SummaryBackgroundRobust evidence of the effectiveness of task shifting of antiretroviral therapy (ART) from doctors to other health workers is scarce. We aimed to assess the effects on mortality, viral suppression, and other health outcomes and quality indicators of the Streamlining Tasks and Roles to Expand Treatment and Care for HIV (STRETCH) programme, which provides educational outreach training of nurses to initiate and represcribe ART, and to decentralise care.MethodsWe undertook a pragmatic, parallel, cluster-randomised trial in South Africa between Jan 28, 2008, and June 30, 2010. We randomly assigned 31 primary-care ART clinics to implement the STRETCH programme (intervention group) or to continue with standard care (control group). The ratio of randomisation depended on how many clinics were in each of nine strata. Two cohorts were enrolled: eligible patients in cohort 1 were adults (aged ≥16 years) with CD4 counts of 350 cells per μL or less who were not receiving ART; those in cohort 2 were adults who had already received ART for at least 6 months and were being treated at enrolment. The primary outcome in cohort 1 was time to death (superiority analysis). The primary outcome in cohort 2 was the proportion with undetectable viral loads (<400 copies per mL) 12 months after enrolment (equivalence analysis, prespecified difference <6%). Patients and clinicians could not be masked to group assignment. The interim analysis was blind, but data analysts were not masked after the database was locked for final analysis. Analyses were done by intention to treat. This trial is registered, number ISRCTN46836853.Findings5390 patients in cohort 1 and 3029 in cohort 2 were in the intervention group, and 3862 in cohort 1 and 3202 in cohort 2 were in the control group. Median follow-up was 16·3 months (IQR 12·2–18·0) in cohort 1 and 18·0 months (18·0–18·0) in cohort 2. In cohort 1, 997 (20%) of 4943 patients analysed in the intervention group and 747 (19%) of 3862 in the control group with known vital status at the end of the trial had died. Time to death did not differ (hazard ratio [HR] 0·94, 95% CI 0·76–1·15). In a preplanned subgroup analysis of patients with baseline CD4 counts of 201–350 cells per μL, mortality was slightly lower in the intervention group than in the control group (0·73, 0·54–1.00; p=0·052), but it did not differ between groups in patients with baseline CD4 of 200 cells per μL or less (0·94, 0·76–1·15; p=0·577). In cohort 2, viral load suppression 12 months after enrolment was equivalent in intervention (2156 [71%] of 3029 patients) and control groups (2230 [70%] of 3202; risk difference 1·1%, 95% CI −2·4 to 4·6).InterpretationExpansion of primary-care nurses' roles to include ART initiation and represcription can be done safely, and improve health outcomes and quality of care, but might not reduce time to ART or mortality.FundingUK Medical Research Council, Development Cooperation Ireland, and Canadian International Development Agency.

Objective In South Africa, many HIV-infected patients experience delays in accessing antiretroviral therapy (ART). We examined pre-treatment mortality and access to treatment in patients waiting for ART. Design Cohort of HIV-infected patients assessed for ART eligibility at 36 facilities participating in the Comprehensive HIV and AIDS Management (CHAM) program in the Free State Province. Methods Proportion of patients initiating ART, pre-ART mortality and risk factors associated with these outcomes were estimated using competing risks survival analysis. Results 44,844 patients enrolled in CHAM between May 2004 and December 2007, of whom 22,083 (49.2%) were eligible for ART; pre-ART mortality was 53.2 per 100 person-years (95% CI 51.8-54.7). Median CD4 count at eligibility increased from 87 cells/mm3 in 2004 to 101 cells/mm3 in 2007. Two years after eligibility an estimated 67.7% (67.1% – 68.4%) of patients had started ART, and 26.2% (25.6% - 26.9%) died before starting ART. Among patients with CD4 counts <25 cells/mm3 at eligibility, 48% died before ART and 51% initiated ART. Men were less likely to start treatment and more likely to die than women. Patients in rural clinics or clinics with low staffing levels had lower rates of starting treatment and higher mortality compared with patients in urban/ peri-urban clinics, or better staffed clinics. Conclusions Mortality is high in eligible patients waiting for ART in the Free State Province. The most immunocompromised patients had the lowest probability of starting ART and the highest risk of pre-ART death. Prioritization of these patients should reduce waiting times and pre-ART mortality.

In recent times, South Africa has been witnessing insurgence of offensive and hate speech along racial and ethnic dispositions on Twitter. Popular among the South African languages used is English. Although, machine learning has been successfully used to detect offensive and hate speech in several English contexts, the distinctiveness of South African tweets and the similarities among offensive, hate and free speeches require domain-specific English corpus and techniques to detect the offensive and hate speech. Thus, we developed an English corpus from South African tweets and evaluated different machine learning techniques to detect offensive and hate speech. Character n-gram, word n-gram, negative sentiment, syntactic-based features and their hybrid were extracted and analyzed using hyper-parameter optimization, ensemble and multi-tier meta-learning models of support vector machine, logistic regression, random forest, gradient boosting algorithms. The results showed that optimized support vector machine with character ngram performed best in detection of hate speech with true positive rate of 0.894, while optimized gradient boosting with word n-gram performed best in detection of hate speech with true positive rate of 0.867. However, their performances in detection of other threatening classes were poor. Multi-tier meta-learning models achieved the most consistent and balanced classification performance with true positive rates of 0.858 and 0.887 for hate speech and offensive speech, respectively as well as true positive rate of 0.646 for free speech and overall accuracy of 0.671. The error analysis showed that multi-tier meta-learning model could reduce the misclassification error rate of the optimized models by 34.26%. INDEX TERMS Machine learning, South Africa, Twitter, hate speech, offensive speech. OLUWAFEMI ORIOLA received the B.Sc. degree (Hons.

Objective. To assess differences in access to antiretroviral treatment (ART) and patient outcomes across public sector treatment facilities in the Free State province, South Africa.Design. Prospective cohort study with retrospective database linkage. We analysed data on patients enrolled in the treatment programme across 36 facilities between May 2004 and December 2007, and assessed percentage initiating ART and percentage dead at 1 year after enrolment. Multivariable logistic regression was used to estimate associations of facility-level and patient-level characteristics with both mortality and treatment status.Results. Of 44 866 patients enrolled, 15 219 initiated treatment within 1 year; 8 778 died within 1 year, 7 286 before accessing ART. Outcomes at 1 year varied greatly across facilities and more variability was explained by facility-level factors than by patient-level factors. The odds of starting treatment within 1 year improved over calendar time. Patients enrolled in facilities with treatment initiation available on site had higher odds of starting treatment and lower odds of death at 1 year compared with those enrolled in facilities that did not offer treatment initiation. Patients were less likely to start treatment if they were male, severely immunosuppressed (CD4 count ≤50 cells/µl), or underweight (<50 kg). Men were also more likely to die in the first year after enrolment.Conclusions. Although increasing numbers of patients started ART between 2004 and 2007, many patients died before accessing ART. Patient outcomes could be improved by decentralisation of treatment services, fast-tracking the most immunodeficient patients and improving access, especially for men.S Afr Med J 2010; 100: 675-681b.

No abstract

No abstract

With the growing amount of data available in the digital age, it has become increasingly important to use automated methods to extract useful information from data. One such application is the extraction of events from news sources for the purpose of a quantitative analysis that does not rely on someone needing to read through thousands of news articles. Overseas, projects such as the Integrated Crisis Early Warning System (ICEWS) monitor news stories and extract events using automated coding. However, not all violent events are reported in the news, and while monitoring only news agencies is sufficient for projects such as ICEWS which have a global focus, more news sources are required when assessing a local situation. We used WhatsApp as a news source to identify the occurrence of violent incidents in South Africa. Using machine learning, we have shown how violent incidents can be coded and recorded, allowing for a local level recording of these events over time. Our experimental results show good performance on both training and testing data sets using a logistic regression classifier with unigrams and Word2vec feature models. Future work will evaluate the inclusion of pre-trained word embedding for both Afrikaans and English words to improve the performance of the machine learning classifier.

No abstract