Background and Aim: The applicability of internationally-accepted infection control guidelines in rural Indonesia is questionable due to differences in resources availability and local contexts. Infection control barriers specific to rural Indonesia therefore need to be identified to assist in developing targeted infection control programs that are resource- and context-appropriate at the institutional, regional and national levels. Methods: This mixed methods study of ten healthcare facilities (hospitals and clinics, public and private) in a rural Indonesian district, all with severely limited resources, explored clinical practices in four areas: intravenous therapy, antibiotic usage, instrument reprocessing and hand hygiene. The quantitative methods used included univariate, multivariate and survival analyses of primary and secondary clinical data. The qualitative methods included a grounded theory analysis of observations, short and in-depth interviews and focus group discussion data. These two components of the study were synthesised to ascertain the magnitude and underlying factors of healthcare-associated infection risks and barriers to infection control programs. Findings: Inappropriate clinical practices caused excessive and unrecognised risks of primary bloodstream infection, surgical site infection, blood borne virus infection, and the development and spread of multi-resistant bacteria. The four diverse clinical practice areas exhibited common and interwoven underlying factors, which were: healthcare workers?? inadequate clinical knowledge, a lack of managerial support, and cultural beliefs shared by the healthcare workers and community members that prevailed over evidence-based knowledge. Non-clinical factors from inside and outside the healthcare facilities were inter-related and cannot be separated from the ensuing clinical practice inadequacies. An analytical framework that categorises infection control barriers into clinical and non-clinical domains as well as internal and external factors is therefore proposed to ensure a comprehensive infection control program design. Conclusion: The current clinically-focused infection control programs are likely to have limited and temporary results in rural Indonesia due to barriers being in the main non-clinical in origin. An effective and sustainable infection control program needs to concurrently address basic clinical practice improvements and the underlying managerial, attitudinal and cultural barriers. This public health aspect of infection control is often neglected yet crucial for the success of any infection control program in rural Indonesia.

Introduction The Breatheasy© Tracheostomy Program based at the Red Cross War Memorial Children's Hospital, Cape Town, manages children mostly from poor socio-economic backgrounds. In our resource-limited setting, it is unclear how these families cope with the demands of a tracheostomised child. We aim to assess the quality of life (QOL) of tracheostomised children and their families as the first study of its kind in a low-resource setting. Methods A descriptive, observational study was done to assess the QOL of tracheostomised children managed by the Breatheasy© Program over 10 months. Children with tracheostomies for longer than 6 months, complex syndromic children, and home ventilated children were included. The validated Paediatric Tracheotomy Health Status Instrument (PTHSI) was utilised, where a higher score, implied a better outcome. Results A total of 68 families were recruited. In 57 (85.1%) of the carers, the highest level of education achieved was primary or high school. Twenty-seven (42%) families reported having an annual household income of less than $675 US Dollars (ZAR10,000). Sixteen (24%) resided within informal housing. The mean scores for the 4 PTHSI domains were: physical symptoms 24.8/35 (70.9%), frequency/financial impact of medical visits 14.2/15 (94.7%), QOL of child 8.8/15 (58.7%), QOL of carer 62.3/85 (73.3%); the overall score was 110.2/150 (73.5%). There was no significant correlation between total PTHSI and annual household income, carer's educational status or type of housing. Children with a concomitant major medical condition had a significantly poorer total PTHSI scores (p-value 0.024). Conclusion Tracheostomy care compounds challenging socio-economic circumstances. In our experience, with adequate training, home-care nursing is not necessary. Despite difficult living conditions, the Breatheasy© Program empowers children and their families to live independently of the hospital system and appear to be thriving. The decision to perform a paediatric tracheostomy should not be influenced by the carer's education level, socioeconomic status, or on the basis of formal or informal housing. Children with major medical comorbidities represent a group that requires more support.

Traditionally, Machine Translation (MT) systems are developed by targeting fluency (i.e. output grammaticality) and adequacy (i.e. semantic equivalence with the source text) criteria that reflect the needs of human end-users. However, recent advancements in Natural Language Processing (NLP) and the introduction of NLP tools in commercial services have opened new opportunities for MT. A particularly relevant one is related to the application of NLP technologies in low-resource language settings, for which the paucity of training data reduces the possibility to train reliable services. In this specific condition, MT can come into play by enabling the so-called “translation-based” workarounds. The idea is simple: first, input texts in the low-resource language are translated into a resource-rich target language; then, the machine-translated text is processed by well-trained NLP tools in the target language; finally, the output of these downstream components is projected back to the source language. This results in a new scenario, in which the end-user of MT technology is no longer a human but another machine. We hypothesize that current MT training approaches are not the optimal ones for this setting, in which the objective is to maximize the performance of a downstream tool fed with machine-translated text rather than human comprehension. Under this hypothesis, this thesis introduces a new research paradigm, which we named “MT for machines”, addressing a number of questions that raise from this novel view of the MT problem. Are there different quality criteria for humans and machines? What makes a good translation from the machine standpoint? What are the trade-offs between the two notions of quality? How to pursue machine-oriented objectives? How to serve different downstream components with a single MT system? How to exploit knowledge transfer to operate in different language settings with a single MT system? Elaborating on these questions, this thesis: i) introduces a novel and challenging MT paradigm, ii) proposes an effective method based on Reinforcement Learning analysing its possible variants, iii) extends the proposed method to multitask and multilingual settings so as to serve different downstream applications and languages with a single MT system, iv) studies the trade-off between machine-oriented and human-oriented criteria, and v) discusses the successful application of the approach in two real-world scenarios.

Stroke survivors in low-resource settings like Rwanda often face high levels of disability, and access to rehabilitation care is limited. To effectively allocate resources, it is crucial to understand and address the most significant concerns of stroke survivors and explore contextually appropriate approaches to post-stroke care. We conducted a needs assessment survey of 337 patients from six hospitals in Rwanda, collecting data at discharge and three months post-stroke. Rwandan stroke survivors have similar unmet functional needs as those in high-resource settings. However, over half of the participants still reported moderate to severe mobility, usual activities, and social/recreational activities needs at three months post-discharge. Stroke survivors indicate that limited access to services was a significant barrier to addressing these needs. Community-based interventions may be an important method for addressing these needs; however, rehabilitation services are limited, warranting the need to consider alternative strategies to address these needs. The thesis explores the potential of involving community health workers (CHWs) in providing rehabilitation interventions for stroke survivors in community settings. Although the effectiveness of CHWs in providing physical rehabilitation interventions in low-resource settings remains uncertain and sustainability of these interventions outside the studied context is also unclear. Subsequent the use of CHWs for post-stroke rehabilitation is a worthwhile endeavor. To facilitate the implementation of CHW-delivered interventions, two crucial steps were considered: the development of an intervention and establishing an operational team that will ensure implementation success. A multi-phased process was used to design a new evidence-informed post-stroke community-level mobility intervention suitable for low-resource settings. The Rehabilitation Treatment Specification System enabled the intervention design and description toward facilitating its accurate replication. The next step is to test the intervention’s feasibility, effectiveness, and implementation in low-resources settings. / Thesis / Doctor of Philosophy (PhD) / The issue of post-stroke disability is significant in Rwanda due to limited resources for stroke survivors. To improve the resources available for stroke survivors in such settings, it is important to understand their specific needs and explore alternative approaches to provision of interventions. We surveyed 337 patients from six hospitals in Rwanda to describe their functional unmet needs after stroke and at three months. Within 90 days of stroke, Rwandan stroke survivors have more moderate to severe functional needs in almost all usual activities, which is twice compared to post-stroke unmet needs for stroke survivors living in areas with post-stroke resources. Three months after leaving the hospital, over half of the participants still have moderate to severe functional needs in mobility and other usual activities, working, and social/recreational activities. Rwandan stroke survivors identified that not being able to use therapy services made it difficult to address these needs. These data indicate it is crucial to focus on community-based interventions to address the needs of stroke survivors. In my research, I investigated whether Community Health Workers (CHWs) could administer rehabilitation treatment in resource-limited areas. Although the effectiveness of physical rehabilitation by CHWs is uncertain, there is potential for CHWs to participate in delivering rehabilitation. It is worth considering the use of CHWs for post-stroke rehabilitation. In order to successfully implement CHW-delivered interventions, two important steps were considered. First, an intervention was developed. Second, an operational team is being established to ensure the intervention’s success. The process of designing a new evidence-informed post-stroke community-level mobility intervention suitable for resource-limited areas, consisted of multiple phases. The Rehabilitation Treatment Specification System was used to design and describe the intervention accurately so that it could be replicated easily. The next phase involves examining if the intervention is practical, efficient, and can be successfully implemented in areas with limited resources.

In recent years, corpus based machine translation systems produce significant results for a number of language pairs. However, for low-resource languages like Urdu the purely statistical or purely example based methods are not performing well. On the other hand, the rule-based approaches require a huge amount of time and resources for the development of rules, which makes it difficult in most scenarios. Hybrid machine translation systems might be one of the solutions to overcome these problems, where we can combine the best of different approaches to achieve quality translation. The goal of the thesis is to explore different combinations of approaches and to evaluate their performance over the standard corpus based methods currently in use. This includes: 1. Use of syntax-based and dependency-based reordering rules with Statistical Machine Translation. 2. Automatic extraction of lexical and syntactic rules using statistical methods to facilitate the Transfer-Based Machine Translation. The novel element in the proposed work is to develop an algorithm to learn automatic reordering rules for English-to-Urdu statistical machine translation. Moreover, this approach can be extended to learn lexical and syntactic rules to build a rule-based machine translation system.

Student Number : 8403267 - PhD thesis - School of Pathology - Faculty of Health Sciences / Sub-Saharan Africa is the eye of the HIV epidemic. This study was conducted when treatment for the majority of HIV-infected patients in low resource settings was considered unattainable and the risks of diagnosing HIV often outweighed the benefits. Coupled with the complexities of HIV diagnosis in infancy, children typically were only diagnosed once already ill or not at all. Key strategies to address the paediatric epidemic focused on preventing mother to child transmission and reducing mortality and morbidity of infected children predominantly with co-trimoxazole prophylaxis. Both strategies required early diagnosis of HIV infection in infancy for monitoring prevention programs and identifying infected children respectively. The diagnostic algorithm for resource limited settings recommended the use of inexpensive, technically simpler HIV antibody detection assays that are unsuitable for use in HIV-exposed children under 12-months of age. Paradoxically this algorithm provided a barrier to HIV diagnosis in children because of high loss to follow-up rates and death in the first year of life. The objective of this study was to establish an accurate, affordable diagnostic algorithm for early diagnosis of HIV infection that could be rapidly implemented in South Africa and benefit other resource limited settings. The HIV infection status of 300 vertically exposed infants was determined according to first world criteria in a prospective, cohort study at Coronation Hospital, Johannesburg over 21 months. This status was used to assess the accuracy of clinical examinations and HIV assays in diagnosing HIV at 6-weeks, 3-, 7- and 12-months of age. The average cost of determining an infant’s HIV infection status was measured. A single HIV DNA PCR test at 6-weeks of age proved highly accurate in determining HIV status at a marginally increased cost to government and was incorporated by the South African Department of Health into national policy. The ultrasensitive p24 antigen assay and HIV antibody detection assays on serum and oral fluid were identified as valuable candidates where PCR testing is unavailable. Dried blood spot samples from heelpricks are critical for policy to be translated into practice since skills to perform venesection in 6-week old babies are limited. The next challenge lies in operationalising these findings at a clinical and laboratory level to the benefit of the 300 000 South African children annually exposed to HIV at birth. The urgency of early diagnosis has been increased by the availability of highly effective antiretroviral therapy.

The World Health Organization has recognised Autism Spectrum Disorder (ASD) as a public health concern and have recommended access to appropriate assessment and interventions. Psychoeducation and parent support soon after diagnosis are considered global best practice. Parent Education & Training (PET) programmes provide education, skills and support to parents. Despite the high need for PET there has been limited research in this field and few programmes are available, particularly in low- and middle-income countries (LMIC). This project aimed to 1) evaluate the evidence-base for a UK-developed PET (EarlyBird/EarlyBird Plus); 2) evaluate the evidence-base for a broader range of PET; 3) generate a framework for evaluating global PET programmes; 4) conduct a comparative feasibility study of two PET programmes in South Africa. The EarlyBIrd/EarlyBird Plus scoping review identified a low level of evidence for the efficacy of the PET, and showed that relatively little implementation science examination of the programmes had been performed to date. Review of a broader range of PET showed very similar findings, suggesting that the field of PET (rather than any specific intervention) was still relatively immature. We proceeded to use an implementation science and participatory approach to generate a multi-stakeholder evaluation framework for PET in future studies. A mixed-methods quasi-experimental design was then used for a comparative feasibility study of two PET in a low-resource South African setting. In the feasibility study parents/carers found both programmes to be acceptable and adaptable for a South African context. Limited efficacy-testing showed positive outcomes for parents, children and families. Application of the Evaluation Framework proved to be a useful structural technique to identify the strengths and weaknesses of PET across the implementation themes of outcomes, processes & procedures, and implementation landscape. Taken together, our results highlighted the relative infancy of this important field of ASD research and identified the need for multi-site, randomized controlled trials of PET, particularly in low-resource settings. In addition, results underlined the importance not only of efficacy of programmes, but of a range of implementation-related factors, that are crucial to ensure sustainable and scalable PET in real-life settings around the globe.

Cross-lingual transfer has been shown effective for dependency parsing of some low-resource languages. It typically requires closely related high-resource languages. Pre-trained deep language models significantly improve model performance in cross-lingual tasks. We evaluate cross-lingual model transfer on parsing Marathi, a low-resource language that does not have a closely related highresource language. In addition, we investigate monolingual modeling for comparison. We experiment with two state-of-the-art language models: mBERT and XLM-R. Our experimental results illustrate that the cross-lingual model transfer approach still holds with distantly related source languages, and models benefit most from XLM-R. We also evaluate the impact of multi-task learning by training all UD tasks simultaneously and find that it yields mixed results for dependency parsing and degrades the transfer performance of the best performing source language Ancient Greek.

Accessible flood modeling for low-resource, data-scarce communities currently does not exist. This paper proposes using drone imagery to compensate for the lack of other flood modeling data (i.e. streamflow measurements). Three flood models were run for Dzaleka Refugee Camp, located in Dowa, Malawi. Two of the models (the Soil and Water Assessment Tool (SWAT) and the Hydrologic Engineering Center River Analysis System (HEC-RAS)) are commonly used hydrological-hydraulic based models. The third model, the Water Caused Erosion Patterns (WCEP) model, was proposed by the author to capitalize on the high-resolution drone imagery using geological-geomorphological information. The drone imagery used in this study has a resolution of 3.5cm and shows erosion patterns throughout the refugee camp. By comparing the erosion patterns to flow direction of the surface, the erosion patterns were determined to be water caused or not water caused, the erosion patterns considered water caused were defined as high-risk flood areas, creating the WCEP model. The three models were compared using locations of collapsed houses throughout the camp. It was found that the WCEP model represents the location of collapsed houses significantly better (misclassification rate below 17%) than the SWAT or HEC-RAS models (misclassification rate below 54%, and 67% respectively). The WCEP model was combined with the best hydrological-hydraulic model (SWAT) to create a hydrogeomorphological model which capitalizes on both the drone imagery and the hydrological process. / Master of Science / The negative impact flooding has on communities can be reduced through flood modeling. But commonly used flood models are not accessible to data-scarce communities because of the historical data the models require. This paper explores using aerial imagery taken by a drone to make-up for the lack of historical data at Dzaleka Refugee Camp located in Dowa, Malawi. Drone imagery has a very high spatial resolution (3.5cm), so it is able to provide a lot of details, including marks that show an increase of flooding in certain areas and elevation information. The flood model presented in this paper is created using the found flood marks in drone imagery. The presented model is then compared to two commonly used flood models, and all three flood models are compared to locations of houses that collapsed from flooding throughout the refugee camp. The model created using drone imagery did the best job predicting high-risk locations with misclassification rates below 17%. The drone imagery model was then combined with a commonly used model to create a more comprehensive flood model, capitalizing on all available data.

Intergenerational relationships refer to relationships between people in alternate generational groups. The increasing numbers of older people all over the world highlight the need for studies on psychological processes that enhance well-being through intergenerational relationships. People from different generations are co-dependent for care and support. Older people constitute a diverse group: some may experience a depletion of energy, physical, financial and emotional resources, while others have good health, financial security and stable social support networks. What all people, including older people, have in common, however, is a need for meaningful relationships. This makes it worthwhile to explore the experiences of relationships with younger people from the perspective of older people. This study is important because research indicates that the relationships between older people (60 years and older) and adolescents and young adults are strained. An initial study on intergenerational care experiences from the perspective of people aged 60 years and older and young people in the middle-adolescent developmental phase (at 16 years) revealed a lack of empathy from the older generation towards the younger generation. It was decided to focus in this study on intergenerational empathy specifically. Mutual empathetic understanding is suggested as a means to overcome challenges between members of generations that have to compete for limited resources, especially in low-resource areas such as the specific rural African community studied in Vaalharts, South Africa. It is suggested that empathy could increase intergenerational understanding between older and younger people dealing with their respective developmental challenges. The research was conducted in a community with high levels of poverty and limited physical and material resources. The study is qualitative in nature and used an interpretive descriptive research design. This design provided the basis to explore the perceptions of empathy by using theoretical constructs to derive at an understanding of experiences in the community and specifically those of the older generation. The 18 participants were all older than 60 and Setswana speaking but were also proficient in Afrikaans and English. Eight of the participants (1 man and 7 women) participated in study using the Mmogo-method®. This is a projective technique that uses visual presentations to gain a deeper insight into meaning that is developed between the participants. For the Mmogo-method®, participants (older people) were invited to make visual representations of how they experienced care in relation to young people in the middle adolescent developmental phase, at around 16 years. A further 10 participants (4 men and 6 woman), who were not part of the Mmogo-method® project, participated in semi-structured interviews. These interviews provided them the freedom to elaborate on their own unique experiences. Semi- structured interviews also provide a certain amount of structure for the interviewer to make sure that detailed information on empathy in arising relevant conversations are obtained. Data were analysed using an inductive approach as well as thematic analysis of experiences about intergenerational care: empathy (and specifically, a lack of empathy) emerged as a recurrent theme. The literature shows a connection between empathy and care, which accordingly served as the focus of this study. The data were further subjected to a directed form of content analysis (deductive category application) which was conducted after literature and theory on intergenerational empathy have been consulted. A visual analysis was conducted on the visual presentations that the participants built as part of the Mmogo- method®. Trustworthiness was ensured by using the principles proposed by Guba’s model of trustworthiness and that include: obtaining thick descriptions, ensuring credibility, member checking and ensuring confirmabillity. Ethical consent to conduct this study was given by the Human Research Ethics committee of the Faculty of Health Sciences in the North-West University (reference number NWU-00053-10-S1). A gatekeeper was used to gain access to the community studied and participants gave their informed consent. The lack of empathy from the perspective or the older people emerged from the way in which older people viewed the relationship with the young solely from their own perspective and from their inability to place themselves in the position of the younger persons. Older people were at first unable to visualize the young in their caring relationships; they expressed negative and judgemental attitudes towards them and perceived that the young people are rejecting the wisdom of their elders. Older people resorted to outsiders to assist in communicating with and disciplining the young. The lack of empathy on the part of the older people for the young can have dire effects on their vital intergenerational relationships. It is recommended that community-based interventions be developed to promote empathy between the two generations. / MA (Research Psychology), North-West University, Potchefstroom Campus, 2015