School of Public Health (ETDs)
Permanent URI for this communityhttps://hdl.handle.net/10539/37953
Browse
156 results
Search Results
Item Early Adolescents’ Knowledge, Beliefs and Behaviours Regarding Gender and Sexuality in Rwanda: Implications for Their Sexual Experiences and Health Outcomes(University of the Witwatersrand, Johannesburg, 2024) Mbarushimana, Valens; Conco, Daphney NozizweBackground: Access to sexual and reproductive health and rights (SRHR) information during adolescence is a fundamental human right and helps adolescents fully develop their potential. Due to developmental changes, early adolescence (10-14 years) offers an opportunity to shape early adolescents' sexual behaviours through various opportunities for education about their sexual and reproductive health before their behaviours become well-established in late adolescence and adulthood. Their sexual and reproductive health and rights knowledge, beliefs and behaviours play a significant role and may affect their sexual experiences and health outcomes. Rwanda has established several policies to promote adolescents' sexual and reproductive health and introduced comprehensive sexuality education (CSE) in schools in 2016 to enhance young people's knowledge, skills, and values to allow them to decide responsibly about their sexual and social relationships, and to promote and sustain risk-reducing behaviours. There has been a significant focus on older adolescents’ sexual and reproductive health and rights and limited focus on early adolescents. Limited evidence on the sexual knowledge, beliefs, and behaviours of young adolescents resulted in framing this research, and there are no similar studies for this age group. This research explores the extent to which early adolescents aged 12-14 years are knowledgeable about their SRHR and assesses whether this knowledge and attitudes have implications for their sexual experiences and health outcomes. This study aimed to assess how SRHR policies in Rwanda address the specific information needs of early adolescents in general and explore the extent to which SRHR knowledge influences sexual behaviours, sexual experiences and health outcomes among this age group (12-14 years). Methods: This research was conducted in three phases, including the desk review, the survey with early adolescents aged 12-14 years (phase 2), and focus group discussions to complement findings from phases one and two and the integration of findings (phase 3). The study used a social- ecological framework adapted for adolescent sexual and reproductive health, whereby multiple factors at the individual, relationships, community and societal levels interact to influence early adolescents' SRHR knowledge, beliefs, behaviours, sexual experiences and health outcomes. The research started with the policy review using the policy triangle framework. This framework identifies four key elements of a policy: context, content, process and actors. The policy review process facilitated engagement with gender and sexuality education stakeholders. This allowed them to share their perspectives on best enabling access to gender and sexuality information for viii early adolescents. Findings from these stakeholders also allowed us to explore the content of SRHR messages conveyed to early adolescents in Rwanda through a review of the CSE as embedded in the competence-based curriculum. The International Technical Guidance on Sexuality Education guided this review. Finally, a survey with 811 early adolescents and 13 focus group discussions with 108 early adolescents allowed us to explore the extent to which this age group were knowledgeable about SRHR, their sexual experiences and health outcomes. Quantitative data were analysed using descriptive and inferential statistics, while qualitative data were analysed thematically. This research obtained ethical and other necessary approvals from research institutions controlling research in Rwanda and from the Human Research Ethics Committee (HREC - Medical) of the University of the Witwatersrand. Furthermore, signed informed consent from SRHR stakeholders and participants' parents or legal guardians and written informed assent from early adolescents were obtained from those who had received parental consent. Findings: SRHR policies generally focus on adolescent health and have little focus on gender and sexuality information and knowledge for early adolescents. SRHR stakeholders indicated that access to gender and sexuality information was determined by factors at the individual level, such as curiosity and age; the relationship level, such as parent-child interaction; the community level, such as the sources of information available, the scope of SRHR programmes, and the general SRHR stakeholders’ misunderstanding about gender, norms; and the societal level such as limited resources, policy-making pathways, unfriendly laws. The CSE review showed that, of the eight areas recommended by UNESCO, CSE for early adolescents in Rwanda focuses mainly on four areas: sexual and reproductive health; human body and development; values, rights, culture and sexuality; and understanding gender, respectively. There was little or no emphasis on the topics of the four other recommended areas of sexuality education, which were violence and staying safe, skills for health and well-being, sexuality and sexual behaviour, and relationships. Findings from stakeholder interviews support these results and suggest other topics to be included in the CSE curriculum because of controversies surrounding these topics in Rwanda. Of the 811 participants involved in the quantitative survey, the majority were females (51%) and 49% males. Their age range was 12-14 years (n=802), and most were 14 (55.4%). Early ix adolescents showed that their knowledge of SRHR was generally low: scores about knowledge of sexuality were below 50% (mean score: 2.8 out of 6, 95% CI: 2.7-2.9), 42% (mean score: 4.9 out of 12, 95% CI:4.8-5.1) about contraceptive methods, 65% (mean score: 8.4 out of 13, 95% CI: 8.2- 8.5) about HIV/AIDS and sexually transmitted infections, and 54% (mean score: 7.1 out of 14, 95% CI: 6.9-7.3) about condoms. Concerning attitudes, 90% of participants exhibited favourable attitudes (>50% score) towards sexual norms, and 97% had a gender-equitable attitude (>50% score). In the survey, about 81% of early adolescents reported a lifetime involvement in any of the nonpenetrative sexual experiences (hugging, holding hands, caressing, kissing on the lips, touching or being touched on genitals, stroking or being stroked genitals to climax). In comparison, 7% of participants reported having experienced penetrative sexual experience involving oral sex, anal and/or vaginal intercourse. Attendance at parties where young people dance and fathers having a lower level of education were associated with higher odds of non-penetrative sexual experience. Factors associated with the experience of penetrative sexual intercourse were being male, drinking alcohol, watching pornographic films three or more times, being a double orphan, discussing sex matters often with one’s father and being forced to have sexual intercourse. High scores on knowledge about condoms and positive attitudes towards sexual norms were associated with both non-penetrative and penetrative sexual experiences. Those with high scores on HIV/AIDS and STDs were less likely to report penetrative sex. Positive gender attitudes were associated with non-penetrative experiences. Concerning participants’ sexual behaviours, periodic abstinence and condoms were the commonly used methods of contraception by the participants, and a significant proportion of them reported having a boy/girlfriend/partner. None of the girls had been pregnant, but four of 311 boys (1.3%) reported impregnating girls or a woman. Nearly 70% of the participants have had an HIV test, and a few reported having had an STD. Knowledge about sexuality, contraceptive methods, and condoms, attitudes towards gender and sexual norms were associated with testing for HIV. Qualitative findings from the focus group discussion revealed that early adolescents benefited from the SRHR information and knowledge that shaped their sexual behaviours and improved their health outcomes. Concerning the relationship between various sources of SRHR information and participants' knowledge, beliefs and behaviours, this research found that schoolteachers were the primary and x preferred sources of SRHR information on puberty and reproduction. Girls preferred their mothers, female schoolteachers, sisters and other extended family members, such as aunts, as sources of SRHR information. In contrast, boys preferred obtaining SRHR information from their fathers and brothers, as well as from films, the internet and television. Conclusion: SRHR knowledge, attitudes, behaviours, experience and health outcomes of early adolescents should be put in their context for interpretation. SRHR policy does not adequately address their specific information needs, and most early adolescents access SRHR information but have a low level of SRHR knowledge. However, this knowledge is vital for their complete and safe sexual health development. Some early adolescents have started engaging in sexual experiences and face challenges related to poor sexual health outcomes. Recommendations include the use of a multi-sectoral approach to promote early adolescents' access to SRHR information and to address their specific needs by involving more stakeholders (e.g. early adolescents and community members) in SRHR policy-making processes, addressing the social-ecological factors hindering proper access to good quality of SRHR information, those associated with sexual experiences; and delivering comprehensive sexuality education as per recommended standards. Future research should focus on integrating the specific needs of early adolescents in the SRHR policies, exploring the challenges related to implementing comprehensive sexuality education (CSE) in school settings and assessing the effects of CSE on the learners’ behaviours and sexual experiences. Future research should also examine the role of faith-based organisations in imparting CSE and explore approaches to delivering high-quality SRHR information to early adolescents. Similar research should be conducted among out-of-school early adolescents aged 10-11 years and those attending schools using international curricula in Rwanda.Item Ageing with HIV: Psychological and Functional Wellbeing and its biopsychosocial determinants at the Kenyan coast(University of the Witwatersrand, Johannesburg, 2024) Mwangala, Patrick Nzivo; Abubakar, AminaBackground: Kenya, and sub-Saharan Africa (SSA) in general, are experiencing rapid demographic transitions, including the ageing of the HIV population. Intuitively, this has created a subgroup of vulnerable older adults living with HIV (OALWH) requiring an urgent response in research, policy, and programming to mitigate their complex and transitioning needs. Existing evidence on HIV and ageing, largely from European cohort studies, demonstrates that OALWH are vulnerable to ill health, especially psychological morbidities (e.g. mental and cognitive impairments), geriatric syndromes (e.g. frailty) and multimorbidity. For the last decade, the SSA region has witnessed noteworthy progress in ageing and health, e.g. the establishment of health and ageing cohorts in South Africa. Yet, current evidence on mental, cognitive, and geriatric syndromes among OALWH is inadequate from the region partly because of heterogeneous findings, few well-designed studies, and significant methodological limitations, e.g. small sample sizes, lack of comparison groups and few adapted/validated measurement tools. In Kenya, HIV and ageing is an emergent subject, and little is known about the burden and determinants of these impairments. To address some of these research gaps, this PhD thesis sets out to: (a) Understand the health challenges faced by OALWH and the coping strategies they use to confront these challenges on the Kenyan coast, (b) Document the burden and associated factors of common mental disorders (CMDs) and frailty among OALWH compared to their HIV-uninfected peers on the Kenyan coast, (c) Examine the acceptability, reliability, and validity of a local (Swahili) measure of cognitive function, the Oxford Cognitive Screen Plus (OCSPlus), among older adults and provide a preliminary understanding of their cognitive performance (by HIV status) and associated factors. v Methodology: We used a mixed methods research design to answer our research questions, employing both qualitative and quantitative approaches. The qualitative phase of the study recruited 34 OALWH receiving routine HIV care and treatment, together with their healthcare providers (n=11) and primary caregivers (n=11). Qualitative data were drawn from semi-structured in-depth interviews that were audio-recorded, transcribed verbatim, and analyzed using the framework approach. Data management was done in Nvivo software. The quantitative phase, a cross-sectional survey, involved 440 older adults aged ≥50 years (58% OALWH). Locally adapted and/or validated tools were used to assess CMDs, frailty and cognitive function. All quantitative analyses were conducted using STATA software. Proportions were used to estimate the prevalence of CMDs and frailty among OALWH and their HIV-uninfected peers. Logistic regression was used to examine the factors associated with CMDs and frailty. Different psychometric tests were used to assess the reliability and validity of the OCSPlus tool. Test-retest reliability was evaluated using intra-class correlations. We computed Pearson correlation coefficients to assess convergent validity between OCSPlus and conventional cognitive tests. Multiple linear regressions were used to examine correlates of cognitive function. Results: Our findings from the qualitative phase showed that OALWH face multiple mental health challenges (e.g. symptoms of depression, anxiety, cognitive complaints) and physical health challenges (e.g. hypertension, diabetes, ulcers, somatic symptoms, and functional impairments), which were often complicated by a host of psychosocial challenges including food insecurity, ageism, HIV-related stigma, loneliness, and financial difficulties. Many of the perceived risk factors for these challenges were observed to overlap across the three health domains. Five major themes emerged from analyzing participants’ narratives on coping strategies, including self-care, positive religiosity, social connectedness, generativity and identity and mastery. Our findings also revealed maladaptive coping strategies, including self-isolation, over-reliance on over-the-counter medication, delayed healthcare seeking and skipping meals and medications. In the cross-sectional dataset, we found no significant differences in the vi prevalence of mild depressive symptoms (24% vs 18%) and anxiety symptoms (12% vs 7%) among OALWH compared to their uninfected peers, respectively. However, the prevalence of frailty was significantly higher among OALWH (24%) than their uninfected peers (13%). After adjusting for biopsychosocial factors, HIV seropositivity was not independently associated with CMDs or frailty. In multivariable analyses, the factors associated with increased odds of CMDs and frailty included HIV-related stigma, ageism, loneliness, functional impairment, sleeping difficulties, increasing household HIV burden, chronic fatigue, increasing medication burden, advanced age (>70 years), higher waist/hip ratio, visiting traditional healers, history of combination antiretroviral therapy (cART) regimen change/interruption, and prolonged illness following HIV diagnosis. In contrast, factors associated with reduced odds of CMDs and frailty included easier access to an HIV clinic, residing in larger households, higher household income, having a social network of friends, and being physically active. Most of these factors were corroborated in the qualitative phase of the study. In our last objective, the OCSPlus measure was found to be acceptable, reliable, and valid. Concerning cognitive performance, OALWH presented with significantly lower mean scores on language and executive function compared to their uninfected peers; however, they (OALWH) performed significantly better on memory. Several factors ranging from behavioural and lifestyle, sociodemographic, medical or treatment factors and psychosocial factors were significantly associated with cognitive performance (using OCSPlus). Discussion: Due to effective antiretroviral therapy (ART), people living with HIV are ageing, and this is witnessed across the world wherever ART is readily available. While this represents the success of HIV care and treatment, those ageing with HIV confront numerous health challenges, some of which are unique to this population, as observed in this study. Older adults living with HIV on the Kenyan coast are vulnerable to mental, cognitive, and functional health challenges, which appear to be complicated by several psychosocial challenges. Despite the observed health challenges, OALWH are able to integrate several positive strategies to promote or address their mental health and well- being. The findings of this study are timely in addressing some of the existing gaps in vii understanding the mental, cognitive, and functional health of OALWH in the country and set the stage for further research and the development of relevant interventions and healthcare strategies to improve the care of these adults. OALWH need to be recognized as an unique subpopulation requiring targeted health and social services, given the unique health challenges observed in this study. Relatedly, there is a need to build the capacity of healthcare providers, for instance, in the screening and management of CMDs and frailty, at least for the OALWH at risk of poor health outcomes, e.g., those facing multiple psychosocial challenges and deteriorating HIV treatment indicators. The majority of healthcare providers interviewed in the qualitative phase lamented their lack of skills and capacity to screen and manage CMDs and cognitive impairments. Additional support may include helping OALWH and their caregivers to mobilize social support and referrals, where possible, to relevant institutions and increase access to basic needs and services. The Kenya HIV prevention and treatment guidelines also need to be updated to highlight the unique needs of OALWH, e.g. multimorbidity. Different models of care have been recommended for the care of OALWH, including comprehensive geriatric assessment and integration of chronic care services into routine HIV care services. Formative work can be conducted to assess the possibility of introducing these models into care. Future work needs to ascertain the cause-and-effect association of the reported correlates and carry out formative work on the feasibility of multicomponent psychological and physical interventions to address identified health problems, e.g. psychosocial challenges, CMDs and frailty.Item A Mixed Methods Analysis of the Implementation and Uptake of the Human Papillomavirus Vaccination of Adolescent Girls in Lusaka, Zambia(University of the Witwatersrand, Johannesburg, 2024) Lubeya, Mwansa KettyBackground and rationale: Cervical cancer is the most common female cancer in sub- Saharan Africa compounded by the high prevalence of HIV among women of reproductive age. The human papillomavirus (HPV), mainly transmitted sexually, and persistent infection with oncogenic strains is a necessary cause of cervical cancer. The HPVs largely cause mild and self-limiting disease in those who are immunocompetent. In 2006, the World Health Organisation (WHO) approved the use of highly effective and immunogenic HPV vaccines for prepubertal girls aged 9 to 14 years, naive to sexual intercourse. Further, in 2020 WHO launched the global call for cervical cancer elimination which includes vaccinating 90% of girls by age 15 by 2023. These policies focus on improving cervical cancer primary prevention. Despite the known vaccine effectiveness and policy pronouncements, uptake remains low in many LMICS including Zambia, where there is a cervical cancer highly prevalent. The low HPV vaccine uptake is partly due to low levels of knowledge about HPV and HPV vaccines, myths and misconceptions, and vaccine mistrust. Implementation barriers, such as low adoption, human resource constraints, and coordination, hinder the successful implementation of the HPV vaccination. Zambia shares a high burden of cervical cancer, yet HPV vaccine uptake is low. The demonstration project conducted between 2013-2017 reported suboptimal coverage. Additional health facility data for 2021 and 2022 after national rollout of the vaccination reported uptake of only 30% among 14-year-old eligible adolescent girls. Factors leading to this low uptake and implementation determinants are not very well understood in the Zambian context. Therefore, this PhD aimed to understand factors associated with the uptake and implementation of the HPV vaccination of adolescent girls in Lusaka and select and tailor strategies based on their acceptability and feasibility to mitigate identified barriers. viii Methodology: This research was embedded in behavioural and implementation research with multiple stakeholders. Based on existing literature, the social-ecological model (SEM) was used to develop the conceptual framework to ensure the study was deeply grounded in theory. The health belief model (HBM) and consolidated framework for implementation research (CFIR) were used to explore participants' behavioural and implementation dimensions, respectively, at the different levels of the SEM. A mixed methods approach was employed, incorporating semi-structured interviews with adolescents, parents, teachers, and healthcare workers and a cross-sectional survey of parents. Thematic analysis and descriptive statistics were used to analyse the data, guided by the CFIR and the HBM. The nominal group technique (NGT) was used to achieve stakeholder consensus on feasible and acceptable implementation strategies to increase HPV vaccine uptake. Secondly, a scoping review was conducted following an a priori protocol to understand the implementation strategies used to increase HPV vaccination of adolescent girls in the broader context of SSA. To have clear and replicable definitions, all the identified implementation strategies were coded according to the Expert Recommendation for Implementation Change (ERIC) taxonomy, grouped into clusters, and recorded their perceived feasibility and importance. Results: This PhD identified low awareness and misinformation as significant barriers to vaccine uptake among adolescents and parents. Cultural beliefs, fears about infertility, and logistical challenges within the healthcare system were substantial obstacles. Teachers and healthcare workers pointed to inadequate training and resources as crucial limitations in program implementation. Frequently used implementation strategies in SSA from the scoping review included building a coalition, changing service sites, distributing educational materials, conducting educational meetings, developing educational materials, using mass media, involving patients/relatives and families, promoting network weaving, staging implementation scale- ix up, accessing new funding, promoting adaptability, and tailoring strategies. According to expert consensus, most of the identified implementation strategies were multifaceted and of high importance and feasibility. Conclusion: The PhD highlights the need for targeted health education campaigns, access to clear information, improved coordination between stakeholders, and the development of flexible delivery models to reach out-of-school girls. Future research should evaluate the effectiveness of the proposed strategies to assess their impact on vaccine uptake in Lusaka, Zambia and within the region.Item Integrated methylome and transcriptome analysis of esophageal squamous cell carcinoma in the South African cohort(University of the Witwatersrand, Johannesburg, 2024) Moodley, MishalanEsophageal squamous cell carcinoma is highly prevalent in Eastern and Southern Africa, characterized by a high incidence and mortality. The etiology of Esophageal Squamous Cell Carcinoma is complex, and multiomic studies to improve our understanding of this deadly disease is severely lacking in this part of the world. DNA methylation is an epigenetic mechanism which plays an important role for gene regulation and frequently displays aberrant patterns in several cancers. This study aimed to identify aberrant methylation patterns in Black South African ESCC patients from matched tumour normal pairs. We designed an efficient workflow for the collection, storage, and processing of high quality ESCC specimens ensuring preservation of tissue architecture and sample quality. A total of 142 ESCC samples were collected. Methylome and transcriptome was investigated in 11 paired tumour and adjacent normal ESCC cases. Differential methylation analysis revealed significant alterations in CpG methylation across the genome. Principal component analysis effectively discriminated the tumour samples from adjacent normal tissues for most samples. The focus was on differentially methylated regions within enhancers and promoters using a methylation difference of ≥20% and an expression fold change difference of ≥2 (p values ≤0.05) between tumour and normal tissue. We identified three critical gene promoters (CDC42EP3, TNC and KRT13) and 1 enhancer (COL6A3) that exhibited significant differential methylation and a significant inverse relationship with gene expression, all of which plays a role in the ECM. From transcriptome data, there were 241 identified differentially expressed genes, of which 170 were upregulated and 71 genes that were downregulated and whose gene promoters were and hypomethylated and hypermethylated respectively. The top three candidate genes exhibited an 8-9-fold increase in expression and had the most significant FDR (<10-9 to 10-16), and are are all involved in the ECM pathway: FN1, POSTN and COL6A3. Transcription data also showed that FN1 and POSTN were also among the top 2 differentially expressed genes across all genes, with COL6A3 also making up the top 3 differentially expressed genes. v In summary this study sheds light on the methylome and gene expression changes in ESCC biased towards the ECM and highlights the importance of epigenetic modifications and their potential role in the disease. Understanding these molecular alterations is critical for developing targeted therapeutic strategies for ESCC.Item Computational approaches to characterizing morbidity and mortality patterns in rural South Africa(University of the Witwatersrand, Johannesburg, 2024) Mapundu, Michael Tondera; Celik, TurgayBackground: Verbal autopsies (VAs) are commonly used in Low to Middle Income Countries, as a way of determining the cause of death in cases where deaths occur outside health facilities and there is no medically certified cause of death. The VA process is usually done by conducting interviews with relatives of the deceased to elicit information about circumstances and events surrounding the death. The compiled VA narratives are then given to two doctors, and supplemented by the full set of responses, both from structured questions for assessment, in order to reach a consensus on the cause of death. In instances where they disagree, a third physician is consulted, a process known as Physician Coded Verbal Autopsy (PCVA). PCVA is the most used process for determining cause of death. However, it is widely criticized because of its lack of robustness, cost, time, inconsistencies, and inaccuracies as it is subjective and prone to errors among many drawbacks. Therefore, these challenges affect the accuracy of verbal autopsy results. Consequently, this results in PCVAs mostly employed for the training and validation of computational approaches. Despite these challenges, VAs have been employed successfully to estimate mortality rates and causes of death in settings where vital registration systems are weak or non-existent. Therefore, efforts are ongoing to improve the validity and reliability of verbal autopsies, including the use of computational approaches for analysing the data. There has been a growing interest from the VA community to apply automated algorithms that are artificially intelligent in order to improve cause of death determination using VA data, thus closing the civil registration gap. It has been proven that the use of machine learning (ML) and natural language processing (NLP) has helped identify patterns and trends in the data that might be missed by manual analysis, but are key in transforming the data into actionable insights that can help improve health outcomes. Study Objectives: The overall aim of this study was to utilize advanced computational methodologies to gain a comprehensive understanding of the complex health dynamics within rural South African communities. The study aims to bridge the gap between traditional statistical and epidemiological approaches and the unique challenges faced by these communities, thereby contributing to more informed public health strategies and interventions. As such, we sought to understand the determinants and circumstances of events leading to cause of death in rural north-east South Africa, using predictive and descriptive analysis. The main focus of this study was to answer the following questions crucial in VA monitoring and decision making: 1) what are the common prevalent topical diseases that led to death at Agincourt Health and Demographic Surveillance System (HDSS) in rural north-east South Africa between 1993 and 2015?; 2) how is mortality clustered by cause of death within households, and what characteristics of households are associated with high mortality at Agincourt HDSS? and 3) to what extent can machine learning and deep learning techniques accurately classify cause of death as compared to physician classification? In the process of our investigation we will also address the following sub questions: 1) how can machine learning and statistical modelling be effectively applied to analyse and predict mortality patterns within rural South African communities?; 2) what are the key determinants and socio-economic factors that contribute to variations in mortality rates?; 3) can computational techniques uncover hidden correlations between specific diseases and socio-economic indicators, providing insights into potential causal relationships?; 4) how can spatial analysis techniques be used to identify clusters of high mortality rates in rural South Africa, and what underlying factors might be driving these patterns?; 5) what role do access to healthcare resources, healthcare infrastructure, and healthcare-seeking behaviour play in shaping mortality outcomes in rural South African communities?; 6) to what extent do traditional health beliefs, cultural practices, and community dynamics influence morbidity and mortality patterns, and how can computational methods account for these factors; 7) how effective are data-driven models in predicting future morbidity and mortality trends in rural South vi Africa, and how might these predictions inform healthcare planning and resource allocation? and 8) what are the potential barriers and opportunities for scaling up successful computational approaches to other rural regions within South Africa or similar global contexts? The specific objectives of this study were; 1. To determine the most prevalent diseases that led to deaths using VA narrative datasets and text mining techniques at Agincourt HDSS in rural north-east South Africa between 1993 and 2015. 2. To identify mortality clusters by cause of death, establish determinants of mortality clusters and investigate the mortality characteristics associated with households at Agincourt HDSS between 1993 and 2015. 3. To establish ML accuracy in automating VA classification and achieve at least the same level of accuracy with that of physician classification on cause of death in rural north-east South Africa between 1993 and 2015. One of the aims of this study was to apply text mining techniques to derive implicit knowledge that is hidden in the unstructured VA narratives and present it in an explicit form. This allowed us to discover, mortality causes and most prevalent diseases which caused the population to succumb to death. Secondly, we sought to establish ML accuracy in automating VA classification and achieve at least the same level of accuracy with that of physician classification on cause of death. This was done through a comparative performance evaluation of ML methods and Computer Coded Verbal Autopsy (CCVA) algorithms on South African VA narratives data. Additionally, we also explored with novel deep learning architectures in order to generate cause of death prediction in a timely, cost effective and error free way. These computational techniques will make us achieve our aim of determining events and circumstances leading to cause of death by identifying morbidity and mortality occurrences in rural north-east South Africa from 1993 to 2015. As such, the study will ease the design, development, implementation and sustainment of tailored health intervention programmes. Consequently, this will improve life expectancy, turnaround time for diagnosis, and enforce a standardised VA reporting approach. This will therefore close the civil registration gap. Method: This study was a secondary data analysis of routinely collected VA data at Agincourt HDSS, for the period of 1993 to 2015. Agincourt HDSS is a surveillance site that specifically provides evidence based health monitoring that seeks to strengthen health priorities, practice and inform policy. In this study, we used three types of datasets. The first dataset is the structured responses from the standard questionnaire, second dataset is the VA narratives, and the third dataset is a combination of the responses and the narratives. The three datasets had 287 columns/features and 16338 records/observations. For the responses only, we took all features that had responses from the standard questionnaire as our predictors and the cause of death assigned by physicians using International Classification of Diseases-10 (ICD-10) code for each record in the dataset as our target variable. Ultimately, we had 231 predictors (all symptoms, age at death and gender) and 1 target variable, and all our features were in English. The predictions using the narratives were done using age at death, gender and the narrative feature and 1 target variable. For the combined VA dataset we used 232 predictors and 1 target variable. We only added the VA narrative feature to the responses dataset in order to have our combined dataset. We further created twelve cause of death categories with corresponding labels, class distribution with number of samples for each class before and after data balancing for our training dataset. The cause of death categories were derived based on InterVA user guide. The text mining and deep learning studies used the narratives only dataset, and the ML study employed all datasets. vii Results: ML models could accurately determine the cause of death from VA narratives, producing results comparable to expert diagnosis, with our optimal models attaining accuracies around 96%, with significant statistical differences in algorithmic performance (p < 0.0001). In the same way our robust novel stacked ensemble deep learning methods (SEDL) performed optimally than conventional DL approaches attaining an accuracy of 82% and employed Local Interpretable Model-agnostic Explanations (LIME) to enhance the interpretability of DL models, thus fostering trust in their use in healthcare. Our empirical results suggest that our automated approaches can be integrated in the CoD pipeline for identifying mortality causes, alongside human annotation, and interpretation. Additionally, through mortality trend and pattern analysis, we discovered that in the first decade of the civil registration system in South Africa, the average life expectancy was approximately 50 years. However, in the second decade, the life expectancy significantly dropped, and the population was dying at a much younger average age of 40 years. This suggests that the HDSS population succumbed to death due to mortality causes such as; vomiting/diarrhoea, chest/stomach pain, fever, coughing and high blood pressure. Interestingly, we found out that the most prevalent diseases entailed human immunodeficiency virus (HIV), tuberculosis (TB), neurological disorders, malaria, diabetes, high blood pressure, chronic ailments (kidney, heart, lung, liver), maternal and accident related deaths. Noteworthy, in the third decade, we see a gradual improvement in life expectancy, possibly attributed to effective health intervention programmes. Our sequential modelling on patient care seeking patterns, suggests that most people in the HDSS seek traditional ways than western ways for their healthcare, when faced with terminal illnesses. Additionally, we noticed that the narratives entail additional variables which can ease the cause of death diagnosis using sequential modelling and semantic and structure analysis, with a retrieval rate of approximately n > 2, per every case where n is number of terms. Through a structure and semantic analysis of narratives where experts disagree, we also demonstrate the most frequent terms of traditional healer consultations and visits. Therefore, this can possibly assist in determining cause of death specifically in the unknown category. Conclusion: This research study explores the utilization of computational techniques to analyse and comprehend morbidity and mortality patterns in rural South Africa, by leveraging large-scale VA data and complex computationa approaches to discern prevalent diseases, health disparities, and mortality trends within this specific context. These computational approaches, avail nuanced insights into disease prevalence, risk factors, and potential correlations between socio-economic factors and health outcomes. The findings of this study flags the potential of computational approaches in uncovering intricate health dynamics in rural settings, shedding light on areas for targeted interventions and policy enhancements. The study enforces the significance of harnessing data-driven strategies to inform public health strategies tailored to the unique challenges faced by communities in rural South Africa. The study represents a significant advancement in the field of CoD determination from VA narratives by introducing innovative ML and DL techniques that offer accurate and interpretable results. The findings suggest the potential for these models to streamline the VA reporting process, ultimately benefiting healthcare systems in LMICs by reducing diagnosis turnaround time, costs and improving the accuracy of CoD determination. Therefore, this research bridges the gap between the amount of data available and conducting research that can lead to practical actions, thus supporting multi- disciplinary research in civil registration systems using VA data. Consequently, it provides a baseline for future studies, generalising these findings to other domains of interest, thus highlighting the importance of improving health intervention programs in LMICs to increase life expectancy, and contribute to the understanding of mortality patterns and prevalent diseases in LMICs by harnessing the power of computational techniques.Item The assessment of patient-centred care among diabetic patients in southern Malawi(University of the Witwatersrand, Johannesburg, 2024) Makwero, Martha; Muula , A; Igumbor, JudeIntroduction: Patient centred care (PCC) is gaining recognition in various quality care reforms, especially in the growing era of chronic care where its effects are palpable yet its functional elements are unknown and hardly measurable. PCC has been shown to improve care processes and patient reported outcomes among chronic care patients such as Diabetes Mellitus (DM). While the Malawi Quality management Directorate (QMD) and the Non-communicable disease and injury (NCDI) policies mention PCC as a strategic quality reform, its functional elements are hardly known and, if it matters, in mediating patient experiences and outcomes. It is clear that the vagueness is thwarting PCC’s implementation, assessment, and advocacy. Thus, as the burden of chronic diseases, including DM grows, there is a need to optimize PCC through the elicitation of its functional elements, its objective assessment, and its recommended implementation strategies. Aim: This study aims to assess the functional elements of PCC and its relationship to outcomes in diabetic patients. Methodology: I conducted a cross-sectional, exploratory mixed methods study in southern Malawi. Sequentially, I employed qualitative and quantitative methods in order to enrich the contextual understanding of the complex PCC construct and its correlates. Therefore, we initially explored the common elements of PCC in LMIC through a scoping review. Building on the findings, in the next phase, I explored PCC conceptualization among three stakeholders (patients, health care providers and policy makers) further, breaking down the PCC construct to its functional elements through a qualitative inquiry. The qualitative themes identified were used to build a tool on important elements of PCC from the same population. A psychometric analysis was done in order to confirm the structure and consistency of the tool to improve the validity of results obtained. Thereafter, I quantitatively measured the current PCC practice ix among patients with DM and its association with baseline adherence to DM care plans, self- efficacy, glycaemic control, and some complications. The study used Covidence to manage the scoping review and Nvivo version 11 for the qualitative data, which was analyzed through the thematic analysis. The quantitative data was analyzed using STATA version 18 and R-studio version 4.2.3. The psychometric analysis employed exploratory and confirmatory factor analysis. Dichotomization of study participants into those who perceived low and high PCC scores was done through K-means clustering was done in R studio. Similarly, correlational studies and Path analysis were conducted to determine the nature and direction of relationships between the perception of PCC during the encounter and adherence, self-efficacy and glycaemic control. Results: The study confirmed the paucity and fragmentation of literature on PCC in LMIC especially in Africa and therefore, the need to situate the elements in our context. The themes identified were interrelated, and differences in expression and emphasis of some of the elements compared to the Eurocentric ones were highlighted. The recurring themes included the facilitating ambience, sharing of information, and patient involvement, the wider organization of care and the macroenvironment in which care takes place. Although they are similar to the Eurocentric themes, our qualitative inquiry identified eight themes and proposed a working definition. Thus, in the Malawian context, PCC was conceptualized as an expected care process that incorporates warm patient reception, where the healthcare provider consciously aims to reduce the patient-HCP power gap to harness a good long-term relationship. This creates a conducive atmosphere that allows the gathering of information that holistically identifies the individual specific problems and all possible interacting factors, ensuring timely access to care and medication. While patients highlighted the value of PCC encounters, particularly shared decision-making, the study documented the x challenges and vulnerabilities faced in an attempt to engage their providers in shared decision making. The psychometric analysis identified three latent themes underlying the PCC elements namely 1) relational aspects of care, 2) individualization and shared decision making and, 3) organizational aspects of care highlighting what is important and the hierarchical nature of the PCC construct. While acknowledging the need for further refinement, the tool exhibited acceptable reliability and validity properties to be used as a measurement framework in the study. The study showed significant deficiencies in the delivery of PCC with 55.8% of patients perceiving low levels, particularly, in the area of individualization and shared decision-making. Higher PCC scores did not have a significant relationship with self-efficacy. Both the perception of PCC and self-efficacy were positive predictors of adherence independently. Having perceived higher PCC scores was associated with a marginal 0.03-point increase in one’s adherence scores (β = 0.03; 95% CI: 0.01 to 0.04, p-value <0.001). Self-efficacy and adherence were both positive predictors of blood sugar control independently. Better self- efficacy was associated with a 0.03 unit decrease in the level of HbA1c (β =-0.03; 95% CI: - 0.04 to -0.022, p-value <0.001). Additionally, as adherence scores went higher, there was a 0.15 unit decrease in HbA1c (β = -0.15; 95% CI: -0.25 to -0.02, p-value <0.05). The results highlight that PCC can affect glycaemic control only indirectly through adherence. Conclusion The study has confirmed the complexity of the PCC construct and the paucity of literature in LMIC. It has highlighted that above and beyond creating a conducive ambiance, medical encounters ought to evoke support and capacitation for patients to undertake self-care behaviors confidently to be able to produce patient outcomes. Even though this study is cross- xi sectional, the initial patterns are promising for the value of PCC in mediating patient-reported experiences and outcomes among DM patients.Item Feasibility and acceptability of a contextualised physical activity and diet intervention for hypertension control in a rural adult population of South Africa(University of the Witwatersrand, Johannesburg, 2024) Sekome, Kganetso; Myezwa, Hellen; Gómez-Olivé , Francesc XavierOver half the population of adults from rural South Africa are hypertensive. Most strokes and ischaemic heart diseases in rural and remote South Africa are because of hypertension, which is a modifiable risk factor. Apart from pharmaceutical treatment, lifestyle changes such as increasing physical activity and reducing dietary salt have been strongly advocated for the control of hypertension. Despite the known benefits of physical activity and diet modifications for hypertension control, adults in rural South African settings still have high levels of uncontrolled hypertension. There is a lack of culturally sensitive, community-based, lifestyle interventions to control hypertension among rural South African adult populations. We designed, implemented, and evaluated a contextual intervention which recommends adjusting daily routine physical activity and dietary behaviour of adults with hypertension - HYPHEN (HYpertension control using PHysical activity and diEt in a rural coNtext). This study aims to evaluate the feasibility and acceptability of HYPHEN in a rural South African community setting. STUDY ONE Methodology: The first study involved establishing a profile of the frequency, intensity, duration, and domain (work or non-work) of self-reported physical activities of rural South African adults living with hypertension. Four hundred and twenty-nine adults diagnosed/self-report with hypertension aged 40 years and above completed the International Physical Activity Questionnaire Long-Form (IPAQ-LF) via telephone interview. Results: Fifty eight percent of participants were women of which 44% were in paid/unpaid work, while 64% of men were in paid/unpaid work. Men reported greater (duration and frequency) vigorous physical activity at work compared to women. Walking frequency as mode of transport was higher for working men and women compared with those not in work. Women reported high frequency of moderate- intensity physical activity inside the house and outside the house compared to men. Non-working men and women spent more time sitting during the week compared to their working counterparts. Conclusion: Sex, work status, and gender roles are related to physical activity among hypertensive adults living in rural South Africa. Further exploration of socioeconomic and cultural factors influencing physical activity in rural African populations is required so that appropriate contextual interventions can be developed and evaluated. vii STUDY TWO Methodology: The second study was a qualitative study to explore the social and cultural beliefs, perceptions and practices regarding physical activity and diet as a hypertension control intervention for adults living in a rural sub-district in South Africa. Nine focus group discussions were conducted with hypertensive adults using a semi-structured interview guide. Results: Participants had a lack of knowledge about blood pressure values. Perceived causes of hypertension were related to psychosocial factors such as family and emotional-related issues. Physical activity practices were influenced by family and community members’ attitudes and gender roles. Factors which influenced dietary practices mainly involved affordability and availability of food. To control their hypertension, participants recommend eating certain foods, emotional control, taking medication, exercising, praying, correct food preparation, and performing house chores. Conclusion: Lifestyle interventions to control hypertension for adults in a rural South African setting using physical activity promotion and dietary control must consider the beliefs related to hypertension control of this population. STUDY THREE Methodology: The third study was the design of a contextualised physical activity and diet intervention using behaviour change models. The intervention comprised of three components: a structured group education session, individualised physical activity education, and individualised dietary education over a period of 10 weeks. Participants were recruited via telephone. Results: Feasibility parameters included recruitment of 30 participants with a self-report hypertension diagnosis over one week, 80% participants retained to completion, 80% engaged with weekly monitoring calls. Acceptability was assessed qualitatively via face-to-face interviews to explore perceived expectations, benefits, motivation, and barriers concerning the intervention. Fidelity parameters included measures of adherence, dosage, quality of intervention delivery, and participant responsiveness. Secondary measures included blood pressure, body mass index, waist- hip ratio, urinary sodium, accelerometer-measured physical activity, and 24-hour diet recall. Conclusion: This study offers a person-centred, sociocultural approach to hypertension control through adaptations to physical activity and dietary intake. This study will determine whether HYPHEN is feasible and acceptable and will inform changes to the protocol/focus that could be tested in a full trial. viii STUDY FOUR Methodology: The fourth study sought to evaluate the feasibility and acceptability parameters of HYPHEN. Results: The evaluation demonstrated high level of feasibility, acceptability, and fidelity. Thirty (100% of target) participants were successfully recruited over two days, 28 (93%) participants were retained, and 28 (93%) provided complete data. Qualitative data demonstrated high acceptability. Fidelity measures demonstrated that adherence was high, dosage of intervention lasted as planned, no deviation was observed from the planned educational content, and all participants engaged with weekly telephonic calls. Conclusion: Although deemed feasible and acceptable, some amendments are required to the intervention procedures that were carried out before a full trial can be rolled out. Implications This study provides the first known evaluation of a contextualized physical activity and diet intervention for hypertension control in rural South Africa. This study shows that a contextually developed intervention for hypertension control has a promise of success if rolled out in a larger trial.Item Pre-exposure prophylaxis (PrEP) for HIV prevention in Eswatini: understanding the barriers, facilitators and opportunities for women(University of the Witwatersrand, Johannesburg, 2024) Bärnighausen, KathrynIn 2015 the World Health Organization confirmed the efficacy and safety of pre-exposure prophylaxis (PrEP) and recommended PrEP for use within populations at high risk of HIV acquisition. In Eswatini – the country with the highest incidence and prevalence globally – the Eswatini Ministry of Health, in partnership with Clinton Health Access Initiative (CHAI), designed a demonstration project which aimed to understand what was needed to successfully introduce PrEP into a national HIV prevention programme for the general population at risk for HIV. Here, using a PrEP prevention cascade to organise our findings, this thesis presents the content of five peer reviewed research articles which stem from a formative qualitative research component built within the demonstration project. Between 2017 and 2020 there were two rounds of rigorous data collection, including 217 semi-structured in-depth interviews, with adult (>18 years) Health Care Workers (HCWs) providing PrEP, relevant stakeholders and PrEP uptake, decline, discontinuation and continuing PrEP clients in Eswatini. Using a constructivist grounded theory approach, this thesis describes qualitatively where – along the cascade – gaps in service provision, demand creation, access, and retention in care for women (and others) are visible. The thesis highlights adaptations made by HCWs and recommendations from participants to address these gaps. The thesis describes the developed and adapted theoretical models we use to demonstrate where resilience can be used to create demand for PrEP, the structural, policy, community, personal and inter-personal levels that influence PrEP uptake, decline, continuance and discontinuance, and how a prevention-effective-adherence approach to PrEP could save resources, limit adherence burdens and mitigate negative perceptions related to stopping and starting PrEP. The work demonstrates the need for community-based demand creation, the inadequacies of the PrEP Promotion Package (PPP), and where changes were made to the PPP based on the analysis and interpretation of our data. While the findings have informed the successful national scale up of PrEP in Eswatini and have contributed to positive programmatic and implementation adaptations, it is also clear how PrEP clients – those most at risk of HIV infection – lie at the intersections of poverty, vulnerability, injustice and inequality. Ensuring the effectiveness of PrEP and the prevention of HIV acquisition requires multilevel approaches that extend far beyond a prevention cascade, and need to consider and incorporate the lived experiences of those in Eswatini and other high incidence settings globally.Item A nexus of student food (in)security, common mental disorders, and academic success in the midst of the covid-19 pandemic(University of the Witwatersrand, Johannesburg, 2024) Wagner, Fezile SthembileBackground: South African Higher Education Institutions (HEIs), akin to their global counterparts, grapple with the challenge of low throughput and dropout rates, as students often extend beyond the minimum required duration to complete their academic programmes. Research has identified the first year of study as the period carrying the highest risk for student failure and attrition. Factors such as prior academic performance, family background, and the student's ability to integrate into various aspects of university life emerge as significant determinants of academic success. Notably absent from these determinants are considerations pertaining to student wellbeing, encompassing mental health and food security status. These aspects assumed heightened awareness during the coronavirus disease of 2019 (COVID-19) pandemic, which posed threats to social, economic physical, and psychological wellbeing. For university students, the pandemic necessitated a swift adoption of a new pedagogical approach - Emergency Remote Teaching and Learning (ERTL) – along with most students being forced to relocate home, while simultaneously facing the harsh realities of mass job loss, illness, and grief brought on by the pandemic. Objective: This PhD aims to measure the impact of the COVID-19 pandemic on first time, first year university students’ food insecurity and mental health status (specifically, the common mental disorders (CMDs): anxiety, depression and mental distress symptoms), as well as to understand the implications of this on academic success. Methods: Taking place at a large, urban South African university, this research made use of a concurrent triangulation research design. Two cross-sectional surveys were administered in the years 2019 (before the pandemic) and 2020 (during the COVID-19 pandemic). Included in the surveys were validated tools used to measure depression (Patient Health Questionnaire 9-item (PHQ 9)), anxiety (Generalized Anxiety Disorder 7-item (GAD-7)), and mental distress symptoms. The Household Food Insecurity Access Scale (HFIAS) was also included in the two cross-sectional surveys to measure food insecurity levels. This tool was validated before being administered. Academic success was evaluated through two distinct approaches: i) the first method involved scrutinizing student failure rates and progress; ii) the second method focused on assessing retention and dropout rates. Qualitative data collection took place in 2020 and took the form of in-depth interviews (IDIs) and focus group discussions (FGDs). ii Results: Due to the lockdown directive, students living at home during studies increased from 29% in 2019 to 88% in 2020. In terms of the student failure rate, a reduction was noted from 23.4% (95% CI: 20.7- 26.3) in 2019 to 14.6% (95% CI: 12.0- 17.7) in 2020. Teaching staff acknowledged that students seemed to have a better grasp of complex content during ERTL, and feedback from students themselves indicated the benefits of being able to access and replay lecture recordings as well as the flexibility introduced by ERTL. Increased dropout levels were found - increasing from 5.5% (95% CI 4.2- 7.2) in 2020 to 10.5% (95% CI 8.2- 13.2) in 2021. Moderate food insecurity status (OR= 2.50; 95% CI: 1.12- 5.55; p=0.025), and severe mental distress symptoms (OR= 7.08; 95% CI: 2.67- 18.81; p<0.001) increased the odds of student dropout. The adjusted prevalence of food security was found to be better during the later time point - 30.3% (95% CI: 27.4- 33.4) in 2019 before the COVID-19 pandemic, to 37.9% (95% CI: 34.1- 41.9) 2020 during the COVID-19 pandemic. While the prevalence of CMD symptoms worsened over time; the prevalence of severe anxiety symptoms increased from 17.5% (95% CI: 15.2- 20.1) in 2019 to 25.4% (95% CI: 22.0- 29.1) in 2020. Conclusions: The study found an increase in student dropout and a decrease in failure rates during the pandemic, findings corroborated by other studies. Analyses suggest mental distress symptoms and food insecurity were important drivers of student dropout during the pandemic. Findings highlighted a clear decline of food insecurity when compared to figures before the pandemic. Literature highlighted how food and eating practices change when students are at home; this may have reduced food insecurity levels. A significant increase in the prevalence of CMD symptoms during the COVID-19 pandemic was observed and is believed to have been influenced by various intricate factors including grief, job loss and confinement. These findings provide important insights to HEIs in the event of future disruptions and as they embrace hybrid teaching and learning approaches post COVID-19. Although the reduction in failure rates may have been possibly due to changing the delivery of content (ERTL), HEIs must consider economic, social and mental health factors that may exclude certain groups of students when designing these approaches. It is crucial to explore ways to facilitate remote learning for students that address epistemological access challenges while maintaining inclusivity and connectedness as this is likely to contribute positively to academic success.Item Utilisation of maternal, newborn and child healthcare services in three sub-Saharan African countries (DRC, Kenya, and Tanzania) using Demographic Health Surveys data from 2007-2016: Application of Generalised Structural Equation and Machine Learning Models(University of the Witwatersrand, Johannesburg, 2024) Mlandu, Chenai; Musenge, EustasiusBackground: The risk of child deaths within the first month of life is elevated than the later stages of childhood. Globally, Sub-Saharan Africa (SSA) has the highest neonatal mortality. Majority of the countries in SSA including the DRC, Kenya and Tanzania are struggling to meet Sustainable Development Goal (SDG) 3.2 of reducing the neonatal mortality rate to 12 deaths per 1,000 live births by 2030 (2). Most causes of neonatal deaths are preventable and treatable. Universal coverage, timely and effective utilisation of maternal, newborn, and child healthcare (MNCH) services during pregnancy, delivery, and postpartum has the potential to save many lives of newborns in high-burden countries. vii Antenatal care (ANC) is the first service offered to pregnant women in MNCH. The timing and frequency of ANC visits is critical for the mother and her unborn child. The WHO recommends that women initiate ANC within 16 weeks of pregnancy and attend a minimum of four ANC visits for timely and optimum care before delivery (3, 4). The WHO also recommends that pregnant women receive assistance from a skilled worker during delivery and get postnatal checks with their newborns within 6 weeks of delivery (5, 6). Furthermore, utilising the Continuum of Care (CoC) for MNCH could significantly reduce maternal and newborn deaths in SSA. In the context of MNCH, the CoC is an approach that ensures continuous care from the period of pregnancy, through to childbirth, postnatal period, infancy, and the childhood period (7). Despite the recognition of the use of vital services in MNCH, timely and adequate uptake of MNCH services remains poor and the coverage of MNCH is far from universal in SSA. Most pregnant women initiate ANC after 16 weeks and hence fail to receive timely ANC interventions (8). Uptake of ANC visits, skilled birth attendance (SBA) and postnatal care (PNC) is suboptimal (8-11). Studies in SSA have explored various factors associated with MNCH services utilisation, however, our understanding of MNCH services utilisation in SSA is still limited. Trends in utilisation of MNCH services over time such as late ANC uptake have not been thoroughly assessed. Late uptake of ANC is still a common problem in SSA. Tracking women’s progress in the timing of ANC will ascertain if they are any changes in women’s late uptake of ANC and the contributing factors. This information will guide future policies and programmes which focus on improving the timely uptake of ANC in the SGD era. There is also a dearth of empirical evidence on the factors associated with the utilisation of ANC, skilled delivery and postnatal care in the CoC using nationally representative data. The CoC views both the mother and child as a collective rather than as separate/ individual entities. Understanding factors that viii contribute to the full utilisation of drop out from the CoC is essential for the formulation of interventions than enhance the CoC. Furthermore, studies which investigated either the individual utilisation of MNCH services such as timing of ANC, ANC visits, SBA and PNC services or the CoC have tended to use more of the traditional analysis methods such as the logistic regression. The application of more versatile analysis methods such as Machine Learning is not common. Machine Learning methods are capable of extracting information that commonly used methods (logistic regression) fail to do by uncovering hidden patterns and relationships, particularly in large data sets (12). The application of Machine Learning methods can offer opportunities of enhancing existing methods (conventional regression methods) for predicting and classifying MNCH utilisation leading to more effective interventions to improve MNCH utilisation. There is also a limited understanding on the interrelationships between MNCH services utilisation and neonatal outcomes. The associations between MNCH services utilisation and newborn outcomes such as neonatal mortality are commonly assessed using traditional approaches that assume direct associations. Specific analytical methods, such as Generalised Structural Equation Modeling (GSEM) can be used to model complex relationships such as interrelated links between utilisation of different MNCH services and neonatal outcomes. GSEM gives a clear understanding of how different services of MNCH are related to one another with neonatal outcomes by estimating both direct and indirect paths associations for more effective targeted interventions. Given the critical role of MNCH in ending preventable neonatal mortality, the overarching aim of this study was to describe the utilisation of MNCH services and their associations with neonatal mortality using GSEM and Machine Learning models in three sub-Saharan African countries: the DRC, Kenya, and Tanzania. ix Methods: The study utilised cross-sectional secondary data of reproductive-age women from the Democratic Republic of Congo (DRC) (2007-2013/14), Kenya (2008-2014) and Tanzania (2010-2015/16) Demographic Health Surveys. Firstly, the multivariate logistic regression analysed factors associated with late ANC initiation accounting for clusters, survey weights and stratification for the different rounds of the Demographic Health Surveys. Trends in late initiation of ANC over time in each country were assessed by comparing the earlier and later surveys using differences in prediction scores (prediction probabilities generated after running the multivariate logistic regression models). Secondly, the study assessed the main predictors of non-utilisation of PNC using the Decision Tree. The model performance of the Decision Tree was compared to the Logistic Regression using Accuracy, Sensitivity, Specificity and area under the Receiver Operating Characteristics. Thirdly, factors associated with the drop out from the MNCH continuum, defined as not fully utilising either ANC, SBA, or PNC services, were analysed using multivariate logistic regression accounting for clusters, survey weights and stratification. Machine Learning analysis was used to predict the drop out from the MNCH continuum using features (predictors) that were found significant in the multivariate logistic regression. Five classification Machine Learning models were built and developed including the Artificial Neural Network, Decision Tree, Logistic Regression, Random Forest and Support Vector Machine to predict the drop out from the MNCH continuum. The prediction accuracies of the models were then compared using parameters including Accuracy, Precision, Recall, Specificity, F1 score and area under the Receiver Operating Characteristics. Fourthly, the Generalised Structural Equation Modeling (GSEM) was used to assess the mediatory role of MNCH services utilisation on neonatal mortality. The endogenous variables x were ANC attendance, SBA and PNC attendance, low birth weight and neonatal mortality. The GSEM analysis also accounted for survey weights and considered cluster random effects. Results: The findings showed a reduction in late ANC initiation (67.8%-60.5%) between 2008-2014 in Kenya as well as in Tanzania (60.9%-49.8%) between 2010-2016, but an increase was observed in the DRC (56.8%-61.0%) between 2007-2014. In the DRC, higher birth order was associated with ANC initiation delays from 2007-2014, whilst rural residency, lower maternal education and household income was linked to ANC initiation delays in 2014. In Kenya, lower maternal education and household income was associated with ANC initiation delays from 2008-2014, whilst rural residency and increased birth order were linked to ANC initiation delays in 2014. In Tanzania, higher birth order and larger households were linked to ANC delays from 2010-2016, whilst ANC initiation delays were associated with lower maternal education in 2010 and lower-income households in 2016. The results also showed that the Decision Tree models had higher prediction accuracy of non- utilisation of PNC than the Logistic Regression models. Using the Decision Tree, low quality of ANC, home deliveries and unemployment were associated with the highest probability of not utilising PNC (92.0%) in the DRC. In Kenya, home deliveries, unemployment and lack of access to mass media were associated the highest likelihood of not utilising PNC (87.0%). In Tanzania, home deliveries, low quality of ANC and unwanted pregnancies exhibited the highest likelihood of not utilising PNC (100.0%). The results also revealed very high rates of dropping out from the MNCH continuum in the DRC (91.0%), in Kenya (72.3%) and Tanzania (93.7%). Rural residence, lower maternal education and non-exposure to mass media were common predictors of dropping out from the MNCH continuum across the three countries. Further, the influence of factors such as xi household wealth, household size, access to money for medication, travel distance to health facilities, and parity and maternal age varied by country. Results from the Machine Learning analysis showed that the Logistic Regression had the least prediction accuracy, while the Random Forest exhibited the highest prediction accuracy. Using the Random Forest, the study further ranked the most important predictors of the drop out from the MNCH continuum. Household wealth, place of residence, maternal education and exposure to mass media were the top four most important predictors. The results also showed direct and indirect associations between MNCH services utilisation and neonatal mortality. ANC attendance mediated the total effects of PNC attendance on neonatal mortality by 8.8% in Kenya and 5.5% in Tanzania. ANC attendance and SBA also sequentially mediated the total effects of PNC attendance on neonatal mortality by 1.9% in Kenya and 1.0% in Tanzania. The results in Tanzania also showed ANC attendance mediated 2.8% of the total effects of LBW on neonatal mortality. No presence of mediation was observed in the DRC; however, ANC attendance moderated the relationship between parity and neonatal mortality. Conclusions: The study found that late uptake of ANC decreased between the two survey rounds in Kenya and Tanzania but increased in the DRC. Women from various geographic, educational, parity, and economic groups showed varying levels of late ANC uptake. Increasing women’s access to information platforms and strengthening initiatives that enhance female education, household incomes, and localise services may enhance early ANC uptake. The Decision Tree models showed higher prediction accuracy of non-utilisation of PNC than the Logistic Regression models in the DRC, Kenya and Tanzania. Using the Decision Tree, women who had poor quality of ANC, home deliveries, unemployment, unplanned pregnancies, and no mass media access were identified as high-risk subpopulations of non- xii utilisation of PNC. Improving access and quality of care, incorporation of TBAs into the formal health systems, government health financing, increasing access to mass media and integrating maternal healthcare services with family planning services should be considered as top priority interventions to improving the utilisation of PNC. Most women and children drop out of the MNCH continuum in the DRC, Kenya and Tanzania. Rural residence, lower maternal education and non-exposure to mass media were common factors linked to the high dropout in the MNCH continuum. The use of Machine Learning can help support evidence-based decisions in MNCH interventions. Rapid response mechanisms such as web-based applications can also be developed through the use Machine Learning whereby a pregnant woman’s future utilisation of the services in CoC is assessed and monitored in real-time. The GSEM findings showed interconnections between MNCH services utilisation such as timing of ANC, ANC visits, SBA, PNC and neonatal mortality. This suggests that more than direct and indirect factors are accountable for the associations between MNCH services utilisation and neonatal mortality. The mediation role of MNCH services on neonatal mortality indicates critical areas for targeted interventions to reduce neonatal mortality. Overall, the study aimed to describe the utilisation of the MNCH services and associations with neonatal mortality in the DRC, Kenya and Tanzania. The study showed declines in late ANC uptake in two countries, however, early uptake of ANC is far is still not universal. The study also showed very low levels of retention in the CoC, and most women and children drop out in the CoC at postpartum period. The findings also showed the existence of social, health system and individual inequalities in MNCH and their impact on early childhood survival. Women who are vulnerable to unequal and poor MNCH services utilisation are characterised by poverty, rural residence, long travel distances to health facilities, unaffordable medical expenses, home deliveries, low quality of xiii care, low education, high parity, younger age, unemployment, limited exposure to mass media, and unplanned pregnancies. Context-specific intervention programs such as female education, government health financing, MNCH promotion programs through mass media and improved accessibility and quality of care in health facilities, particularly for the most vulnerable groups of the populations such as women of low socioeconomic status and women from underserved rural areas are essential to improve the overall health of mothers and children and meeting the SDG-3 goals. Modern biostatistical models like Machine Learning provide essential tools to understand public health problems. These techniques should be applied to complement the conventional statistical methods, particularly the tree-based models like the Decision Tree and Random Forest for predicting and classifying the utilisation of MNCH services. The GSEM established interconnections between timing of ANC, ANC visits, SBA and PNC and neonatal mortality. The timing of the first ANC contact is an important starting point to a continuation through the COC. It makes women better informed about pregnancy and the subsequent use of MNCH services. All stakeholders should work more on promoting early uptake of ANC by setting up initiatives that increase women’s access to information platforms, enhance female education, improve household incomes, and bring services closer to communities.