38981 research outputs found
Sort by
Credit scorecards in retail banking: enhancing interpretability through shapley values and evaluating the effectiveness of alternative data for improved accuracy
Get PDFThis research addresses the dual challenges of improving credit scorecard accuracy and maintaining interpretability. While machine learning algorithms like random forest and eXtreme gradient boosting outperform traditional logistic regression in accuracy, their complex predictor variable representation hinders interpretability. To reconcile this, the study discretizes numerical variables, applies one-hot encoding, and employs Shapley values to derive interpretable credit scores for random forest, eXtreme gradient boosting, light gradient boosting machine, and categorical boosting models. This approach produces credit scorecards that align with industry standards. Additionally, the investigation into the role of alternative data in credit scoring reveals its impact on model accuracy. By analysing unique predictor variables such as an applicant's social circle default status, regional ratings, and local population size, the significance of alternative data is demonstrated. Leveraging the model-X knockoffs framework for predictor variable selection contributes to superior model performance, achieving the highest area under the curve on the Kaggle home credit data
Analysing fuel transactions of government vehicles in the Eastern Cape, South Africa
Get PDFFuel management and fraud detection in government fleets are critical issues that have far-reaching financial and operational implications. To address these challenges, an investigation of fuel usage patterns and anomalies in the Eastern Cape Province government fleet in South Africa from April 2021 to January 2022 was conducted. Through the application of exploratory data analysis, clustering techniques, and predictive modelling, the research uncovers valuable insights that can be used to optimise fuel consumption and detect fraudulent activities within the fleet. Univariate and bivariate analyses reveal distinct patterns in fleet composition, transaction volumes, and fuel eJiciency across various vehicle makes, model derivatives, and departments. The use of clustering techniques enables the identification of distinct vehicle segments and transaction patterns, emphasising the importance of considering contextual factors when analysing fuel usage. To detect potential fraud, three key indicators are developed: abnormally large transactions, frequent transactions, and fuel price diJerences. Predictive models, including XGBoost, Multi-layer Perceptron, and Random Forest, are employed to automate the classification of transactions based on these fraud indicators. The Multi-layer Perceptron demonstrates the best performance, achieving an accuracy of 87% on the test set. The dissertation acknowledges limitations due to the scope of the data and missing information for certain geographic variables such as district and site. Future research could expand the geographical and temporal range, incorporate qualitative data, explore real-time monitoring systems, and investigate vehicle maintenance and fuel eJiciency. The present research makes a noteworthy contribution to the knowledge of fuel management and fraud detection in government fleets by oJering a data-driven approach to expose ineJiciencies and anomalies. The insights and methodologies presented serve as a foundation for future research and practical applications, ultimately leading to more eJicient, cost-eJective, and transparent fleet operations
The influence of normative whiteness in the South African private sector
Get PDFThirty years into democracy and notwithstanding the requirements of the Employment Equity Act (Act 55 of 1998), the South African private sector remains largely under the control of White executives, while Black African people constitute less than 20% of senior and top management (Commission for Employment Equity, 2023). Academic research (Myeza & April, 2021) as well as frequent anecdotal reports in mass and social media, indicate that Black professionals continue to experience racism in the private sector workplace, although this racism is often covert in nature and thus escapes the everyday understanding of the term. Studies by scholars of covert racism have established how normative Whiteness in society has operated to exclude or reduce the socio-economic prospects of Black people (Sue, 2010), but is rendered invisible through the establishment of ‘colour-blind' norms (Bonilla-Silva, 2018). However, there has been limited organisational research into how normative Whiteness, by maintaining an appearance of professional neutrality, perpetuates racism in the workplace (Nkomo, 1992, 2021). In addition, few studies on racism in organisations include the testimony of White executives as the primary data source. In this thesis I attempt to address this gap in the literature by using hermeneutic phenomenology within a critical theory framework to analyse how 35 White male and female executives at senior and top management level perceive: (i) Black professionals and their lived experiences and (ii) themselves and other White professionals; in the context of the employment equity imperatives of the South African private sector. As a White South African woman with over 15 years of experience in executive management, I include a self-reflexive approach both in the conducting of the interviews with White executives, as well as in the phenomenological analysis of the interview data. The significance of my research is that it shifts the study of racism to the locus of workplace power- White executives - and assesses how the normalisation of covert racism and Whiteness influences the perpetuation of organisational discrimination against Black professionals. My research shows that, notwithstanding ostensible support for the idea of employment equity, White executives' perceptions of Black people are shaped by stereotypes which perpetuate covert racism and the marginalisation of Black (particularly Black African) professionals. Concomitantly, White executives unknowingly establish Whiteness as a neutral standard for professional advancement and success in the South African private sector workplace
Growth and determinants at two years of age in a South African birth cohort
Get PDFAim: Early childhood is a critical period for optimal growth but there are limited data from low- and middle-income countries (LMICs) on the determinants of growth during this time. This study aimed to investigate growth and its determinants in children at 2 years of age in a South African birth cohort in a resource-limited region. Methods: Mother-child pairs enrolled in a South African birth cohort, the Drakenstein Child Health Study (DCHS), were followed from birth to two years. Comprehensive socio-demographic and psychosocial data collected during the antenatal period and nutritional information, intercurrent illness and immunisation data collected longitudinally in the first 2 years was used. Anthropometric data, measured at 2 years, were analysed as z-scores adjusted for child age, sex, and prematurity, and used to classify undernutrition (wasting, underweight for age and or stunting), overweight for age or obesity. Logistic regression was done to identify predictors of undernutrition, overweight or obesity at 2 years. Results: Anthropometric data from 897 children (51% male; 22% HIV-exposed, uninfected) were included. 69 (8%) were underweight, 20 (2%) were wasted and stunting occurred in 167 (19%) children; 116 children were overweight (13%), and 42 were (5%) obese. Overall, 351 (39%) had some form of growth impairment. Higher birth weight, or length, higher maternal height or better maternal education were associated with lower odds of undernutrition or stunting at 2 years. Male sex, antenatal maternal smoking or prematurity were associated with increased odd of undernutrition or stunting. Higher birth weight was associated with increased odds of being overweight for age. Better maternal education was associated with increased odds of obesity, but antenatal smoke exposure reduced the risk. Conclusion: There is a substantial burden of malnutrition in South African children during early childhood encompassing both undernutrition and obesity. Stunting, indicative of chronic growth impairment, is most prevalent. Antenatal risk factors are important predictors of poor growth outcomes. Strengthened strategies to improve childhood nutrition and address modifiable factors in the antenatal and early childhood periods are needed
Pulmonary aspergillomata: an assessment of clinical outcome after surgical resection and of the most immune response to aspergillus infection at cellular level
Get PDFA pulmonary aspergilloma, one of the forms of pulmonary disease caused by the ubiquitous fungus Aspergil/us species, is a secondary fungal infection imposed on pre-existing pulmonary disease. It is commonly associated with haemoptysis, which may be life-threatening 8•9•12•17- 19 •30. Colonisation of pre-existing cavities due to previous infection with mycobacterium tuberculosis is well described 1•2. In South Africa, due to the high incidence of pulmonary tuberculosis 73, pulmonary aspergillomata are frequently encountered
Postpartum contraceptive use in HIV-positive women who started antiretroviral therapy in pregnancy in Cape Town
Get PDFObjective: This study described factors associated with at least one episode of unmet need for contraception up to 24 months postpartum in women who began ART during pregnancy in Gugulethu, Cape Town. Methods: This was a retrospective review of data from a broader trial of antiretroviral drug delivery to women living with HIV during the postpartum period. Between 11 January 2016 and 10 November 2017, a total of 409 women living with HIV who had been on ART for at least three months, were at least 18 years old and within 28 days of delivery were participated in the study. Interviews were conducted to get information on family planning use and pregnancy intentions. The proportion of women who do not want to get pregnant in the future and in the next 12 months and are not using contraception was used to estimate the unmet demand for contraception. The factors associated with unmet need for contraception up to 24 months postpartum in women who began ART during pregnancy were investigated using binary logistic regression. The outcome of interest is the episode of unmet need for contraception. Results: Contraceptive use was high at each visit with the uptake of 97.80% at the first visit (enrolment). The overall prevalence of unmet family planning need was 19.60%. Women who are not married in this study were (OR=2.09, 95% CI: 1.26-3.47) more likely to have an unmet need for family planning and women who had been pregnant twice or more were (OR=2.24, 95% CI: 1.66-3.03) more likely to have an unmet need for family planning compared to the once who had been pregnant once. Conclusion: Contraceptive use among these women was high. The findings highlight that unmet need remains a source of concern in unmarried women and those who have been pregnant more than once. The findings can help drive national public health strategies to meet unmet contraceptive needs in women living with HIV through improving family planning initiatives and more effectively integrating family planning services into HIV treatment facilities
The impact of the Covid-19 pandemic and various lockdown levels on emergency medical services response times to trauma incidents in the Western Cape Province, South Africa
Get PDFThe outbreak of the COVID-19 crisis presented an immense challenge to both private and public health institutions and brought to the fore the importance of emergency medical services. The COVID-19 crisis provided evidence of the importance of out-of-hospital care for the lives of the patients depending on access to prompt medical care. Emergency Medical Services (EMS) serve as the first step in accessing the healthcare system for many within South Africa and as such played a pivotal role in the continuation of initiating medical care during the COVID-19 pandemic. The importance of EMS within the healthcare system was highlighted during the pandemic due to increased pressures placed on an already overstrained healthcare system
The impact of overconfidence bias on firm performance: an empirical analysis of S&P 1500 firms
Get PDFIn the context of public firms competing for profitable growth, there is abundant evidence that a high proportion of strategic decisions are value destructive. When managers make strategic decisions to position their firms for competitive advantage, they do so in a context of uncertainty and, in so doing, they apply heuristics or intuition, which makes them vulnerable to cognitive biases. We focused on overconfidence bias and its impact on firm performance. Overconfidence is considered to manifest in three distinct forms, namely overestimation, overprecision, and overplacement. While there are numerous gaps and conflicts in the literature, the most notable issues are conflicting results on the overall impact of overestimation, a paucity of studies on the impact of overprecision, and a focus on the overconfidence of the CEO, rather than on the management team making the strategic decisions. Our review of theory and empirical studies led us to hypothesise a negative relationship between firm performance and overestimation, and a positive relationship with overprecision. We examined the impact of overestimation and overprecision by analysing the performance of the largest public firms, excluding financial services, in the United States over the period 2009 to 2019. We measured firm performance using industry-adjusted return on assets. We derived proxies for overestimation and overprecision from management earnings guidance. Since strategic decisions play out over various time horizons, we considered one to five-year lags between our treatment and dependent variables. We estimated our panel data using a year and firm fixed-effects model. We found support for our hypotheses with a sizeable effect size and statistical significance. We also considered the combined impact of overestimation and overprecision and found that the “specific pessimist” management teams, those that express the greatest underestimation (negative overestimation) and the greatest overprecision, deliver the best firm performance. In contrast, the “vague optimist” management teams deliver the worst firm performance. Our theoretical contribution enriches the debate on the overall impact of group-level overestimation and overprecision on firm performance. Our findings also have practical implications in corporate governance, investment strategies, and the recruitment and promotion of senior executive
Mapping perioperative randomised controlled trials in Sub-Saharan Africa: a scoping review
No full textBackground: Surgical and anesthesia care is an integral component of universal healthcare coverage. In Sub-Saharan Africa (SSA), 93% of the population lacks access to essential surgical services. Post-surgery mortality in Africa is double the global average. The involvement of anesthesiologists is crucial for improved outcomes. Perioperative research can produce context-specific solutions to challenges faced in the perioperative period. SSA conducts fewer randomized controlled trials (RCTs) than high-income countries, limiting its contribution to global evidence. Objectives: Our primary objectives were to document the geographical distribution of included RCTs, describe their characteristics, and evaluate the reporting quality using the Consolidated Standards of Reporting Trials (CONSORT)-2010-checklist. Methods: We followed the PRISMA Scoping Reviews (PRISMAScR) Checklist. We searched MEDLINE, the Cochrane Library, and Scopus. We identified anesthesia-related RCTs within SSA published from 2000 to 2022. Two independent reviewers screened potential studies, and extracted data in duplicate, with disagreements resolved through consensus or a third reviewer. Quantitative analysis was done with STATA 16 and we compared RCT quality pre-CONSORT-2010 to post-CONSORT-2010, using Pearson's chi-squared test or Fisher's exact test (as applicable), considering p<0.05 as statistically significant. data were summarized narratively Results: Of 3319 records, 169 eligible RCTs were identified, randomizing 45376 participants, with a mean sample size of 98. Between the years 2000 and 2022, there was an exponential trend towards an increasing number of RCTs in SSA (y = 1,5619e0,1051x). The RCTs were from 16 countries in SSA. Most studies were single-country, single-center, led by authors from Nigeria (63/169, 37.3 %) and South Africa (41/169, 24.3%). Most interventions were conducted intraoperatively (n = 125/169, 74%). Pharmacotherapy interventions were most investigated (n = 64/169, 37.9%), followed by analgesic interventions (n = 42/169, 24.9%). The surgical discipline most investigated was obstetrics (n = 51/169, 30.2%). The reporting quality was generally poor, with most RCTs not adhering to CONSORT guidelines and failing to register on a trial registry. Conclusion: This scoping review provides a comprehensive overview of anesthesia-related RCTs in SSA, highlighting limitations such as small sample sizes, under-representation of high surgical burden disciplines, and poor reporting of study outcomes. The clinical trial capacity is limited to a few countries and institutions, and the methodological quality remains poor despite reporting guidelines. There is an opportunity to enhance context-appropriate RCTs in SSA by prioritizing high-quality research through collaborative efforts. Our findings serve as a resource for researchers, funders, and policymakers in anesthesia research in Africa to improve future RCT designs and reporting
The incidence of subsequent pregnancies in women enrolled in the DolPHIN 2 studies
Get PDFBackground Inadequate safety and efficacy data for medications during pregnancy, particularly for antiretroviral therapy (ART) regimens, stems from the exclusion of pregnant and postpartum women from clinical trials. Despite global efforts, challenges persist in enrolling this population. In the context of antiretroviral therapy, (ART) the shift from efavirenz (EFV) to dolutegravir (DTG) as the World Health Organization-recommended (first-line regimen) occurred with limited pregnancy data. In some clinical trials, becoming pregnant resulted from withdrawal from the trial, emphasizing the ongoing challenges in including pregnant women in trials. Furthermore, longer term follow-up beyond the duration of trials has been lacking. The objective of this secondary analysis is to provide an interim pathway, of retaining women who subsequently become pregnant and obtaining comprehensive pregnancy data, allowing pregnant women to continue participating in clinical trials. However, evidence on the incidence of subsequent pregnancies among women in such trials is limited and thus we aim to describe it. Methodology This secondary analysis included data from the 250 women (South Africa n=114, Uganda n=135) who met the intention to treat (ITT) criteria in the DolPHIN-2 study and were actively followed up over time. This study focuses on describing the incidence of subsequent pregnancies in a cohort of women enrolled in the DoLPHIN-2 randomized trial and its observational extension study D2 TRIO. The aim of this trial was to explore the safety and efficacy of DTG in pregnant women aged ≥18 years of age, presenting late (≥28 weeks gestational age) for antenatal care. Women were randomised 1:1 to receive DTG or EFV. Data from South Africa and Uganda collected between January 22, 2018, and September 5, 2023, were analysed descriptively. Results This secondary analysis included data from the 250 women (South Africa n=114, Uganda n=135) who met the intention to treat (ITT) criteria in the DolPHIN-2 study and were actively followed up over time. Within this cohort of women, 21% (n=53) experienced at least one subsequent pregnancy, with 7% (n=18) during the clinical trial and 14% (n=34) during the observational extension study. A higher proportion of subsequent pregnancies occurred in the EFV arm (57%, n=30) and at the Uganda site (58%, n=31) compared to the South African site (42 %, n=22). Characteristics of women with subsequent pregnancies revealed a majority being multigravida (87%, n=46), unmarried (90%, n=40), with a high school education (70%, n=33), and unemployed (57%, n=25). Among women with outcome data, 66% (n=23) had a live birth, 11% (n=4) had a termination of pregnancy, and most infants had a normal birth weight (90%, n=19). Conclusion This secondary analysis demonstrated that a high proportion of women had a subsequent pregnancy during the study period. There are calls for pregnant women to be involved in more safety and efficacy medications studies to broaden medication availability to this population. In the interim, instead of excluding women who because subsequently pregnant from clinical trials we recommend that they are retained. This study demonstrates that there may be sufficient numbers who have subsequent pregnancies, who typically would have been excluded, who can shed light on valuable early pregnancy information we would not otherwise have access too