Page 212 - HIVMED_v21_i1.indb

Page 3 of 9 Original Research

with curtains drawn where possible. Given the lack of an Inc., Redmond, WA, USA), and then imported into Stata v.14
EMR or patient tracking system, HCT staff placed a small dot (StataCorp, TX, USA) for analysis.
on the folders of all patients who were approached and
offered HCT. Every 4 h, the study supervisors audited the The outcome of interest, declining HCT, was measured as a
folders of patients located within the ED to ensure that all binary variable (‘no’ = 0 and ‘yes’ = 1). The independent
eligible patients had been approached. variables measured were age (18–30, 31–50, 51–70, 70+),
sex (male, female), presenting complaint (trauma, medical),
Based on recent survey data from the 2017 South African South African triage score (death, routine visit, urgent,
national HIV prevalence study, HIV prevalence amongst very urgent, emergent), access to primary care (yes, no),
2
South Africans of all ages was estimated at 14%. Our study past medical history (hypertension, coronary artery disease,
aimed to recruit a sample size of 700 patients at each site. This tuberculosis, diabetes, asthma, chronic obstructive
would present a large enough sample to capture the variation pulmonary disorder, cancer), visit time (within regular
in testing preferences in the study setting, allowing us to operating hours, 9 am to 5 pm, or out of regular operating
detect a difference of greater than 5% from the baseline hours), visit reason (new complaint, return visit, referral),
estimate of 14%, assuming a two-sided α of 0.05 and 80% mode of transport (self-transport, ambulance, police),
power, for a period of 7 weeks at each site. presenting symptoms (pain, fever) and disposition (death,
intensive care unit admission, general admission, emergent
Intervention surgery, transfer, discharge, absconded).
Patients were offered point-of-care HIV testing following the
South African national HIV testing guidelines. Patients who Data analysis and statistics
7
consented to the test provided a blood sample obtained Analysis was conducted on patients unaware of their
through a lancet finger prick. Following the recommended status, to examine the relationship between the outcome of
testing algorithm, patients were first tested using the interest and all other independent variables. Chi-square
Advanced Quality Anti-HIV 1&2 rapid test (InTec Products, tests were used to explore individual variable associations
Inc., Fujian, China). Non-reactive samples were reported as with declining HCT. Logistic regression analysis was
an HIV-negative result. Reactive samples were confirmed conducted to assess the contribution of each variable to
with an HIV 1/2/O Tri-line HIV rapid test (ABON Biopharm, declining HCT. Bivariate analysis was conducted to estimate
Hangzhou, China). Confirmed reactive samples were the association between the outcome and each predictor
reported as an HIV-positive result, and patients were variable, as well as multivariate analysis to estimate the
provided with a referral letter to a local ARV clinic. Confirmed independent effect of each predictor variable, adjusting for
non-reactive samples were reported as an indeterminate all others. All variables were included in the final model,
result, and patients were counselled to repeat the test in following checks for collinearity and goodness of fit and
4–6 weeks. Counselling preceded and followed all tests performing a best-subsets variable selection. Sub-group
and included education on HIV transmission, prevention, analysis was completed on the top reasons for accepting
and management. Results were available within 10–15 min and declining HCT by gender.
of testing, whereas counselling required an additional
10–15 min, depending on the HIV test result. A reference level was selected for categorical variables with
multiple responses, and other levels were accordingly
Data collection compared. Associations were assessed using odds ratios
(ORs), 95% confidence intervals (CIs) and p-values. A p-value
Ten local research assistants were hired and trained in rapid of ≤ 0.05 was regarded as statistically significant. A pooled
point-of-care HCT, good clinical practice and data collection, analysis of data collected from both sites is presented; no
and were familiarised with the study protocol before the start significant differences were observed between the two sites
of the study. Research assistants and study staff worked in (Table 1).
shifts to ensure 24-h coverage of the ED.
Ethical consideration
In tandem with offering HIV testing, HCT staff administered
a brief survey. Patient responses to questions about their The study was approved by the Johns Hopkins University
gender, past medical history, mode of arrival, reason for School of Medicine Institutional Review Board (reference
visit, presenting complaint, and symptoms were recorded as number IRB00105801), the Human Research Ethics
pre-determined binary or categorical options, age was Committee from the University of Cape Town (MREC
recorded as free text, and reasons for accepting or declining reference number 856/2015), the Human Research
testing were captured via pre-determined categorical Committee of Walter Sisulu University (reference number
options derived from the literature or as free text. Data were 069/2015) and the Eastern Cape Department of Health.
recorded on case report forms. These forms were scanned Written consent was obtained from all participants who
and uploaded onto iDatafax (DF/Net Research, Inc., Seattle, enrolled in the study and was required for the collection of
WA, USA) by trained study staff. Following validation and demographic data, HCT and a follow-up call for newly
cleaning, data were exported into Excel v.16.9 (Microsoft, diagnosed patients, separately.

http://www.sajhivmed.org.za 204 Open Access