A cohort study to assess a communication intervention to improve linkage to HIV care in Nakivale Refugee Settlement, Uganda

Abstract

Objectives:

Communication interventions to enhance linkage to HIV care have been successful in sub-Saharan Africa but have not been assessed among refugees.

Methods:

Refugees and Ugandan nationals participating in HIV testing in Nakivale Refugee Settlement were offered weekly phone call and short message service (SMS) reminders. We assessed linkage to care and predictors of linkage within 90 days of testing, comparing Intervention participants to those unwilling or ineligible to participate (Non-Intervention).

Results:

Of 208 individuals diagnosed with HIV, 101 (49%) participated in the intervention. No difference existed between Intervention and Non-intervention groups in linkage to care (73 [72%] vs. 76 [71%], p=0.88). Excluding those who linked prior to receipt of intervention, the intervention improved linkage (69 [68%] vs. 50 [47%], p=0.002). Participants were more likely to link if they were older (aOR 2.39 [1.31, 4.37], p=0.005) or Ugandan nationals (aOR 3.76 [1.12,12.66], p=0.033).

Conclusions:

Although the communication intervention did not significantly improve linkage to HIV care, linkage was improved when excluding those with same-day linkage. Excluding participants without a phone was a significant limitation; this data is meant to inform more rigorous designs moving forward. Innovative methods to improve linkage to HIV care for this vulnerable population are urgently needed.

INTRODUCTION

Diagnosing and treating refugees living with HIV is critical to curbing HIV transmission worldwide and achieving the UNAIDS 90-90-90 goal (UNAIDS, n.d., 2014). Sub-Saharan Africa hosts 25.6 million (70%) of the people living with HIV worldwide (UNAIDS, 2013), and 18 million people of concern to the United Nations High Commissioner for Refugees (UNHCR), including refugees, internally displaced and stateless persons, and asylum seekers (UNHCR, n.d.). Refugees suffer adversities including reduced access to basic needs, disrupted social networks, limited livelihood opportunities, threats to their security, and increased susceptibility to mental health conditions. These hardships increase their vulnerability to HIV infection as they may lead to sexual violence, transactional sex, poor access to condoms, and difficulty accessing medical services (O’Laughlin et al., 2013; O’Laughlin et al., 2018; Tanaka et al., 2008; UNAIDS 2007; UNAIDS and UNHCR, 2005; UNHCR, 2013). Additionally, these hardships impact refugees’ engagement in HIV testing, HIV clinic attendance, antiretroviral therapy (ART) initiation and adherence.

An evaluation of the HIV care continuum in the Nakivale Refugee Settlement in southwestern Uganda revealed that only 54% of individuals newly diagnosed with HIV participating in voluntary, routine clinic-based testing initiated clinical care within 3 months of diagnosis (O’Laughlin et al., 2017). Studies in other resource limited settings have demonstrated that cell phone technology—the primary source of communication in resource-limited settings—has enhanced communication between people living with HIV and clinic staff and improved linkage to care and ART adherence (Campbell et al., 2015). Short message service (SMS or text messages) are acceptable and effective in sub-Saharan Africa, improving linkage and retention in care (Horvath et al., 2012; Lester et al, 2010; Mills et al., 2014; Sutton et al., 2017). A study of children living with HIV in Cameroon found that phone calls, SMS, and the combination of both improved clinic attendance (Bigna et al., 2014). In the general population in Uganda, investigators found that phone access was high, privacy and confidentiality concerns rarely deterred participation (Siedner et al., 2012), and mobile phone interventions significantly increased clinic attendance (Kunutsor et al., 2010).

We evaluated a communication intervention for people newly diagnosed with HIV in a refugee settlement in Uganda to determine the feasibility and effectiveness of phone calls and text messages for clinic reminders in this unique population and setting.

METHODS

Study design

We defined those who partook in the communication program as the Intervention cohort. As part of the Intervention, participants received weekly phone calls for 12 weeks from a research assistant trained in HIV counseling. If the client was not reached initially, a maximum of 3 attempted calls were made per week. For all calls, the identity of the client was verified by confirming their full name and age prior to any conversation. Additionally, literate patients received weekly SMS reminders for 12 weeks. For purposes of comparison, while recognizing its inherent biases, we also defined a Non-Intervention cohort which consisted of those who tested HIV positive during the same period but were either unwilling or ineligible to participate in the communication program; these individuals consented to having their clinic data followed as a part of the HIV testing component of the research.

The time for intervention was chosen as 12 weeks. The Centers for Disease Control and Prevention (CDC) defines linkage to care as those who have “visited an HIV heath care provider within 1 month (30 days) after learning they were HIV positive” (Centers for Disease Control and Prevention, 2019). The time was extended to 12 weeks given the increased barriers refugees face in accessing HIV care. Additionally, a prior study evaluating the HIV care continuum in the Nakivale Refugee Settlement defined linkage to care as initiating HIV clinical care within 90 days of diagnosis (O’Laughlin et al., 2017).

The study was approved by the Makerere University School of Health Sciences Institutional Review Board (Kampala, Uganda), the Uganda National Council of Science and Technology (Kampala, Uganda), and the Partners Human Research Committee (Boston, MA, USA).

Study setting

We enrolled participants at the Nakivale Refugee Settlement, home to approximately 68,000 refugees and asylum-seekers from 11 countries. These 68,000 refugees were primarily from the Democratic Republic of Congo (52%), Somalia (17%), Burundi (15%), and Rwanda (15%) (UNHCR, 2014). Health services in the settlement were delivered by the non-profit organization Medical Teams International, and overseen by the UNHCR.

The study was conducted at the Nakivale Health Center, where HIV services were free of charge for refugees and Ugandan nationals including HIV testing (using serial rapid HIV tests as per the Uganda HIV Rapid Test Algorithm (Ugandan Ministry of Health, 2009), pre-ART co-trimoxazole prophylaxis, and ART. Additionally, clients in HIV care were followed with monthly clinic visits either at the main facility, Nakivale Health Center, or at one of the three satellite clinics in the refugee settlement. The HIV prevalence among refugees and Ugandan nationals accessing routine clinic-based HIV testing at Nakivale Health Center was between 3.3–4.5% (O’Laughlin et al., 2014) in 2013, and 7.4% in the surrounding region of Uganda in 2014 (The Republic of Uganda, 2015).

Participant recruitment

Given the concern for poor linkage to HIV care, from November 2014 to July 2016, all refugees and Ugandan nationals participating in routine voluntary clinic-based HIV testing at Nakivale Health Center were offered a communication intervention prior to testing. During the intake survey, we asked all individuals if they would be willing to participate in the communication intervention if deemed eligible. The intake survey was administered to individuals prior to their exclusion from the intervention study. The survey was also used to collect demographic information (sex, age, refugee status, marital status, education), to document self-reported access to a mobile phone and literacy, and to assess time and cost to reach clinic.

Eligibility criteria included: adults ≥ 18 years, able to provide informed consent (in Kiswahili, Runyankore, Kinyarwanda, or English), no prior HIV diagnosis, and mobile phone access.

Classification of endpoints, data collection, and implementation considerations

The primary outcome was HIV clinic attendance (i.e. “linkage”) within 90-days of HIV diagnosis. Participants were considered linked if one of the following criteria were met within 90 days of HIV testing: 1) at least one HIV clinic visit on or after the diagnosis date; 2) initiated ART; and/or 3) had a CD4 test. Data were prospectively obtained from written clinical registers at Nakivale Health Center and the three satellite clinics. After initial study design but prior to study implementation, the Nakivale clinics instituted a policy to enhance HIV clinic attendance after a positive diagnosis—clients were escorted to the HIV clinic immediately after testing to facilitate “linkage.” Because we hypothesized the intervention would be most helpful for those who did not immediately link, we established a secondary outcome, changing the criteria for “linkage” as attending clinic any day after the initial testing day.

Statistical methods

For comparisons between the Intervention and Non-Intervention cohorts, we used the Wilcoxon Rank Sum test for continuous variables and Fisher’s Exact test or the Pearson chi-square test for categorical variables. We calculated an unadjusted and adjusted odds ratios (OR and aOR, respectively) and the corresponding p-values for predictors of linkage to care after the day of diagnosis for those in the Intervention group. Variables were included in the multivariable analysis if they were considered important a priori (sex, age, refugee status) or if the p-value was less than 0.10 (currently married, number of times reached by phone). Statistical analyses were performed using SAS software (version 9.4, SAS Institute, Cary, NC).

RESULTS

Participant Characteristics

Of 5,583 participants surveyed and tested for HIV, prior to assessment of eligibility, 2,901 (52%) responded that they would be willing to participate in the communication intervention if invited. Willingness to participate was lower among females versus males (1,067/2,851 [37%] vs. 1,833/2,729 [67%], p <0.0001), among refugees versus Ugandan nationals (1,830/3,738 [49%] vs. 1,066/1,834 [58%], p <0.0001), and among those living inside versus outside Nakivale (2,383/4,750 [50%] vs. 517/815 [63%], p<0.0001).

There were 208 newly diagnosed individuals living with HIV, 101 (49%) of whom participated in the intervention. The Non-Intervention cohort consisted of 107 individuals newly diagnosed with HIV who were not eligible or not willing to participate in the intervention. Of those in the Non-Intervention group, 94 (88%) were unable to participate due to lack of mobile phone access. The Intervention cohort included fewer refugees (21% vs. 41%; p=0.003), participants with higher school education (p<0.001), and participants who spent more money to travel to the clinic (p=0.006) (Table 1). No significant differences in proportion female (p=0.324), age (p=0.364), marital status (p=0.069), or longer travel-time to clinic (p=0.068) were detected (Table 1).

Table 1.

Baseline characteristics of participants in the Intervention and Non-Intervention cohorts.

Variable	Intervention (N=101) % (n)	Non-Intervention (N=107) % (n)	P-value*
Female	55 (56)	63 (67)	0.324
Age, median years (IQR)	30 (25, 36)	30 (25, 40)	0.364
Refugee^	21 (21)	41 (43)	0.003
Currently married, yes^	62 (63)	50 (52)	0.069
Education^			<0.001
No school	12 (12)	37 (39)
Some primary school	48 (48)	49 (51)
Completed primary school	41 (41)	14 (15)
Time to clinic^, median hours (IQR)	0.8 (0.4, 1)	1 (0.5, 2)	0.068
Cost to clinic^, Ugandan schillings			0.006
0	52 (45)	76 (73)
>0-5,000	16 (14)	9 (9)
>5,000-10,000	13 (11)	4 (4)
≥10,000	20 (17)	10 (10)

Note: Denominators vary due to participant non-response.

^{^}Missing data (intervention group / non-intervention group): refugee status (0/1), currently married (0/2), education (0/2), time to clinic (0/2), cost to clinic (14/11)

^*P-value based on Wilcoxon test (continuous variables) or Fisher's Exact Test or Pearson chi-square test (categorical variables depending on number of categories).

Evaluating the primary outcome for this study, we found no difference in linkage to clinic between the Intervention and Non-intervention groups (73 [72%] vs. 76 [71%], p=0.879). Of individuals living with HIV in both cohorts, 37% attended the HIV clinic on day 0, prior to receiving the intervention. Evaluating the secondary outcome, linkage to HIV clinical care within 90 days excluding the day of HIV diagnosis, we found increased linkage in the Intervention group compared to the Non-intervention group (69 [68%] vs. 50 [47%], p=0.002).

In the Intervention group, 56 (55%) self-reported being literate and received SMS and phones calls—the remainder received calls without SMS. For those in the Intervention cohort who linked to care within 90 days of testing excluding those who linked on the day of diagnosis, the median number of phone call attempts prior to linkage was 6, and the median number of successful calls was 2. For those in the Intervention group who did not link to care, the median number of call attempts within 90 days of testing was 25 with a median of 3 successful calls per person. Research Assistants reported that some attempted calls occurred when phones were answered by family or friends and others did not go through because the phone was off or the connection was poor.

Because the intervention had a greater impact on linkage when participants who linked on the day of HIV testing were excluded, we evaluated the predictors of linkage to HIV clinic after the day of diagnosis in the Intervention group (Table 2). In the univariate analysis, Intervention participants were more likely to link to the HIV clinic after the initial day of testing if they were older (OR 2.31 [1.28, 4.18] per 10 years, p=0.006), married (OR 2.59 [1.09, 6.14], p=0.031), and if they were reached by phone more often (OR 1.11 [0.99, 1.23] per contact, p=0.067). Ugandan nationals were more likely to link compared to refugees (OR 3.09 [1.15,8.32], p=0.026). Among 83 participants in the Intervention group reached by phone, 56 (67%) linked (after the testing date) and 27 (33%) did not link. Of the 18 in the intervention cohort not reached by phone, 13 (72%) linked after the day of diagnosis (p=0.786 between groups). In the multivariable analysis, significant variables of linkage after the day of HIV testing in the Intervention cohort included older age as a predictor of linkage (aOR 2.39 [1.31, 4.37], p=0.005) and Ugandan nationals compared to refugees (aOR 3.76 [1.12,12.66], p=0.033) as predictors of linkage.

Table 2.

Predictors of linkage to HIV clinic after day of diagnosis in the Intervention group.

	Unadjusted		Adjusted
Variable	Odds Ratio [95% CI]	P-value	Ods Ratio [95% CI]	P-value
Female	0.95 [0.41, 2.22]	0.912	1.48 [0.55, 3.99]	0.433
Age, 10 years	2.31 [1.28, 4.18]	0.006	2.39 [1.31, 4.37]	0.005
National (Ugandan) vs Refugee	3.09 [1.15-8.32]	0.026	3.76 [1.12-12.66]	0.033
Currently married, yes	2.59 [1.09, 6.14]		2.27 [0.83, 6.24]
Completed primary school, yes	0.57 [0.24, 1.33]	0.031		0.111
Time to clinic, hours	1.69 [0.84, 3.41]	0.192
Cost to clinic, Ugandan schillings		0.142
>0-5,000 vs 0	0.60 [0.18, 2.06]	0.704
>5,000-10,000 vs 0	1.20 [0.28, 5.23]
≥10,000 vs 0	1.47 [0.41, 5.31]
Ever reached by phone, yes	0.80 [0.26, 2.47]
Number of times reached by phone	1.11 [0.99, 1.23]	0.695	1.11 [0.98, 1.25]
		0.067		0.114

Note: The overall p-value is provided for variables with more than 2 categories.

DISCUSSION

Among individuals newly diagnosed with HIV in Nakivale Refugee Settlement in Uganda, a phone call/SMS intervention to encourage linkage to HIV clinical care did not significantly improve linkage for all comers. Unrelated to this study, the HIV testing center in Nakivale developed a practice of walking newly diagnosed individuals to the HIV clinic to link on day 0. However, after excluding participants who linked to care on the day of diagnosis prior to receipt of the intervention, the communication intervention did significantly increase linkage to care for Intervention cohort participants (68% vs 47%, p=0.002). Most ineligible individuals were excluded because they did not have a mobile phone. Only about half of participants were literate and could participate in the text message component of the intervention. Among the Intervention cohort, important characteristics of successful linkage to care included older age and being a Ugandan national (vs refugee). The impact of this mobile phone HIV linkage intervention in the refugee settlement and the barriers to successful intervention implementation merit further examination.

A crucial barrier to successful implementation of this phone call/text message intervention in this humanitarian setting was disparities in access to a mobile phone. The majority of those who did not meet inclusion criteria were excluded because they did not have access to a phone. This problem of limited mobile phone access among potential study participants was also recognized in a HIV study in South Africa where 43% of patients at a recruitment site had mobile phones (Venter et al., 2018). A study in Kampala, Uganda, evaluated the use of mobile phones to communicate about tuberculosis care and reported that 75% of clinic attendees owned a phone (Ggita et al., 2018). Data is not available regarding the availability of mobile phones among people accessing services in Nakivale Refugee Settlement. Given the severe poverty and limited livelihood opportunities in the refugee settlement, it is likely that mobile phone availability is even more limited than in other areas of Uganda. This is concerning because it is likely that those with less resources, including lack of mobile phones, need more assistance to overcome barriers to engage in clinical care. This is a significant limitation of the study; however, there are few studies focused on refugee populations, as significant barriers often prevent testing and assessing such a vulnerable and isolated population. This data is meant to inform more rigorous designs moving forward.

Our study may reflect that those with more resources better overcame barriers to linking to HIV clinical care in this austere setting. There was increased linkage to HIV clinical care within 90 days excluding the day of HIV diagnosis in the Intervention group compared to the Non-intervention group. The Intervention cohort included individuals who had more years of education, were Ugandan national (vs. refugee), and were more likely to have access to a mobile phone. While these characteristics are not necessarily indicative of more wealth, it is likely that is the case. Furthermore, in a recent study in Uganda on mobile phones and mental well-being, survey data demonstrated that households with mobile phones had higher levels of wealth (Pearson et al., 2017). Given that socio-economic status is a predictor of linkage to care, this selection bias is an important limitation of the study.

While mobile phone technology interventions have improved engagement in HIV care in sub-Saharan Africa (Horvath et al., 2012; Lester et al, 2010; Mills et al., 2014; Sutton et al., 2017), several unique barriers affected the intervention effectiveness and should help inform future mobile phone interventions in refugee settlements or other austere settings. It would be helpful to provide a phone stipend to people newly diagnosed with HIV to facilitate their successful engagement in HIV care. This would ensure people could participate and potentially benefit from this intervention even if they did not possess their own mobile phone. Providing money instead of a phone could also minimize stigma associated with recognizing phones dispensed by the study. Additionally, more than half (55%) of the participants in the Intervention group were illiterate and could not receive text messages. Future mobile phone interventions among refugee populations might use audio messages for reminders instead of texts. It would be important to ensure the phone was passcode protected or that there was some way that only the study recipient could access the audio message in case the phone was being shared with other people. Given the number of times study participants were not reached because phones were powered off presumedly to preserve power, the intervention may have enhanced penetrance if participants had facilitated access to charge phones.

Finally, this type of program would need to be carefully designed to avoid introducing stigma or unintentional disclosure of HIV status. Willingness to participate was lower among females and refugees. Given these populations are often more vulnerable and rely more heavily on social support to survive, the unwillingness to participate may have stemmed from concern that phone calls would inadvertently result in disclosure of HIV status and thus stigmatization.

CONCLUSIONS

Interventions to engage people in HIV care in humanitarian settings are needed. Phone calls and SMS reminders may represent a promising strategy, however, ensuring access to a mobile phone and alternatives to text messages including audio messages are critical to the success of mobile phone technology interventions in this environment. This is of critical importance as it is plausible that individuals lacking access to these resources, and thus of a lower socio-economic status, are likely to face more barriers to engagement in HIV clinical care. The options for communication between healthcare providers and clients are limited in humanitarian settings due to the lack of landlines, paper mail, and internet access; thus, communication might be enhanced if critical barriers to effective mobile technology implementation are mitigated. This may be especially important for home-based testers, who are not likely to link to HIV clinic the day of testing and may benefit from communication interventions. Interventions relying on technology to enhance communication and improve engagement in care may be feasible and useful in refugee settlements if barriers to accessing and optimizing the safe use of mobile phones are first addressed.