Brief counselling after home‐based HIV counselling and testing strongly increases linkage to care: a cluster‐randomized trial in Uganda

Abstract Introduction The aim of this study was to determine whether counselling provided subsequent to HIV testing and referral for care increases linkage to care among HIV‐positive persons identified through home‐based HIV counselling and testing (HBHCT) in Masaka, Uganda. Methods The study was an open‐label cluster‐randomized trial. 28 rural communities were randomly allocated (1:1) to intervention (HBHCT, referral and counselling at one and two months) or control (HBHCT and referral only). HIV‐positive care‐naïve adults (≥18 years) were enrolled. To conceal participants’ HIV status, one HIV‐negative person was recruited for every three HIV‐positive participants. Primary outcomes were linkage to care (clinic‐verified registration for care) status at six months, and time to linkage. Primary analyses were intention‐to‐treat using random effects logistic regression or Cox regression with shared frailty, as appropriate. Results Three hundred and two(intervention, n = 149; control, n = 153) HIV‐positive participants were enrolled. Except for travel time to the nearest HIV clinic, baseline participant characteristics were generally balanced between trial arms. Retention was similar across trial arms (92% overall). One hundred and twenty‐seven (42.1%) participants linked to care: 76 (51.0%) in the intervention arm versus 51 (33.3%) in the control arm [odds ratio = 2.18, 95% confidence interval (CI) = 1.26–3.78; p = 0.008)]. There was evidence of interaction between trial arm and follow‐up time (p = 0.009). The probability of linkage to care, did not differ between arms in the first two months of follow‐up, but was subsequently higher in the intervention arm versus the control arm [hazard ratio = 4.87, 95% CI = 1.79–13.27, p = 0.002]. Conclusions Counselling substantially increases linkage to care among HIV‐positive adults identified through HBHCT and may enhance efforts to increase antiretroviral therapy coverage in sub‐Saharan Africa.


| INTRODUCTION
Home-based HIV counselling and testing (HBHCT) is highly acceptable [1][2][3] and has the potential to substantially increase knowledge of HIV status in sub-Saharan Africa (SSA) [4]. However, linkage to care following HBHCT is often inadequate. In the absence of interventions to facilitate linkage, less than one-third of HIV-positive persons identified through HBHCT in SSA link to care [5].
In order for the populations targeted by HBHCT to benefit from HIV prevention and care services, interventions that can link them to these services need to be identified and utilized. Results from observational studies suggest that counselling provided after referral could increase linkage to care among HIV-positive persons identified through HBHCT in SSA [6][7][8][9][10]. A major limitation of observational studies, however, is that it is difficult to account for the effects of confounding factors. Therefore, randomized trials are required to determine the potential effects of counselling on linkage to care.
In this study, we evaluated the effectiveness of counselling after HIV diagnosis and referral to care, compared to referral to care only, on linkage to care among HIV-positive individuals identified through HBHCT in Masaka, Uganda. We used a cluster-randomized design to reduce the risk of contamination between trial arms and increase acceptability of the interventions.

| Study design and participants
The study was an open-label cluster-randomized controlled trial. A description of the methods has been previously reported [11]. Briefly, 28 communities (clusters) were randomly allocated (1:1) to intervention (HBHCT, referral and counselling at one and two months) or control (HBHCT and referral only). HBHCT was offered to all adults (≥18 years). Newly and previously identified HIV-positive persons were eligible to participate if they were able to consent, not previously or currently in care, available for follow-up, and not participating in other health-related research. To reduce the possibility of revealing participants' HIV status to other community members, one HIV-negative adult was recruited for every three HIV-positive participants.

| Study setting
The study area comprised three rural subcounties of Masaka, a district in south-western Uganda. The subcounties have clinics that offer free HIV care and antiretroviral therapy (ART). Typically, patients newly presenting for care were issued a clinic number and card (registration), underwent a clinical examination and CD4 count testing (results were usually available after two to four weeks), started on cotrimoxazole prophylaxis (CTXp), and if eligible, ART. At the time of the study, the CD4 count threshold for ART eligibility in Uganda was ≤500 cells/mm 3 [12].

| Content
Counselling sessions addressed the following generic points: acceptance of HIV diagnosis, plans to seek care, fear or experience of stigma, importance of HIV status disclosure and availability of psychosocial support for linkage to care; and information about available care services, antiretroviral drugs and the rationale for early linkage to care. In addition, the sessions were used to address personal issues, e.g. marital discord arising after disclosure of HIV status. HIV-negative participants in the intervention arm received HIV risk reduction counselling and information on the importance of regular HCT.

| Staff training
Counsellors did not have medical training but were trained in HCT and had experience in conducting community-based HCT. Training on the intervention was conducted in the four weeks prior to study initiation and consisted of: i) one group seminar that comprised a didactic 30-min presentation and 60-min interactive session; ii) a single 30-min one-on-one session with each of the counsellors that was conducted one to two weeks after the seminar.

| Quality assurance
Counsellors were provided with a checklist of the counselling content to refer to during the sessions. In addition, counsellors took notes of the issues discussed in each session. The study investigator reviewed these notes and discussed any omissions with individual counsellors.

| Outcomes
Primary outcomes were linkage to care (clinic-verified registration for care), determined six months after HIV diagnosis, and time to linkage. Secondary outcomes were time to clinicverified receipt of a CD4 count and ART initiation, and selfreported adherence to CTXp. A further secondary outcome was uptake of repeat HCT at the six month visit among HIVnegative participants.

| Randomization
A cluster was defined as a village or a set of villages with at least 400 adults. Clusters were separated by a buffer zone of ≥1 non-participating villages to minimize the risk of contamination. Clusters were randomly allocated to intervention and control arms. Stratification and restricted randomization were used to minimize between-cluster variation and achieve overall balance between trial arms with regard to key variables [11]. Stratification was defined by distance (≤10 or >10 km) from the district capital, and by cluster make-up, i.e. whether the cluster was composed of a single village or several villages. Restricted randomization was then applied to achieve balance on the following cluster-level variables: size (total number of adults); presence of a trading centre; location along a major road; lakeshore location and presence of an HIV clinic within 5 km.
2.6 | Procedures 2.6.1 | HCT Following community mobilization, standard HCT (including pre-test and post-test counselling) was provided to all adults in each randomized community. Married/cohabiting individuals had the option of receiving HCT as a couple. Blood obtained by finger-prick was tested using HIV rapid test kits: Alere Determine HIV-1/HIV-2 (Alere Medical, Japan) for screening, Stat-Pak HIV 1/2 (Chembio Diagnostic systems, USA) for confirmation of positive results and Uni-Gold HIV 1/2 (Trinity Biotech, Ireland) as tie-breaker. Post-test counselling included a discussion of HIV risk reduction strategies, disclosure plans, partner testing, care and support services and referral options.

| Enrolment
Eligible individuals received information about the study including that their community could be in either of the trial arms and invited to consent. After enrolment, individuals were informed about which trial arm their community was allocated to. Participants completed a counsellor-administered questionnaire on socio-demographic characteristics and HCT history.

| Referral
HIV-positive participants received a referral form a copy of which was retained in a centrally stored folder. Data recorded on the form included the participant's names, age, sex, study number and address.

| Counselling sessions
These were conducted at participants' homes one and two months after HBHCT; each lasted approximately 45 min. Counselling was conducted using a client-centred approach, i.e. it was tailored to the participant's needs and circumstances [29]. Issues identified in the first session were followed up in the second session (for 92% of the participants, both sessions were conducted by the same counsellor). Counsellors worked across clusters.

| Collection of outcome data
Six months after HBHCT, all participants were visited by a counsellor (for 58% of the participants in the intervention arm, this was the same counsellor who had conducted the counselling sessions) to collect information on study outcomes. To assess the potential of contamination between trial arms, counsellors asked participants if they had discussed the study with fellow participants from other villages.
For each participant who reported that they linked to care, a field team member visited the clinic to verify whether the participant had indeed been registered for care. Participants whose records could not be found were re-interviewed to clarify whether they had actually linked or not. Only participants whose records were found were categorized as "linkers". Participants' clinic records were checked to confirm CD4 count testing and receipt of the results, and initiation of CTXp and ART. Adherence to CTXp was measured by self-reported responses to a question on missed doses. Participants who missed ≤5 of 30 doses in the past month were categorized as "adherers".

| HIV-negative study component
Similar enrolment and follow-up counselling procedures were applied to HIV-negative participants. Repeat HBHCT was offered to participants in both arms at six months.

| Ethical considerations
The trial was approved by Uganda Virus Research Institute Research Ethics Committee, the Uganda National Council for Science and Technology, and the Ethics Committee of the London School of Hygiene and Tropical Medicine. Written informed consent was obtained from each participant. The trial is registered at ClinicalTrials.gov (NCT02497456).

| Sample size
On the basis of findings from HBHCT studies in which only referral was provided following HIV diagnosis [30][31][32][33] and those in which counselling was provided after referral [6,7], we assumed linkage of 35% in the control arm and that this would increase to 60% in the intervention arm. We aimed to have 80% power to detect this increase at a significance level of 5%. On the basis of data from settings similar to Masaka [11], we assumed that the between-cluster coefficient of variation (k) for linkage in the absence of intervention was 0.25. Based on these assumptions and an estimated harmonic mean of seven participants per cluster, 28 clusters would be needed. This sample size would also provide >80% power to detect a hazard ratio of 1.7 for the effect of the intervention on time to linkage [34].

| Analysis
All analyses were pre-specified. The primary analyses were intention-to-treat and based on individual-level data, since there was a sufficient number of clusters per arm and the cluster size varied considerably [35]. Random effects logistic regression and Cox regression with shared frailty were used to estimate the effect of the intervention on linkage and time to linkage respectively. Participants who were lost to followup were assumed not to have linked. Those who were reported at the six month follow-up visit to have moved were censored midway between enrolment and that visit. Those who were in the study area but did not attend the six month visit were censored at that visit. The primary analyses of the intervention effect were adjusted for randomization stratum as a fixed effect. Secondary analyses were adjusted for age and sex a priori, and other characteristics that showed substantial baseline imbalance. Due to the nature of the intervention (repeated counselling), it was expected that its effect might change over time. Therefore, the proportional hazards assumption was examined by splitting follow-up time into two intervals (0-2 months and >two months) at a point corresponding with the time of the second counselling session, and testing for an interaction between trial arm and time. Similar methods were used to estimate the effect of the intervention on the proportions that adhered to CTXp, and on the time to obtaining CD4 counts, and ART initiation among HIV-positive participants, and on the proportion that accepted repeat HCT among HIV-negative participants. Analyses based on cluster-level summaries (Supplementary File) were also performed to check the robustness of the individual-level analysis, using methods for stratified cluster-randomized trials [36]. Intervention effects on binary outcomes were measured using prevalence ratios (PR), calculated as the ratio of the arithmetic mean of the cluster-specific prevalence of the outcome in each arm. The 95% confidence interval (CI) was calculated with variance estimated from the residual mean square from a two-way analysis of variance (ANOVA) of cluster-specific prevalence on stratum and trial arm. Adjusted PR were calculated as the arithmetic mean ratio of observed to expected prevalences in each cluster, with logistic regression used to estimate the expected prevalence, adjusted for age, sex, stratum and other variables that showed baseline imbalance. CIs were obtained from an ANOVA of the observed/expected prevalences on stratum and trial arm, as described above. Intervention effects on time-to-event outcomes were measured using rate ratios (RR) by similar methods; Poisson regression was used to estimate expected number of events for the adjusted analyses.
Data from the control arm were used to estimate k for linkage to care. STATA version 12.0 (StataCorp, College Station, TX, USA) was used for all analyses.

| Contamination
Two (<1%) individuals reported that they had discussed the trial with participants from villages other than their own.

| Coefficient of variation
k values for linkage to care and rate of linkage were 0.33 and 0.52 respectively.

| DISCUSSION
This trial showed that an intervention comprising two brief counselling sessions for adults identified with HIV through HBHCT strongly increases linkage to care. In the individuallevel analysis, the intervention was associated with a twofold increase in the proportion of persons linking to care, and approximately 5-fold increase in the hazard of linkage after the second counselling visit; the estimates did not change after adjusting for age, sex and travel time to the clinic. Results from the cluster-level analysis were also consistent with an intervention effect on linkage to care and time to linkage. The trial also showed that counselling significantly increased the rate of obtaining CD4 counts and ART initiation, and adherence to CTXp. For all time-to-event outcomes, the intervention effect increased with time, and became apparent after the second counselling visit. Although the two counselling sessions covered similar content, the second session also provided an opportunity to follow-up on issues identified in the first session as well as address new concerns. The effect estimates observed in our trial were stronger than those observed in the only other trial that previously investigated the impact of counselling on linkage to care among HIV-positive persons identified through communitybased HCT (mobile HCT and HBHCT) in SSA [37]. Counselling in that trial increased the proportion of individuals linking to care by 4% only. This was probably due to the already high (89%) level of linkage to care in the standard-ofcare (referral only) arm of that trial. In contrast, linkage to care in our standard-of-care arm was 33%, a figure consistent with that observed after routine referral in previous studies [20,30,31,[38][39][40]. Linkage to care did not differ between trial arms in the month between HIV diagnosis and the first counselling session, suggesting that on its own, knowledge of trial arm had no effect on linkage. The probability of linking to care was also higher in the first month than in subsequent periods. This has been observed previously [10,38] and may mean that persons who accept their status, and are motivated to get help, promptly link to care following HIV diagnosis [41]. The finding also suggests that one month after HIV diagnosis may be the optimal time for targeting persons who are less motivated to seek care.
Whilst the overall proportion of participants who initiated ART was low, ART initiation among those who linked was high. This suggests that linkage to care is the main bottleneck to uptake of ART in this population. As the new WHO guidelines recommending ART initiation regardless of CD4 count become widely adopted, the focus of linkage efforts is likely to shift from just registering for care to ART initiation [42]. Even so, in the absence of home-based ART initiation, HIV-positive patients identified through HBHCT will still have to visit and register with an HIV clinic before receiving ART.
Although linkage to care was significantly higher in the intervention arm, it was still low [25]. This finding points to the existence of additional barriers that cannot be addressed by counselling and the need to identify and evaluate interventions that target these barriers. Such barriers may be programmatic; socio-cultural or structural [24]. Since no single intervention is likely to address all linkage barriers, research will also be required to identify how best to combine interventions that are found to be effective. Higher linkage rates than ours have been observed in some studies that used counselling to facilitate linkage to care [43]. However, most of these studies were prone to confounding, used counselling alongside other interventions [6,7,9], or were conducted in settings where linkage was high even in the absence of any intervention [37].
Consistent with previous studies in similar settings [42,44], uptake of repeat HCT among HIV-negative individuals was high. Previous HCT may reduce the psychological stress associated with testing hence making it easier to accept repeat HCT. Counselling had no measurable effect on uptake of repeat HCT. This may partly be due to the high (79%) uptake of HCT in the control arm, and partly due to the small sample size.
Despite the strong overall intervention effect, outcomes varied substantially between clusters with some intervention clusters showing poorer outcomes than control clusters (Supplementary File). This may have been due to chance or interaction between the intervention and key cluster-level characteristics such as access to health facilities. Variations in implementation of the intervention are unlikely to have affected response to intervention since the same counsellors worked across all clusters.
The study had a number of strengths. It is the first randomized trial to demonstrate that counselling provided after referral substantially increases linkage to care among HIVpositive persons identified through HBHCT in SSA. The trial was conducted under real-world conditions in a relatively large geographical area with facilities that provided standard services. Participation and retention rates were high and did not differ by treatment arm thereby reducing risk of selection bias. The cluster-randomized design further minimized selection bias, and ensured that the risk of contamination between treatment arms, and consequently that of diluting intervention effects, was low. Referrals were tracked to verify outcomes for all participants who reported having linked to care. This minimized the risk of bias that would have resulted from reliance on self-reported data or incomplete tracking of referrals.
The study also had some limitations. First, it was an openlabel study: knowledge of the interventions may have influenced participants' decisions to remain in the study. However, since there was no differential loss to follow-up between arms, this is unlikely to have been a major issue. Second, whereas registering for care is a critical step in the HIV care cascade [27,45], it may not result in sustained engagement in care [46]. Therefore, more research may be necessary to demonstrate the effectiveness of counselling on retention in care and other long-term care outcomes. Third, the counselling sessions and collection of outcome data were conducted by the same person for some participants. This may have increased the risk of reporting bias. In general, reporting bias is unlikely to have been a major issue since only data on adherence to CTXp were based on self-reports. Fourth, the observed k was higher than that assumed for the sample size estimation implying that the trial could potentially have been underpowered.

| CONCLUSIONS
The results of this trial suggest that counselling strongly increases linkage to care and uptake of other services including ART among HIV-positive adults identified through HBHCT. In settings such as ours, counselling may enhance efforts aimed at achieving the second UNAIDS target (i.e. receipt of ART for 90% of the people who know their HIV status) [47] and should be considered for scaling up.

F U N D I N G
This study was jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/ DFID Concordat agreement and is also part of the EDCTP2 program supported by the European Union. The International AIDS Vaccine Initiative provided funds for HIV test kits. The Department of Infectious Disease Epidemiology, London School of Hygiene & Tropical Medicine provided funds to cover CD4 count tests.