Using repeated home-based HIV testing services to reach and diagnose HIV infection among persons who have never tested for HIV, Chókwè health demographic surveillance system, Chókwè district, Mozambique, 2014–2017

Background HIV prevalence in Mozambique (12.6%) is one of the highest in the world, yet ~40% of people living with HIV (PLHIV) do not know their HIV status. Strategies to increase HIV testing uptake and diagnosis among PLHIV are urgently needed. Home-based HIV testing services (HBHTS) have been evaluated primarily as a 1-time campaign strategy. Little is known about the potential of repeating HBHTS to diagnose HIV infection among persons who have never been tested (NTs), nor about factors/reasons associated with never testing in a generalized epidemic setting. Methods During 2014–2017, counselors visited all households annually in the Chókwè Health and Demographic Surveillance System (CHDSS) and offered HBHTS. Cross-sectional surveys were administered to randomly selected 10% or 20% samples of CHDSS households with participants aged 15–59 years before HBHTS were conducted during the visit. Descriptive statistics and logistic regression were used to assess the proportion of NTs, factors/reasons associated with never having been tested, HBHTS acceptance, and HIV-positive diagnosis among NTs. Results The proportion of NTs decreased from 25% (95% confidence interval [CI]:23%–26%) during 2014 to 12% (95% CI:11% –13%), 7% (95% CI:6%–8%), and 7% (95% CI:6%–8%) during 2015, 2016, and 2017, respectively. Adolescent boys and girls and adult men were more likely than adult women to be NTs. In each of the four years, the majority of NTs (87%–90%) accepted HBHTS. HIV-positive yield among NTs subsequently accepting HBHTS was highest (13%, 95% CI:10%–15%) during 2014 and gradually reduced to 11% (95% CI:8%–15%), 9% (95% CI:6%–12%), and 2% (95% CI:0%–4%) during 2015, 2016, and 2017, respectively. Conclusions Repeated HBHTS was helpful in increasing HIV testing coverage and identifying PLHIV in Chókwè. In high HIV-prevalence settings with low testing coverage, repeated HBHTS can be considered to increase HIV testing uptake and diagnosis among NTs.

were provided additional HIV counseling (e.g., referrals, linkages to care, information about the benefits 118 of early treatment and adherence, disclosure, partner or family HIV testing, and condom use).

119
Counselors conducted up to 5 follow-up home visits to encourage HIV-positive participants to enroll in 120 and adhere to HIV care. HIV-negative and HIV-indeterminate participants were provided risk-reduction 121 counseling, including recommendations for periodic HIV testing and behavioral prevention 122 strategies/service; uncircumcised men were referred to voluntary medical male circumcision services.

123
Pregnant women were referred for antenatal care when needed.

124
Outcomes and other HIV testing-related variables 125 The primary outcome variable was never having been tested for HIV. All persons surveyed were asked,

126
"Have you ever been tested for HIV?" Additional outcome variables included acceptance of HBHTS 127 among NTs, and HIV-positive yield among NTs who tested for HIV after their survey interview. For 128 participants who responded "Yes" to ever testing for HIV, information about location of their most 129 recent HIV test was collected. For participants who responded "No" to ever having been tested for HIV, 130 the reasons for not being tested and intention to test for HIV during the next 12 months were assessed.

151
To assess the changes among NTs, the number and proportion, including 95% Wald confidence intervals 152 (CIs), of NTs in each round were calculated. To identify the factors associated with never having been 153 tested, bivariate and multivariable analyses were conducted for each survey round using the annual data.

154
Logistic regression analyses were used to identify factors associated with never having been tested. A 155 backward-elimination procedure was used to select the final model. Two-way interaction terms between 156 the factors were also evaluated.  To test the robustness of the observed findings, we conducted a sensitivity analysis. We re-analyzed the 165 data using a weighted approach with a survey weight calculated by age, gender and region (urban or 166 rural).

184
Factors and reasons associated with never having been tested 185 Factors associated with NT, after adjusting for all the other variables in the model, were similar for all 4 186 survey rounds (Table 2). During survey rounds 1-3, adolescent (ages 15-17 years) boys and girls and 187 adult (ages ≥18) men were more likely than adult women to be NTs (P < .001). Adolescent girls had 15.41) times the adjusted odds of never having been tested, compared with adults (ages≥25).

194
Participants who had ever had sex, who had higher HIV knowledge scores, or who had ever asked 195 partners about HIV status were less likely to be NTs for each of the 4 years.

219
The sensitivity analysis results using the weighted approach are given in Table 1 and Table S2-S4. The 220 estimates using the weighted approach and the observed data were similar (i.e., the counts were slightly 221 different; the proportions were very close; and significant factors identified were the same).

223
During 2014-2017, we used 4 rounds of HPS and HBHTS data to investigate the potential of repeating 224 HBHTS annually for reaching NTs, and to determine factors or reasons associated with never having 225 been tested previously, acceptance of HBHTS and yield of new HIV diagnosis among NTs who 226 participated in these surveys. The results revealed that the proportion of participants who reported never 227 having been tested previously for HIV decreased substantially after HBHTS was implemented, from 228 25% to 7% after 2 rounds of HBHTS. During these first two rounds, 87% to 90% of survey participants 229 who had never tested previously for HIV subsequently accepted HBHTS, and the yield of new HIV 230 diagnoses among those who accepted was high (11%-13%). Additionally, home rather than a hospital in 2014 to 65%, 57%, and 33% in subsequent years) and an increasing proportion of NTs who were aged 244 <18 (from 30% during 2014 to 46%, 58%, and 80% in subsequent years). Among NTs aged ≥25 years 245 (97% of whom had sex), the positive yield was ~20% during the first 3 years, decreasing to 6% during 246 2017. These findings suggest that when HBHTS is repeated, screening for sexual risk behavior should 247 be considered to avoid unnecessary costs associated with testing persons who are pre-sexual debut and 248 are likely not to have not been exposed to HIV.  Men were more likely to be NTs than women during all four survey rounds. This is concerning because For the sensitivity analysis, the estimates using the weighted approach and the observed data were very 291 similar because of random sampling. This confirms the distributions of race, gender and region of the 292 10% or 20% random samples of CHDSS households with members aged 15-59 years were similar to the 293 underlying population. When comparing the demographic distributions (e.g., age, gender) of the survey 294 respondents to baseline census data, there were a slightly higher proportion (3-5 percentage points) of 295 older persons (ages 45-59) and a lower proportion (1-10 percentage points) of males among survey 296 respondents. This might be because men are more likely to be away from home during the day and older 297 persons are more likely to be at home. Although the weighted results were similar to the results using 298 observed data, given that 40%-48% of eligible participants were not reached at home or refused to 299 respond during the 4 survey rounds, and we do not know if these not-reached or refused-to-respond 300 eligible participants can be represented by the HPS respondents, we do not generalize the weighted 301 results to all CHDSS participants in the district. HIV prevalence will be helpful. Despite these limitations, to our knowledge, this is the first study to 319 investigate repeated HBHTS in reaching, testing, and diagnosing HIV infection among NTs and fills a 320 crucial gap in the literature.

322
The repeated HBHTS approach was helpful to increase HIV testing coverage and identifying PLHIV in 323 Chókwè. HBHTS acceptance rates were high across all sex, gender and NTs with different barriers.

324
HIV-positive yields among NTs who accepted HBHTS exceeded 10%. In high HIV-prevalence settings 325 with low testing coverage, repeated HBHTS can be considered to increase testing uptake and HIV 326 diagnosis among NTs.

328
The authors would like to thank Key Smith for her helpful editorial review to enhance the manuscript.

329
The authors also would like to acknowledge Dr. Alfredo Vergara, Dawud Ujamaa and Judite Cardoso