of both the last menstrual period approach (LMP – selfreported)and ultrasound techniques (FoixL‟Helias and Blondel, 2000). North African mothers have fewer children with low birth weight when they migrate. This is considered to be an epidemiological paradox.
Yang et al. (2002) note that gestational ages measured through an Endoscopic Ultrasound Scan (EUSscan) are lower on average than those measured via the LMPmethod,resulting in a higher rate of preterm birth.
Low birth weight can be caused either by premature delivery (short gestation) or by foetal growth retardation. Known factors for preterm delivery and foetal growth retardation which are associated with LBW include low maternal food intake, hard physical work during pregnancy, and illness, especially infections (Klingenberg et al. 2003).
Low birthweight (LBW) relates to intrauterine growth retardation (IUGR) and preterm birth (PTB) which are two distinct processes with differing etiology and consequences. LBW is defined by WHO as a birthweight less than 2500g. IUGR has no generally accepted standard definition, but the following are commonly used: birthweight less than 10th percentile for gestational age; birthweight less than 2500g and gestational age greater than 37 weeks; and birthweight less than 2 standard deviations below the mean value for gestational age. PTB is defined as gestational age less than 37 weeks at delivery. It is important to bear in mind that only a proportion of infants with IUGR or PTB will be classified as having LBW as shown in Figure 1.1.
22
Figure 1.1 Relationships between PTB, LBW AND IUGR
Subtypes of IUGR are described according to whether restriction in growth relates more to weight or length of the fetus. Infants who are less than 10th percentile for both weight and ponderal index are classified as „asymmetric‟ IUGR. Infants who are lessthan10th percentile for weight but not ponderal index (PI) are classified as„symmetric‟ IUGR.
The existence of subtypes of IUGR has led to inferences about the timing of growth restriction, based on Tanner‟s schematic curves of fetal and neonatal length growth which suggested that the velocity of growth in length was greater during early pregnancy. This had led to the view that symmetric IUGR is related to earlier and more persistent impairment of growth. However real data on fetal length growth suggest that growth in length during pregnancy is linear. A more accurate interpretation of the different subtypes of IUGR may therefore be that „asymmetric‟IUGR reflects more severe degrees of growth restriction.
PTB and IUGR are associated with early mortality and morbidity and with adverse long term outcomes. The longterm sequelae of PTB include: Neurological complications such as periventricular leucomalacia, Cerebral Palsy, Seizures, delayed development, and learning difficulties: Pulmonary outcomes such as bronchopulmonary dysplasia, recurrent wheezing with respiratory infection: Opthalmologic complications such as retinopathy
23
and blindness. The longterm sequelae of IUGR include small permanent deficits in weight and length, mild neurocognitive deficits and increased risk of hypertension, coronary heart disease and diabetes in adult life.
2.2.1LEVELS, PATTERNS AND DETERMINANTS OF LBW IN DEVELOPING COUNTRIES
Table 1.1 Incidence of low birth weight in selected African Countries

Country 
% of low 
Year 


Birth weight infants 






Ghana 
9 
2006 

Gabon 
14 
2000 

Botswana 
10 
2000 

Nigeria 
14 
2003 

Tanzania 
10 
20042005 

Angola 
12 
2000 

Rwanda 
6 
2005 

Mali 
23 
2001 




SOURCE: UNICEF 2008
At least 20 million infants are born every year with LBW, representing about 16% of all newborns in developing countries. Nearly 80% of all affected newborns with LBW at term are born in Asia (mainly southcentral Asia, with Bangladesh having the highest incident rate in the world (Arifeen , 1997); about 15% and 11% are born LBW at term in middle and western Africa respectively, and approximately 7% in the Latin American and Caribbean region( ACC/SCN, 2000 ). The geographical incidence of LBW at term in selected Asian and African countries (Figures 2.1 and 2.2 respectively) and table 1.1 confirm that many developing countries exceed the internationally recommended cutoff
24
levels which should trigger public health action. Incident rates of >15% for LBW and >20% for IUGR indicate that LBW at term is a major public health problem. Population wide interventions aimed at preventing LBW at term are therefore urgently required (de Onis et al. 1998).
Figure 2.1 Incidence of LBW at term in selected Asian countries
Source: de Onis et al. (1998) Eur J Cl Nutr 52(S1):S5.
Figure 2.2. Incidence of LBW at term in selected African countries
Source: de Onis et al. (1998) Eur J CI Nutr 52(S1):S5.
25
Historically, because valid assessment of gestational age is often not available in developing countries, evidence of LBW has often been used as a proxy to quantify the magnitude of IUGR. The incident rates for LBW at term conservatively estimate IUGR because when all infants below the 10th percentile of the birthweightforgestationalagereference are considered, approximately 24% or 30 million newborns in developing countries would be affected each year. Major constraints to deriving this estimate include both the quantitative and qualitative limitations of the available birthweight data (de Onis et al. 1998). Most of the data available from different parts of the world are from clinic or hospital deliveries, whereas, in some regions of Africa and southeast Asia most infants are born at home and are not measured. There is a need to determine whether data from hospitalborn infants in developing countries are representative of the large population born at home.
Prematurity and IUGR are the two main causes of LBW. The majority of LBW in developing countries is due to IUGR, while most LBW in industrialized countries is due to preterm birth (Villar and Belizán , 1982).
In many cases, the causes of prematurity are unknown; they may include high maternal blood pressure, acute infections, hard physical work, multiple births, stress, anxiety, and other psychological factors. Causes of IUGR are complex and multiple, but center on the foetus, the placenta, the mother, and combinations of all three. For instance, growth will be retarded in utero if the placenta is abnormally small or blocked causing insufficient nutrients to reach the foetus. The maternal environment is the most important determinant of birthweight, and factors that prevent normal circulation across the placenta cause poor nutrient and oxygen supply to the foetus, restricting growth. These factors may include maternal undernutrition, malaria (where it is endemic), anaemia, and acute and chronic
26
infections (such as sexually transmitted diseases and urinary tract infections) (Cameron and Hofvander, 1983). Also associated with IUGR are primiparity; multiple gestation; foetal, genetic or chromosomal anomalies; as well as maternal disorders such as renal diseases and hypertension (Prada and Tsang, 1998). Cigarette smoking and preeclampsiacause the highest relative risks for IUGR in industrialized countries, while alcohol and drug use may also restrict foetal growth (Kramer ,1998).
Major determinants for LBW in developing countries, however, are poor maternal nutritional status at conception, low gestational weight gain due to inadequate dietary intake, and short maternal stature due to the mother‟s own childhood undernutrition and/or infection (Kramer, 1987). Because maternal undernutrition is a major determinant of LBW in developing countries, high rates of LBW should be interpreted not merely as an indicator of undernutrition, morbidity and mortality for the newborn, but as an urgent public health warning that women of childbearing age are undernourished as well. Countries with higher percentages of LBW infants generally have a higher percentage of women with low body mass index (BMI) and a higher percentage of underweight children (WHO,1997)
To address these issues successfully, the underlying and basic causes of LBW in developing countries such as household food security, maternal and child care, access to and quality of antenatal and other health services, sanitation and hygiene, education, gender discrimination and poverty must be included in any longterm strategies for prevention.
Based on a metaanalysis and critical assessment of the English and French language medical literature published over the period 197084, (Kramer, 1987) presents a synthesis comprising 43 potential maternal risk factors for low birth weight grouped into 7
27
categories: (1) genetic and constitutional factors; (2) demographic and psychosocial factors; (3) obstetric factors; (4) nutritional factors; (5) maternal morbidity during pregnancy; (6) toxic exposures; and (7) antenatal care.
The Kallan (1993) framework comprises 11 ‘intervening variables’ (maternal characteristics) grouped into 4 categories: (1) sociodemographic (age; education; marital status); (2) healthrelated (parity; prior history of foetal loss or low birth weight; hypertension; diabetes; pelvic infectious disease); (3) attitudinal (wantedness of the pregnancy); and (4) behavioural (smoking; prenatal care).
Both frameworks deserve a careful assessment – even though Kramer, for instance, admits that nearly 75% of the causal factors for prematurity cannot be identified (Kramer, 1987). Both frameworks provide a comprehensive overview of risk factors and group them into suitable categories, while also acknowledging that the causes for prematurity and IUGR respectively are different. However, the direct and indirect effects of certain factors (e.g. the direct biological effect of smoking and smoking as a marker of social disadvantage) are not well explained.
Furthermore, neither framework distinguishes between risk factors by birth weight category (low vs. very low vs. extremely low birth weight). Little account is also taken of combined maternal characteristics, such as low parity at high age. Paternal characteristics or the timing of the onset of menarche are also disregarded (Forssas et al. 1999). Both authors have been able to assess the relative importance of each risk factor. According to Kramer (1987) the ranking is as follows for IUGR: cigarette smoking; low caloric intake or weight gain; low prepregnancy weight; primiparity; female sex of the child; short stature; nonwhite race; maternal LBW and prior LBW history; general morbidity; other (together accounting for an almost complete explanation). And the ranking is as follows for prematurity: cigarette smoking; prior prematurity and spontaneous abortion; low pre
28
pregnancy weight; and in utero exposure to diethylstilbestrol; unknown (together accounting for a 1/3 explanation)
According to Kallan (1993) the most important intervening variables between race and preterm birth are: parity; prior LBW; hypertension; and pelvic inflammatory disease (PID). The most important intervening variables between race and IUGR are: marital status; prior LBW; pregnancy wantedness; and smoking.
In short, the risk factors for IUGR are fairly well understood, while that is less the case for prematurity. There exists more clarity when it comes to the role of behavioural and health factors, while the role of sociodemographic factors is still part of the big unknown.
2.2.2BEHAVIOURAL RISK FACTORS 2.2.2.1 Smoking:
As already pointed out, smoking is an important risk factor. Jaakkola and Gissler (2004), in their attempt to assess whether there exists a causal relation between maternal smoking and asthma at age 7, studied the 1987 Finnish cohort of singleton births, and found that the risks for low birth weight, SGA, and preterm delivery were strongly related to maternal smoking during pregnancy. Zarén et al. (1997), examining the effect of maternal smoking on the relationship between maternal haemoglobin levels and pregnancy outcome, analysed data on Swedish and Norwegian women coming to antenatal care from before 17 to after 37 weeks between January 1986 and March 1988. They found that maternal smoking negatively affects foetal growth. In another article, based on the same population and study design, Zarén et al. (2000), found that maternal smoking affects male and female foetuses differently. Male and female foetuses of smoking mothers
29
2.2.2.2. Alcohol
(Larroque et al. 1995), in their article on moderate prenatal alcohol exposure and psychomotor development at preschool age in France, refer to a number of studies finding associations between moderate alcohol consumption and decreases in birth weight.
2.2.2.3. Stress
(Copper et al. 1996), in a psychosocial assessment study at 25 to 29 weeks of gestation, and based on an analysis of data from the MaternalFetal Medicine Units Network of the National Institute of Child Health and Human Development between October 1992 and July 1994, found that after adjusting for maternal demographic and behavioural characteristics, stress was associated with spontaneous preterm birth and low birth weight.
2.2.2.4. Coffee
(Vik et al. 2003), in their study in Norway and Sweden of enrolments that took place between 1 January 1986 and 31 March 1988 of women at either parity 1 or 2, of Caucasian origin, singleton pregnancy and registered before the 20th week of gestation, found that a high caffeine intake in the third trimester of pregnancy was associated with an increased risk of SGA. The increased risk was only found for male foetuses, not for female foetuses. The association held even after controlling for smoking.
30
2.2.3HEALTHRELATED RISK FACTORS
2.2.3.1 Abortion
A distinction has to be made between spontaneous and induced abortion. Based on an analysis of the Danish population registry over the period 19801992,( Basso et al.1998) found that a prior spontaneous abortion constitutes a risk factor for both very preterm and preterm delivery, even though spontaneous abortion and preterm delivery themselves also appear to share some risk factors. Another study by Henriet and Kaminski (2001), on the basis of the 1995 French national perinatal survey, found that women who had experienced induced abortion had a higher risk of preterm delivery.
2.2.3.2 Subfecundity and fertility treatment:
A clear link has been established between fertility treatments, also called Artificial Reproductive Technology (ART), and multiple births. In countries such as Canada, France and the US, it has been found that “estimates based solely or partly on data from surveys or registers showed that in the late 1980s and in the 1990s, between 20% and 40% of triplet deliveries followed ART, and all in all, that about three quarters of triplet deliveries occurred after procedures for subfertility”(Blondel et al. 2002). Similar evidence that fertility treatments increase the likelihood of multiple gestation (twins, triplets), or of sole surviving singletons from an initially larger group of foetuses, was found (Schieve et al. 2002). Also Macfarlane et al. (1990) found in their study on the UK in the years 1980 and 198285 that there exists a strong association between infertility treatment and the occurrence of multiple births: 55 percent of mothers of quadruplets and above and 31 percent of the mothers of triplets had undergone some form of
31
gynaecological investigation for infertility compared to only seven percent of mothers of twins and under three percent of mothers of singletons.
Multiple gestation increases the likelihood of low birth weight. This is the case for twins. For example, “the risks of very preterm delivery and very low birth weight attributable to twins were higher than the risks of overall preterm delivery or low birth weight” (Blondel et al. 2002). But this is also the case for triplets, quadruplets and higher order multiples.
Schieve et al. (2002) found that the proportion of low birth weight among those who underwent assisted reproductive technologies in the US 199697 was 13.2 percent among singletons and nearly 100 (%) percent among quadruplets or higher order multiples. Singletons were more likely to be low birth weight when originally more than 2 foetuses were found by ultrasonography.
2.2.3.3 Hypertension:
Hypertension is also considered to be a risk factor mainly for IUGR, although more study is required to verify this. Studying French data for the years 1991 and 1993, Haelterman et al. (1997) found that chronic hypertension has an impact on having a small child. In their analysis, the mean birth weight decreased by 161g for women with chronic hypertension compared to other women.
2.2.4 SOCIODEMOGRAPHIC RISK FACTORS
Finally, attention has to be paid to sociodemographic factors such as age, marital status and education. These are highly complex. They have their own predictor effect. They also interact among themselves, however (e.g. age, education and class interact with each other). Furthermore, they interact with behavioural and health related factors.
32
2.2.4.1 Age:
Maternal age constitutes an important risk factor. FoixL‟Helias and Blondel (2000) found that in France a maternal age above 34 years old constituted an important risk factor in both 1981 and 1995 while a maternal age below 20 years constituted an important risk factor in 1981 but not in 1995. Similar results were obtained by Dičkute et al. (2004), who in their study of Lithuania in 20012 identified a Ushaped relationship between maternal age and LBW risk (younger than 20 years and 35 years and older), although other factors also clearly played a role.
The importance of a young maternal age appears to be decreasing in European countries. In countries where the proportion of young mothers is small (below 3 percent), examples being France, Finland and Sweden in the 1990s, there is no (or only a weak) association between young maternal age and preterm birth (FoixL’Helias and Blondel, 2000, referring to Hemminki and Gissler, 1996).
On the other hand, as FoixL‟Helias and Blondel and others have demonstrated, a higher maternal age (e.g. above 34 years old) remains an important risk factor. Other authors have also found a strong relation between a higher maternal age and the risk for premature birth. An Italian study (Astolfi and Zonta, 2002) has found that the 35 plus age group always has a significantly higher risk of still or preterm birth, or low birth weight at term, even when parity and education variables are controlled for. A French study (Bréart, 1997) has similarly found negative consequences of delayed childbearing (35 plus) in the format of higher featal death rates.
Similar findings were also reported for Alberta, Canada, where Tough et al. (2002) reported that “when only those age 35 and older of first parity were considered, the contribution to the population increase in LBW and preterm delivery was less than half of the total contribution for this age group (36%). The negative effect of a higher maternal
33
age can be weakened by education and parity, but only to a rather small extent (Astolfi and Zonta, 2002, referring also to the Danish casestudy of Basso et al, 1997).
2.2.4.2 Education:
Another sociodemographic factor is education. Most studies conclude that higher levels of maternal educational attainment lead to better pregnancy outcomes.
Raum et al. (2001) examined the influence of maternal education on IUGR in two different political and social systems, West and East Germany in 1987/88 and 1990/91 respectively. They found that mothers among the lowest category of education in both West and East Germany had an unadjusted relative risk of 2.5 for delivering an SGA child compared to high education.
In France, two methods are used for legal abortions: surgical abortion (vacuum aspiration followed by curettage) and prescription of abortion pills mifepristone (RU486) and prostaglandins for < 7 weeks gestational age. Intervention at high gestational age is low (4% after 11 weeks) in France (Henriet and Kaminski, 2001) Those of the highest education category, This indicates that the education of the mother is a proxy for the socioeconomic and lifestyle status, including smoking behaviour, frequency of prenatal care visits (health seeking behaviour), etc.
2.2.4.3 Marital status:
Another sociodemographic factor is marital status. FoixL‟Helias and Blondel (2000) found that risk factors included being single in 1981 but not in 1995. So at the country level, the importance of being single as a risk factor appears to decrease over time in France. In countries with a high proportion of nonmarried mothers there is no association
34
between marital status and preterm birth. Examples of such countries are Finland, Sweden and France in 1995 (FoixL’Helias and Blondel, 2000).
2.2.4.4 Social class:
The sociodemographic risk factor of class interacts with behavioural risk factors such as smoking. This is demonstrated by inter alia Kramer et al. (2000) Many times, the socio demographic risk factor of class is treated together with education and marital status. Dičkute and his colleagues (2004), based on a study of Lithuania in 20012, found that there exist a Ushaped relationship between maternal age and LBW risk (younger than 20 years and 35 years and older) but clearly say that low birth weight is associated with other factors: low levels of education, unstable marital status, and low income. However, the role of socioeconomic status is very complex.
A good example is its interaction with selection and screening processes. On the one hand, Carlson et al. (1999) and Carlson and Hoem (1999), by focusing on trajectories of foetal loss in the Czech republic, identify pregnancy outcomes that are more positive for low than for high socioeconomic status women, in the format of the ‘survival paradox’. On the other hand, Khoshnood et al. (2004) report in their study based on the French National Perinatal Survey of 1998 on the better access of high socioeconomic status women to maternal serum screening for Down syndrome, and thus, to more informed decisionmaking as compared to low socioeconomic status women.
The above clearly shows that most studies of low birth weight and SGA have focused on the relationship between perinatal factors and cognitive and neurological development. Relatively little is known about other factors which are important to child health development, such as the social and home/family environment (Andersson et al. 1997).
35
Nevertheless it has been suggested that the environment, especially the early home environment, plays a central role in child health and (cognitive) development. Andersson et al. (1997), for instance, have noted how the quality and amount of stimulation received by children in different environments varies greatly, caretaking practices of lower and middleclass parents differ substantially, and children from lowerclass families seem to develop more slowly. Angelsen et al. (2001) have highlighted the importance of environmental factors – such as adequate nutrition and the parental ability to create a good and stimulating home environment – and have focused on the positive correlation between breastfeeding and cognitive development in children. Further, in the study by Smith et al. (2003), where the effect of breastfeeding (which was believed to be associated with confounding socioeconomic factors) on childhood cognitive development among VLBW children was examined, it was found that both expressed milk feedings and direct breastfeedings were associated with improved visualmotor function. Breastfeeding mothers were more likely to be older and married, and to smoke less.
So far, however, the role of the environment in shaping low birth weight children’s health outcomes had not been studied sufficiently, or incorrectly. Andersson et al. (1997) have noted that “the relationship between cognitive development and social conditions among infants born SGA had been sparsely studied”. On the other hand, Saigal et al. (2003) have noted that “socioeconomic factors, racial and ethnic differences, the nature of funding of health care … may further contribute to differences in the reported outcomes” ( but that quite often these are not sufficiently controlled for.”
Another problem can be a strong selection effect generating an overrepresentation of lower socioeconomic groups. In one study on the 20 year outcome of infants weighing 600 grams or less at birth (Sweet et al. 2003), for instance, all mothers were of low socio
36
economic origin, whether measured by race, marital status, the receipt of social assistance, or tobacco use. This selection effect may stem from the fact that higher socio economic groups are more likely to engage less in behaviour possibly giving rise to low birth weight (smoking, drinking, drugs, etc.). Furthermore, even if a low birth weight child is born to women in these groups, they may prefer not to pursue aggressive health treatments since they are better able to assess the possible negative impacts of low birth weight on later health and on their own lives. Those findings that are available thus far on the role of the socioeconomic environment tend to contradict each other. At least one study found that the socioeconomic background does not matter much for child health outcomes (Hack et al. 2002). On the other hand, Resnick et al. (1999) concluded that the impact of sociodemographic factors on adverse educational outcomes is greater than that of perinatal factors. Singer et al. (1997) noted that mental outcomes at age 3 were affected by socioeconomic factors such as minority status, race and lower social class. The relation with Bronchopulmonary Dysplasia (BPD) and higher rates of learning disabilities at school age in VLBW cohorts was not clear, however.
2.3MORBIDITY AND MORTALITY CONSEQUENCES OF LBW IN NEONATES
LBW is generally associated with increased morbidity and mortality, impaired immune function, and poor cognitive development for neonates (newborns 128 days of age) and infants. Infants born LBW are at risk to develop acute diarrhoea or to be hospitalized for diarrhoeal episodes at a rate almost two to four times greater than their normal birthweight counterparts (Bukenya et al. 1991)
Infants who are LBW risk contracting pneumonia or acute lower respiratory infections (ALRI) at a rate almost twice that of infants with normal birthweight; and more than three
37
times greater if their weight is less than 2000 g (Cerqueiro et al. 1990)). LBW is also implicated as a contributor to impaired immune function which may be sustained throughout childhood.
The risk of neonatal death for infants who are LBW weighing 20002499 g at birth is estimated to be four times higher than for infants weighing 25002999 g, and ten times higher than for infants weighing 30003499 g (Ashworth,1998)). In Brazil, 67% of all infants dying during their first week of life are LBW infants; in Indonesia the rate is 40%; and in the Sudan the rate is 35%. Infant mortality (less than one year of age) due to LBW was slightly lower: 47% in Brazil and 19% in Indonesia (Barros, et al. (1992). LBW infants during the postneonatal period (>28 days of age) also have high mortality rates – and in some cases their risk may be greater than those for LBW infants during the neonatal period (Ashworth A and Feachem RG,1985). LBW accounted for 69% of the ALRI deaths in India, and it is estimated that in Bangladesh, almost half of the infant deaths from pneumonia or ALRI and diarrhoea could be prevented if LBW were eliminated (Datta et al. 1987).
Over the past few decades, the chances of survival for low birth weight children have improved significantly. This is reflected in the decrease in overall infant mortality rates, which is mainly due to decreasing mortality for the lowest birth weights. Data from the WHO ‘Health for All’ database show that foetal, neonatal and infant mortality rates have been decreasing over the past few decades. However, while neonatal and infant mortality rates are clearly converging, this does not yet appear to be the case for foetal mortality rates.25 In the 10 new EU Member States too all three mortality rates are broadly decreasing, but at higher levels than the EU15, and with as yet much less convergence.
38
Reductions in low birth weight mortality have greatly contributed to the decrease in overall mortality. Hack et al. (1994) note that “during the past decade, advances in perinatal care have resulted in increases in the survival of extremely small and immature infants. Whereas few infants with birth weights below 750g were actively treated before the 1980s, treatment is now accepted practice for most infants born in North America with birth weights of at least 500g, those born at 24 or more weeks’ gestation, or both”. Forssas et al. (1998) note that the largest decline in stillbirth mortality occurred among those weighing below 1,000g, while for early neonatal deaths the group most affected weighed 1,0001,499g at birth.
One has to be careful when discussing death rates, however. There is still substantial underreporting of stillbirths and early neonatal deaths. This was convincingly demonstrated though the recent PERISTAT project a lower rate of postterm birth. The usual explanation for this systematic difference includes inaccurate reporting of the last normal menstrual period and the more frequent occurrence of delayed rather than early ovulation. While its conclusions are not completely definite, this study indicates that “EUSscanbased estimates are more valid on average than LNMPbased estimates. According to Lack et al. (2003) the use of ultrasound decreases the proportion of reported postterm pregnancies and increases the proportion of preterm infants, implying the underestimation of gestational age.
2.4 GROWTH IN LBW CHILDREN
Do LBW infants grow normally? What are the consequences of LBW on body size, composition, strength and cognitive development? Attaining full growth potential is
39
especially important for women and girls in order to break the intergenerational cycle of LBW and have fewer delivery complications. Maternal height is not only a reflection of genetic makeup, but also reflects her dietary history. From societal, community and individual standpoints, adolescents and adults born with LBW generally have less strength and lower lean body mass resulting in decreased work capacity and lost productivity, which may cost nations billions of dollars (Martorell et al. 1998)
When growth restriction in utero occurs early in pregnancy, infants exhibit symmetrical (or proportional) growth with length, weight, head and abdominal circumference all below the 10th percentile reference for a given gestational age (stunting). When growth restriction in utero occurs late in pregnancy, the infant exhibits asymmetrical (or disproportionate) growth with a normal length and head circumference, but low weight due mainly to a lower proportion of visceral and fat tissue (wasting) (Bakketeig ,1998). Neonatal mortality rates are reported to be higher among asymmetrical IUGR infants, but if they survive, they have a better prognoses for longterm growth and development than that for symmetrical IUGR infants. IUGR infants catchup partially in growth relative to their appropriate birthweight counterparts during their first one or two years of life. Thereafter, IUGR children maintain their place in the distribution and neither catchupnor fall further behind. They remain about 5 cm shorter and 5 kg lighter as adults. Premature infants (who are usually asymmetric LBW), who survive their first year, have a much better prognosis in terms of future growth than IUGR infants. Despite their earlier disadvantage, preterm children gradually catchup with their appropriate birthweight, term counterparts. Premature infants and IUGR infants should be studied as separate groups because they show different patterns of growth, morbidity and mortality. From a programmatic viewpoint these differences have enormous implications for intervention
40
strategies and limitations of the approach of nutritional recovery of IUGR infants in early childhood.( Hass, et al.1987)
Neurological dysfunction is often associated with attention deficit disorders, hyperactivity, clumsiness, and poor school performance. Neurologic dysfunction, when present, seems to affect IUGR boys more than girls, and children of lower socioeconomic circumstances. If IUGR infants are symmetrical and head growth is affected, there seems to be more of an impact on neurological function and it is not clear whether interventions directed toward these infants will improve their outcome. For asymmetric IUGR infants, preventing asphyxia should reduce the prevalence of major and minor handicaps, especially cerebral palsy and mental impairment frequently seen in these infants. IUGR is a much larger public health problem in developing countries than in industrialized countries and the outcomes are more likely to be aggravated by obstetric complications and perinatal problems, and later by poor health and nutrition as well as psychosocialdeprivation (Goldenberg et al.1998).
In developing countries children are exposed to poor nutrition, high levels of infections, and other conditions of poverty, thus, their long term development is dependent to a large extent on the quality of their environment. It is difficult to isolate the effects of IUGR from these factors in relation to cognitive development. Cognitive deficits appear to change over time. For instance, when IUGR infants were examined, no differences were found during the first year of life, but differences emerged during two and three years of age; and then differences disappeared at four to five years. Deficits have been found in children with very low birthweights, the smallest size, or with early IUGR (growth restriction prior to 26 weeks gestation). Since LBW occurs more often in deprived environments, it can serve as a marker for the associated poor outcomes throughout life.
41
A length deficit at an early age (stunting) would be the best predictor of motor and mental
development deficits (GranthamMcGregor, 1998 )
42
CHAPTER THREE
METHODOLOGY
3.0 INTRODUCTION
This chapter deals with the study area, method and formula of Multiple logistic regression model, concept of Multiple logistic regression models, preliminary test, linear regression model and probit regression model.
3.1STUDY AREA
St Martin‟s hospital is a nonprofit health facility which was founded in 1957. It was elevated to the status of a district hospital in 1990. The hospital is the property of the Obuasi Catholic Diocese, under the general supervision of his Lordship Rt. Rev. Justice Yaw Anokye, the Bishop. It is the only hospital facility in the Amansie West District of the Ashanti and serves a population of 144,924 people within 160 communities.
The hospital‟s humble beginning is traceable to the untiring efforts of Rev. Fr. L Bekema (SMA), Dutch who first settled at Agroyesum in 1953 as a missionary. He was later joined by Rev. Fr. Dr. Fevers (both missionary and medical practitioners) and the two of them were able to convince the traditional authorities of the need to establish a health facility in the Manso area to cater for the health needs of the rural dwellers. Having been given the mandate, the hospital was founded in October, 1957, first as Maternity Home in the small Mission House of St Anthony‟s ParishAgroyesum. In 1960, Dr Degrootanother Dutch medical practitioner took over the running of the new facility as the first Medical Officer. He later acquired the land for the hospital‟s permanent site, and through communal efforts, together with support from the Dutch Missionaries, the first permanent
43
structures consisting of a ward with small attachment for surgical procedure was built in 1965. The hospital grew steadily afterwards, amidst grave difficulties. It was elevated to the status of a District Hospital in 1990, and has since remained a focal point in the health care delivery system of the district. The hospital is well acclaimed both locally and internationally for its painstaking devotion to the treatment of Buruli Ulcer, mycobacterium ulcerous infections, endemic in the catchment area of the hospital, and other parts of the country.
3.1.2RANGE OF SERVICE OF THE HOSPITAL
The Hospital is a general Hospital and provides a wide range of diagnostic, curative and preventive services befitting its status as the District Referral facility. These include; 24 hour Out and Inpatient Care, Laboratory Services, XRay, Ultra Sound, Surgical Services, (both emergency and elective), Ophthalmic Service (Eye Clinic) Child Welfare and Primary Health care facility, Physiotherapy, Reproductive Health Care and Save Motherhood, as well as Specialized Clinic for Diabetic, HIV/AIDS, TB, and other patients with chronic diseases or conditions.
Though not well equipped for the treatment and management of Buruli Ulcer cases yet the Hospital is renowned, both locally and internationally for services it offers to the people with this ailment. The Hospital treats Buruli Ulcer referral cases from across the length and breadth of the country.
The hospital has a bed compliment of 116 distribute over the four admitting wards as follows; males 41, females26; maternity 16 and children‟s wards – 33.
3.2 MEASUREMENT AND DATA ANALYSIS
This was a health facility based retrospective review of data of pregnant mothers who delivered in the health facility in Amansie West district. Records of women who
44
delivered a singleton in the hospital in 2010 were extracted from the registers and those met were asked mothers to show the health card on which the birth weight was recorded. Birth weights were recorded in grams to the nearest 10g. Low birth weight, one of the response variables in our analysis, was defined as a weight at birth of less than 2,500g. The biological data collected include age of mother at the time of delivery, birth weight and sex of the child. Birth weights of newborns were measured without clothes within 15 30 minutes of birth to the nearest 10g, using an infant beam balance.
The socioeconomic data available from hospital case cards related to the occupation of the mother were also taken. Data on socioeconomic and biological determinants of birth weight in Ghana is generally incomplete.
We used multiple logistic regression analysis to assess the strength of association of each explanatory variable with the likelihood of a child having low birth weight, after adjusting for other covariates. Results of the regression analyses are presented in the form of odds ratios (OR), with 95% confidence intervals (CI) and pvalues. We used Hosmer Lemeshow and Pearson‟s goodnessoffit tests to check the fit of the logistic regression models.
The multiple regression analyses presented are based on a final sample size of 212 neonates for low birth weight. A level of p <0.05 and 0.1 were considered statistically significant, and all statistical analyses were conducted using SAS.
3.2.1 COVARIATES
Biological attributes at birth: The biological attributes at birth are those that cause poor developmental outcomes for the fetus. These include maternal attributes as well as the child‟s. In the analysis, childlevel covariates include birth weight (<2,500g, ≥2,500g) and
45
sex (boy, girl). Maternal covariates include mother‟s age at each child‟s birth, Gestation age and mother‟s weight and Hypertension status.
3.3 CHISQUARE TEST FOR INDEPENDENCE
A chisquare test for independence was used to determine whether there was a significant association between occupation, hypertension and sex that were categorical in the data and the LBW.
TABLE 3.1: CONTINGENCY TABLE



J 






1 
2 
3 … 
j 








1 
011 
012 
013 … 01j 

01+ 


2 
021 
022 
023 … 02j 

02+ 

I 
. . 
. 
. 
. 

. 


. . 
. 
. 
. 

. 


. . 
. 
. 
. 

. 


i 
0i1 
0i2 
0i3 … 0ij 

0i+ 



0+1 
0+2 
0+3 … 0+j 

N 




The test statistic is a chisquare random variable (χ2) defined by the following equation.




χ2 
= 
� 
� 
(��� − ���� )2 
∼ χ2 
(3.1) 




� � 








���� 
�−1 (�−1) 













Where the expected frequency ���� 
is defined as 



�� 

= 
��+× �+� 






(3.2) 








�� 

� 

















The ��� 
is the observed frequency count at level ��+ of Variable j 
and level �+� of 
Variable i and (i1)(j1) is the degrees of freedom.
46
3. 4 MULTIPLE LINEAR REGRESSION MODEL
Multiple regression is used to account for (predict) the variance in an interval dependent, based on linear combinations of interval, dichotomous, or dummy predictor variables. Power terms can be added as predictor variables to explore curvilinear effects. Cross product terms can be added as predictor variables to explore interaction effects. Using hierarchical regression, one can see how much variance in the dependent can be explained by one or a set of new predictor variables, over and above that explained by an earlier set. The estimates (β coefficients and constant) can be used to construct a prediction equation and generate predicted scores on a variable for further analysis.
We have k variables that we control, or know in advance of outcome, that are used to predict Y, the response (dependent variable). The k predictor variables are labeled X1, X2,…,Xk. The levels of these variables for the ith case are labeled X1i ,…, Xki . Note that simple linear regression is a special case where k=1, thus the methods used are just basic extensions of what we have previously done.
E(Y  X ) = βo + β1x1 + β2 x2 +… + βK xK + ε 
(3.3) 
where βj is the change in mean for Y when variable Xj increases by 1 unit, while holding
the k1 remaining predictor variables constant (partial regression coefficient). This is also referred to as the slope of Y with variable Xj holding the other predictors constant.
i. Y is the continuous and random variable we wish to estimate or predict x1, x2,…,xk.
ii.x1, x2,…,xk are the predictor variables that are measured without error. iii.β0, β1, β2,…, βk are the parameters to be determined.,
47
iv.ε is the random error for any given set values of x1, x2,…,xk.
3.4.1 Assumptions
i.E(ε) = 0 for i = 1,2,…,n that is a given set values x1, x2,…,xk.
E(yi) = β0 + β1x1 + β2xi2 +… + βkxik = 0
ii.Var(εi) = σ2 which is constant for all the values of x. and Var(yi) = σ2 , a constant for
each recorded value of yi.
iii.The random error εi is independent, that is their covariance is equal to zero. .
cov(εi ,εk ) = 0 = E(εi ,εk ) = 0,i ≠ k .
iv. The random error ε 
i 
N (0,δ 2 ), that is 


yN (β0 + β1xi1 + β2 xi 2 + … + βk xik ,δ 2 ) .
3 .4.2 Inferential Analysis
Analysis of Variance (ANOVA) is an inferential method that is used to test the equality of three or more population means. The application of ANOVA in regression analysis is based on the partitioning of the total variation and its degree of freedom into components.
We have:
( yi − yˆ) = ( yi − yˆi ) + ( yˆi − y)
Where
( yi − y) = total variation
( yi − yˆi ) = unexplained variation
( yˆi − yi ) = explained variation
48
n 
n 


− y )]2 

∑( y − y)2 
= ∑[( y − yˆ 
+ ( yˆ 

i 
i 
i 
1 
i 
(3.4) 
i=1 
i=1 




n 
n 

n 


∑( y − y)2 
= ∑( y − yˆ 
)2 + ∑( yˆ 
− y)2 

i 
i i 


i 

i=1 
i=1 

i=1 

Where:
n
∑( yi − y)2 = total variation (SSyy )
i=1
∑( yi − yˆi )2 = unexplained variation (SSE)n
i=1
n
∑( yˆ1 − y)2 = explained variation (SSR)
i=1
The three quantities are measures of dispersion. SSyy is a measure of dispersion of the total variation in the observed values, yi. SSR measures of amount of total variation in the observed values of yi that is accounted for by the linear relationship between the observed values of x and y. The unexplained sum of squares (SSE) is a measure of dispersion of the observed yi values about the regression line which is sometimes called the error or residual sum of squares.
3.4.3 Analysis of Variance Table (Based on k Predictor Variables)
The decomposition of the total sum of squares and its degrees of freedom (df) are displayed in the table below:

n 









SST = ∑ 
(Y − 

)2 


= n − 1 

Total Sum of Squares and df: 
Y 
df 
T 


i =1 
i 














n 
^ 







Regression Sum of Squares: 
SSR = ∑ (Y i − 
Y 
)2 
df R 
= k 
i =1
49

n 
^ 




SSE = ∑ 


= n − k − 1 

Error Sum of Squares: 
(Y − Y )2 
df 
E 


i =1 
i i 








TABLE 3.2 Analysis of variance (based on k predictor variables)
Source 
df 
SS 
MS 
F 
Regression 
dfR 
SSR 
MSR=SSR/k 
Fobs=MSR/MSE 
Error 
dfE 
SSE 
MSE=SSE/(nk1) 
— 
Total 
dfT 
SST 
— 
— 
3.4.4 FTEST FOR OVERALL MODEL
Ftest is used to test whether any of the predictor variables are linearly associated with Y.
H0: β1 = … = βk = 0 (Y is not linearly associated with any of the predictor variables)
HA: Not all βj = 0 (At least one of the predictor variables is associated with Y )
MSR
TS: Fobs = MSE
RR:Fobs ≥ Fα ,k ,n− k −1 that is reject H0: β1 = … = βk = 0
Pvalue: Area in the Fdistribution to the right of Fobs
3.4.5 MULTIPLE COEFFICIENT OF DETERMINATION
Proportion of variation in Y “explained” by the regression on the k predictor variables is called multiple coefficient of determination denoted by R2.
If R 2 = 1, the fitted model passes through all the data point and also
εˆi = 0 for all i = 1, 2, 3, …, n. If R2 = 0, βˆ0 = y and β1 = β2 = …βˆk = 0
In this case the predictor variables xi 1 , xi 2 , xi 3 ,…, xik have no imfluence on the response
50
variable yˆi . If more predictor variables are introduced in the model R2 tend increase it value.
The coefficient of determination R2 
is given as 




n 
⎛ ^ 


⎞ 2 







∑ 









⎜ 


⎟ 








⎝ Y i − Y⎠ 


SSR 

SSE 

R2 = r 2 
= 
i =1 





= 
= 1 − 

Y∙1,,k 

n 




2 

SST 
SST 



∑ (Yi 
− Y) 









i=1
Since R2 is often by including larger number of predictor variables, it is sometime
suggested to modified R2 to account for the number of independent variables in the model,
The adjusted multiple coefficient of determination, denoted, R 2 is defined by 


⎛ 
n −1 ⎞ SSE 


Ra2 = 1− ⎜ 

⎟ 







⎝ n − ρ ⎠ SST 
(3.5) 






=1 (1− R2 )(n −1) n − k −1
3.4.6Confidence Interval for Estimated Coefficients in Multiple Regressions
If the regression assumptions on the residuals are satisfied, including the normality assumption, then the sampling distribution of an estimated regression coefficient is normal with a variance proportional to the residual mean square (MSE). The variance of the estimator also depends on the variances and covariances of the predictors. For more than one predictor, the confidence intervals for regression can be computed similarly, but the equation is more complicated. The equation for the variances and covariances of estimated coefficients is expressed in matrix terms by
ˆ 
(3.6) 
Var( β) = Se2( XTX)1 = MSE( XTX)1 
51
Where Se2 is the residual mean square and X is the time series matrix of predictors. Equation (3.7) returns a matrix, with the variances of the parameters along the diagonal, and the covariances as the offdiagonal elements. The appropriate degrees of freedom of the t distribution is df = nk1 , where K is the number of predictors in the model, and n is the sample size.
The 95% Confidence Interval for β provides the lower and upper bounds for the unstandardized regression coefficients, based on the previously computed standard error.
3.5THE MULTIPLE LOGISTIC REGRESSION MODEL
Logistic regression analysis extends the techniques of multiple regression analysis to research situations in which the outcome variable is categorical, that is, taking on two possible values. In this study, the risk factors for the low birth weight are identified using logistic regression analysis.
Logistic regression is a useful way of describing the relationship between one or more predictor variables (e.g., age, sex, etc.) and a binary response variable, expressed as a probability, that has only two values, such as having low birth weight (“has low birth weight” or “doesn’t have low birth weight”)
In setting up the logistic regression model, we must first establish the fundamental model for any multiple regression analysis.
To explain the popularity of logistic regression, we show here the logistic function, f(z)which describes the mathematical form on which the logistic model is based and Consider
acollection of k predictor variables denoted by the vector �′ = (�1, �2 … �� ). It is assumed that the outcome variable z is a linear combination of a set of predictors.
Z = β0 + β1x1 + β2x2 +… + βKxK 
(3.7) 
52
and β0 and the βi are constant terms representing unknown parameters. This function, called f(z), is given by
f (z) = 
1 
(3.8) 
k 


−( β0 +∑βi xi ) 

1 + e 
i =1 

The logistic regression model indirectly models the response variable based on probabilities associated with the values of Y. We will use ��(x) to represent the probability that Y =1, which is the presence of low birth weight similarly, we will define 1 ��(x) to be the probability that Y =0, which is absence of low birth weight. in which case the logistic regression model is:
The general formular for the logistic regression model with single variable is;


k 


e 
βo +∑βi xi 

π (x) = 
i=1 
(3.9) 







k 



βo +∑βi xi 

1+ e 
i=1 

From equation (3.9), k = the number of independent variables and P(Y=1/x) = conditional probability that the outcome is present. The equation(3.9) above as:
P(Y=1x1, x2 ,…, xk ) = 

1 




k 


e 
−(β0 +∑β1xi ) 

i=1 



�� (x) = the can be put
(3.10)
The logits, natural logs of the odds, of the unknown binomial probabilities are modeled as a linear function of the XK.
Logit (��(x)) = ln[ 
��(x) 
] = β 

+ β x + β x 
+… + β x 
(3.11) 

0 


1−��(x) 
1 1 2 2 
k k 







53
Note that a particular element of XK can be set to 1 for all K to yield an intercept in the model. The unknown parameters βK are usually estimated by maximum likelihood using a method common to all generalized linear models.
The interpretation of the βK parameter estimates is as the additive effect on the log of the odds for a unit change in the Kth explanatory variable. In the case of a dichotomous explanatory variable, for instance gender, eβ is the estimate of the odds of having the outcome for, say, males compared with females.
3.5.2 FITTING THE MULTIPLE LOGISTIC REGRESSION MODEL
Assume that we have a sample of n predictor observations ���� , � = 1,2, … , �. As in the univariate case, fitting the model requires that we obtain estimates of the vector�′ = �°, �1,…,�� . The method of estimation used in the univariate situation will be employed
in the multivariate case – maximum likelihood. There will be k+1 likelihood equations that are obtained by differentiating the log likelihood function with respect to the
k+1 coefficients. The likelihood equations that result may be expressed as follows:
�
�� − �� �� = 0
�=1
and
�
��� �� − �� �� = 0 ���� � = 1,2, … , �
�=1
As in the univariate model, the solution of the likelihood equations requires special
software that is available in most, if not all, statistical packages. Let β denote the solution to these equations. Thus, the fitted values for the multiple logistic regression model are �� �� , the value of the expression in equation (3.11) computed using β and ��.
54
In the previous section only a brief mention was made of the method for estimating the standard errors of the estimated coefficients. Now that the logistic regression model has been generalized both in concept and notation to the multivariate case, we consider estimation of the standard errors in more detail.
The estimators are obtained from the matrix of second partial derivatives of the log likelihood function. These partial derivative have the following general form
��2��(�) 

� 




= − 
�2 
�� (1 − �� ) 
(3.12) 

���2 


�� 
� 
� 


� 

�=1 









and

��2�� � 
� 









= − 
��� ��� ���(1 − ���) 
(3.13) 

���2 



� 


�=1 







���� �, � = 0,1,2, … , � 
where 
��� denotes �� �� . Let the 
� + 1 × � + 1 matrix 
containing the negative of the terms given in equations (3.12) and (3.13) be denoted as I (β ) . This matrix is called the observed information matrix. The variances and
covariances of the estimated coefficients are obtained from the inverse of this matrix which we denote as ���� � = I−1 β . Except in very special cases it is not possible to write down an explicit expression for the elements in this matrix. Hence, we will use the
notation ���� � to denote the ���� 
diagonal element of this matrix, which is the variance 
� 

of �� , and covariance ����� �� , �� 
to denote an arbitrary offdiagonal element, which is 
the covariance of βj and βl. The estimators of the variances and covariances, which will be
denoted 
by ����(�) 
are obtained by evaluating ���� � ��� �. We will use 
���� �� 
��� ����� �� , �� 
, �, � = 0,1,2, … , � to denote the values in this matrix. 


55 
For the most part, we will have occasion to use only the estimated standard errors of the standard coefficients, which we will denote as
��� �� = ���� �� 
3.14 
for � = 0,1,2, … , �. We will use this notation in developing methods for coefficient testing and confidence interval estimation.
A formulation of the information matrix which will be useful when discussing model fitting and assessment of fit is � � = �′�� where X is an � �� � + 1 matrix containing the data for each subject, and V is an n by n diagonal matrix with general element ���(1 − ���). That is, the matrix X is

1 
�11 
… 
� 



… 
1� 

1 
� 
� 

� = 
21 
… 
2� 

⋮ 
⋮ 
⋮ 
1��1 … ���
and the matrix V is

��1(1 − ��1) 
0 
… 
0 

� = 
0 
��2(1 − ��2) 
… 
0 

⋱ 

⋮ 
0 
⋮ 







0 
… 
0 
��� (1 − ��� ) 
3.5.3 Testing for the Significance of the Model
Before concluding that any or all of the coefficients are nonzero, we may wish to look at the univariate Wald test statistics,
�� = 
�� 

(3.15) 

���(�� ) 



The multivariable analog of the Wald test is obtained from the following vectormatrixcalculation:
56
W = β′ Var β −1β 

W = β′ (X′ VX)β, 
(3.16) 
which will be distributed as chisquare with � + 1 degrees of freedom under the hypothesis that each of the � + 1 coefficients is equal to zero.
Then multivariable analog of the Score test for the significance of the model is based on the distribution of the p derivatives of ��(�) with respect to �. The computation of this test is of the same order of complication as the Wald test.
3.5.3HOSMERLEMESHOW STATISTIC
A summary measure verifies the fit of a model by first computing the difference between the observed and predicted scores throughout the entire collection of observations, then examining the individual contribution of the difference from each pair relative to the error structure of the whole model, and, finally, determining the suitability and efficiency of the model by a single statistic. Empirical consistency, which refers to the observed outcome behaving in accordance with the model prediction, is revealed by a chisquare statistic that is computed with a contingency table that collapses the observations into a fixed number of groups.
Hosmer and Lemeshow (1980) proposed a grouping strategy that is based on the value of estimated probabilities. Supposing that there are n observations in the probability model, Group 1 would consist of the n/g observations with the lowest predicted probabilities, Group 2 would consist of the n/g observations with the nextlowest predicted probabilities, and so on. Once all of the groups are created, a Pearson chisquare statistic is estimated based on the observed and expected number of observations in the interest
57
category of every group. When the Pearson chisquare statistic is significant, the null hypothesis that the proposed model sufficiently describes the empirical association is rejected.
Let nk be the number of observations in the kth group and yi be the response of the ith
observation. Then, the number of observed responses in interest for the kth group would be

nk 

ok 
= ∑ yi . 
(3.17) 

i =1 

In contrast, if πˆ i denotes the predicted 
probability for the 
ith observation, then the 
average estimated probability for this group would be
nk πˆ
πk = ∑ n i .i =1 k
Eventually, the Pearson chisquare statistic is determined by
ˆ 
g 
(ok − nk πk )2 

C = ∑ n π 
k 
(1 − π 
k 
) , 

k 
= 
k 



1 





(3.18)
(3.19)
with a (g2) degree of freedom.
The HosmerLemeshow statistic is usually computed using g = 10 groups. However, because the restriction of g > p +1, where p is the number of covariates in the proposed model, arose in the simulation process that was demonstrated by Lemeshow and Hosmer (1982), g = 50 is used in the current study, as more than 40 covariates are considered in the proposed logistic regression model.
58
3.5.4 CONFIDENCE INTERVAL ESTIMATION
The confidence interval estimator for the logit is a bit more complicated for the multiple variable mode. The general expression for the estimator of the logit for a model
containing k covariates is � � 
= �′� 
(3.20) 
where the vector�′ = � , � 
� denotes the estimator of the � + 1 coefficients and the 

° 1,…, 
� 

vector �′ = �0, �1, �2,…,�� represents the constant and a set of values of the k – covariates
in the model, where �0 = 1.The expression of the variance of the estimator of the logit in (3.21) is
�� �
����� � � = �2 
����� 
� + 
2� � 
���� � � 
(3.21) 
� 

� 
� � 
� � 

� =0 


� =0 �=�+1 


We can express this result much more concisely by using the matrix expression for the estimator of the variance of the estimator of the coefficients. From the expression for the observed information matrix, we have that
����� � 
= X′ VX −1 
(3.22) 

The 
equivalent expression for the estimator in (3.21) is 


����� 
� � 
= �′ ����� � � 
(3.23) 
3.5.5 ODDS RATIO
In order to interpret this result we need to introduce and discuss measure of association termed the odds ratio.
The possible values of the logistic probabilities may be conveniently displayed in a 2 × 2as shown in Table 3.3.
59
Table 3.3 Values of the Logistic Regression Model When the Predictor Variable Is
Dichotomous
Predictor Variable (X)
Outcome Variable (Y) 

� = 1 
� = 0 




� = 1 
�� 1 = 

��0+�1 

�� 0 = 

��0 





1 + ��0+�1 
1 + ��0 



� = 0 

1 



1 




1 − �� 1 = 

1 − �� 0 = 




1 + ��0+�1 
1 + ��0 


Total 

1.0 

1.0 















The odds of the outcome being present among individuals with � = 1 is defined as
�� 1 1 − �� 1 
. Similarly, the odds of the outcome being present among individuals 

with x = 0 is defined as π 0 
1 − π 0 . The odds ratio, denoted OR, is defined as the 

ratio of the odds for � = 1 to the odds for � = 0, and is given by the equation 

�� = 
�� 
1 
1 − �� 
1 

(3.24) 






�� 
0 
1 − �� 
0 

Substituting the expression for the logistic regression model shown in Table 3.3 into
(3.24) we obtain:


��0+�1 
1 



�� = 

1 + ��0+�1 

1 + ��0+�1 




��0 

1 












1 + ��0 

1 + ��0 

��0+�1
=��0
=� �0+�1 −�0
60
=� �1
Hence, for logistic regression with dichotomous predictor variable coded 1 and 0, the relationship between the odds ratio and the regression coefficient is
�� = ��1 
3.25 
Nevertheless, if the coding scheme is different from the (0, 1) then the odds ratio formula needs to be modified, but for the purpose of this study all the dichotomous variables will be coded using the (0, 1) coding scheme.
The simple relationship between the coefficient and the odds ratio is the fundamental reason why logistic regression has proven to be such a powerful analytic research tool. The odds ratio is a measure of association which has found a wide use, especially in epidemiology, as it approximates how much more likely (or unlikely) it is for the outcome to be present among those with � = 1 than among those with � = 0.
The interpretation given for the odds ratio is based on the fact that in many instances it
approximates a quantity called the relative risk. This parameter is equal to the ratio����(1)(0). It
follows from (3.25) 
that the odds ratio 
approximates the relative risk 

if 1 − �� 0 
1 − �� 1 
≈ 1. This holds when �� � 
is small for both � = 1 and 0. 
A 100 1 − � % confidence interval (CI) estimate for the odds ratio is obtained by first calculating the endpoint of a confidence interval for the coefficient,�1, and then exponentiating these values. In general, the endpoints are given by the expression
��� �1 ± �1−�2 × ���(�1)
Because of the importance of the odd ratio as a measure of association, many software packages automatically provide point and confidence interval estimates based on the
61
exponentiation of each coefficient in a fitted logistic regression model. These quantities provide estimates of odds ratios of interest in only few special cases (e.g. a dichotomous variable coded zero or one that is not involved in any interactions with other variables).
3.6 PROBIT REGRESSION MODEL
Probit regression, also called a probit model, is used to model dichotomous or binary outcome variables. Probit models simply use the cumulative Gaussian normal distribution rather
than the logistic function for calculating the probability of being in one category or not. In the
probit model, the inverse standard normal distribution of the probability is modeled as a linear combination of the predictors. The probit link is another function suitable for fitting regression and ANOVA models to binomial data.
Probit regression models the probability that Y=1 using the cumulative standard normal distribution function, evaluated at
Z= β0 + β1x1 + … + βk xk
Thus, the probit linear regression model can be written as
Pr(Y =1 X ) = Φ(β0 + β1x1 +… + βk xk ) 
(3.26) 

Equation (3.26) can be transformed to 






1 

∫−∞β0 +β1x1 +…+βk xk e − 
1 


= 


2 
z2 dz 








2π 










Where Φ is the cumulative normal distribution function.
∙z = β0 + β1 x1 + … + βk xk is the “zvalue” or “z– index” of the probit model.
62
where β0; β1,… βk are the parameters to be determined and the symbol Φ is simply the cumulative standard distribution, while the lower case symbol, φ, as before, represents the standard normal density function. We maximize the same loglikelihood function.
Note that this gives the model in the form of the inverse link. You can write the probit
model in terms of the link function as 

Probit (π) =Φ−1(π) = β0 + β1X1 + β2X2 + …+ βkXk, 
(3.27) 
where Φ1(π) means the Z value such that the area under the curve less than Z is π. Interpretation of the probit coefficients is, in some senses, rather easier than the logit. The regression coefficients of the probit model are effects on a cumulative normal function of the probabilities that Y = 1 (i.e. the probability that a neonate is a having LBW).
63
CHAPTER FOUR
4.0 RESULTS AND DISCUSSION
Logistic regression and probit models were used to identify the risk factors associated with LBW using SAS. Frequency, mean and standard deviation were obtained for continuous variables while the categorical variable was assessed by computing frequencies. Odds ratio (OR) and 95% confidence interval (CI) for each variable of interest were calculated. Pvalues were calculated by likelihood ratio test for the significance of the beta coefficients; Pvalues of 0.05 and 0.1 were considered significant for all the predictor variables in the model. Linear regression was also computed to identify factors associated with mean birth weight. Data gathered from the study were analysed according to WHO definition of LBW (< 2.5 kg).
4.1 DESCRIPTIVE ANALYSIS
The descriptive analysis examines the incidence of birth weight and low birth weight on selected variables. A total of 221 infants were eligible and all satisfied the inclusion criteria and so were recruited from January 2010 to December, 2010.
From Table 4.1, the Mean birth weight of the infants was 2.812 kg. Mean gestation age of the sample is 36.39 weeks and mean weight of women whose infants were studied was 57.633kg. Majority of the LBW infants 45 (78.9%) were preterm gestational age infants and 12 (21.1%) were atterm gestational age. The mean age of women whose infants were studied is 24.57 years ranging from 14 years to 44 years.
From Table 4.2, there were 95 male infants which represent about 43% and 126 female infants which also represent about 57% of the study population The birth weight were categorised into high and low based on the WHO standards. There were a total of 57 LBW infants which represents about 26%, with males accounting for 23 representing
52
about 10% and 34 Females representing 15% and 164 infants representing about 74% were normal birth weight as shown from Table 4.2.
Most of the women whose infants were studied were in the age group 2025 years (39%) followed by women aged <20 years (23.2%),with 20.1% aged between 2630 years and above 30 years (17.7%) (Appendix A.9).
Out of the total number of low birth weight infants seen, 31 (54.4%) of them were delivered by mothers in the age group 20 30 years, 11(19.3%) of LBW infants were delivered by women in age group less than 20 years and about 15(26.3%)LBW infants were delivered by women of age above 30 years .
Many of the by mothers whose occupation was farming had the highest LBW infants (17.65%), followed by mothers who were students (5.43%). Petty traders were in the minority (2.71%). (Appendix A.2)
TABLE 4.1 Summary statistics of selected variables
Variable 
Mean 
Minimum 
Maximum 
Std Dev 
Lower 95% 
Upper 95% 





CL for Mean 
CL for 






Mean 
AGE(years) 
24.57 
14.000 
44.000 
6.237 
22.929 
26.210 
BWT 
2.812 
1.000 
4.500 
0.642 
2.727 
2.897 
GAGE 
36.398 
26.000 
48.000 
3.131 
35.983 
36.813 
MWT 
57.633 
42.000 
75.000 
5.539 
56.899 
58.368 







53
TABLE 4.2 Test for independence (sex and low birth weight)
SEX 
LOW 
Total 


>=2500g 
<2500g 

Male 
72 
23 
95 
% 
32.58 
10.41 
42.99 




Female 
92 
34 
126 
% 
41.63 
15.38 
57.01 




Total 
164 
57 
221 
% 
74.21 
25.79 
100.00 



The chisquare test for independence was performed to assess for the association between the outcome of Low and some variables such as Sex, Occupation and Hypertension. In other words, it is seen whether the proportion of low birth weight is the same in each category. From Table 4.2, Chisquare test indicates that the outcome of LBW and Sex is predictor (P=0.6408)
Also from the contingency tables in appendix A, the usual Chisquare tests were performed to see if the outcome of low birth weights is associated with hypertension and occupation. In both cases there were no significant results between low birth weight and the predictors.
54
TABLE 4.3 Parameter estimates and Standard errors with associated pvalues for Multiple regression model
Variable 
DF 
Parameter 
Standard 
t Value 
Pr > t 


Estimate 
Error 


Intercept 
1 
0.46648 
0.61467 
0.76 
0.4487 
AGE 
1 
0.03027 
0.03414 
0.89 
0.3763 
SEX 
1 
0.03862 
0.08034 
0.48 
0.6312 
GAGE 
1 
0.08362 
0.01311 
6.38 
<.0001 
FARMING 
1 
0.09163 
0.13730 
0.67 
0.5053 
STUDENT 
1 
0.34321 
0.17011 
2.02 
0.1149 
MWT 
1 
0.00771 
0.00725 
1.06 
0.2889 
HT 
1 
0.10105 
0.20324 
0.50 
0.6196 
4.2.1 MULTIPLE LINEAR REGRESSION RESULTS
Tables 4.3 show the result of linear regression analysis investigating the factors affecting mean Birth weight among neonates in St Martins Hospital. The data set contains six predictor variables: Mothers Age(Age), categories of Occupation( farming and student), Hypertension(HT), Sex, Mothers weight(MWT) and gestation age(GAGE). The response variable was the mean actual birth weight(BWT).
The R2 value of 0.20 shows that about 20% of the variability in birth weight data is accounted for (appendix A.4). This reveals that the regression model may not be appropriate for prediction purpose. This may mean that some important variables were not considered.
The result of the regression analysis as shown in appendix A.4 the global test for model significant is statistically significant (F= 7.29, P=<0.0001). The variable that showed significant with mean birth weight is Gestation Age (p value of <0.0001) as shown in Table 4.1. However, Age of the mother, sex of the baby, mother`s age ,weight of the mother at last menstrual period as well as the history of hypertension were not seen as
55
having an effect on the mean actual Birth weight in this model as shown in table 4.3 above. The fitted multiple regression model is:
BŴT = 0.46648 – 0.03027*AGE – 0.03862*SEX + 0.08362*GAGE – 0.09163 *FARMING – 0.34321* STUDENT + 0.00771* MWT + 0.10105 *HT
TABLE 4.4 Parameter estimates and Standard errors with associated pvalues and odd
ratios 
for Multiple Logistic regression model 












Parameter 
Estimate 
Standard 
Wald 95% Confidence 
Pr > Chi 
Exp(Est) 




Error 
Limits 

Sq 

Intercept 
17.5752 
4.8173 
8.1334 
27.0170 
0.0003 
60771156 

AGE 

0.4119 
0.2526 
0.0832 
0.9069 
0.1030 
1.510 
SEX 

0.1460 
0.5344 
0.9014 
1.1934 
0.7847 
1.157 
GAGE 

0.2589 
0.1080 
0.4706 
0.0472 
0.0165 
0.772 
OCCU 
FARMIG 
0.1331 
1.2605 
2.3374 
2.6035 
0.9159 
0.807 
OCCU 
STUDENT 
0.9094 
1.4053 
1.8450 
3.6638 
0.5176 
1.754 
OCCU 
TRADER 
0.0000 
0.0000 
0.0000 
0.0000 
. 
– 
MWT 

0.0691 
0.0505 
0.0298 
0.1681 
0.1707 
1.072 
HT 

0.9862 
1.2934 
1.5487 
3.5212 
0.4457 
2.681 








4.2.2 MULTIPLE LOGISTIC REGRESSION RESULTS
Logistic regression was performed to investigate the factors that contribute to the risk of Low Birth Weight among neonates. The model contained six predictor variables, however only gestation age was statistically significant at 5% level of significance with p value of 0.0165 as depicted from table 4.4 above. Age of the mother is however slightly significant at 10% level of significance (p =0.103, OR= 1.510). From table 4.4 above, the non significant variables are: sex of the child, Mother‟s weight, occupation of the mother and Hypertension status of the mother.
56
This means that a mother with gestation age less than 37 weeks has 1.3 times chance giving birth to an infant with low birth weight than atterm (gestation age greater than or equal to 37 weeks gestation age).
The fitted multiple logistic regression model is:
Logit( π ) = 17.5752 + 0.4119*AGE + 0.1460*SEX – 0.2589*GAGE +0.1331 *FARMING + 0.9094* STUDENT + 0.0691* MWT + 0.9862 *HT
TABLE 4.5 Parameter estimates and Standard errors with associated pvalues for Multiple Probit regression model
Parameter 

Estimate 
Standard 
Wald 95% 
Chi 
Pr > ChiSq 




Error 
Confidence Limits 
Square 










Intercept 

7.7412 
2.1975 
3.4342 
12.0482 
12.41 
0.0004 
AGE 

0.2630 
0.1289 
0.0103 
0.5157 
4.16 
0.0414 
SEX 

0.2141 
0.2843 
0.3432 
0.7714 
0.57 
0.4514 
GAGE 

0.1492 
0.0514 
0.2499 
0.0485 
8.44 
0.0037 
OCCU 
FARMING 
0.3718 
0.5967 
0.7977 
1.5413 
0.39 
0.5332 
OCCU 
STUDENT 
0.8129 
0.6843 
0.5282 
2.1541 
1.41 
0.2348 
OCCU 
TRADER 
0.0000 
0.0000 
0.0000 
0.0000 
. 
. 
MWT 

0.0457 
0.0268 
0.0068 
0.0982 
2.91 
0.0883 
HT 

0.6358 
0.7311 
0.7971 
2.0686 
0.76 
0.3845 
4.2.3 MULTIPLE PROBIT REGRESSION RESULTS
The multiple probit model was also used to examine the relationship between the likelihood of low birth weight and other variables. The results of the probit model were similar to the logit model except that this model gave more predictor variables to be significant at 5% and 10% levels of significance.
57
Gestation age and Age of the mother were statistically significant at 5% as depicted from table 4.5 above. The weight of the mother gained after last menstrual period (MWT) in table 4.5, although was not significant at 5% would be significant at 10%. From table 4.5, it can also be seen that the nonsignificant predictor variables at both 5% and 10% level of significance are: sex of the child, levels of occupation of the mothers and Hypertension status of the mother.
The fitted probit model is
Probit( π ) = φ−1 (π ) = 7.7412 + 0.2630*AGE + 0.2141*SEX – 0.1492*GAGE +0.3718
*FARMING + 0.8129* STUDENT + 0.0457* MWT + 0.6358 *HT
58
CHAPTER FIVE
SUMMARY, CONCLUSION AND RECOMMENDATIONS
5.1 SUMMARY
The study results highlight some factors that are significantly associated with mean actual birth weight and LBW. The data used for this study consist of eight predictor variables; age, occupation, hypertension, actual birth weight, low birth weight, sex, mothers weight and gestation age.
The preliminary results showed the results from our independent test of LBW with some predictor variables (hypertension, sex and occupation). It was reported that the predictors hypertension, sex and occupation were not statistically significant to our outcome variable, LBW.
Findings from this study show that the modal age group of the women who gave birth at the St Martin‟s Hospital was 20 to 25 years age (39%). Although this is the recommended reproductive age group, was responsible for the about 45.6% of low birth weight infants followed by Mothers aged < 20 years with 19.3% (appendix A.9.). This indicates that LBW incidence is largely a young adult phenomenon.
The study also indicates that about 60% of LBW cases were found among female infants. The mean actual birth weight of 2812 grams observed in this study. In the study it was found out that more than three quarters (78.9%) of all low birth infants to be preterm
59
(gestation age < 37 weeks) and 21.1 % of low birth weight deliveries were atterm(gestation age >= 37 weeks).
The findings also suggest that 68.4% of low birth weight infants observed among mothers who were farmers.
From the Multiple linear Regression model, Gestation age was significant with the mean actual birth weight. The regression coefficient of gestation age shows that there is a positive relationship between gestation age and mean birth weight. The regression coefficient estimate of gestation age ( 0.084) means that for one unit increase in gestation age, the mean actual birth weight would increase by 0.084 units, when other predictor variables remains identical or unchanged.
The other predictor variables age, categories of occupation, sex, mothers weight and hypertension were not statistically significant. This may mean that the mean actual birth weight do not depend lineally on these predictor variables.
From the Logistic regression model, Gestation age was significant with the Low birth weight. The regression coefficient of gestation age shows that there is a negative relationship between gestation age and Low birth weight indicating that the higher the gestation age the lesser the chance giving birth to an infant with low birth weight. This may mean that preterm infants were 1.3 times more likely to have low birth weights than full term infants.
Maternal age also showed positive association with low birth weight in the logistic model at 10% level of significance. The findings suggest that as maternal age increase, the risk of LBW also increase. The other predictor variables categories of occupation, sex, mothers‟ weight and hypertension were not statistically significant. This may mean that the Low birth weight do not depend on these predictor variables.
60
According to the probit model results of this study, gestation age also shows a significant negative relationship with LBW. Age of the mother as expected shows a significant positive relationship with LBW implying that the infants of younger mothers have lesser chance to get LBW than that of older mothers. Low birth weight and maternal weight also showed relationship in the probit model. The other predictor variables categories of occupation, sex and hypertension were not statistically significant. This may mean that the Low birth weight do not depend on these predictor variables.
5.2 CONCLUSION
From the whole study, the following conclusions are drawn from both our preliminary results and the results from our model in achieving our objectives;
The independent test of LBW versus some predictor variables showed that occupational status of the mother, sex of the infant and hypertension status were not associated with an infant‟s risk of being LBW.
The results from the three models indicate that gestation age was the predictor that has an influence on birth outcome. This may mean that the mean actual birth weight depends lineally on gestation age.
We also conclude from the probit and logit models that gestation age, mother‟s age and weight of the mother at last menstrual period are risk factors of low birth weight (LBW). It was observed that predictors like sex, occupation and hypertension status of the mother were not significant to the study.
61
5.3 RECOMMENDATIONS
Based on the findings from this study, the following recommendations are made to reduce the risk of an infant getting low birth weight.
It is recommended that Adolescent nutrition and reproductive health behaviours, maternal malnutrition for improved pregnancy outcomes and reversal of negative implications of LBW incidence in Amansie west district and for that matter Ghana, must be effectively addressed.
In addition, there is a need to explore the extent to which the incidence of low birth weight could be reduced by giving to all prospective mothers‟ access to antenatal care but keeping constant their other socioeconomic characteristics.
There is also the need for further research on first, an indepth study on the impact of several factors like multiple births, parity, number of antenatal visits, mother‟s educational levels qualitative studies on mothers‟ cultural and behavioural factors determining on birth size is necessary.
62
REFERENCES
1.Adelusi B, Ladipo O.A (1976). Preterm and other Infants with low birth weight in Ibadan. Trop. Geogr. Med., 28: 316322.
2.ACC/SCN (1999) Prevention of Foetal and Infant Malnutrition. Geneva: UN ACC/SCN Working Group.
3.ACC/SCN (2000) Fourth Report on the World Nutrition Situation. Geneva: ACC/SCN in collaboration with IFPRI.
4.Andersson, H. W., S. J. Gotlieb and K. G. Nelson (1997) Home environment and cognitive abilities in infants born smallforgestationalage. In: Acta Obstetrica et Gynecologica Scandinavia Supplement. 165, 76,826.
5.Angelsen, N. K., T. Vik, G. Jacobsen and L. S. Bakketeig (2001) Breast feeding and cognitive development at age 1 and 5 years. In. Archives of Disease in Childhood.
85,18388.
6.Arifeen S.E. (1997) Birth weight, intrauterine growth retardation and prematurity: a prospective study of infant growth and survival in the slums of Dhaka, Bangladesh Doctor of Public Health dissertation, Johns Hopkins University, Baltimore MD.
7.Ashworth A, Feachem R.G. (1985) Interventions for the control of diarrhoeal diseases among young children: prevention of low birth weight. Bulletin of the World Health Organization 63:165184.
63
8.Ashworth A. (1998) Effects of intrauterine growth retardation on mortality and morbidity in infants and young children. European Journal of Clinical Nutrition
52(Supplement 1):S34S42.
9. Astolfi, P. and L. A. Zonta (2002) Delayed maternity and risk at delivery. In: Paediatric and Perinatal Epidemiology. 16, 6772.
10.Bakketeig L.S. (1998) Current growth standards, definitions, diagnosis and classification of fetal growth retardation. European Journal of Clinical Nutrition
52(Supplement 1):S1S4.
11.Barker D.J.P, Osmond C. (1986) Infant mortality, childhood nutrition, and ischaemic heart disease in England and Wales. Lancet I:10771081.
12.Barker D.J.P. (1998) Mothers, Infants and Health in Adult Life. Edinburgh: Churchill Livingstone.
13.Barros F.C., Huttly S.R.A., Victoria C.G., Kirkwood B.R., Vaughan J.P. (1992) Comparison of the causes and consequences of prematurity and intrauterine growth retardation: a longitudinal study in southern Brazil. Pediatrics 90:238244.
14.Basso, O., J. Olsen, A. M. T. Johansen and K. Christensen (1997) Change in social status and risk of low birth weight in Denmark: population based cohort study. In: British Medical Journal. 315: 14981502.
15.Basso, O., J. Olsen and K. Christensen (1998) Risk of preterm delivery, low birth weight and growth retardation following spontaneous abortion: a registrybased study in Denmark. In: International Journal of Epidemiology. 27, 64246.
64
16.Blondel, B., M. D. Kogan, G. R. Alexander, N. Dattani, M. S. Kramer, A. Macfarlane and S. W. Wen (2002) The impact of the increasing number of multiple births on the rates of preterm birth and low birth weight: an international study. In: American Journal of Public Health. 92, 8, 132330.
17.Bréart, G. (1997) Delayed childbearing. In: European Journal of Obstetrics and Genecology and Reproductive Biology. 75, 7173.
18.Bukenya G., Barnes T., Nwokolo N. (1991) Low birthweight and acute childhood diarrhoea: evidence of their association in an urban settlement of Papua New Guinea.
Annals of Tropical Paediatrics 11(4):357362.
19.Cameron M., Hofvander Y. (1983) Manual on Feeding Infants and Young Children, ed. Series, Third ed. Oxford: Oxford University Press.
20.Carlson, E., J. M. Hoem and J. Rychtarikova (1999) Trajectories of fetal loss in the Czech Republic. In: Demography. 36, 3, 32737.
21.Carlson, E. and J. M. Hoem (1999) Lowweight neonatal survival paradox in the Czech Republic. In: American Journal of Epidemiology. 149, 5, 44753.
22.Cerqueiro M., Murtagh P., Halac A., Avila M., Weissenbacher M (1990) Epidemiologic risk factors for children with acute lower respiratory tract infections in Buenos Aires, Argentina: a matched casecontrol study. Reviews of Infectious Diseases
12(S8):S10211028.
23.Copper, R. L., R. L. Goldenberg, A. Das, N. Elder, M. Swain, G. Norman, R. Ramsey, P. Cotroneo, B. A. Collins, F. Johnson, P. Jones, A. Meir and the National Institute of Child Health and Human Development MaternalFetal Medicine Units
65
Network (1996) The preterm prediction study: maternal stress is associated with
spontaneous preterm birth at less than thirty five weeks gestation. In: American Journal of Obstetrics and Gynecology. 175, 5, 128692.
24.Datta N, Kumar V, Kumar L, Singhi S (1987) Application of case management to the control of acute respiratory infections in lowbirthweight infants: a feasibility study.
Bulletin of the World Health Organization 65:7782.
25. Dičkutė, J., Ž. Padaiga, V. Grabauskas, R. J. Nadišauskienė, V. Basys and A. Gaižaukienė (2004) Maternal socioeconomic factors and the risk of low birth weight in Lithuania. In: Medicina (Kaunas). 40, 5, 47582.
26.de Onis M., Blössner M. (1997) WHO Global Database on Child Growth and Malnutrition. Geneva: WHO.
27.de Onis M, Blössner M, Villar J (1998) Levels and patterns of intrauterine growth retardation in developing countries. European Journal of Clinical Nutrition 52(Suppl. 1):S5S15.
28.Dreyfuss, M.L., Msamanga, G.I., Spiegelman, D., Hunter, D.J., Urassa, E.J.N., Hertzmark, E. & Fawzi, W.W. (2001) Determinants of low birth weight among HIV infected pregnant women in Tanzania. American Journal of Clinical Nutrition 74, 814–
826.
29.FoixL‟Hélias and B. Blondel (2000) Changes in risk factors of preterm delivery in France between 1981 and 1995. In: Paediatric and Perinatal Epidemiology. 14, 314 23.
66
30.Fonseca W, Kirkwood BR, Victoria CG, Fuchs SR, Flores JA, Misago C (1996) Risk factors for childhood pneumonia among the urban poor in Fortaleza, Brazil: a case control study. Bulletin of the World Health Organization 74:199208.
31.Forsdahl A. (1977) Are poor living conditions in childhood and adolescence an important risk factor for arteriosclerotic heart disease? British Journal of Preventive Social Medicine 31:9195..
32.Forssas, E., M. Gissler, M. Sihvonen and E. Hemminki (1999) Maternal predictors of perinatal mortality: the role of birth weight. In: International Journal of Epidemiology.
28, 47578.
33.Goldenberg R., Hoffman H., Cliver S. (1998) Neurodevelopmental outcome of smallforgestational age infants. European Journal of Clinical Nutrition 52(S1):S54S58.
34.GranthamMcGregor S.M. (1998) Small for gestational age, term infants, in the first six years of life. European Journal of Clinical Nutrition 52(Supplement 1):S59S64.
35.Hack, M., D. J. Flannery, M. Schluchter, L. Cartar, E. Borawski and N. Klein (2002) Outcomes in young adulthood for very low birth weight infants. In: New England Journal of Medicine. 346, 3, 14957.
36.Haelterman, E., G. Bréart, J. ParisLlado, M. Dramaix and C. Tchobroutsky (1997) Effect of uncomplicated chronic hypertension on the risk of SmallforGestational Age Birth. In: American Journal of Epidemiology. 68995.s
37.Hales C.N., Barker D.J.P., Clark P.M.S., Cox L.J., Fall G., Osmond C. et al. (1991) Fetal and infant growth and impaired glucose tolerance at age 64. British Medical Journal
303:10191022.
67
38.Hass J, Balcazar H, Caulfield L (1987) Variations in early neonatal mortality for different types of fetal growth retardation. American Journal of Physical Anthropology
76:467473
39.Henriet, L. and M. Kaminski (2001) Impact of induced abortions on subsequent pregnancy outcome: the 1995 French national perinatal survey. In: British Journal of Obstetrics and Gynaecology. 108, 103642.
40.Hosmer, D.W., and Lemershow, S. (1980). A goodnessoffit test for the multiple logistic regression model. Communications in Statistics, A10, 10431069.
41.Jaakkola, J. J. K. and M. Gissler (2004) Maternal smoking in pregnancy, fetal development and childhood asthma. In: American Journal of Public Health. 94, 13640.
42.Jurjus, A.R. (1995) Low birth wight in Lebanon: morphological parameter and a health status indicator. Eastern Mediterranean Health Journal 1, 194–200.
43.Kallan, J. E. (1993) Race, intervening variables and two components of low birth weight. In: Demography. 30, 3, 489506.
44.Kelly, Y. J., J. Y. Nazroo, A. McMunn, R. Boreham and M. Marmot (2001) Birth weight and behavioural problems in children: a modifiable effect? In: International Journal of Epidemiology. 30, 8894.
45.Khoshnood, B., B. Blondel, C. de Vigan and G. Bréart (2004) Socioeconomic barriers to informed decision making: regarding maternal serum screening for down syndrome: results of the French National Perinatal Survey of 1998. In: American Journal of Public Health. 94, 3, 48491.
46.Klingenberg, C., Olomi, R., Oneko, M., Sam, N. & Langeland, N. (2003) Neonatal morbidity and mortality in a Tanzanian tertiary care referral hospital. Annals Paediatrics 23, 293299.
68
47.Kramer, M. S. (1987) Determinants of low birth weight: methodological assessment and metaanalysis. In: Bulletin of the World Health Organizations. 65, 5, 663737.
48.Kramer M. S. (1998) Socioeconomic determinants of intrauterine growth retardation.
European Journal of Clinical Nutrition 52(S1):S29S33.
49. Lack, N., J. Zeitlin, L. Krebs, W. Kűnzel and S. Alexander (2003) Methodological difficulties in the comparison of indicators of perinatal health across Europe. In: European Journal of Obstetrics and Genecology and Reproductive Biology. 111, S3344.
50.Larroque, B., M. Kaminski, P. Dehaene, D. Subtil, M. J. Delfosse and D. Querleu (1995) Moderate prenatal alcohol exposure and Psychomotor Development at Preschool Age. In: American Journal of Public Health. 85, 12, 165461.
51.Lawoyin T.O, Oyediran A.B.O. (1992). A Prospective Study on some Factors which Influence the Delivery of Low Birth Weight Infants in a Developing Country. Afr. J. Med. Sci., 21(1): 3339.
52.Lemeshow, S., Teres, D., Avrunin, J. S., and Pastides, H. (1988). Predicting the outcome of intensive care uniy patients. Journal of the American Statistical Association,
83, 348356.
53.Macfarlane, A. J., F. V. Price and E. G. Daw (1990) Infertility drugs and procedures.
In: B. J. Botting, A. J. Macfarlane and F. V. Price, eds, Three, four and more – A study of triplet and higher order births. 4658.
54.Martorell R, Ramakrishnan U, Schroeder DG, Melgar P, Neufeld L (1998) Intrauterine growth retardation, body size, body composition and physical performance in adolescence. European Journal of Clinical Nutrition 52(Supplement 1):S43S53.
55.Murthy H.V.V. (1991). Birth weight and the associated factorsa review. Ind J Soc
69
Work L11, 151156.
56.Prada J., Tsang R. (1998) Biological mechanisms of environmentally induced causes in IUGR. European Journal of Clinical Nutrition 52(S1):S21S28
57.Raum, E., B. Arabin, M. Schlaud, U. Walter and F. W. Schwartz (2001) The impact of maternal education on intrauterine growth: a comparison of former West and East Germany. In: International Journal of Epidemiology. 30, 8187.
58.Resnick, M. B., R. V. Gueorguieva, R. L. Carter, M. Ariet, Y. Sun, J. Roth, R. L. Bucciarelli, J. S. Curran and C. S. Mahan (1999) The impact of low birth weight, perinatal conditions and socio demographic factors on educational outcome in kindergarten. In: Paediatrics. 104, 6, e74.
59.Saigal, S., L. den Ouden, D. Wolke, L. Hoult, N. Paneth, D. L. Streiner, A. Whitaker and J. PintoMartin (2003) Schoolage outcomes in children who were extremely low birth weight from four international populationbased cohorts. In: Pediatrics. 112, 4, 943
60.Schieve, L., A., S. F. Meikle, C. Ferre, H. B. Peterson, G. Jeng and L. S. Wilcox (2002) Low and very low birth weight in infants conceived with use of assisted reproductive technology. In: New England Journal of Medicine. 346, 10, 73137.
61.Singer, L., T. Yamashita, L. Lilien, M. Collin and J. Baley (1997) A longitudinal study of developmental outcome of infants with Bronchopulmonary Dysplasia and very low birth weight. In: Pediatrics. 100, 6, 98793.
62.Smith, M. M., M. Durkin, V. J. Hinton, D. Bellinger and L. Kuhn (2003) Influence of breastfeeding on cognitive outcomes at age 68 years: followup of very low birth weight infants. In: American Journal of Epidemiology. 158, 11, 107582.
70
63.StevensSimon, C. and Orleans, M. (1999). Low birthweight prevention programs: the enigma of failure. Birth 26: 18491.
64.Sweet, M. P., J. E. Hodgman, I. Pena, L. Barton, Z. Pavlova and R. Ramanathan (2003) Two yearoutcome of infants weighing 600 grams or less at birth and born 1994 through 1998. In: Obstetrics and Gynaecology. 101, 1, 1823.
65.Tough, S. C., C. NewburnCook, D. W. Johnston, L. W. Svenson, S. Rose and J. Belik. (2002) Delayed childbearing and its impact on population rate changes in lower birth weight, multiple birth, andpreterm delivery. In: Paediatrics. 100, 3.
66.UNICEFWHO (2004) United Nations Children‟s Fund and World Health
Organisation, (2004) Low Birth Weight; Country and Global Estimates. UNICEF, New York
67.Vangen, S., C. Stoltenberg, R. Skjaerven, P. Magnus, J. R. Harris and B. Stray Pedersen (2002) The heavier the better? Birth weight and perinatal mortality in different ethnic groups. In: International Journal of Epidemiology. 31, 654660.
68.Vik, T., L. S. Bakketeig, K. U. Trygg, K. LundLarsen and G. Jacobsen (2003) High caffeine consumption in the third trimester of pregnancy: genderspecific effects on fetal growth. In: Paediatric and Perinatal Epidemiology. 17, 32431.
69.Villar J, Belizán J.M (1982). The relative contribution of prematurity and fetal growth retardation to low birth weight in developing and developed societies. American Journal of Obstetrics and Gynecology 143:793798.
70.Westwood, M., M. S. Kramer, D. Munz, J. M. Lovett, and G. V. Watters (1983) Growth and development of fullterm nonasphyxiated smallfor gestationalagenewborns: Followup through adolescence. In: Pediatrics. 71, 37682.
71
71. WHO (1997) National reports on the third evaluation of the implementation of “Health for All” strategies. New Delhi: WHO Global Database.
72.Yang, H., M. S. Kramer, R. W. Platt, B. Blondel, G. Bréart, I. Morin, R. Wilkins and R. Usher (2002) How does early ultrasound scan estimation of gestational age lead to higher rates of preterm birth?. In: American Journal of Obstetrics and Gynaecology. 1, 186, 3, 433 7.
73.Zarén, B., G. Lindmark and P. Bergsjø (1997) Hemoconcentration in smoking mothers is associated with impaired fetal growth. In: Actan Obstetricia et Gynecologica Scandinavica. 76, 93341.
74.Zarén, B., G. Lindmark and L. Bakketeig (2000) Maternal smoking affects fetal growth more in the male fetus. In: Paediatric and Perinatal Epidemiology. 14, 11826.
75.Zeitlin, J. A., M. J. SaurelCubizolles, P. Y. Ancel and the EUROPOP Group (2002) Msarital status, cohabitation, and the risk of preterm birth in Europe: where births outside marriage are common and uncommon. In: Paediatric and Perinatal Epidemiology. 16, 124130.
72
APPENDICES
APPENDIX A
Table A.1 Test for independence (hypertension vs low birth weight)


LOW 





NO 
YES 



HT 


Total 


NO 
157 
55 
212 


% 
71.04 
24.89 
95.93 














YES 
7 
2 
9 


% 
3.17 
0.90 
4.07 














Total 
164 
57 
221 



74.21 
25.79 
100.00 

Statistics for Table of HT by LOW
Statistic 
DF 
Value 
Prob 
ChiSquare 
1 
0.0625 
0.8026 
Likelihood Ratio ChiSquare 
1 
0.0644 
0.7996 
73
Table A.2 Test for independence (occupation vs low birth weight)
OCCU 
LOW 
Total 


no 
yes 

FARMING 
131 
39 
170 
% 
59.28 
17.65 
76.92 




STUDENT 
18 
12 
30 
% 
8.14 
5.43 
13.57 




TRADER 
15 
6 
21 
% 
6.79 
2.71 
9.50 




Total 
164 
57 
221 
%74.21 25.79 100.00
Statistics for Table of OCCU by LOW
Statistic 
DF 
Value 
Prob 
ChiSquare 
2 
3.9707 
0.1373 
Likelihood Ratio ChiSquare 
2 
3.7039 
0.1569 
Table A.3 Statistics for Table of SEX by LOW
Statistic 
DF 
Value 
Prob 




ChiSquare 
1 
0.2177 
0.6408 




Likelihood Ratio ChiSquare 
1 
0.2185 
0.6402 




74
TABLE A.4 Analysis of variance table for regression model

Analysis of Variance 



Source 
DF 
Sum of 
Mean 
F Value 
Pr > F 


Squares 
Square 


Model 
7 
17.53962 
2.50566 
7.29 
<.0001 
Error 
213 
73.22825 
0.34379 


Corrected Total 
220 
90.76787 



RSQUARE=0.20 





Table A.5: Model fit Statistics for Logistic regression
Model Fit Statistics
Criterion 
Intercept 
Intercept 

Only 
and 


Covariates 
AIC 
254.324 
115.204 
SC 
257.722 
145.788 
2 Log L 
252.324 
97.204 
Testing Global Null Hypothesis: BETA=0
Test 
ChiSquare 
DF 
Pr > ChiSq 
Likelihood Ratio 
155.1194 
8 
<.0001 
Score 
110.5747 
8 
<.0001 
Wald 
43.4891 
8 
<.0001 
75
TABLE A.6: Odd ratio estimates of logistic regression model
Odds Ratio Estimates
Effect 
Point Estimate 
95% Wald 



Confidence Limits 

AGE 
1.510 
0.920 
2.477 
SEX 
1.157 
0.406 
3.298 
BWT 
0.003 
<0.001 
0.020 
GAGE 
0.772 
0.625 
0.954 
OCCU FARMING vs TRADER 
1.142 
0.097 
13.511 
OCCU STUDENT vs TRADER 
2.483 
0.158 
39.008 
MWT 
1.072 
0.971 
1.183 
HT 
2.681 
0.213 
33.826 
TABLE A.7 criteria for assessing goodness of fit for logistic regression
Criterion 
DF 
Value 
Value/DF 
Deviance 
212 
97.2045 
0.4585 
Scaled Deviance 
212 
97.2045 
0.4585 
Pearson ChiSquare 
212 
2065.5977 
9.7434 
Scaled Pearson X2 
212 
2065.5977 
9.7434 
Log Likelihood 

48.6022 

TABLE A.8 criteria for assessing goodness of fit for probit (normit) regression
Criterion 
DF 
Value 
Value/DF 
Deviance 
212 
104.7543 
0.4941 
Scaled Deviance 
212 
104.7543 
0.4941 
Pearson ChiSquare 
212 
1836.6472 
8.6634 
Scaled Pearson X2 
212 
1836.6472 
8.6634 
Log Likelihood 

52.3772 

76
Table A.9. Summary of dataset

Birth weight 


Variable 



Birth 
`Birth weight < 2.5 kg 



weight>=2.5kg 

Total 




Sex of the baby Male 
72(32.6%) 
23(10.4%) 
95(43.0%) 




Female 
92(41.6%) 
34(15.4%) 
126(57.0%) 








Gestation Age < 37 wks 
81(36.7%) 
45(20.4%) 
126(57.0%) 
>= 37wks 



83(37.6%) 
12(5.4%) 
95(43.0%) 





Hypertension No 
157(71.0%) 
55(2.5%) 
212(95.9%) 
Yes 



7(3.2%) 
2(1.0%) 
9(4.1%) 





Occupation farmer 
131(59.3%) 
39(17.6%) 
170(76.9%) 
Trader 



15(6.8%) 
6(2.7%) 
21(9.5%) 

Student 



18(8.1%) 
12(5.4%) 
30(13.6%) 





Age of mother < 20 yrs 
38(17.2%) 
11(5.0%) 
49(22.2%) 
2025 



64(30.0%) 
26(11.8%) 
90(40.7%) 

2630 



33(14.9%) 
5(2.3%) 
38(17.2%) 

3134 



12(5.4%) 
8(3.6%) 
20(9.0%) 

> =35 



17(7.7%) 
7(3.2%) 
24(10.9%) 





77
Check
KNUST Main Page  www.knust.edu.gh
Related Post:
 KNUST Wassce Grading System
 KNUST World Ranking
 KNUST Weekend Masters Programmes
 KNUST Weekend Courses
 KNUST Academic Year
 KNUST Final Year Results
 KNUST First Year Fees
See Also:
2020/2021 Admission Forms For Public Universities
 2020/2021 UPSA Admission Forms
 2020/2021 GIMPA Admission Forms
 2020/2021 KNUST Admission Forms
 2020/2021 Ghana School Of Law Admission Forms
 2020/2021 University of Cape Coast UCC Admission Forms
 2020/2021 University Of Ghana UG Legon Admission Forms
 2020/2021 Ghana Institute Of Languages GIL Admission Forms
 2020/2021 Ghana Institute of Journalism GIJ Admission Forms
 2020/2021 University of Education Winneba UEW Admission Forms
 2020/2021 University for Development Studies UDS Admission Forms
 2020/2021 University Of Mines and Technology UMaT Admission Forms
 2020/2021 University of Health and Allied Sciences UHAS Admission Forms
 2020/2021 University of Energy and Natural Resources UENR Admission Forms