Sentiment Analysis of Social Media Data on Ebola Outbreak Using Deep Learning Classifiers.

The Ebola virus disease (EVD) is an extremely contagious and fatal illness caused by the Ebola virus. Recently, Uganda witnessed an outbreak of EVD, which generated much attention on various social media platforms. To ensure effective communication and implementation of targeted health interventions, it is crucial for stakeholders to comprehend the sentiments expressed in the posts and discussions on these online platforms. In this study, we used deep learning techniques to analyse the sentiments expressed in Ebola-related tweets during the outbreak. We explored the application of three deep learning techniques to classify the sentiments in 8395 tweets as positive, neutral, or negative. The techniques examined included a 6-layer convolutional neural network (CNN), a 6-layer long short-term memory model (LSTM), and an 8-layer Bidirectional Encoder Representations from Transformers (BERT) model. The study found that the BERT model outperformed both the CNN and LSTM-based models across all the evaluation metrics, achieving a remarkable classification accuracy of 95%. These findings confirm the reported effectiveness of Transformer-based architectures in tasks related to natural language processing, such as sentiment analysis.

Infection Prevention and Control Initiatives to Prevent Healthcare-Associated Transmission of SARS-CoV-2, East Africa.

The coronavirus disease pandemic has highlighted the need to establish and maintain strong infection prevention and control (IPC) practices, not only to prevent healthcare-associated transmission of SARS-CoV-2 to healthcare workers and patients but also to prevent disruptions of essential healthcare services. In East Africa, where basic IPC capacity in healthcare facilities is limited, the US Centers for Disease Control and Prevention (CDC) supported rapid IPC capacity building in healthcare facilities in 4 target countries: Tanzania, Ethiopia, Kenya, and Uganda. CDC supported IPC capacity-building initiatives at the healthcare facility and national levels according to each country's specific needs, priorities, available resources, and existing IPC capacity and systems. In addition, CDC established a multicountry learning network to strengthen hospital level IPC, with an emphasis on peer-to-peer learning. We present an overview of the key strategies used to strengthen IPC in these countries and lessons learned from implementation.

Capture-Recapture Among Men Who Have Sex With Men and Among Female Sex Workers in 11 Towns in Uganda.

BACKGROUND: Key populations at higher risk for HIV infection, including people who inject drugs, men who have sex with men (MSM), and female sex workers (FSWs), are disproportionately affected by the HIV/AIDS epidemic. Empirical estimates of their population sizes are necessary for HIV program planning and monitoring. Such estimates, however, are lacking for most of Uganda's urban centers. OBJECTIVE: The aim of this study was to estimate the number of FSWs and MSM in select locations in Uganda. METHODS: We utilized conventional 2-source capture-recapture (CRC) to estimate the population of FSWs in Mbale, Jinja, Wakiso, Mbarara, Gulu, Kabarole, Busia, Tororo, Masaka, and Kabale and the population of MSM in Mbale, Jinja, Wakiso, Mbarara, Gulu, Kabarole, and Mukono from June to August 2017. Hand mirrors and key chains were distributed to FSWs and MSM, respectively, by peers during capture 1. A week later, different FSWs and MSM distributors went to the same towns to collect data for the second capture. Population size estimates and 95% CIs were calculated using the CRC Simple Interactive Statistical Analysis. RESULTS: We estimated the population of FSWs and MSM using 2 different recapture definitions: those who could present the object or identify the object from a set of photos. The most credible (closer to global estimates of MSM; 3%-5%) estimates came from those who presented the objects only. The FSW population in Mbale was estimated to be 693 (95% CI 474-912). For Jinja, Mukono, Busia, and Tororo, we estimated the number of FSWs to be 802 (95% CI 534-1069), 322 (95% CI 300-343), 961 (95% CI 592-1330), and 2872 (95% CI 0-6005), respectively. For Masaka, Mbarara, Kabale, and Wakiso, we estimated the FSWs population to be 512 (95% CI 384-639), 1904 (95% CI 1058-2749), 377 (95% CI 247-506), and 828 (95% CI 502-1152), respectively. For Kabarole and Gulu, we estimated the FSWs population to be 397 (95% CI 325-469) and 1425 (95% CI 893-1958), respectively. MSM estimates were 381 (95% CI 299-462) for Mbale, 1100 (95% CI 351-1849) for Jinja, 368 (95% CI 281-455) for Wakiso, 322 (95% CI 253-390) for Mbarara, 180 (95% CI 170-189) for Gulu, 335 (95% CI 258-412) for Kabarole, and 264 (95% CI 228-301) for Mukono. CONCLUSIONS: The CRC activity was one of the first to be carried out in Uganda to obtain small town-level population sizes for FSWs and MSM. We found that it is feasible to use FSW and MSM peers for this activity, but proper training and standardized data collection tools are essential to minimize bias.

Improving the ability of districts in Uganda to monitor their HIV programs.

BACKGROUND: Although district health teams (DHT) in Uganda are supposed to monitor and support facilities to ensure quality HIV data collection, reporting and use, they are often ill-equipped to do so. We implemented a program designed to build the capacity of districts to manage and use their own HIV-related program data and to assist facilities to collect and evaluate their own data. METHODS: We conducted a baseline assessment of the monitoring and evaluation (M&E) capacity of 38 districts. In the 10 worst-performing districts, we identified and trained district-level staff to become M&E mentors who in turn trained and supervised facility-level staff. We collected information on action plans developed by facilities to address major issues of concern. Following the intervention, we reassessed M&E capacity of the 10 targeted districts. RESULTS: Among the 38 districts assessed, one-half did not have a biostatistician, less than one-quarter had staff trained in the basics of M&E or data analysis, and less than one-quarter had an M&E plan. The main concerns of facilities included lack of updated data collection tools, lack of supervision, inaccurate data recording, and limited ability to analyze and use data. In the 10 targeted districts, comparison before and after the intervention showed that the number of districts with trained M&E staff increased (4 to 9), the number of M&E plans increased (3 to 6), and the number using data for programming increased (4 to 8). Implementation of action plans by facilities successfully addressed many issues and led to improved programming. CONCLUSION: Challenges of district M&E in Uganda mainly result from a lack of skilled human resources. On-the-job training and direct involvement of district staff to provide support to facilities can lead to improvements in data quality and use.

Retention in care among HIV-infected pregnant and breastfeeding women on lifelong antiretroviral therapy in Uganda: A retrospective cohort study.

BACKGROUND: In 2013, Uganda updated its prevention of maternal-to-child transmission of HIV program to Option B+, which requires that all HIV-infected pregnant and breastfeeding women be started on lifelong antiretroviral therapy (ART) regardless of CD4 count. We describe retention in care and factors associated with loss to follow-up (LTFU) among women initiated on Option B+ as part of an evaluation of the effectiveness of the national program. METHODS: We conducted a retrospective cohort analysis of data abstracted from records of 2,169 women enrolled on Option B+ between January and March 2013 from a representative sample of 145 health facilities in all 24 districts of the Central region of Uganda. We defined retention as "being alive and receiving ART at the last clinic visit". We used Kaplan-Meier analysis to estimate retention in care and compared differences between women retained in care and those LTFU using the chi-squared test for dichotomized or categorical variables. RESULTS: The median follow-up time was 20.2 months (IQR 4.2-22.5). The proportion of women retained in HIV care at 6, 12 and 18 months post-ART initiation was 74.2%, 66.7% and 62.0%, respectively. Retention at 18 months varied significantly by level of health facility and ranged from 70.0% among those seen at hospitals to 56.6% among those seen at lower level health facilities. LTFU was higher among women aged less than 25 years, 59.3% compared to those aged 25 years and above, 40.7% (p = 0.02); among those attending care at lower level facilities, 44.0% compared to those attending care at hospitals, 34.1% (p = 0.01), and among those who were not tested for CD4 cell count at ART initiation, 69.4% compared to those who were tested, 30.9% (p = 0.002). CONCLUSION: Retention of women who were initiated on Option B+ during the early phases of roll-out was only moderate, and could undermine the effectiveness of the program. Identifying reasons why women drop out and designing targeted interventions for improved retention should be a priority.

An early assessment of Uganda's roll-out of Option B+: Service capacity and infant outcomes.

BACKGROUND: Uganda was one of the first countries in sub-Saharan Africa to implement Option B+ as its national strategy for prevention-of-mother-to-child transmission (PMTCT) of HIV, doing so in 2013. We report on two evaluations designed to assess the capacity of the health care system to implement Option B+, and to obtain preliminary information on the maternal-to-child-transmission rate of HIV. METHODS: We performed: 1) a cross-sectional assessment in 2014 of 505 health care facilities (49 district hospitals, 83 Health Center [HC] IVs, and 373 HCIIIs) in 62 of Uganda's 112 districts to evaluate whether services and commodities required for Option B+ were being provided; and 2) a retrospective record review of 283 HIV-exposed infants enrolled in post-natal care in 2013 in the Central Region to evaluate infant outcomes at 18-months of age. RESULTS: Less than 50% of HCIIIs performed routine diagnostics, including syphilis, hemoglobin, and urinalysis testing, required at all ante-natal clinics; almost all facilities performed a baseline CD4 cell count, but only 44.5% of HCIIIs and 60.2% of HCIVs performed follow-up testing. The proportion of facilities monitoring antiretroviral therapy (ART) (47.2-69.4%) and clinic adherence (50.0-67.3%) was low. Many facilities (20.4-45.8%) reported stock-outs of ART and HIV test kits in the prior month. At 18 months, 53.7% of HIV-exposed infants were lost to follow-up (LTFU). Among those retained, 6.5% were HIV infected. CONCLUSION: Significant shortcomings in service provision and high LTFU of HIV-exposed infants are barriers to Uganda's ability to implement Option B+ successfully.