Abstract
Background: HIV-infected refugees resettled in the United States face many challenges. Longitudinal data regarding HIV-specific outcomes in this population are limited. Methods: We reviewed charts of 51 HIV-infected sub-Saharan African refugees matched to 102 nonrefugees. Outcomes analyzed included CD4 counts, viral loads (VLs), antiretroviral treatment (ART) use, appointment adherence, opportunistic infections, and resistance mutations. Results: The ART initiation was similar. Appointment adherence was similar in year 1, but refugees were significantly less adherent beyond year 3. Refugees and nonrefugees spent similar amounts of time in care suppressed (83% vs 80%, P = .93). Refugees had higher odds of viremia following undetectable VL (OR 2.3, P < .05). Discussion: Initially, sub-Saharan African HIV-infected refugees have comparable appointment adherence, ART use, and VL suppression to nonrefugees. Overtime refugees were less adherent to appointments and more likely to have postsuppression viremia. The support services provided to refugees early in care may be important for retention in care and treatment success.
Introduction
The United Nations High Commission for Refugees (UNHCR) estimates that there are over 15 million refugees worldwide; of these, 2 million come from sub-Saharan Africa, the region with the highest HIV prevalence. 1 Different from foreign-born immigrants who have relocated by choice, refugees by definition have fled their home countries because of fear of persecution, typically within the context of war and violence. 2 Many have experienced difficult living conditions, poverty, malnutrition, overcrowding, a lack of health care infrastructure and have suffered physical and psychological trauma. These factors are thought to increase their risk for HIV infection, a risk magnified by refugees’ often limited knowledge of HIV transmission and prevention methods. 3 –5
Until 1999, US immigration law excluded HIV-positive persons from resettling in the United States. This policy was revised in 1999, allowing the first HIV-positive refugees to enter by waiver. 6 In January 2010, HIV infection was removed as cause of inadmissibility. 7 Since that time, the number of refugees entering the United States has increased, with close to 75 000 resettled in 2009 and 2010. 8 Considering the high prevalence of HIV in the home countries of many refugees and the circumstances they have survived, it is likely that even more HIV-infected refugees will resettle in the United States.
There are many challenges faced by refugees following resettlement including language and cultural differences. 9 Most refugees are unfamiliar with the American medical system, the concept of preventative care, and the management of chronic diseases. Resettled refugees often have poor nutrition and carry a high burden of tuberculosis, malaria, hepatitis, and intestinal parasites. 4,10,11 Rates of mental health disorders, including posttraumatic stress disorder (PTSD), are high. 4,12 Difficulties adjusting to a diagnosis of HIV and the associated stigma can magnify these challenges.
Previous work has been inconclusive regarding the long-term HIV treatment outcomes in refugees. In 2009, Beckwith et al found that refugees in Rhode Island (largely sub-Saharan African), compared with HIV-infected nonrefugees, had similar stages of HIV infection upon presentation to initial HIV care. However, refugees were less likely to start antiretroviral treatment (ART), attend clinic appointments, and enroll in clinical trial studies. 13 Long-term outcomes were not evaluated. In their report on HIV-infected refugees in 2003, Moreno et al highlighted that discussing indications for ART was a challenge in this population. While the authors suggested that most refugees who were started on ART did well in follow-up, longitudinal data was limited, with a mean follow-up of only 8.6 months. 14
To clarify the impact of refugee status and experiences on success in HIV care, we performed a retrospective study to compare demographic and longitudinal clinical data from HIV-infected refugees from sub-Saharan Africa, with nonrefugees who received care at our institution. Our goal was to identify differences in the long-term outcomes of viral load (VL) suppression and immune restoration between refugees and nonrefugees and to identify the contributing factors that highlight potential targets for interventions to improve care for this population.
Methods
HIV-1-infected persons from sub-Saharan Africa classified as refugees or asylees according to the UNHCR (classed together as refugees for this analysis) who established care at the Miriam Hospital Immunology Center between 2000 and 2008 were analyzed. Each refugee was matched with 2 HIV-infected nonrefugees by gender, CD4 category (<200, 200-350, and >350 copies/µL), and closest date of their initial appointment using an optimal variable matching algorithm. 15 The study was approved by the Miriam Hospital Institutional Review Board.
Baseline data abstracted from medical records included demographics (marital status, country of nationality, education), substance use (any alcohol, tobacco, or illicit drug use), primary and additional risk factors for HIV infection and significant medical comorbidities (active mycobacterium tuberculosis [MTB], latent TB infection [LTBI; defined as a tuberculin skin test >5 mm with a negative chest X-ray], hepatitis B infection [positive hepatitis B surface antigen], hepatitis C antibody status, presence of opportunistic infections, depressive symptoms, PTSD symptoms), and HIV status including diagnosis date, history of AIDS defining illnesses, date of AIDS diagnosis, initial CD4 count, and initial HIV VL. Laboratory data were obtained from the clinic database and verified against the paper chart and hospital electronic records.
Longitudinal data were collected from the first point of care to last point of care prior to May 6, 2010, and included medical comorbidities, AIDS defining illnesses, dates of attended clinical appointments, documented missed appointments, clinical trial referrals and enrollment, CD4 counts, HIV VLs, medications, documented HIV antiviral resistance mutations, and status (active patient, transferred care, lost to follow-up, deceased) at the time of the last service received. Time in care was estimated as the time from first point of care to the last attended visit. First point of care was considered either the registration date or the first visit, whichever came first. In cases where individuals were seen in the hospital or jail and started on medication within 6 months prior to the initial outpatient clinic visit (n = 5), the ART start date was used as the first point of care.
Missed visits were recorded based on the notation in the chart. Four longitudinal clinic adherence outcomes were considered: number of appointments scheduled per year, number of appointments kept per year, proportion of scheduled appointments kept per year, and whether the participant kept 3 appointments per year as a minimum clinical standard of appointments. For the final year of follow-up, data from participants in care for less than 6 months were excluded, and for more than 6 months, but less than 1 year, a threshold of 1 appointment was set as the minimum standard. Generalized linear mixed-effects models for Poisson or binomial data were used to analyze these clinic adherence outcomes, with random effects both for patient stratum and for patient within stratum.
Regarding HIV treatment, we analyzed time to initiation of ART, CD4 count recovery, time to virologic suppression following ART initiation, occurrence of viremia following virologic suppression, and proportion of time for which virologic suppression was maintained following initial suppression. For the latter assessment, virologic suppression was assumed to be maintained during the interval between tests only if the VLs were both undetectable. Accumulation of antiviral resistance mutations was considered as a secondary indicator of treatment failure. The frequency of opportunistic infections in care was also compared as a measure of treatment failure.
Conditional logistic regression was used to compare demographic factors and clinical outcomes between refugees and nonrefugees. Stratified Cox models were used to assess differences in time to ART initiation and time to viral suppression following initiation. A Markov transition model, 16 with random effects for both patient stratum and patient within stratum, was used to determine the likelihood of transitioning from undetectable VL to viremia at the current visit (defined as >400 copies/mL). Predictors included refugee status, viremia at the prior visit, and their interaction. The longitudinal and survival outcomes were adjusted for 2 factors: foreign-born status and any enrollment in a clinical trial. Lastly, a sensitivity analysis was performed on all models to determine the impact of those with initial undetectable VLs, by removing these participants and then re-running the analysis.
Results
Matching and Countries of Origin
Fifty-one HIV-infected refugees from sub-Saharan Africa and 102 HIV-infected nonrefugees matched for gender, date of initial care, and initial CD4 count were included. Sixty-nine percent of participants were female. The median initial CD4 count in the refugee group was 374 cells/µL (range 41-1252), as compared with 337 cells/µL (range 5-1480) in the controls.
The majority of refugees were from Liberia (69%; see Table 1). Forty-one percent of the nonrefugees were foreign born; and 11% were from sub-Saharan Africa.
Country of Origin.
a Central/South America (14), Caribbean (12), Pakistan (1), Portugal (1), Spain (1), Not documented (2).
Sociodemographics, Substance Use, and HIV Transmission
The mean ages at entry to care in the refugee and nonrefugee groups were similar, with medians of 32 years and 35 years, respectively (see Table 2). A similar proportion of refugees and nonrefugees were single (39%) or partnered (53%). The nonrefugee group was more educated. Refugees were less likely than nonrefugees to report tobacco or alcohol use and no refugees reported illicit drug use. Thirty-seven percent of the nonrefugees reported illicit drug use, most commonly cocaine followed by heroin and marijuana. The most common HIV transmission risk factor was heterosexual sex in both groups, reported for 94% of refugees and 68% of nonrefugees (P < .01). Nineteen percent of nonrefugees reported “men who have sex with men” (MSM) sexual activity and 11% reported intravenous drug use (IDU) as primary risk factors, neither of which was reported among the refugee group. Almost half (47%) of refugees reported an additional risk factor of exposure to contaminated medical instruments, as compared with only 3 nonrefugees, all foreign born.
Demographics, Substance Use, and HIV Transmission Risk.
Abbreviation: HS, high school.
a Partnered: married or partnered.
b Single: single, divorced, widowed.
c Illicit drugs by frequency: cocaine (25%), heroin (13%), marijuana (13%), methadone (as marker or prior illicit drug use; 8%), methamphetamine (6%), and crack (4%).
d Individuals with dual risk of MSM/IDU are reported under both routes.
e Vertical, transfusion, contaminated needles/instruments.
Conditions on Presentation
AIDS diagnoses at presentation were recorded for roughly one-third of refugees and nonrefugees (see Table 3). The most common opportunistic infection was MTB, present actively or by history in 12% of refugees and 6% of nonrefugees. Refugees were more likely to be diagnosed with LTBI, both at presentation (20% of refugees and 4% of nonrefugees, P = .14) and once in care (24% of refugees vs 1% of nonrefugees, P < .01). Excluding TB, there was no significant difference in opportunistic infections (OIs) on presentation including Pneumocystis jirovecii pneumonia (PJP; 3 refugees and 6 nonrefugees), toxoplasmosis (1 refugee and 3 nonrefugees), esophageal candidiasis (2 nonrefugees), Cryptococcus (1 refugee), and Kaposi sarcoma (KS; 1 nonrefugee). Refugees were significantly more likely to have active hepatitis B (20% vs 4%, P < .05). Hepatitis C was present in 8% of refugees and 15% of nonrefugees. Symptoms of PTSD and depression were common among both refugees and nonrefugees, with similar rates of depressive symptoms but more PTSD symptoms in refugees (see Table 3).
Conditions at Presentation.
Abbreviations: TB, tuberculosis; MTB, mycobacterium tuberculosis; PTSD, posttraumatic stress disorder; KS, Kaposi sarcoma.
a Excluding TB, in order of frequency: Pneumocystis jirovecii pneumonia, toxoplasmosis, esophageal candidiasis, Cryptococcus, and KS.
Course in Care
Time from diagnosis to first point of care was longer for refugees than nonrefugees (median 5 months vs 1 month, P < .01). The average duration in care was 64 months (range, 1-116 months), with no significant difference between groups. A similar number of refugees (78%) and nonrefugees (84%) started ART during the study period. Nonrefugees started ART sooner than refugees (median time in care 2 months [range, 0-83 months] versus 4 months [range, 0-66 months]), but the difference did not reach statistical significance (hazard ratio [HR] = 0.58, 95% confidence interval [CI] 0.19-1.83, P = .09). Refugees were less likely than nonrefugees to enroll in clinical trials (18%-46%, P < .05). At the end of the study period, the majority of both refugees (75%) and nonrefugees (81%) remained active patients in clinic. More refugees (17%) than nonrefugees (8%) had transferred their care to another clinic (both out of state and in state). In all, 3 (6%) refugees and 10 (10%) nonrefugees were lost to follow-up or died.
Clinic Appointment Adherence
Overall clinic appointment adherence was good in both groups, with refugees and nonrefugees each averaging 5.3 attended appointments per year. In the first year, refugees scheduled more appointments than nonrefugees (rate ratio [RR] 1.33, 95% CI 1.17-1.52, P < .01, Table 4), however they attended a lower percentage of scheduled appointments (RR 0.85, 95% CI 0.77-0.95, P < .01). This pattern continued into years 2 and 3; but in year 4 and beyond, there was no obvious difference in the percentage of appointments scheduled or kept. In years 1 through 3, there was no significant difference between groups in meeting a minimum clinical standard of 3 visits a year. However, refugees had significantly lower odds in year 4 and beyond (odds ratio [OR] 0.28, 95% CI 0.10-0.74, P < .05) of meeting this threshold; 70% of refugees compared with 90% of nonrefugees. These results were adjusted for foreign-born status and clinical trial enrollment.
Yearly Clinic Appointment Adherence, Refugees versus Nonrefugees.
Abbreviation: CI, confidence interval.
a Data adjusted for clinical trial enrollment and foreign-born status.
b P < .01.
c P < .05.
HIV Treatment Outcomes
For those started on ART, the CD4 count at ART initiation was similar for refugees and nonrefugees (see Table 5). Peak CD4 counts recovered after ART initiation were not significantly different.
HIV Treatment Outcomes.
Abbreviations: ART, antiretroviral therapy; CI, confidence interval; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI, protease inhibitor.
a Two cases without CD4 data after antiretroviral therapy initiation.
b As defined in Johnson et al. 17
Three individuals had undetectable HIV VLs at the first point of care, 2 refugees and 1 nonrefugee. All were excluded from analysis related to time to viral suppression and postsuppression outcomes. Their removal had no significant effect on our survival and longitudinal estimates.
There was no significant difference found in the time from ART initiation to observed VL suppression between refugees and nonrefugees (median 65 days vs 63 days, HR = 0.79, 95% CI 0.17-3.70, P = .50). Of those who started on ART, 3 refugees and 5 nonrefugees did not reach viral suppression at any point during the study period. From initial suppression till their final day in care, refugees and nonrefugees spent a similar amount of time, 83% versus 80%, with suppressed VL (P = .93). However, refugees had significantly more variability in VL following suppression, having on average a 2.3 higher odds of viremia following an undetectable VL compared to nonrefugees (95% CI 1.18-4.48, P < .05), adjusted for foreign-born status and clinical trial enrollment.
Comparing those that had genotype testing during care (39 refugees, 69 nonrefugees), neither group was significantly more likely to have major resistance mutations (see Table 5). There were very few OI diagnoses once in care for either group. In all, 1 refugee and 4 nonrefugees were diagnosed with active MTB, 2 refugees and 1 nonrefugee was diagnosed with PJP, 1 refugee and 1 nonrefugee were diagnosed with KS, and 1 nonrefugee was diagnosed with esophageal candidiasis.
Discussion
Overall, both sub-Saharan refugees and nonrefugees in care at our center have done well, over 90% achieving viral suppression at some point, and after initial suppression, both groups maintained viral suppression 80% of the time in care.
Among study participants, the average time in care was over 5 years, providing a substantial amount of time to evaluate outcomes. Entry into care after HIV diagnosis occurred quickly with a median time of 1 month for nonrefugees, 86% of whom were in care by 3 months. This time was a few months longer for refugees than nonrefugees, in part reflecting the time involved in resettlement. The number lost to follow-up was low and similar in both groups. More refugees transferred their care elsewhere, highlighting the transient nature of this population.
Clinic adherence varied between the groups over time. In year one, over 90% of refugees and nonrefugees met the clinical standards for frequency of care visits. During years 1 to 3, refugees made more appointments and attended a smaller percentage of appointments than nonrefugees, and after year 3 the percentage of refugees meeting the clinical standard was significantly lower than nonrefugees. This difference may be a reflection of the additional services, including case management, provided to the refugee population early after resettlement for a limited time. This significant decrease among refugees in adherence to the clinical standard compared with nonrefugees implies the potential for long-term medical consequences; however, our results did not demonstrate this.
We found that a similar percentage of refugees and nonrefugees initiated ART, at a similar CD4 count, and with similar timing. This is somewhat contrary to previous work from our group, which showed refugees were less likely to start ART. 12 As CD4 criteria for initiating ART have changed over the study period, delays in treatment based on CD4 count at initiation could not be assessed, however the similarity in CD4 counts between the groups at ART initiation does suggest that no specific delays were experienced based on refugee status.
We found no difference in the time to achieve viral suppression between the groups, and, similarly, both groups had prolonged periods of viral suppression. This suggests that despite the varied challenges faced by refugees, treatment goals of viral suppression were achievable. However, our transition analysis shows refugees have higher odds of recurrent viremia postviral suppression. This suggests a potential increased risk for poor outcomes including treatment failure and resistance, progression of HIV, and development of opportunistic infections. During the period of our follow-up, this did not translate to clinical consequences of higher rates of major resistance mutations or OIs. However, our numbers of such outcomes were low; larger studies are necessary to further investigate the significance of the high odds of recurrent viremia in refugees.
There are limitations in our study. This study reviews the experience of a relatively small population of refugees in Rhode Island. As such, the power to distinguish small differences in outcomes is low. As noted previously, we have a diverse clinic population which includes many foreign-born persons. A significant percentage of our nonrefugees were thus foreign born, many from sub-Saharan Africa, and therefore potentially share some of the cultural challenges faced by the refugee population. To account for this, we adjusted for foreign-born status in our clinic adherence, survival, and postsuppression viremia transition models. Doing this did not noticeably or significantly change the estimates and intervals for the refugee indicator, thus not altering our findings. Our refugee and nonrefugee groups had differences in educational level, substance abuse, hepatitis B virus (HBV) and hepatitis C virus (HCV) status, and symptoms of PTSD, which may impact HIV outcomes. 18 –21 We were not able to control for all differences because of our small sample size. As this was a retrospective observational study, we were limited in our assessments by the available documentation in the clinical record. As the guidelines for HIV clinical care, specifically indications of ART initiation, have changed over the years, it was not possible to evaluate the timing of ART initiation against current guidelines. The inconsistent documentation of support services prevented detailed analysis of the role of these services in supporting refugee adherence to treatment and retention in care. Despite these limitations, overall we found few clinical markers of poor outcomes suggesting that the panel of services provided led to successful treatment of this at risk population.
In conclusion, refugees in our clinic are doing as well as the nonrefugees, having achieved similar rates of overall viral suppression. Refugees had higher odds of recurrent viremia once suppressed. This finding suggests the need to closely monitor engagement in care among refugees over time. Refugee retention in care was highest early on, a period of time when they received the most intensive case management services. Further studies are needed to better assess the role of these case management and other support services, their impacts on retention in care, and their potential beneficial impacts with regard to adherence and the risk of recurrent viremia on treatment.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Lifespan/Tufts/Brown Center for AIDS Research (grant P30AI42853). Dr Winston and Dr Montague were supported by the National Institute on Drug Abuse (grant 5T32DA013911). MJ Lopez was supported by the National Institute for Health (IMSD grant R25GM083270).