Identifying Patterns of Retention in Care and Viral Suppression using Latent Class Analysis among Women Living with HIV in Florida 2015-2017

Abstract

The study objective was to classify women with newly diagnosed HIV into patterns of retention in care (≥2 HIV care visits ≥3 months apart) and viral suppression over time and identify factors associated with class membership. Florida HIV/AIDS surveillance data were used to conduct Latent Class Analysis to classify women into patterns, and multinomial regression was used to compare prevalence of class membership by demographic and clinical factors. Four classes were selected based on model fit parameters: (Class 1) consistently retained and suppressed (>90% probability of being retained and suppressed), (Class 2) not consistently retained or suppressed (≤10% probability of being retained and suppressed), (Class 3) increasingly retained and suppressed, and (Class 4) decreasingly retained and suppressed. The proportion of women in each class was 48.6%, 24.9%, 14.3%, and 12.2%, respectively. Women aged 25–34 compared to 35-49 years old, injection drug use mode of exposure, US born, and not linked to care three months post diagnosis had a lower prevalence of belonging to the consistently retained and suppressed class. Findings may be useful in tailoring and targeting interventions to increase the prevalence of women who are consistently retained in care and virally suppressed.

Introduction

Retention in continuous HIV care and achieving viral suppression are critical components of the HIV care continuum (NHAS, 2017). Nationally, 57.2% and 59.8% of persons living with HIV diagnosis (PLWH) in 2015 were retained in care and achieved viral suppression respectively (CDC, 2018). Prior research has shown that the likelihood of achieving viral suppression greatly improves when PLWH are retained in care, (Crawford and Thornton, 2017), and the risk of HIV transmission and disease progression to AIDS decreases when viral load is suppressed (Crepaz et al., 2016; Nwangwu-Ike, Frazier, Crepaz, Tie, & Sutton, 2018). Furthermore, having multiple measures of viral load facilitates better understanding of viral load over time and may identify periods of non-viral suppression (Marks et al., 2015; Mugavero et al., 2011; Terzian et al., 2012).

In 2015, women composed 28.4% of PLWH in Florida (Florida AIDS Vu, 2015). Women with HIV face sociodemographic, behavioral, and clinical barriers to achieve viral suppression (Nwangwu-Ike, et al., 2018). Compared to men, women are less likely to be engaged in care, prescribed antiretroviral therapy (ART), and become virally suppressed (Bradley, Mattson, Beer, Huang, & Shouse, 2016; Muthulingam, Chin, Hsu, Scheer, & Schwarcz, 2013; Wiewel, Borrell, Jones, Maroko, & Torian, 2017). Hispanic women are more likely to live in poverty and encounter language barriers compared to Hispanic men (Luna-Gierke et al., 2018), while Black women face added barriers of poverty, low education, and lack of insurance compared to White women (Beer, Mattson, Bradley, & Skarbinski, 2016). Therefore, the objective of this study was to classify women with newly diagnosed HIV into patterns of retention in HIV care and viral suppression over time and determine factors associated with belonging to these classifications.

Methods

De-identified surveillance data were obtained from Florida Department of Health enhanced HIV/AIDS Reporting System for women ≥13 years old who received a HIV diagnosis in 2014 and were living in Florida from 2015 through 2017. Retention in care was defined as being engaged in care ≥2 times at least 3 months apart within a year, while viral suppression was defined as having a viral load test of <200 copies/ml in the last viral load test of the year (CDC, 2019). Demographic variables included age (13-24, 25-34, 35-49 and 50+ years old), race (Non-Hispanic White, Non-Hispanic Black and Hispanic), and US nativity including territories. Clinical variables included self-reported HIV exposure (history of injection drug use [IDU], non-IDU), laboratory or provider confirmed AIDS diagnosis by the end of 2014 (having CD4 count <200 cells/μL or an AIDS-defining illness), and linkage to care within three months of diagnosis, which was reported through laboratory testing, viral load, or CD4 results. Other race/ethnicity category (n =14), transgender women (n = 22), <13 years old (n = 9), or prison diagnosis (n=11) in 2014 were excluded due to small sample size.

Descriptive analyses were conducted using SAS statistical software (SAS, 2014). We modelled latent classes according to patterns of retention in care and viral suppression for each year from 2015-2017. The best fit model was selected from 2-6 latent classes using the lowest Bayesian Information Criterion (BIC) (Schwarz, 1978) and a higher entropy, suggesting a better distinction between classes (Celeux and Soromenho, 1996). Multivariate multinomial regression models followed by the use of the NLMeans macro were used to estimate prevalence ratios and 95% confidence intervals of class membership by demographic and clinical variables.

Results

Of the 950 women diagnosed with HIV during 2014 (Table 1), 14.2% were Hispanic, 67.5% were Non-Hispanic Black and 18.3% were Non-Hispanic White. 91.8% acquired HIV through non-IDU; 65.4% were US-born, and 33.7% were 35-49 years old. Having an AIDS diagnosis and not being linked to care within three months of diagnosis accounted for 25.4% and 17.7% of women, respectively.

Table 1:

Characteristics of women who received an HIV diagnosis in 2014 and class membership according to retention and viral suppression patterns 2015-2017, Florida

Characteristics	Total (2015-2017) N= 950 n (%)	Class 1: consistently retained and suppressed (>90% probability of both outcomes) N= 481 n (%)	Class 2: not consistently retained and suppressed (≤10% probability of both outcomes) N= 238 n (%)	Class 3: increasingly retained and suppressed N= 119 n (%)	Class 4: decreasingly retained and suppressed N= 112 n (%)	P-Value
Race/Ethnicity						0.2940#
Hispanic	135 (14.2)	73 (54.1)	34 (25.2)	12 (8.9)	16 (11.9)
Non-Hispanic Black	641 (67.5)	325 (50.7)	162 (25.3)	87 (13.6)	67 (10.5)
Non-Hispanic White	174 (18.3)	83 (47.7)	42 (24.1)	20 (11.5)	29 (16.8)
Age Categories (years)						0.0180*#
13-24	139 (14.6)	60 (43.2)	34 (24.5)	25 (18.0)	20 (14.4)
25-34	243 (25.6)	105 (43.2)	75 (30.9)	33 (13.9)	30 (12.4)
35-49	320 (33.7)	177 (55.3)	67 (20.9)	38 (11.9)	38 (11.9)
50+	248 (26.1)	139 (56.1)	62 (25.0)	24 (9.3)	24 (9.7)
HIV Exposure						0.0033*
IDU exposure	78 (8.2)	24 (30.8)	26 (33.3)	15 (19.2)	13 (16.7)
Other exposure	872 (91.8)	457 (52.4)	212 (24.3)	104 (11.9)	99 (11.4)
US-Born						0.0513
Yes	621 (65.4)	294 (47.3)	166 (26.7)	83 (13.4)	78 (12.6)
No	329 (34.6)	187 (56.8)	72 (21.9)	36 (10.9)	34 (10.3)
AIDS Diagnosis by the end of 2014						0.5577
Yes	241 (25.4)	130 (53.9)	56 (23.2)	31 (12.8)	24 (9.7)
No	709 (74.6)	351 (49.5)	182 (25.6)	88 (12.4)	88 (12.4)
Linked to HIV care within 3 months of diagnosis						0.0001*
Yes	782 (82.3)	443 (56.7)	142 (18.2)	99 (12.7)	98 (12.5)
No	168 (17.7)	38 (22.6)	96 (57.1)	20 (11.9)	14 (8.3)

^#= Bonferroni adjusted p-value

Race: [Hispanic vs. NHB= 1.000; Hispanic vs NHW= 1.0000; NHB vs NHW= 0.4625];

Age: [13-24 vs 25-34=1.0000; 13-24 vs 35-49= 0.5532; 13-25 vs 50+=0.1077; 25-34 vs 35-49= 0.1233; 25-34 vs 50+= 0.2333; 35-49 vs 50+= 1.0000]

^*= Significance

A 4-class model was selected based on BIC value (303.12) and entropy (0.89) (data not shown). Class 1 (48.6%) are women who were “consistently retained and suppressed” (>90% probability of both retention in care and viral suppression for each of the three years of follow-up); Class 2 (24.6%) are women who were “not consistently retained and suppressed” (≤10% probability of retention in care and viral suppression for each of the three years of follow-up); Class 3 (14.3%) are women who were “increasingly retained and suppressed” (an increasing probability of being retained/suppressed over the study time); and Class 4 (12.2%) are women who were “decreasingly retained and suppressed” (decreasing probability of being retained/suppressed over the study time). The prevalence of being in Class 1 was lower among 25-34 year-old women compared to 35-49 year olds, HIV exposure of IDU compared to other HIV exposure, US-born compared to foreign-born, and not being linked to care within 3 months of diagnosis compared to being linked to care within 3 months (Table 2). A significantly higher prevalence of being in Class 2 was observed among those not linked to care within 3 months of diagnosis compared to those linked to care within 3 months (Table 2). The prevalence of being in Class 4 was significantly lower among non-Hispanic Blacks than among non-Hispanic Whites and among women not linked to care within 3 months of diagnosis compared to those linked to care within 3 months of diagnosis.

Table 2.

Adjusted prevalence ratios and 95% confidence intervals of characteristics of women diagnosed with HIV during 2014 in Florida with retention in care and viral suppression class membership obtained from multinomial regression output, 2015-2017

Characteristics	Adjusted Prevalence Ratio (PR) and 95% CI Class 1: consistently retained and suppressed (>90% probability of both outcomes)	Adjusted PR and 95% CI Class 2: not consistently retained and suppressed (≤10% probability of both outcomes)	Adjusted PR and 95% CI Class 3: increasingly retained and suppressed	Adjusted PR and 95% CI Class 4: decreasingly retained and suppressed
Race
Hispanic vs. Non-Hispanic White	0.88 (0.55, 1.19)	1.20 (0.80, 1.60)	0.97 (0.28, 1.67)	0.74 (0.26, 1.22)
Non-Hispanic Black vs. Non-Hispanic White	0.91 (0.67, 1.16)	1.03 (0.74, 1.31)	1.49 (0.74, 2.24)	0.68 (0.36, 0.99)*
Age Group
13-24 vs. 35-49	0.79 (0.54, 1.04)	0.98 (0.67, 1.29)	1.45 (0.76, 2.13)	1.12 (0.52, 1.72)
25-34 vs. 35-49	0.76 (0.56, 0.96)*	1.20 (0.92, 1.49)	1.05 (0.59, 1.52)	0.93 (0.49, 1.37)
50+ vs. 35-49	1.00 (0.76, 1.24)	1.15 (0.87, 1.42)	0.74 (0.37, 1.10)	0.83 (0.42, 1.23)
HIV Exposure
IDU exposure vs. Other exposure	0.52 (0.30, 0.75)*	1.27 (0.88, 1.67)	1.77 (0.78, 2.76)	1.06 (0.40, 1.72)
US-Born
Yes vs. No	0.82 (0.65, 0.99)*	1.12 (0.89, 1.35)	1.03 (0.64, 1.43)	1.05 (0.61, 1.48)
AIDS Diagnosis by the end of 2014
Yes vs. No	0.83 (0.63, 1.04)	1.22 (0.96, 1.48)	1.03 (0.61, 1.45)	0.73 (0.39, 1.06)
Linkage to Care within 3 months of diagnosis
No vs. Yes	0.34(0.22, 0.46)*	2.80 (2.23, 3.38)*	0.73 (0.37, 1.09)	0.53 (0.23, 0.84)*

^*= Significance

Discussion

Our findings revealed four patterns of retention in care and viral suppression, with approximately 38% of women with patterns of consistently low or decreasing probabilities of retention and viral suppression. This finding suggests that there may be factors that prevent women from attending HIV care appointments and additional resources are needed to improve care outcomes.

Women aged 25-34 vs 35-49 years, with an HIV transmission mode of IDU vs. non-IDU, who were US-born vs. foreign-born, and not linked to care within 3 months of diagnosis vs. timely linkage to care were less likely to be consistently retained and virally suppressed. These findings are consistent with research demonstrating that younger PLWH have lower rates of retention in care (Ulett et al., 2009) and viral suppression (Beer, et al., 2016; Muthulingam, et al., 2013), while persons who inject drugs have poor engagement in continuous care, either due to lack of medication adherence or difficulties with managing health care appointments (Olatosi, Probst, Stoskopf, Martin, & Duffus, 2009). The findings for US-born women were in contrast to other studies that have reported lower odds of retention in care and viral suppression for foreign-born compared to US-born PLWH (Sheehan et al., 2017). However, another study found US-born Blacks had lower rates of retention in care and viral suppression compared to foreign-born Blacks (Demeke et al., 2018). In a post-hoc analysis, we found that US-born Non-Hispanic Blacks were less likely to be consistently retained and suppressed compared to foreign-born Blacks. Further studies should be conducted to determine factors that challenge retention in care and viral suppression for US-born women. Additionally, lack of timely linkage to care is commonly associated with poor care outcomes especially among minority populations including women (Ulett, et al., 2009) consistent with our findings of lower prevalence of being consistently retained and suppressed for women who were not timely linked to care.

Women in the not consistently retained and suppressed class could either have a variable degree of retention and suppression over time or were not retained or suppressed throughout the observation period. However, results from the latent class analysis did not provide a distinction between these groups. Notably, women who were not linked to care within 3 months of diagnosis had a higher prevalence of being not consistently retained and suppressed. Surprisingly, non-Hispanic Black women had a lower prevalence of being in the decreasingly retained and suppressed class when compared to their non-Hispanic White counterparts. This could be attributed to increased uptake of ART over the years for non-Hispanic Blacks (Bradley, et al., 2016; Crepaz, Dong, Wang, Hernandez, & Hall, 2018). Furthermore, in our data, only 48% of non-Hispanic White women were classified in the consistently retained and suppressed class, underscoring the need for improving care outcomes for White women (data not shown). Results from our LCA showed that women in the decreasingly retained and suppressed group had good retention and viral suppression rates during the first year of diagnosis followed by a decline over time (data not shown). However, women who did not receive timely linkage to care started with poor outcomes at the beginning of this study, and over time were less likely to have declining retention and viral suppression (Table 2).

This study has several limitations. First the analysis was conducted using surveillance data which has limited information on psychosocial factors that are important for assessing HIV care and treatment. In addition, the surveillance data only includes care information for women seeking care within Florida. We do not have data on care that is transferred or received outside Florida during the study period. Second, the small sample size within some of the classes may affect the ability to detect associations of characteristics with those classes.

Conclusions

Overall, we found that women who were younger, US born, had a IDU mode of exposure, or delayed linkage to care were less likely to belong to the class with patterns of consistent retention in care and viral suppression. Our results are generalizable to groups of women with these care patterns rather than individual women within each class. Retention in care requires ongoing interaction with the health care system which could pose obstacles for women who face various challenges. The patterns identified by this analysis demonstrate the need for tailored interventions that target women that show patterns of low rates of retention in care and viral suppression, which will ultimately help improve their overall health and aid in decreasing HIV transmission.