A hybrid mobile approach for population-wide HIV testing in rural east Africa: an observational study.

BACKGROUND: Despite large investments in HIV testing, only an estimated 45% of HIV-infected people in sub-Saharan Africa know their HIV status. Optimum methods for maximising population-level testing remain unknown. We sought to show the effectiveness of a hybrid mobile HIV testing approach at achieving population-wide testing coverage.
METHODS: We enumerated adult (≥15 years) residents of 32 communities in Uganda (n=20) and Kenya (n=12) using a door-to-door census. Stable residence was defined as living in the community for at least 6 months in the past year. In each community, we did 2 week multiple-disease community health campaigns (CHCs) that included HIV testing, counselling, and referral to care if HIV infected; people who did not participate in the CHCs were approached for home-based testing (HBT) for 1-2 months within the 1-6 months after the CHC. We measured population HIV testing coverage and predictors of testing via HBT rather than CHC and non-testing.
FINDINGS: From April 2, 2013, to June 8, 2014, 168,772 adult residents were enumerated in the door-to-door census. HIV testing was achieved in 131,307 (89%) of 146,906 adults with stable residence. 13,043 of 136,033 (9·6%, 95% CI 9·4-9·8) adults with and without stable residence had HIV; median CD4 count was 514 cells per μL (IQR 355-703). Among 131,307 adults with stable residence tested, 56,106 (43%) reported no previous testing. Among 13,043 HIV-infected adults, 4932 (38%) were unaware of their status. Among 105,170 CHC attendees with stable residence 104,635 (99%) accepted HIV testing. Of 131,307 adults with stable residence tested, 104,635 (80%; range 60-93% across communities) tested via CHCs. In multivariable analyses of adults with stable residence, predictors of non-testing included being male (risk ratio [RR] 1·52, 95% CI 1·48-1·56), single marital status (1·70, 1·66-1·75), age 30-39 years (1·58, 1·52-1·65 vs 15-19 years), residence in Kenya (1·46, 1·41-1·50), and migration out of the community for at least 1 month in the past year (1·60, 1·53-1·68). Compared with unemployed people, testing for HIV was more common among farmers (RR 0·73, 95% CI 0·67-0·79) and students (0·73, 0·69-0·77); and compared with people with no education, testing was more common in those with primary education (0·84, 0·80-0·89).
INTERPRETATION: A hybrid, mobile approach of multiple-disease CHCs followed by HBT allowed for flexibility at the community and individual level to help reach testing coverage goals. Men and mobile populations remain challenges for universal testing. FUNDING: National Institutes of Health and President's Emergency Plan for AIDS Relief.

Introduction

Despite large investments in HIV testing, only 45% of people living with HIV in sub-Saharan Africa are estimated to know their status.(1, 2) To take full advantage of recent advances in treatment as prevention,(3) there is a compelling need to increase HIV testing at the population level. On this basis, UNAIDS has established an ambitious global target of 90% HIV testing coverage by 2020.(1) However, how best to maximize population-wide testing coverage is unknown. Barriers to HIV testing are multiple, and include lack of awareness of HIV risk, minimally symptomatic early HIV disease, stigma, and challenges with access, costs and waiting times associated with health facility-based testing.(4–6) Moving HIV testing out of health facilities and into communities can overcome some of these barriers.(7)

Out-of-facility HIV testing approaches include home-based,(8–10) work-based,(11) index testing,(12) self-testing,(13) and community health campaigns.(14, 15) Each of these approaches has advantages, however no single approach is likely to work across diverse settings in sub-Saharan Africa. Of these, home-based testing and mobile health campaigns have achieved the highest levels of population coverage.(7) Large-scale mobile health campaigns achieve high levels of coverage rapidly.(14–16) By incorporating multi-disease services, campaigns may normalize HIV testing as routine care, create a mechanism for coping with stigma, improve access, and reduce transport costs and waiting times.

Home-based testing (HBT) also improves access, and has proved effective in various settings.(8, 17) Unlike campaigns, HBT allows for couple counseling and reaches those who do not seek venue-based testing.(18, 19) Technologic improvements in data management, geographic information systems, and digital biometric identification now offer increasingly simple methods to enumerate large populations. This allows for a clearer understanding of who is not reached by campaigns,(16) and thus selective use of HBT. Based on the relative advantages of each mobile approach, we hypothesized that a combination of large-scale health campaigns followed by HBT of campaign non-participants could rapidly achieve 90% population testing coverage.

We sought to demonstrate the effectiveness at achieving population-wide testing coverage of a hybrid mobile HIV testing approach of multi-disease community health campaigns (CHC) followed by HBT of campaign non-participants during rapid testing scale-up in an HIV “test and treat” trial in Uganda and Kenya. We also sought to identify baseline predictors of HBT (vs. CHC-testing) among adults who tested, and of non-testing for HIV, in order to characterize populations that did not engage in campaigns and that are “hard to reach” for testing, respectively.

Methods

Study Design

The hybrid mobile HIV testing approach is the primary testing strategy in the Sustainable East Africa Research in Community Health (SEARCH) HIV test and treat cluster-randomized controlled Trial (NCT:01864603: https://clinicaltrials.gov/ct2/show/NCT01864603). The SEARCH Trial consists of 32 communities (Figure 1) selected from 54 candidate communities that met initial eligibility criteria of a rural community (defined as one or more national geopolitical units, just above the village level: i.e. a “parish” in Uganda, and a “sub-location” in Kenya), with population 10,000, within the catchment area of a President’s Emergency Plan For AIDS Relief (PEPFAR)-supported HIV clinic in southwestern Uganda, eastern Uganda or western Kenya. We performed ethnographic mapping, reviewed national census and epidemiologic data for each candidate community, and then selected 16 matched pairs based on region, population density, occupational mix, access to transport routes, and number of trading centers.(20) All 32 communities underwent census enumeration followed by the hybrid mobile HIV testing approach.

Figure 1

East African Map of 32 SEARCH communities in 3 regions: Southwestern Uganda (study community names: 1. Nsiika; 2. Bugamba; 3. Rugazi; 4. Mitooma; 5. Kitwe; 6. Rubaare; 7. Rwashamaire; 8. Ruhoko; 9. Kazo; 10. Nyamuyanja), Eastern Uganda (1. Nsiinze; 2. Nankoma; 3. Kiyunga; 4. Kamuge; 5. Bugono; 6. Muyembe; 7. Merikit; 8. Kiyeyi; 9. Kameke; 10. Kadama) and Western Kenya (1. Nyatoto; 2. Nyamrisra; 3. Ogongo; 4. Kitare; 5. Magunga; 6. Kisegi; 7. Tom Mboya; 8. Sena; 9. Ongo; 10. Othoro; 11. Sibuoche; 12. Bware).

Procedures

Study staff performed baseline resident enumeration and trial enrollment in all communities using a 2–4 week per community, door-to-door census. Census staff, working with village leaders, visited all residential structures within each community. The census interview consisted of: 1) enumeration of all persons who lived on the property for ≥1 month in the year preceding the census visit; 2) digital biometric fingerprint measurement (U.are.u 4500 reader, Digital Persona, Crossmatch, Florida, USA) of all available household members; 3) measurement of geographic positioning system coordinates of the home; and 4) an interview to obtain demographic, household socioeconomic, and migration data. “Stable” residence was defined as living in the community for ≥6 months over the past year.

Before initiating mobile HIV testing, study staff met with local leaders to solicit advice on CHCs and HBT implementation. Local leaders then reached out to their communities to provide information about the multi-disease CHC. Study staff co-implemented mobilization activities with local leaders one month before the CHC. Information was disseminated using posters and pamphlets, announcements during religious services and community events, question and answer sessions at gathering places (e.g. bars and markets), and during the census. Small non-monetary prizes were awarded to randomly selected CHC participants as a way of promoting attendance (prizes totaled ≤US$ 2,000/community).

Two-week mobile, multi-disease CHCs were conducted in partnership with the Uganda and Kenya Ministries of Health (MoH), at well-known, convenient community locations. Services included rapid, finger prick-blood based HIV antibody testing and counseling (HTC) for all persons ≥18 months of age (regardless of self-reported HIV status) using MoH test kits and testing algorithms, followed by point-of-care CD4+ T cell count measurement (PIMA, Inverness), provision of a 30-day supply of trimethoprim-sulfamethoxazole, and referral to HIV care if HIV-infected. Non-HIV services varied by community, and included services such as hypertension and diabetes screening, malaria rapid diagnostic testing for participants with fever, male condom distribution, referral for medical male circumcision, family planning and cervical cancer screening, and Vitamin A and albendazole treatment for young children. Residence status was defined by baseline census enumeration. Fingerprint biometrics were used to verify resident status and record CHC attendance on-site at the CHC entrance prior to HIV testing, using USB-enabled fingerprint scanners (U.are.u 4500 reader, Digital Persona) connected to tablet computers that each contained the census database; if fingerprint matching failed, name-based matching to the census database was used, with verification of name-matched participants’ household members as an added cross-check in the event of multiple similar names in the same community. Self-reported residence at time of CHC participation was also accepted to define resident status, provided self-reported residents could be linked to census-enumerated households. Daily reports on the number of residents seen each CHC day were reviewed to monitor testing coverage in real-time and identify demographic groups for additional mobilization efforts.

Using census enumeration and CHC attendance data, we identified residents who did not engage in HIV testing at CHCs. These residents were approached for testing at their homes, or a place of their choosing, over 1–2 months, in order to reach minimum testing coverage of 80% among stable men and women residents (age 15–50). HTC and referral services offered during HBT were identical to the CHC; however, non-HIV services were not provided. Resident identity was verified using fingerprint biometrics in the same fashion as CHC identification. If a CHC non-participant was not home during initial HBT visit, staff attempted to contact them by phone and/or return up to three times.

All HIV-infected persons identified at CHCs or HBT received one-on-one post-test counseling that included information on living with HIV, preventing transmission, the benefits of linking to care and treatment, and the logistics of attending the local ART-providing clinic. Face-to-face introductions to local clinic staff occurred at CHCs. Specific appointment dates within two months of testing were provided to each HIV-infected person. HIV-infected persons with CD4 counts <200 cells/μL or pregnant at time of testing were given priority appointments, within one week. Staff provided transport vouchers to all HIV-infected persons, for reimbursement upon linking to care.

Statistical Analyses

Predictors of no prior HIV testing and testing at home rather than CHC among stable adults who tested, and non-testing for HIV among all stable adults, were estimated using collaborative targeted maximum likelihood estimation (C-TMLE).(21) Specifically, we estimated the marginal relative risk associated with each predictor, after controlling for the other predictor variables. C-TMLE was implemented instead of standard logistic regression to avoid the modeling assumptions inherent in parametric regression, and to help alleviate problems due to collinearity of multiple predictor variables. All analyses were adjusted for clustering by household. A household wealth index across all communities was calculated using principal components analysis based on ownership of livestock (cows, goats and poultry) and household items (clock, radio, television, phone, refrigerator, bicycle, motorcycle and electricity).(22)

Geospatial Analysis

One community per region was selected for geospatial mapping in order to visually demonstrate changes in HIV testing coverage among stable adults at three time periods: 1) prior to the hybrid approach (i.e. self-reported prior testing in the preceding year); 2) after CHC implementation; and 3) after implementing the hybrid approach. Maps were created using ArcGIS (Esri, Redlands, CA), by determining the density of persons meeting an outcome criterion per km2, and then standardizing color intensity scales of the outcome densities to allow for cross-community comparison.

The Makerere University School of Medicine Research and Ethics Committee and the Ugandan National Council on Science and Technology (Uganda), and the Kenya Medical Research Institute Ethical Review Committee (Kenya), and the University of California San Francisco Committee on Human Research approved the consent procedures and the study. All participants provided informed consent in their preferred language.

Role of the funding source

The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

Between April 2013-June 2014, the SEARCH Trial enumerated 335,005 people, including 168,772 adults (≥15 years), during study censuses: 103,580 persons in southwestern Uganda, 110,113 in eastern Uganda, and 121,312 in Kenya (Figure 1). National census projections estimated a population of 345,181 persons in the 32 communities.(23, 24) Average duration of study census enumeration was 19 (range: 8–31) days/community. Stable residents represented 87% (146,906 persons) of enumerated adults. Baseline characteristics of enumerated adults are shown in Table 1.

Table 1

Baseline SEARCH Trial community adult (≥15 years) resident demographic characteristics, by study region.

	Eastern Uganda	Southwestern Uganda	Western Kenya	Total
Uganda (2002) & Kenya (2009) National Census Projections 2012	54,108	51,850	66,633	172,591
N (Population by Study Census)	51,561	54,654	62,557	168,772
Stable Residents	47,445 (92%)	47,074 (86%)	52,387 (84%)	146,906 (87%)
Median Age (IQR)	29 (20–44)	29 (21–43)	29 (20–43)	29 (20–43)
Female	27,639 (54%)	28,699 (53%)	33,203 (53%)	89,541 (53%)
Marital Status
Single	15,483 (30%)	18,308 (34%)	20,324 (32%)	54,115 (32%)
Married	30,156 (59%)	30,015 (55%)	35,480 (57%)	95,651 (57%)
Widowed/Divorced/Separated	5,714 (11%)	6,226 (11%)	6,558 (11%)	18,528 (11%)
Polygamy (% of married adults)	7,452 (25%)	3,711 (12%)	8,813 (25%)	19,976 (21%)
Occupation
Farmer	30,439 (59%)	27,685 (51%)	21,991 (35%)	80,115 (48%)
Fisher	116 (0.2%)	147 (0.3%)	6,211 (10%)	6,474 (4%)
Student	11,404 (22%)	11,097 (20%)	15,425 (25%)	37,926 (22%)
No job	1,794 (4%)	2,451 (5%)	4,950 (8%)	9,195 (5%)
Other	7,808 (15%)	13,274 (24%)	13,980 (22%)	35,062 (21%)
Education
No Education	7,581 (15%)	9,217 (17%)	3,680 (6%)	20,478 (12%)
Primary School only	30,112 (58%)	29,323 (54%)	40,534 (65%)	99,969 (59%)
Any Secondary School	13,766 (27%)	16,077 (29%)	18,099 (29%)	47,942 (28%)
Household: median number of acres owned (IQR)	1 (0.5–2)	2 (1–3)	1.5 (0.5–3)	1.5 (0.5–3)
Households with phone ownership	9,859/19,437 (51%)	13,131/19,959 (66%)	17,174/23,267 (74%)	40,164/62,663 (64%)
Households with electricity in home	719/19,437 (3.7%)	936/19,959 (4.7%)	596/23,267 (2.6%)	2,251/62,663 (3.6%)

Over one year, 89% (131,307/146,906) of stable adult residents tested for HIV using a hybrid strategy of CHC followed by HBT of CHC non-participants. HIV testing coverage of stable adult residents by testing modality and demographic sub-group are shown in Table 2. Testing coverage at CHCs ranged from 52–82% across the 32 communities. Testing coverage using the hybrid approach was 22% (4,726/21,866) among non-stable adult residents, and 81% (136,033/168,772) among all adult residents (stable and non-stable). HIV prevalence was 9.6% (13,043/136,033 adults; 95% CI: 9.4–9.8%), with a median CD4+ count of 514 (IQR: 355–703) cells/μL among HIV-infected adults. Adult HIV prevalence was 6.5% (2,861/43,942; 95% CI: 6.3–6.7%) in southwestern Uganda, 3.4% (1,539/45,175; 95% CI: 3.2–3.6%) in eastern Uganda, and 18.4% (8,643/46,916; 95% CI: 18.1–18.8%) in western Kenya.

Table 2

Stable adult resident population HIV testing coverage by mobile testing modality, and by country of residence, gender, and age. HIV prevalence, CD4 cell count, and self-reported new HIV diagnosis by mobile testing modality.

	Enumerated Population (Stable Adults)	Community Health Campaign (CHC)-based Testing Coverage	Home-based Testing (HBT) Coverage	Hybrid Testing (CHC+HBT) Coverage
Stable Adult Residents	146,906	104,635 (71%)	26,672 (18%)	131,307 (89%)
Coverage by sub-group
Uganda	94,519	71,867 (76%)	13,748 (15%)	85,615 (91%)
Kenya	52,387	32,768 (62%)	12,924 (25%)	45,692 (87%)
Men	66,726	42,622 (64%)	14,771 (22%)	57,393 (86%)
Women	80,180	62,013 (77%)	11,901 (15%)	73,914 (92%)
Age, in years
15–19	28,738	19,753 (69%)	5,952 (21%)	25,705 (89%)
20–49	88,415	62,435 (71%)	16,211 (18%)	78,646 (89%)
≥50	29,753	22,447 (75%)	4,509 (15%)	26,956 (91%)
HIV prevalence, stable adults who tested	-	9,781/104,635 (9.4%)	3,004/26,672 (11.3%)	12,785/131,307 (9.7%)
Median CD4 (IQR) cells/μL	-	522 (359–714)	503 (347–681)	518 (356–707)
New HIV diagnosis*	-	3,612/9,781 (37%)	1,202/3,004 (40%)	4,814/12,785 (38%)

New HIV diagnosis was defined at the time of testing for HIV at CHC or HBT, by self-report of either a) no prior HIV testing, or b) last prior HIV test was negative or unknown.

CHC-based testing was the most common mode of HIV testing. Among stable adult residents who tested for HIV, 104,635/131,307 (80%, range 60–93% across communities) tested at CHCs. Among adult CHC attendees, 99.5% (104,635/105,170) accepted HIV testing. The average number of CHC days/community was 12.5 (range 12–17). Average daily CHC participation by both adults and children was 590 residents/day (724 residents/day in eastern Uganda, 597/day in southwestern Uganda, and 484/day in Kenya) and an overall mean of 290 adult residents/day. The median duration of time spent participating in CHC activities was 43 (IQR: 31–61) minutes/person. CHC-based testing identified 76% (9,967/13,043) of all HIV-infected adults diagnosed with the hybrid approach. Characteristics of HIV-infected stable adults diagnosed at CHC are shown in Table 2.

Among stable adult residents who tested for HIV, 26,672/131,307 (20%) tested via HBT, with a range of 7–40% across communities. Among adults encountered during HBT, 79% (26,672/33,697) accepted HIV testing. The average number of home visits was 1.6/person. HBT identified 24% (3,076/13,043) of all HIV-infected adults (stable and non-stable) diagnosed with the hybrid approach. Characteristics of HIV-infected stable adults diagnosed at HBT are shown in Table 2. An average of 26 (range: 11–62) days/community was spent conducting HBT.

Of stable adults tested, 43% (56,106/131,307) reported no prior HIV testing vs. 51% (2,423/4,726) of non-stable adults. Among HIV-infected adults, 38% (4,932/13,043) reported being unaware of their status prior to testing (21% [2,754/13,043] reported their last HIV test was negative or unknown, and 17% [2,178/13,043] reported no prior testing). Of HIV-uninfected stable adults reporting prior testing, 41% (26,269/64,336) reported testing >1 year ago. Predictors of no prior testing among stable adults who tested for HIV are shown in Table 3. In multivariable analyses, risk factors with the largest risk ratios included male gender (relative risk [RR]: 1.28, 95% CI: 1.26–1.29), single marital status (RR: 1.33, 95% CI: 1.31–1.34 vs. non-single), and student occupation (RR: 1.21, 95% CI: 1.18–1.25 vs. jobless).

Table 3

Multivariable analysis evaluating predictors of: A) No prior HIV testing (by self-report) among stable adult residents who tested for HIV with the hybrid mobile approach. B) Requiring home-based HIV testing (HBT: i.e. not participating in testing at a community health campaign [CHC]) among stable adult residents who tested for HIV with the hybrid mobile testing approach; and C) Not testing for HIV among all stable adult residents (including persons who refused HIV testing at a CHC or during HBT), despite the hybrid mobile testing approach.

	A) Relative Risk (95% CI) of no prior HIV testing	B) Relative Risk (95% CI) of requiring home-based HIV testing	C) Relative Risk (95% CI) of not testing for HIV
Uganda resident	Ref.	Ref.	Ref.
Kenya resident	0.52 (0.51–0.53)	1.82 (1.77–1.87)	1.46 (1.41–1.50)
Female	Ref.	Ref.	Ref.
Male	1.28 (1.26–1.29)	1.48 (1.45–1.51)	1.52 (1.48–1.56)
Non-single marital status	Ref.	Ref.	Ref.
Single	1.33 (1.31–1.34)	1.39 (1.36–1.42)	1.70 (1.66–1.75)
Age, in years
15–19	Ref.	Ref.	Ref.
20–29	0.77 (0.76–0.78)	1.26 (1.21–1.32)	1.35 (1.27–1.43)
30–39	0.71 (0.70–0.73)	1.11 (1.05–1.17)	1.58 (1.52–1.65)
40–49	0.78 (0.77–0.79)	1.00 (0.96–1.04)	0.85 (0.77–0.94)
≥50	1.02 (1.01–1.04)	0.97 (0.94–1.00)	1.18 (1.12–1.24)
Occupation
Unemployed	Ref.	Ref.	Ref.
Farmer	0.92 (0.89–0.95)	0.61 (0.58–0.64)	0.73 (0.67 –0.79)
Fisher	0.81 (0.78–0.85)	0.80 (0.75–0.85)	0.98 (0.90–1.08)
Student	1.21 (1.18–1.25)	0.82 (0.79–0.85)	0.73 (0.69–0.77)
Other employment	0.87 (0.84–0.90)	0.91 (0.86–0.96)	1.10 (1.02–1.19)
Education
No education	Ref.	Ref.	Ref.
Primary education only	0.86 (0.84–0.88)	0.85 (0.83–0.88)	0.84 (0.80–0.89)
Any secondary education, or more	0.67 (0.65–0.69)	0.97 (0.94–1.00)	1.08 (1.01–1.17)
Wealth quintile
1	Ref.	Ref.	Ref.
2	0.90 (0.89–0.92)	0.96 (0.92–1.00)	0.94 (0.89–0.99)
3	0.86 (0.84–0.88)	0.96 (0.92–1.00)	0.92 (0.87–0.97)
4	0.85 (0.83–0.86)	0.99 (0.95–1.03)	0.89 (0.84–0.94)
5	0.83 (0.82–0.85)	1.13 (1.09–1.17)	0.97 (0.91–1.03)
Months away from community in year prior to enrollment (up to 6 months)
None	N/A	Ref.	Ref.
≥1 Month		1.36 (1.31–1.40)	1.60 (1.53–1.68)
HIV-uninfected	N/A	Ref.	N/A
HIV-infected		1.12 (1.08–1.16)

Predictors of HBT (i.e. CHC non-participation) among stable adults who tested are shown in Table 3. In multivariable analyses, risk factors with the largest risk ratios included Kenya residence (RR: 1.82, 95% CI: 1.77–1.87 vs. Uganda), male gender (RR: 1.48, 95% CI: 1.45–1.51), single marital status (RR: 1.39, 95% CI: 1.36–1.42), and migration out of the community for ≥1 month (RR: 1.36, 95% CI: 1.31–1.40, vs. no migration). HIV-infected status was also an independent predictor of increased probability of HBT (RR: 1.12, 95% CI: 1.08–1.16, vs. HIV-uninfected).

Predictors of non-testing for HIV among stable adults are shown in Table 3. Risk factors with the largest risk ratios included male gender (RR: 1.52, 95% CI: 1.48–1.56), single marital status (RR: 1.70, 95% CI: 1.66–1.75), 30–39 year old age group (RR: 1.58, 95% CI: 1.52–1.65, vs. 15–19 years), Kenya residence (RR: 1.46, 95% CI: 1.41–1.50), and migration out of the community for ≥1 month (RR: 1.60, 95% CI: 1.53–1.68, vs. no migration).

HIV testing coverage before, during, and after implementing the hybrid approach is shown in three selected communities (one/region) with variable CHC-based testing coverage (Nyatoto having the lowest CHC testing coverage of all 32 communities) in Figure 2.

Figure 2

Density of HIV Un-Tested Persons Over Time

Three selected communities (one per region: Nyamuyanja in southwestern Uganda, Nankoma in eastern Uganda, and Nyatoto in western Kenya), with density of stable adult residents who have not participated in HIV testing from the year prior to study start through the end of the hybrid mobile testing approach, viewed at three time points: A) In the one year before implementing the hybrid mobile HIV testing approach, based on self-report; B) Upon completing community health campaign (CHC) implementation; C) After the hybrid mobile testing approach (combined CHC-based and home-based testing). Color intensity ranges from blue (HIV tested) to red (HIV untested), based on density of untested persons (population/square kilometer). Red crosses indicate location of government-run health facilities, and stars indicate locations of CHCs.

Discussion

We achieved 89% HIV testing coverage of enumerated stable adult residents across 32 communities in Uganda and Kenya using a novel, hybrid mobile HIV testing approach of multi-disease CHCs, followed by HBT of CHC non-participants. This hybrid approach allowed for flexibility in testing modality use across multiple communities with heterogeneous HIV prevalence and prior testing rates. The findings are important in light of recent UNAIDS targets for HIV treatment scale-up, including an ambitious target that 90% of HIV-infected persons will know their status by 2020.(1) We show that rapidly achieving UNAIDS testing coverage goals across a variety of rural settings is feasible using this hybrid approach.

Our hybrid mobile testing approach demonstrates flexibility and efficiency in reaching HIV testing targets by allowing the balance between campaigns and HBT to vary in response to each community’s level of testing coverage at campaigns. Campaign-based testing coverage ranged from 52–82% of stable adult residents, and the hybrid strategy allowed us to adapt the amount of HBT accordingly. The hybrid approach also allows for community input on location of mobile testing sites, and for individuals to self-select the modality best suited to them. Even in western Kenya, an area with high adult HIV prevalence, the testing goal was achieved with increased HBT following CHCs. With data from rapidly conducted censuses, sub-groups that test at low rates can be targeted for more intensive mobilization and testing efforts, including the selective use of incentives for testing. This built-in efficiency may reduce implementation costs.

This approach has several novel features. To our knowledge, it is the first to combine out-of-facility testing interventions strategically to maximize testing coverage. Unlike prior estimates of population testing coverage, our enumeration of a large, diverse target population with fingerprint biometric measures prior to mobile testing, allowed for rigorous measurement of population coverage and identification of persons who fail to test.(9, 14, 25) Lastly, the use of CHCs as the initial modality for rapid testing scale-up is a novel feature of our approach.

Integrating multi-disease services at CHCs demonstrates how HIV testing interventions can complement and enhance other public health priorities. While achieving HIV testing targets, CHCs can be leveraged to screen for communicable and non-communicable diseases, promote children’s health, and provide referral to treatment and preventive services. The presence of non-HIV health services may normalize HIV testing and provide a mechanism to cope with stigma for people seeking HIV testing. Multi-disease services may also serve as an incentive for repeat testing among persons who have tested HIV negative in the past (indeed, 21% of HIV-infected adults we identified reported a prior negative test), and for counseling on linkage to care among HIV-infected persons aware of their status but not in care.

Despite high HIV testing coverage, men, single adults, and mobile persons remain challenging sub-populations for achieving universal testing. Evidence of a gender disparity in testing was observed in both increased need for HBT, and increased risk of non-testing, among men. Across sub-Saharan Africa, men test for HIV at substantially lower rates than women.(2) Consequently, HIV-infected men are diagnosed later in disease, and are less likely to link to care and start ART than women.(26–28) Low testing uptake among men therefore poses an enormous barrier to HIV prevention strategies. This gender disparity may explain, in part, the persistently high HIV incidence rates among 15–20 year old women in sub-Saharan Africa, who often have older sexual partners and acquire HIV through heterosexual transmission.(29) Despite the increased risk of non-testing among men, our approach achieved high testing coverage among men through increased HBT, with subsequent reduction in the gender disparity seen in CHC participation.

Mobile populations are likely to be a major challenge to achieving population HIV testing coverage. Coverage for our stable population was 89% vs. 22% among non-stable adults. Although men represented over half (56%) of stable adults who spent >1 month away from the community in the year prior to the census, after adjusting for gender, mobility remained a predictor of not testing for HIV. Our hybrid strategy took place over a rapid time frame, and low coverage among migrants may simply result from this sub-population being away from the community when testing was offered. Whether mobile persons are testing elsewhere is not clear. However, non-stable adults who tested in our study did report lower rates of prior testing than stable adults, and others have observed an association between migration and increased HIV risk.(30, 31) Therefore, ensuring access to testing among mobile persons remains a challenge to HIV “test and treat” approaches.

The study has several limitations. Estimates of prior HIV testing rely on self-report, and are subject to reporting bias. In enumerating our study population, we may have missed residents resulting in over-estimation of testing coverage, or misclassified some non-residents as residents. However, we conducted robust enumeration efforts and our population measurements were similar to national population projections. A potential limitation to the generalizability of our approach is that targeting HBT to CHC non-participants requires community enumeration. However, in 2014 we performed a community-led CHC that utilized existing village infrastructure (i.e. local leaders and clinical staff) to implement population enumeration before the CHC, demonstrating that low-cost, community-run enumeration is feasible.(32) Despite these limitations, our findings demonstrate an effective, flexible approach to achieving high testing coverage and characterizing who remains untested, in large, well-enumerated rural populations spanning two countries.

The hybrid, mobile HIV testing approach was effective in rapidly achieving high levels of population HIV testing coverage that are essential for the success of recent advances in HIV treatment and prevention. The hybrid approach allowed for flexibility in choice of testing modality and in how coverage goals were met, multi-disease service delivery, and rigorous identification of hard-to-reach populations for universal HIV testing scale-up. Future research on cost-effectiveness, and on how best to engage hard-to-reach populations, including men and migrants, will be necessary to maximize the policy implications of this mobile HIV testing strategy.