Utility of a machine-guided tool for assessing risk behaviour associated with contracting HIV in three sites in South Africa.

Introduction: Digital data collection and the associated mobile health technologies have allowed for the recent exploration of artificial intelligence as a tool for combatting the HIV epidemic. Machine learning has been found to be useful both in HIV risk prediction and as a decision support tool for guiding pre-exposure prophylaxis (PrEP) treatment. This paper reports data from two sequential studies evaluating the viability of using machine learning to predict the susceptibility of adults to HIV infection using responses from a digital survey deployed in a high burden, low-resource setting. Methods: 1036 and 593 participants were recruited across two trials. The first trial was a cross-sectional study in one location and the second trial was a cohort study across three trial sites. The data from the studies were merged, partitioned using standard techniques, and then used to train and evaluate multiple different machine learning models and select and evaluate a final model. Variable importance estimates were calculated using the PIMP and SHAP methodologies. Results: Characteristics associated with HIV were consistent across both studies. Overall, HIV positive patients had a higher median age (34 [IQR: 29-39] vs 26 [IQR 22-33], p < 0.001), and were more likely to be female (155/703 [22%] vs 107/927 [12%], p < 0.001). HIV positive participants also had more commonly gone a year or more since their last HIV test (183/262 [70%] vs 540/1368 [39%], p < 0.001) and were less likely to report consistent condom usage (113/262 [43%] vs 758/1368 [55%], p < 0.001). Patients who reported TB symptoms were more likely to be HIV positive. The trained models had accuracy values (AUROCs) ranging from 78.5% to 82.8%. A boosted tree model performed best with a sensitivity of 84% (95% CI 72-92), specificity of 71% (95% CI 67-76), and a negative predictive value of 95% (95% CI 93-96) in a hold-out dataset. Age, duration since last HIV test, and number of male sexual partners were consistently three of the four most important variables across both variable importance estimates. Conclusions: This study has highlighted the synergies present between mobile health and machine learning in HIV. It has been demonstrated that a viable ML model can be built using digital survey data from an low-middle income setting with potential utility in directing health resources.

The readability of informed consent forms for research studies conducted in South Africa.

BACKGROUND: Informed consent forms (ICFs) are used to obtain consent from participants. However the complexity and comprehensiveness of these forms may not be appropriate. Readability can be quantified by formulas in Microsoft (MS) Word, such as the Flesch Reading Ease test. The South African (SA) ethics guidelines suggest that the MS Word Flesch-Kincaid Reading Grade score should be used to assess the complexity of ICFs and should be the equivalent of grade 8 level, or lower. OBJECTIVES: To use readability formulas to determine whether current SA ICFs are appropriate for the general population. METHODS: This was a descriptive study of a sample of English ICFs (solicited from our studies, as well as from local researchers) which received approval from local ethical review boards during the past 5 years, for prospective (≥6 months) drug studies that explored treatment and prevention of HIV, tuberculosis, diabetes or cardiovascular disease. ICFs were evaluated in MS Word for Flesch Reading Ease and Flesch-Kincaid Reading Grade, with the Simple Measure of Gobbledygook (SMOG) index calculated using www.readabilityformulas.com. Recommended targets for easy readability are above 60 for the Flesch Reading Ease score, and less than or equal to a grade 8 reading level for the Flesch-Kincaid Reading Grade and SMOG. RESULTS:   A total of 75 consent forms from 35 individual research studies conducted in SA over the last 5 years were included. The consent forms had been approved by six ethics committees across seven of the SA provinces. The median (interquartile range (IQR)) Flesch Reading Ease score was 55.8 (48.7 - 59.7) and 18 (25.0%) of the ICFs had easy or standard readability, while the median (IQR) Flesch-Kincaid Grade was 10.2 (8.8 - 11.4), with 23 (30.6%) at least a grade 8 level or lower. The median (IQR) SMOG index was 9.8 (9.0 - 11.1) and 4 (5.3%) scored below grade 8 level. CONCLUSIONS: Two-thirds of the ICFs from this study fail to meet the SA readability standard, a result matched by using alternative readability formulas. Readability can be improved with simple techniques and by actively monitoring readability metrics.

Descriptive analysis of World Health Organization-recommended second-line antiretroviral treatment: A retrospective cohort data analysis.

BACKGROUND: World Health Organization guidelines recommend that HIV patients who do not achieve viral suppression on efavirenz-based first-line antiretroviral therapy (ART) should be changed to a protease inhibitor (PI)-based regimen. In South Africa (SA), ~200 000 people are on second-line treatment, but little is known about these patients. OBJECTIVES: To describe second-line black African patients in a large urban area. METHODS: A quantitative retrospective study of 825 second-line patients in central Johannesburg, SA (subdistrict F), was performed with data extracted from government databases. Demographic characteristics, treatment status and laboratory information were gathered, then analysed with CD4+ cell count, viral load (VL) and retention-in-care data as outcome variables. RESULTS: The average recorded time to VL measurement after the switch to a PI-based ART regimen was 20 months, and 83.1% (570/686) of patients with a recent VL achieved viral suppression while on second-line treatment. The most recent median CD4+ cell count for the cohort was 286 cells/µL (interquartile range 160 - 478), which represented a 177 cells/µL increase from the baseline count at the start of first-line ART. Slightly less than three-quarters (72.4%) of the population remained active in care in the study clinics from initiation on first-line ART. Demographic characteristics such as being <25 years of age, male sex and geographical transfer (started initial treatment in a different region) independently predicted low CD4+ cell counts and virological failure on second-line treatment. Patients with virological failure were most likely (odds ratio (OR) 3.13, 95% confidence interval (CI) 1.50 - 6.56) to be lost to follow-up after the switch, while patients from Hillbrow Community Health Centre (OR 0.27, 95% CI 0.16 - 0.44), South Rand Hospital (OR 0.24, 95% CI 0.12 - 0.47) and Jeppe Clinic (OR 0.38, 95% CI 0.16 - 0.88), three larger sites, were most likely to remain active in care. CONCLUSIONS: VL suppression was high in patients on second-line treatment, but one-fifth of patients were lost to follow-up. Younger age, male sex and transfer from other treatment sites predicted poor treatment outcomes, highlighting opportunities for prioritisation of adherence interventions.

Access to and utilisation of healthcare services by sex workers at truck-stop clinics in South Africa: A case study.

BACKGROUND: Sex worker-specific health services aim to respond to the challenges that this key population faces in accessing healthcare. These services aim to integrate primary healthcare (PHC) interventions, yet most services tend to focus on prevention of HIV and sexually transmitted infections (STIs). North Star Alliance (North Star) is a public-private partnership providing a healthcare service package in roadside wellness clinics (RWCs) to at-risk populations along transport corridors in sub-Saharan Africa. OBJECTIVES: To inform future service development for sex workers and describe North Star's contribution to healthcare provision to this population in South Africa, we describe services provided to and utilised by sex workers, and their views of these services. METHODS: Using a mixed-methods approach, we present quantitative analyses of anonymised North Star routine data for sex workers for October 2013 - September 2015, covering nine sites in seven provinces. Clinic visits were disaggregated by type of service accessed. We performed thematic analysis of 25 semi-structured interviews conducted at five clinics. RESULTS: A total of 2 794 sex workers accessed RWCs during the 2 years. Sex workers attending clinics were almost exclusively female (98.2%) and aged <40 years (83.8%). The majority were South African (83.8%), except at Musina, where the majority of clients were Zimbabwean. On average, sex workers visited the clinics 1.5 times per person. However, in most cases only one service was accessed per visit. PHC services other than for HIV and STIs were accessed more commonly than HIV-specific services and STI treatment. There was an increase in the number of services accessed over time, the figure almost doubling from 1 489 during the first year to 2 936 during the second year. Although during recruitment participants reported having had sex in exchange for goods or money during the past 3 months, not all participants self-identified as sex workers during interviews; however, all reported feeling at higher risk of poor health than the general population owing to their involvement in sex work. Participants reported satisfaction with site accessibility, location and operating hours. Sex workers accessing sites described services as being suitable and accessible, with friendly staff. CONCLUSIONS: RWCs were highly appreciated by the users, as they are suitable and accessible. The sex workers who used the clinics visited them irregularly, mostly for PHC services other than HIV and STIs. Services other than the one for which the sex worker came to the clinic rarely appeared to be offered. We recommend areas for service expansion.