Assessing the Information-Motivation-Behavioral Skills Model to predict pre-exposure prophylaxis adherence among Black men who have sex with men and transgender women in a community setting in New York City

Abstract

The Information-Motivation-Behavioral Skills (IMB) Model has been used to understand adherence to medications and intentions to uptake pre-exposure prophylaxis (PrEP) to prevent HIV infection. In the current study, the IMB Model was used to understand factors that predict adherence to PrEP among a community-based cohort of 204 Black men who have sex with men (MSM) and transgender women (TGW) using structural equation modeling. PrEP motivation was directly associated with PrEP behavioral skills (β = .320, p = .009), and PrEP behavioral skills were directly associated with PrEP adherence (β = .416, p = .001). PrEP knowledge and PrEP motivation were not associated with PrEP adherence, directly or indirectly. The analysis identified intervenable factors that predicted PrEP adherence. Screening for motivation and behavioral skills could be used to identify where additional support to improve PrEP adherence is needed, or whether to offer alternative PrEP modalities or delivery strategies.

Resumen

El Modelo de Información-Motivación-Habilidades Conductuales (IMB) ha sido utilizado para comprender la adherencia a los medicamentos y la intención de tomar la profilaxis pre-exposición (PrEP) para prevenir la infección por el VIH. En el estudio actual, se usó el modelo IMB para comprender los factores que predicen la adherencia a la PrEP entre una cohorte reclutada en la comunidad de 204 hombres que tienen sexo con hombres (HSH) y mujeres transgénero (TGW) de raza negra, usando modelos de ecuaciones estructurales. La motivación de adherir a la PrEP se asoció directamente con las habilidades conductuales de la PrEP adherencia (β = 0,320, p = 0,009), y las habilidades conductuales de la PrEP adherencia se asociaron directamente con la adherencia a la PrEP (β = 0,416, p = 0,001). El conocimiento de PrEP y la motivación de adherir a la PrEPno se asociaron con la adherencia a la PrEP, ni directa o indirectamente. El análisis identificó factores intervenibles que predijeron la adherencia a la PrEP. La evaluación de la motivación de adherir a la PrEP y las habilidades conductuales de la PrEP adherencia podría ser usado para identificar situaciones en que se necesita apoyo adicional para mejorar la adherencia a la PrEP, o si se deben ofrecer modalidades alternativas de recibir PrEP o estrategias alternativas para entregar PrEP.

Introduction

Gay, bisexual and other cisgender men who have sex with men (MSM) comprise the largest group living with HIV in the United States (US), accounting for 75% of all new infections annually (1–3). Stark racial disparities in HIV prevalence and new infections exist among MSM, especially among Black MSM (2–6). Nationally and in New York City (NYC), nearly one in three MSM living with HIV are Black (7, 8). Furthermore, rates of HIV infection are increasing among Black MSM, while decreasing among White MSM (6–9). Black MSM account for one out of every four new HIV diagnoses (8). The Centers for Disease Control and Prevention (CDC) estimates that, based on current infection rates, one in two Black MSM living in the US.will be infected with HIV during their lifetimes (10).

Transgender women who have sex with men (TGW) have also been identified as a population severely impacted by HIV (11–14), with stark racial disparities characterizing the epidemic among TGW in the U.S (12, 15). In NYC, where new infections among TGW have increased since surveillance including this population began in 2001 (16), Black TGW accounted for 41% of new HIV diagnoses among TGW compared to 4% for white TGW; these racial groups represent 24% and 43% of the population of NYC, respectively (16). Despite their increased HIV risk and burden, TGW are not sufficiently reached by current HIV prevention efforts (17–19), and thus are in critical need of increased attention.

Pre-exposure prophylaxis (PrEP) is a critical tool for curbing the HIV epidemic. The effectiveness of daily oral PrEP has been confirmed across numerous studies (20). Studies show that high-coverage PrEP with effective levels of adherence in populations heavily affected by HIV, such as Black MSM and TGW, can rapidly reduce new HIV acquisition rates (21, 22). Thus, equitable expansion of PrEP use with effective levels of adherence and persistence will accelerate progress towards meeting the US Ending the HIV Epidemic goals (23). While PrEP awareness and utilization has been increasing in recent years (21, 24–26), concerns remain about PrEP uptake and adherence among populations at risk (27–29), especially Black MSM and TGW, as PrEP awareness, use and adherence remain low in these groups (19, 30–32).

Among those who initiate PrEP, suboptimal persistence and adherence to PrEP limit its utility (33, 34). Several recent studies have found that the majority of those prescribed PrEP have discontinued use or have been lost to follow-up within one year (35–38). Studies have also found lower rates of PrEP adherence and persistence among African Americans (38, 39). Access to trusted health and HIV prevention services for MSM (40) and TGW (41, 42) has been associated with greater acceptability and PrEP use among these groups. Multiple other barriers also exist, including discrimination and stigma (43), and the limited availability of transgender-competent health services and providers (43–45). Further research is needed on strategies to increase PrEP adherence among Black MSM and TGW (46).

Health behavior theories offer a means for understanding the complexity of PrEP adherence. Screening for behavioral factors that could help predict PrEP adherence would offer patients and providers a tool for assessing whether additional support should be provided to improve PrEP adherence or whether to offer other emerging alternative PrEP modalities, including long-acting PrEP agents. The IMB model is among the most comprehensive and adaptable models available for understanding HIV-related behaviors (47), including ART adherence (48, 49). The IMB model posits that to the extent that individuals have adequate salient information about a health behavior, are motivated to act on their knowledge, and have necessary behavioral skills to perform that health behavior, they will successfully overcome obstacles to perform tit. Some research has explored the IMB model in relation to PrEP outcomes. One study found empirical support for the IMB model in predicting willingness to use PrEP among 400 HIV-negative individuals enrolled in a community-based methadone maintenance treatment who reported recent HIV risk behaviors (50). A review article concluded that there is empirical support for the utility of an adapted IMB model for understanding PrEP uptake (51). A subsequent study found empirical support for the IMB model in predicting PrEP use among a sample of 357 MSM and TGW in the Midwestern US (52). However, the IMB model, to our knowledge, has not been assessed in terms of predicting PrEP adherence. The current study was designed to assess the utility of the IMB Model to predict an effective level of PrEP adherence among Black MSM and TGW in a community setting. We used structural equation modeling (SEM) to to assess whether the IMB model could be used to predict PrEP adherence, as SEM allows for tests of complex, theory-based hypotheses with regards to associations among multi-dimensional constructs, accounting both for measurement error and model fit (53).

Methods

Participants

The data were drawn from an implementation science study that utilized a two-arm randomized controlled trial design to assess the effect of an experimental PrEP adherence support intervention compared to standard of care support on PrEP adherence among Black MSM and TGW in a community setting in New York City over the course of 12 months, which has been fully described previously (54). PrEP-related medical assessment and follow-up services were provided by Harlem United (HU), a community-based organization operating three medical clinics in Central and East Harlem in close proximity to the research site.

Potential study participants were identified through: referrals from HU and other studies at the research site; online advertising, or in community settings. MSM and TGW were considered for study participation if they reported being African American or Black, assigned male at birth, and 18 years of age or older; residing inNew York City with no plan to relocate during the study period; having condomless anal or neovaginal intercourse with a male or TGW sexual partner of unknown or positive HIV status in the past six months; contracting any STI in the past 6 months; or engaging in an on-going sexual relationship with an HIV-positive male or TGW partner. Exclusion criteria were not being eligible to take PrEP for medical reasons or currently participating in another PrEP adherence study. Eligible individuals were referred to HU for clinical assessment with collection of biological samples followed by determination of eligibility to start PrEP. Participants were enrolled if they completed the evaluation for PrEP eligibility and were started on PrEP.

Procedures

Individuals who provided written informed consent were randomized to receive enhanced or standard support services, and were provided with a 30-day supply of emtricitabine/tenofovir (FTC/TDF) for PrEP. SAS version 9.4 was used to generate block randomization using blocks of 10; study and HU staff were not aware of block size. Consistent with PrEP prescribing guidelines, participants returned to their clinicians for follow-up medical assessment after one month and were provided with a two-month supply of FTC/TDF. At 3, 6, 9, and 12 months following enrollment, participants returned for follow-up medical assessments and received a three-month supply of FTC/TDF at each timepoint, as well as condoms, lubricant, and quarterly testing for HIV and STIs.

Participants were administered face-to-face interviews by a trained research assistant at baseline, 3, 6, 9, and 12 months (the same schedule as their medical assessments). Interview questions solicited information on demographic characteristics, HIV behavioral risk behaviors, PrEP knowledge and attitudes, substance use, and included a mental health assessment. Participants were reimbursed for their time upon completion of all research interviews and provided with transportation assistance. The study was reviewed and approved by the Columbia University Irving Medical Center Institutional Review Board (AAAO0852). All study participants provided written informed consent. The study was registered at ClinicalTrials.gov (NCT02167386).

Measurement

Supplementary Table I details all of the measures used in the analyses. The primary outcome, self-reported PrEP adherence, was measured at all follow-up visits using the 3-item Wilson adherence scale (55). Participants were asked the number of days in the last 30 days in which they missed taking PrEP, how well they did in the last 3 months taking PrEP the way they were supposed to (1–6 scale of “Very Poor” to “Excellent”), and how often they took their PrEP in the way they were supposed to in the last 3 months (1–6 scale of “Never” to “Always”). These three items were linearly transformed to create a scale of 0–100, and good adherence was defined as ≥ 57 out of 100. This cut-point was used as a proxy for taking PrEP for an average of 4 out of 7 days (57%), which has been identified in prior research as a benchmark for a minimally effective dose (56, 57). Participants lost to follow-up were assumed to not be taking PrEP (0 out of 100 on the Wilson scale). It was not possible to determine if people were active in PrEP care elsewhere.

The study assessed several aspects of the IMB model determinants, all of which were based on previously published instruments (58), and tailored for PrEP by the study team using the following multi-step approach: 1) conceptualizing the applicability of IMB constructs to PrEP based on familiarity with that model; 2) modifying an existing instrument used previously by the authors which measures knowledge, attitudes and beliefs about tuberculosis prevention (59). At baseline, participants were asked 12 true/false items on PrEP knowledge. Items which were relevant to PrEP adherence included: “PrEP is a pill you can take to prevent HIV”, “PrEP can be used for HIV negative people whose partners have HIV”, and “Taking PrEP can be very effective for people who are high risk for HIV.” A PrEP knowledge index was created by tallying correct answers to each of these items (range = 0–3). Participants were also asked twenty-three items on PrEP motivation. PrEP motivation was considered multi-dimensional (51, 58), and domains included: HIV risk perception, anticipated stigma, attitudes towards PrEP, attitudes towards healthcare, attitudes towards condoms, personal intentions, safer sex intentions, sex expectancies, and perceived social norms. Response options ranged from 1 = ‘Strongly disagree’ to 4 = ‘Strongly agree.’ Items were phrased as expressions of low motivation (e.g., “It’s too difficult to take PrEP regularly,” “You don’t want people to know that you’re taking PrEP”), and thus were reverse coded so that a higher score represented higher motivation. Lastly, at three months, after participants had started taking PrEP, they were asked eleven items on behavioral skills (i.e., self-efficacy). Domains asked about included managing potential side effects; ability to take PrEP in privacy; and obtaining, refilling and remembering to take PrEP. Response options ranged from 1 = ‘Very difficult to do’ to 5 = ‘Very easy to do.’

Other covariates were also measured, including sociodemographic characteristics, substance use (60, 61), mistrust in healthcare providers (62), and self-reported symptoms of depression, using the Center for Epidemiologic Studies Depression (CES-D) scale (63).

Statistical Analyses

SEM analyses were conducted using Mplus (version 8) (64), using a common 2-step procedure (65). Confirmatory factor analysis was conducted to assess the fit of the measurement model comprising all latent variables. Path analysis was then conducted with model specification based on theory and informed by data-driven modification indices. A weighted least squares multivariate estimator was used with a diagonal weight matrix, with theta parameterization. In the model, exogenous variables (i.e. Information and Motivation) were allowed to covary. Depression symptoms, which were previously found to be associated with PrEP adherence in this sample (66), was the only covariate that met criteria for confounding (p < .05) and therefore was included in the model (67). Indirect effects were calculated as part of the standard statistical methods used by Mplus. To assess model goodness of fit, the comparative fit index (CFI), the Tucker-Lewis index (TLI) and the root mean squared error of approximation (RMSEA) were examined. CFI and TLI values above 0.9 and RMSEA values below 0.05 represent a good model fit (68, 69). Missing data were handled using a Full Information Maximum Likelihood (FIML) approach in order to maintain the full sample size. We also conducted sensitivity analyses where participants lost to follow-up were excluded completely.

Results

Participant socio-demographic characteristics

Of the 204 participants who participated in the baseline interview, 131 (64%) completed the 3-month interview in which factors hypothesized to influence PrEP use were assessed, and 132 completed the 12-month interview, when the final assessment of PrEP adherence was performed. Nearly all participants, 194 (95.1%) identified as male and 10 (4.9%) identified as TGW. According to study criteria, all participants reported their race as Black (using a variety of terms), and 42 (20.6%) reported their ethnicity as Latinx. The median age was 31 years (IQR = 20 years) and 91 participants (44.6%) were between 18 and 29 years of age. Thirty-six (17.6%) did not complete high school, 90 (44.1%) held a high school diploma or equivalent, 51 (25.0%) had some college, and 27 (13.2%) had graduated from college and/or taken further coursework. At the time of enrollment, 90 participants (44.1%) reported income of less than $10,000 per year; the median income was $12,000 per year. One hundred twenty participants (58.8%) lived in marginal housing. Twenty-five participants (12.3%) reported being in a relationship. Ninety-two participants (45.1%) met the CES-D threshold for having symptoms of depression. The sample has been described further previously (66).

IMB Constructs and PrEP Adherence

The mean PrEP knowledge score was 2.70 (SD=0.55, Range: 1–3). The mean PrEP motivation score was 2.15 (SD=0.37, Range: 1.13–3.13). The mean PrEP self-efficacy score was 2.12 (SD=1.94, Range: 0.10–4.80).

In the trial, loss to follow-up was considerable (35% at 12-months) and PrEP adherence suboptimal (32.4% at 12-months) with no intervention effect observed (33.8% in the intervention arm vs. 30.3% in the control arm, p=0.69) (54). Nearly a third (32.4%) of participants were adherent to PrEP at 12 months. Non-adherent participants comprised roughly equally-sized groups of participants who either reported little or no use of PrEP or were lost to follow-up at 12-months (35.3% of participants).

Measurement Model

The measurement model comprised three latent factors: PrEP motivation, PrEP behavioral skills, and symptoms of depression. A single symptom of depression (“I felt sad”) did not load onto the latent factor of depression and was therefore removed. The subsequent measurement model converged normally with acceptable fit: RMSEA = .053 (90% CI: .046 − .060, probability of RMSEA < .05 = .222); CFI = .97; TLI = .97. Standardized factor loadings (λ) were all significant (P < .001).

Structural Equation Modeling: IMB model and PrEP adherence

Our structural equation model converged normally with acceptable fit: RMSEA = .050 (90% CI: .043 − .057, probability of RMSEA < .05 = .471); CFI = .97; TLI = .97. However, modification indices suggested additional correlations between motivation items and depression items, all consistent with theory. With these additions, a final model (Figure 1) converged normally with improved fit: RMSEA = .037 (95% CI: .028 − .045, probability of RMSEA < .05 = .998); CFI = .98; TLI = .98.

Figure 1.

Structural equation model of direct and indirect pathways. Significant effects are bolded. Dashed lines indicate adjustment by a covariate.

Table I details the direct and indirect effects of pathways between IMB constructs and PrEP adherence. To summarize: (1) In terms of total effects, neither PrEP knowledge nor PrEP motivation were associated with PrEP adherence; (2) PrEP behavioral skills were directly associated with PrEP adherence (β = .416, p = .001); (3) PrEP motivation was directly associated with PrEP behavioral skills (β = .320, p = .009); (4) The indirect effect of PrEP motivation on PrEP adherence via PrEP behavioral skills was β = .133, although this did not achieve statistical significance (p = .072); (5) The model accounted for 19% of the variance in PrEP adherence (p = .033). Table II includes a correlation matrix for the IMB measures and the covariate.

Table I.

Standardized β from structural equation modeling to test direct, indirect and total effects of IMB Model for PrEP adherence among Black MSM and TGW (N = 204)

Dependent variables		Explanatory variables			Covariate
		Information	Motivation	Behavioral Skills	Depression Symptoms
Direct Effects	Information	—			.260**
	Motivation	—	—		.383***
	Behavioral Skills	.195	.320**	—	.138
	Adherence	.113	−.051	.416**	−.016
Indirect Effects	Information	—			—
	Motivation	—	—		—
	Behavioral Skills	—	—	—	.173**
	Adherence	.081	.133	—	.139*
Total Effects	Information	—			.260**
	Motivation	—	—		.383***
	Behavioral Skills	.195	.320**	—	.311***
	Adherence	.195	.082	.416**	.124

R² in Adherence = 19% (p = .033); RMSEA = .037 (95% CI: .028 − .045, probability of RMSEA < .05 = .998); CFI = .98; TLI = .98

^*p < .05,

^**p < .01,

^***p < .001

Table II.

Correlation Matrix of Latent Variables and Covariates for IMB Model for PrEP adherence among Black MSM and TGW (N = 204)

	Explanatory variables			Outcome	Covariate
	Information	Motivation	Behavioral Skills	Adherence	Depression Symptoms
Information	—
Motivation	.280*	—
Behavioral Skills	.321*	.428**	—
Adherence	.228	.153	.426**	—
Depression Symptoms	.260*	.383**	.311**	.124	—

^*p < .05,

^**p < .01,

^***p < .001

Sensitivity Analyses

Supplementary Table II and Supplementary Figure 1 detail the results of sensitivity analyses where participants lost to follow-up (n=72) were excluded completely. Results and model fit were very similar to those of the original model, and R² for PrEP adherence increased from 19% to 35%.

Discussion

We found that intervenable factors from a well-studied behavioral model -- the IMB model -- for HIV-related behaviors, including ART adherence, predicted PrEP adherence among Black MSM and TGW receiving PrEP in a community health clinic. Those who endorsed having higher levels of behavioral skills were more likely to report being adherent to PrEP. Those who endorsed having higher levels of motivation were more likely to have higher levels of behavioral skills. There was also a marginally significant mediating effect of motivation on PrEP adherence via behavioral skills. Collectively, our results suggest that assessing motivation and behavioral skills could be of value to inform whether additional support for PrEP adherence is needed, or whether to offer alternative PrEP modalities or delivery strategies, as they become increasingly available. The findings from this study can also inform the development of further interventions that target motivation and behavioral skills as a means to improve PrEP adherence.

Previous research provides support for our findings, albeit in relationship to IMB Model and adherence to HIV treatment (58, 70, 71). However, it should be noted that the IMB Model did not perform exactly as hypothesized. PrEP knowledge was not associated with any IMB constructs nor with PrEP adherence. Results from this research are similar to other findings (71, 72), which suggest that assessing knowledge may not be useful in differentiating between those who will be adherent to a medication and those who will not. Simply providing information to Black MSM and TGW about PrEP may not be sufficient to improve PrEP adherence, while attaining skills with regard to adherence behavior may be more important than being knowledgeable about PrEP. One explanation for the lack of association between information and the other constructs in the IMB model and adherence is that the knowledge items measured by our instrument may not have included the kinds of knowledge that would be most relevant to PrEP adherence in this population. For example, many of the items to assess PrEP knowledge were general, and not about the relationship of PrEP adherence to its effectiveness in preventing HIV infection.

It is also noteworthy that PrEP knowledge among the participants was very high. This is likely due to the fact that during screening for PrEP, participants received information on PrEP. In this context, assessment of motivation and self-efficacy, rather than knowledge, could be used to help identify those who might need additional support for future PrEP adherence.

Strategies to increase PrEP adherence to enable achieving its potential for HIV prevention are urgently needed in order to meet the US Ending the HIV Epidemic goals (23). As PrEP awareness and utilization are increasing (24–26, 73), information about how to provide support for PrEP adherence is increasingly important (27–29). Achieving such levels of adherence to PrEP has already been identified as a challenge (74–76), thus limiting its potential impact (33, 34). As the first study to explore the utility of the IMB model to predict PrEP adherence, these findings help address a critical gap in our understanding of how to optimally deliver and support PrEP among high-risk populations. This study also builds off of the findings about the utility of the IMB model for understanding PrEP uptake (50–52).

There are several limitations to this study. First, the use of self-reported adherence as the primary outcome is a subjective measure of adherence rather than biological tests, such as measurement of tenofovir-diphosphate levels (TFV-DP) in the blood. However, other research has shown that self-reported PrEP adherence could be used to distinguish between participants with and without protective levels of PrEP measured using TFV-DP levels in the blood (77). Other researchers have pointed out the need for more nuanced methods to assess PrEP adherence that account for variable risk for HIV infection and use of other HIV prevention methods (78). Second, we were not able to determine if participants who were lost to follow-up were active in PrEP care elsewhere; thus we assumed them to not be taking PrEP. In this sense, our outcome, PrEP adherence, is potentially confounded with retention in the study and PrEP persistence. Therefore, we ran sensitivity analyses excluding participants who were lost to follow-up. In those analyses, the results were similar to those of the original model, except that model fit worsened and knowledge was significantly associated with adherence. Relatedly, self-efficacy was assessed at 3-months after considerable loss to follow-up, and thus we were not able to include the entire sample in these analyses, potentially limiting the generalizability of these findings. Along these lines, this sample was not intended to be representative of Black MSM and TGW in Harlem or other areas, thus also potentially limiting generalizability of the findings. Third, we controlled for symptoms of depression because they met the criteria as a potential confounder, but further analyses could additionally take into account important contextual factors that potentially moderate IMB constructs (79). Lastly, we used the same recruitment strategy for both MSM and TGW and only 4.9% of study participants identified as TGW. Thus, the study was under-powered to be able to assess possible differences between MSM and TGW, even though we acknowledge the unique HIV prevention needs of TGW (42, 80).

In summary, the results of this analysis suggest that motivation and behavioral skills to adhere to PrEP may be useful in targeting individuals who need PrEP adherence support. Future research on how to effectively assess these constructs in community settings and to develop interventions that can effectively improve these constructs is needed in order to achieve the promises of this essential HIV prevention tool.