Abstract
Current literature on retention in HIV care fails to account for patients who continually/simultaneously access different providers. This statewide study examined retention in early HIV medical care and its impact on viro-immunological improvement and survival outcomes. It was a retrospective study of South Carolina residents ≥13 years old who were diagnosed with HIV infection in 2004–2007 and initially entered in care. CD4 count/percent and viral load (VL) tests that must be reported to the South Carolina HIV surveillance database were used as a proxy for a clinical visit. Retention was defined as at least one visit in each of four 6-month periods over 2 years postlinkage. Retention rates were categorized as “optimal” (visits in four intervals), “suboptimal” (visits in three intervals), sporadic (visits in two or one intervals), and “dropout” (no visits). Logistic regression and Cox proportional analyses were used to examine retention. Of the 2197 persons, about 50% failed to maintain optimal retention in care postlinkage. Male gender, nonwhite race/ethnicity, younger age, delayed linkage, and HIV-only status were significant predictors of lower rate of retention. Mean decrease in baseline log10 VL was greater among those with optimal compared to suboptimal (−1.81 vs. −1.42; p < 0.001) and sporadic retention (−1.81 vs. −0.70; p < 0.001). Mean increase in baseline CD4 count was greater in optimal retention compared to suboptimal (169.70 vs. 107.5; p < 0.001) and sporadic retention (169.70 vs. 2.43; p < 0.001). Increased risk of mortality was associated with sporadic retention (aHR 2.91; 95% CI 1.54–5.50) and “dropout” (aHR 4.00; 95% CI 1.50–10.65). Rate of poor retention in early HIV medical care was relatively higher than reported in clinic-based data. Increasing the rate of retention in early HIV care could substantially improve viro-immunological parameters and survival outcomes.
Introduction
E
Pertinent literature indicates that a substantial proportion of people initially engaged in care may cycle in and out of regular HIV medical care. 6,8,9 In a recent study by Ulett et al. 6 about 40% of the participants failed to remain fully engaged over 2 years after initial linkage. Various sociodemographic, structural, behavioral, and clinical factors could be associated with lower rates of retention in HIV care. 4
The central premise of retention is continuous engagement with HIV care without interruption of access to cART. Many prior studies on retention in HIV care are mostly involve clinic-based data from large urban centers. 6,8,9 Horstmann et al. 4 suggest that most previous studies on retention may not identify those persons who may simultaneously and/or continually access different healthcare settings, which could potentially lead to misclassification bias. Therefore, it is important to study retention based on comprehensive state-level data that cover all healthcare providers in a state.
In this report, an objective was to determine retention rates, after initial linkage to medical care was established, in a cohort of newly diagnosed HIV-infected persons in South Carolina (SC) and characterize factors associated with lower retention after initial entry into HIV care. Another objective was to examine the association of retention in the first 2 years after linkage with clinical outcomes and survival.
Materials and Methods
Study population and procedure
Data for this study were obtained from the SC enhanced HIV/AIDS Reporting System (eHARS) surveillance database. The study cohort included SC residents ≥13 years old who were newly diagnosed with HIV infection on or after January 1, 2004 and who have been linked to HIV care by December 31, 2007. Inclusion of follow-up eHARS data through December 31, 2009 permitted the observation of retention in care for at least 2 years postlinkage for each eligible individual. There were 719 persons diagnosed with HIV infection from January 1, 2004 through December 31, 2007 who were not linked to care and thereby excluded from these analyses. Since we observed rate of retention over 2 years postlinkage, persons who died in less than 2 years after linkage to care (n = 180) were excluded. This resulted in a total study cohort of 2197 persons.
In SC, since January 2004, state legislation has made it mandatory for all licensed laboratories to electronically report all viral loads (VL) and CD4+ T cell counts test results to the SC Department of Health and Environmental Control (SC-DHEC). These data are continually entered in the eHARS surveillance database. SC-eHARS data quality ratings exceed the Centers of Disease Control and Prevention (CDC) minimum standards of reporting timeliness with 95% of cases being reported within 6 months of diagnosis and 98% completeness of reporting (by comparison with external databases) (SC-DHEC unpublished data, December 2008).
Variable definitions
According to the current Department of Health and Human Services (DHHS) guidelines for adolescents and adults with HIV, primary care visits should be made by newly diagnosed persons at least every 3 to 4 months until initial patient evaluation is completed and stable clinical and immunological status is achieved for 2 to 3 years. Thereafter, at least one visit every 6 months is recommended for monitoring optimal health outcomes. 10 Based on these minimum recommended standards of engagement in HIV primary care, retention was defined as at least one visit in each of the four 6-month intervals over 2 years postlinkage to care. Similar methodology of defining retention as at least one visit in each 6-month period has been used in previous studies. 6,11 The dates of a laboratory test of CD4+ T cell count and/or VL in SC-eHARS were used as a surrogate marker of primary care visits since these test are ordered by an HIV care provider. 12 We used the date of first CD4+ T cell count or VL laboratory test reported after a 30-day washout period from the date of HIV diagnosis as a proxy for linkage to care. The washout period of 30 days from the date of HIV diagnosis was used to capture and eliminate laboratory records potentially associated with testing the site of initial HIV diagnosis and therefore was not indicative of linkage to medical care.
Rate of retention over 2 years postlinkage was categorized into four groups: optimal retention was defined as visits in four out of the four 6-month intervals; suboptimal retention was defined as visits in three out of four 6-month intervals; sporadic retention was defined as visits in two or one out of four 6-month intervals; and “dropout” defined as no visit recorded over 2 years. We did not examine visit in two or one out of the four 6-month intervals as separate categories because of small numbers.
Predictor variables requested from SC-eHARS included gender, age at time of diagnosis, place of residence at time of diagnosis (rural/urban), and race/ethnicity (black, white, and “Other” including nonwhite Hispanics, Pacific Islanders, and Asians). For multivariable analyses “Other” race/ethnicity” and black categories were merged as a “nonwhite” category because of the small number of “Other” race/ethnicity (n = 104). The self-reported mode of HIV exposure risk was categorized as men who have sex with men (MSM), injecting drug users (IDU), heterosexuals, and unidentified risk. Men who reported both MSM and IDU were grouped together with IDU because of their small numbers. The reported source/site of HIV diagnosis was categorized as county health department, hospitals, private physician/medical group practice, state facilities (Department of Corrections and Mental Health), and “Other” locations, which included laboratories, blood banks, private businesses, and federal institutions (e.g., Veterans Administration clinics or hospitals). Disease stage at time of diagnosis was defined as HIV (CD4+ T cell count >200 cells/mm3) or concurrent AIDS (CD4 ≤200 cells/mm3 or >200 cells/mm3 with clinical disease) defined as diagnosis of AIDS within 3 months of the first positive HIV test. Time between date of HIV diagnosis and initial linkage to care was categorized as ≤3 months and >3 months. We created a proxy variable for participants' probable sociodemographic challenges to access care at the time of diagnosis by using the 2006 U.S. Census data of the proportion of the population below the poverty level in the county of residence. 13
Statistical analysis
Preliminary descriptive analyses were performed to determine data distribution and assess univariate association of sociodemographic and clinical stage variables with the retention categories (optimal, suboptimal, sporadic, and “dropout”). Statistical differences were assessed by the chi-square test for categorical variables and Mann–Whitney test for continuous variables (Table 1). Multinomial logistic regression analysis was used to examine the variables associated with suboptimal and sporadic retention and “dropout” as compared to optimal retention (referent outcome variable).
Data are N (%) for categorical variables and the median [interquartile range (IQR)] for continuous variables. Retention in care was defined as having at least one CD4+ T cell count and/or viral load test reported to eHARS in every 6-month period over the 2 year follow-up from the date of linkage to care. “Optimal retention” represents persons who had a visit in four out of four 6-month intervals; “Suboptimal retention” represent persons who had a visit in three out of four 6-month intervals; whereas “Sporadic retention” represents persons who had one or two visits out of four 6-month intervals. “Dropout” is defined as no CD4 cell count or viral load test reported to eHARS during 2 years after the initial linkage to medical care. This table includes persons newly diagnosed with HIV on or after January 1, 2004 who initially linked to care by December 31, 2007. Inclusion of eHARS data through December 31, 2009 permitted the observation of retention in care for 2 years postlinkage for each eligible individual.
The “other” category is composed of Hispanic, Asian, and Pacific Islander groups.
This category includes heterosexuals who had sex with a high-risk individual (such as an injection drug user, bisexual male, transfused individual, or an individual who is HIV+).
State facilities are largely composed of persons diagnosed positive at the state department of corrections and at mental health facilities.
The others category includes South Carolina residents who were reported from other states, from blood banks/businesses, federal facilities, laboratories, and from unknown sources.
Concurrent AIDS is defined as the diagnosis of AIDS within 3 months of the first positive HIV test.
MSM, men who have sex with men.
We evaluated the association of retention in care with improvement in baseline CD4+ T cell count and log10 VL over a 2-year follow-up period after linkage to care. A general linear model was used to separately evaluate the association of retention with change in baseline CD4+ T cell count and log10 VL over a 2-year follow-up period after linkage. The first log10 VL and CD4+ T cell count obtained within a 6-month period after the 2-year observation period of retention in care were used to calculate the change in baseline parameters. A similar methodology was used for examining change in immunovirological parameters by Giordano et al. 9 Least square mean changes in log10 VL and CD4+ T cell count from the final parsimonious model are reported. Persons with no reported CD4+ T cell count or VL in 2 years after linkage to care (“dropout”) were not included in this analysis.
Survival analysis was used to investigate the association of retention in care with time to death. The SC-eHARS database is linked with the SC DHEC vital records database. Data used in this study were up to date for mortality records at the end of the study period—December 2009. The survival clock for each individual was started 2 years after linkage to care, when the 2-year rate of retention has been observed and assigned. For these analyses, persons were censored at the end of the study observation period—December 31, 2009. The proportionality assumption and univariate associations were assessed by Kaplan–Meier curves. Cox proportional hazards models were used to obtain unadjusted and adjusted hazard ratios of factors associated with the long-term survival.
The initial full models were reduced by a backward elimination procedure (p-value for elimination defined as >0.08) to achieve a parsimonious model. p-values <0.05 were considered statistically significant and all statistical analyses were performed with SAS software, version 9.1.3 (SAS Institute, Cary, NC).
Results
Mean age (±standard deviation: SD) of the study cohort (N = 2247) was 37 ± 12 years and the majority were men (66%), black (72%), and urban residents (71%). More than one-third (36%) of the study cohort reported MSM as their mode of exposure to HIV. First CD4+ T cell count and VL results were reported for 100% and 97% of the eligible persons, respectively. The baseline median CD4+ T cell count was 328 cells/mm3 [interquartile range (IQR): 148–512] with the majority (54%) having ≤350 cells/mm3. The median initial VL was 25,119 copies/ml (IQR 5,012–100,000) with a majority (75%) having values <100,000 copies/ml. Approximately 36% of these persons had concurrent AIDS.
Factors associated with a lower rate of retention in care
In our cohort, only 1092 (50%) had optimal retention in the 2 years after linkage to care with at least one visit in each 6-month period (Table 1). Multinomial logistic regression analysis (Table 2) suggests that suboptimal retention (visits in three out of the four 6-month intervals) was significantly more likely among men compared to women [adjusted odds ratios (aOR) 1.46; 95% confidence interval (95% CI) 1.08–1.96] and among urban residents as compared to rural residents (aOR 1.33; 95% CI 1.02–1.74). Suboptimal retention was less likely among those of older age at time of diagnosis (aOR 0.92; 95% CI 0.87–0.97), among those who reported MSM as a risk of HIV transmission compared to heterosexual transmission (aOR 0.67; 95% CI 0.47–0.94), and among those who had concurrent AIDS (aOR 0.56; 95% CI 0.43–0.71).
The “black and other” (nonwhite) category is composed of blacks and nonwhite Hispanic, Asian, and Pacific Islander groups (n = 104).
Estimate is reported for a 5-year increase in age.
This category includes heterosexuals who had sex with a high-risk individual (such as an injection drug user, bisexual male, transfused individual, or an individual who is HIV+).
Concurrent AIDS is defined as a diagnosis of AIDS within 3 months of the first positive HIV test.
MSM, men who have sex with men; IDU, injecting drug users; alpha level, 0.05; Ref., referent group; CI, confidence interval; aOR, adjusted odds ratio;
Sporadic retention (visits in one or two out of the four 6-month periods) was more likely among men (aOR 1.54; 95% CI 1.14–2.06), among those with nonwhite race/ethnicity (aOR 1.44; 95% CI 1.09–1.90), and among those who had an initial linkage to care in more than 3 months after HIV diagnosis (aOR 1.35; 95% CI 1.08–1.69). Sporadic retention was less likely among those of older age at time of HIV diagnosis (aOR 0.84; 95% CI 0.80–0.89) and among those with concurrent AIDS (aOR 0.52; 95% CI 0.41–0.65).
Failure to engage in care over 2 years after initial entry in care (“dropout”) as compared to optimal retention was significantly more likely in men (aOR 2.24; 95% CI 1.49–3.36), among those with nonwhite race ethnicity (aOR 2.40; 95% CI 1.54–3.74), and among those who had linkage to care delayed by more than 3 months after HIV diagnosis (aOR 1.96; 95% CI 1.42–2.69); The odds of failure to engage in care were significantly less likely in those of older age at the time of diagnosis (aOR 0.87; 95% CI 0.81–0.94) and among those who had concurrent AIDS (aOR 0.37; 95% CI 0.26–0.52).
Association of retention with changes from baseline CD4+ T cell count and VL
Results from the multivariable general linear model suggested that the mean decrease from baseline log10 VL over 2 years after linkage was greater among those persons who had optimal retention as compared to those with suboptimal retention (mean log10 VL: −1.81 vs. −1.42; p-value <0.001), among whites as compared to nonwhite race/ethnicity (mean log10 VL: −1.45 vs. −1.18; p-value = 0.003), and in those with concurrent AIDS as compared to HIV only at the time of diagnosis (mean log10 VL: −1.75 vs. −0.87; p-value <0.001). A 5-year increase in age at diagnosis was associated with a mean decrease of 0.07 log10 VL (p-value <0.001). Gender, mode of exposure, residence in rural/urban area, and time to linkage to care after HIV diagnosis were not found to be significant predictors of change in log10 VL.
The final model for change in CD4+ T cell count indicated that the mean increase from the baseline CD4+ T cell count was significantly higher by 170 cells/mm3 among those with optimal retention intervals as compared those with suboptimal retention (mean CD4 count: 169.70 vs. 107.5; p-value <0.001) and sporadic retention (mean CD4 count: 169.70 vs. 2.43; p-value <0.001). Results also indicated that the mean increase was greater among whites compared to nonwhite race/ethnicity (mean CD4 count: 111.0 vs. 75.5; p-value = 0.019) and among those with concurrent AIDS by 157 cells/mm3 as compared to those with HIV only at the time of diagnosis (mean CD4 count: 171.5 vs. 14.9; p-value <0.001). No other significant differences were found.
Association of survival with retention in care
In this population, 58/2197 (2.6%) persons died in the follow-up period of up to 35 months after a 2-year observation period for retention in care. Mean survival time for the cohort was 30.4 months. The multivariable proportional hazards analyses (Table 3) indicated that after adjusting for other variables, an increased risk of mortality was associated with sporadic retention (aHR 2.91; 95% CI 1.54–5.50) and “dropout” (aHR 4.00; 95% CI 1.50–10.65) as compared to optimal retention. Adjusted survival curves by categories of retention are presented in Fig. 1. An increased risk of mortality was also associated with older age at HIV diagnosis (aHR 1.05; 95% CI 1.03–1.08) and among those with concurrent AIDS (aHR 2.76; 95% CI 1.59–4.78). No other significant association was observed.
Adjusted survival curves by retention in care among 2197 newly HIV-diagnosed persons in South Carolina. Retention in care was defined as having at least one CD4+ T cell count and/or viral load test reported to the enhanced HIV/AIDS Reporting System (eHARS) in every 6-month period over the 2-year follow-up from the date of linkage to care. "Optimal retention" represents persons who had a visit in four out of four 6-month intervals, “suboptimal retention" represent persons who had a visit in three out of four 6-month intervals, whereas "sporadic retention” represents persons who had one or two visits out of four 6-month intervals. “Dropout” is defined as no CD4 cell count or viral load test reported to eHARS during 2 years after initial linkage to medical care. This figure includes persons newly diagnosed with HIV on or after January 1, 2004, who initially entered care by December 31, 2007. Inclusion of eHARS data through December 31, 2009 permitted the observation of retention in care for 2 years postlinkage for each eligible individual. The adjusted survival curve is based on survival estimates obtained from the final Cox proportional hazards model, which included retention categories, race/ethnicity, age at diagnosis, and concurrent AIDS.
Retention in care was defined as having at least one CD4+ T cell count and/or viral load test reported to eHARS in every 6-month period over the 2-year follow-up from the date of linkage to care. “Optimal retention” represents persons who had a visit in four out of four 6-month intervals; “Suboptimal retention” represent persons who had a visit in three out of four 6-month intervals; “Sporadic retention” represents persons who had one or two visits out of four 6-month intervals. “Dropout” is defined as no CD4 cell count or viral load test reported to eHARS during 2 years after initial linkage to medical care. This table includes persons newly diagnosed with HIV on or after January 1, 2004 who initially entered care by December 31, 2007. Inclusion of eHARS data through December 31, 2009 permitted the observation of retention in care for 2 years postlinkage for each eligible individual.
The “black and other” (nonwhite) category is composed of blacks and nonwhite Hispanic, Asian, and Pacific Islander groups (n = 104).
Estimate is reported for a 1-year increase in age.
This category includes heterosexuals who had sex with a high-risk individual (such as an injection drug user, bisexual male, transfused individual, or an individual who is HIV+).
State facilities are largely composed of individuals who were diagnosed at the state department of corrections and at mental health facilities.
The “other” category is composed of SC residents who were reported from other states, from blood banks/businesses, from federal facilities, from laboratories, and from unknown sources.
Concurrent AIDS is defined as a diagnosis of AIDS within 3 months of the first positive HIV test.
MSM, men who have sex with men; alpha level, 0.05; Ref., referent group;
Discussion
This study demonstrates the impact of retention in the early period of HIV care on clinical outcomes and explores the factors associated with retention in care using statewide surveillance data. In our cohort, 40% persons had lower rates of retention and another 10% failed to remain engaged in care over 2 years after initial linkage. These results are worrying since they suggest that even after taking into account persons who may simultaneously and/or continually access different HIV care providers within the state due to various reasons such as relocation, incarceration, insurance status changes, convenience, and/or choice, about half of newly diagnosed persons fail to establish adequate HIV care after the initial linkage. These rates are relatively higher as compared to that reported in a recent clinic-based study from the South that found that about 40% of patients fail to have a visit in each 6-month period over 2 years after initial linkage to care. 6 Other studies that have examined retention after the initial entry in care at primarily urban settings have also reported lower rates of retention. 4,8,9
Lower retention in HIV primary care was found to be significantly associated with sociodemographic and clinical factors. In this SC population, male gender, younger age, and nonwhite race/ethnicity were significant predictors of lower retention or failure to remain in care. These findings are reflective of well-documented racial, gender, age, socioeconomic, and sexual behavior disparities in HIV-related healthcare access and utilization. 7,14 –17
Our results suggest that men are at higher risk of a lower rate of retention or failure to remain engaged in care as compared to women, whereas other clinic-based data suggest women are at a higher risk of suboptimal retention. 8,18 A plausible explanation for this finding could be that in this SC population, the majority of the women (n = 527/740; 71%) were of childbearing age of 15–44 years old and were more likely to have routine HIV testing as part of prenatal care. 19,20 In a multicenter study by Fernandez et al. 20 on acceptance of HIV testing during prenatal care, 86% of the participants reported having been tested or consented to be tested. Consequently, heightened awareness about vertical transmission and receipt of outpatient HIV care at the same healthcare facility as that providing prenatal and postnatal care may promote early retention in care among HIV-infected women.
The majority of the SC HIV-infected population was diagnosed at an advanced disease stage with 54% of eligible persons having a CD4 count ≤350 cells/mm3 and 36% with concurrent AIDS. It has been shown that that those with concurrent AIDS at the time of diagnosis were more likely to have a higher rate of retention in care compared to those with HIV only. 6,8 However, failure to engage in regular care among persons with HIV disease stage increases the risk of poor health outcomes and on-going transmission of HIV.
In this study cohort, more than half failed to establish a linkage with HIV medical care within 3 months of diagnosis. Previous studies have also indicated that a substantial proportion of people with known seropositive status may delay entering care by several months. 5,6,12,21,22 In a recent study, HIV surveillance data from Philadelphia showed that the median time to entry into care of newly diagnosed persons was 8 months. 23 After controlling for demographic variables such as race and gender, linkage to care was found to be a significant independent predictor of retention, with those having delayed linkage being more likely to have lower rates of retention or dropout after initial linkage. These results reiterate the importance of promoting early linkage to care after HIV diagnosis, which may, in turn, facilitate better retention in care by providing prompt support to mitigate various sociodemographic, psychological, structural, and clinical barriers that may impede retention in care.
Ancillary support programs such as case management for unmet needs, mental health, and substance use services could be successful measures to improve retention in care. 7,24,25 In a multisite study by Cabral et al., participants who received nine or more contacts during the first 3 months of their programs were 50% less likely to have substantial gaps in care during the first 12 months of follow-up. 25 A seven-site qualitative study that enrolled underserved HIV-infected participants who cycled in and out of care found that an outreach program could be greatly beneficial in reconnecting these persons to HIV care. 26 Moreover, targeted intervention toward vulnerable groups may be an effective strategy in improving the rates of engagement in regular care. 4,18,27
In this SC population, a lower rate of retention, after adjusting for potential sociodemographic confounders and disease stage, was associated with worse clinical outcomes. Those persons with a lower rate retention, over 2 years of follow-up after linkage with care, had a significantly lower viral load suppression and less improvement from baseline CD4 count. Furthermore, a lower rate of retention was associated with a three times higher risk of mortality among persons who had visits in one or two of the four 6-month periods over 2 years postlinkage. The persons who failed to remain in care had a four times higher risk of mortality as compared to those who had visits in all four intervals. However, the difference in risk of mortality for those who had visits in three out of the four intervals did not reach significance. A recent study by Giordano et al. 28 found a dose–response relationship between retention in the first year of care and improved clinical outcomes, such as decreased VL, increased CD4 count, and better survival. Other published studies have shown that missed HIV clinical care appointments could be associated with adverse health outcomes. 6,8,28 One plausible explanation of this finding is that a lower rate of retention suggests lower adherence to therapeutic interventions, especially cART, leading to both poor clinical improvement and increased risk of mortality. 9 Some recent studies have indicated that adherence to cART in the initial phase of treatment could be a predictor of long-term virologic response. 29 Retention in the initial phase of treatment provides an early opportunity for intervention to mitigate barriers to treatment adherence such as mental health or substance abuse. 4,25
Limitations of our study include the inability to evaluate individual reasons for lower rates of retention in care because this information is not collected as part of routine HIV surveillance. In addition, cause of death was not captured and so mortality directly attributable to HIV infection could not be ascertained. Because laboratory tests were used as a surrogate for demonstrating retention in care, this may lead to some misclassification. However, this misclassification should be minimal because all SC licensed laboratories participate in the electronic reporting system. Another limitation of our data is that some laboratory results could have originated from nonprimary HIV care settings, such as inpatient care and emergency departments, which could not be filtered out because of a lack of a source for CD4/VL reports in the current data set. We are unable to adjust for physician experience/specialty, insurance status, and the presence of comorbid conditions that may have an impact on retention because such information is also not routinely collected in the eHARS database.
This statewide study demonstrates a high rate of poor retention among newly HIV-diagnosed persons who initially enter care. Our data take into account patients who continually/simultaneously access different HIV medical care providers within the state and therefore fill an important gap in the literature on retention in HIV care. Sociodemographic factors, clinical stage at diagnosis, and site of HIV testing are significant predictors of retention in care. Our results show that an increasing rate of retention in the initial 2 years of HIV care could substantially improve the viro-immunological parameters and survival outcomes.
Footnotes
Acknowledgments
The authors express gratitude to Dr. Lytt I. Gardner for careful review and editorial advice during the preparation of this article. We also thank Terri Stephens for valuable support in the provision of data for this analysis.
Author Disclosure Statement
No competing financial interests exist.