Significance

Black single mothers have the highest risk of early termination from home visiting services. Outpatient mental health studies have linked strong alliance with improved retention and the preference specifically of Black clients for clinician racial matching. This study aims to evaluate associations with alliance in home visiting services and whether racial bias or racial matching relate to outcomes and retention of clients.

Introduction

The Infant Mental Health-Home Visiting (IMH-HV) model is an intensive, needs-driven, relationship-focused home visiting intervention serving parents and their infants or toddlers (Lawler et al., 2017; Mckelvey et al., 2015; Weatherston & Tableman, 2015). Based on well-established clinical and developmental theories and manualized, IMH-HV serves families with children at risk for a variety of emotional, behavioral, and developmental concerns. Core components include infant-parent psychotherapy, developmental guidance, emotional support, case management, life course planning, and reflective supervision (Shea et al., 2020).

Therapeutic Alliance

Therapeutic alliance (TA) is defined as the bond and collaboration between a practitioner and their client and has been linked to therapeutic outcomes (Martin et al., 2000; McGuire-Snieckus et al., 2007). TA has been associated with client perceptions of therapists as trustworthy, experienced, supportive, and affirming (Ackerman & Hilsenroth, 2003), and predicts a range of outcomes, yet limited data are available regarding factors associated with the strength of the alliance (Martin et al., 2000).

Therapeutic Alliance, Race, and Home Visiting Services

Home visiting is a well-supported approach to intervention for at-risk families; many effective programs emphasize delivering emotional support, enhancing positive parenting, and promoting child development (Roggman et al., 2008). A number of factors predict home visiting retention; for example, single parenthood and frequent changes in residence are associated with greater risk for drop-out (Roggman et al., 2008), while strong TA between clinician and client appears to promote retention (McCurdy et al., 2003). Prior studies have indicated that Black clients prefer to be matched with a racially similar clinician (Cabral & Smith, 2011; Horst et al., 2012), and that racial matching predicted program retention among Black participants, which is particularly relevant as Black clients are disproportionately represented in home visiting programs (McCurdy et al., 2003). Systemic racism and racialized trauma unique to the Black community likely explains the preference for racial matching among Black clients (Helms et al., 2010). From the clinician’s perspective, clinicians whose race matched their client’s perceived greater understanding of client problems and better treatment outcomes compared to clinicians who treated clients of different racial backgrounds (Murphy et al., 2004).

Black clients are at greatest risk for premature termination of home visiting (Raikes et al., 2006), perpetuating racial disparities in care. Inadequate mental health care received by racially and ethnically diverse patients can be attributed to, for example, lack of shared understanding, improper treatment due to a higher risk of misdiagnosis, and stereotyping (Williams, 2002). One recent study found that Black men were recommended for and received more testing when their physician was Black, which may be explained by increased communication and reporting of symptoms (Alsan et al., 2018). Despite this handful of studies suggesting that race and racial matching may play an important role in program retention, TA, and outcomes, little is known about this dynamic within home visiting. Our purpose is to examine how alliance among clinicians and clients may differ depending on racial match/mismatch and other predictors of alliance. We hypothesize TA will be lower between white clinicians and Black clients, compared to racially matched dyads. Additionally, we predicted that clients with lower TA would have lower retention in the IMH-HV program.

Method

The current study examined the effects of a community-delivered, Medicaid-funded, infant mental health home visiting model (IMH-HV). The study was approved by the University of Michigan Review Board (ID no. HUM00096040).

Participants

Participants in the study included mothers, infants, and the clinicians delivering the service. Clinicians from 12 partnering community mental health agencies providing IMH-HV recruited clients from their caseloads. Eligible clients were pregnant women or parents of children ages 0–24 months (M = 9.8, SD = 8.4) who had initiated IMH-HV within three months of recruitment (modal period since initiation = 4 weeks). Of 123 clients approached by clinicians, 116 were eligible. Of these, 91 clients and their 80 children (11 children had two parents or caregivers enrolled in the study) were enrolled. Clients included biological or foster mothers and fathers, and all children enrolled were Medicaid recipients. Clients were incentivized for their participation in data collection and could receive up to $280 USD over the course of the study. All participants were volunteers, and all clients and clinicians provided written informed consent. One mother voluntarily withdrew from the study, leaving a sample size of 90. Given the small number of fathers enrolled (n = 12), and considering that all but one of the children of enrolled fathers also had a mother in the study, we included data only from mothers (n = 78). Data from two mothers was removed due to cognitive impairments preventing their understanding of study measures, and two left treatment prior to the 3-month assessment. Seventeen clients changed clinicians, for instance, due to a clinician’s maternity leave or leaving the agency; we omitted data from cases where a new clinician had 4 or fewer meetings with the client prior to the TA assessments. At the 3-month assessment, this was true in 3 of the 74 cases, leaving 71 cases for analysis.

All IMH-HV clinicians (N = 50) attended a brief training on data collection and study procedures; no incentives were provided for data collection from clinicians.

Procedure

Caregivers completed assessments at five time points: baseline (corresponding to study entry), then again at 3-, 6-, 9-, and 12-months after baseline. Assessments included self-report questionnaires measuring maternal mental health, child social-emotional wellbeing, and parenting. Clinicians reported at baseline on their experience in IMH-HV, their education, caseload, and frequency of reflective supervision.

Scale to Assess Therapeutic Relationships (STAR)

STAR-P/C is a 12-item questionnaire measuring perceptions of TA between clinician and clients, including positive collaboration, positive clinician input, emotional difficulties for the client, and non-supportive clinician input (McGuire-Snieckus et al., 2007). STAR-P was completed by clients and STAR-C by clinicians, starting at the 3-month time point to allow time to develop their relationship. Items use a 5-point Likert scale (0 = “Never”; 4 = “Always") from which a total score (max = 48) and three subscales are calculated. For STAR-P, subscales include positive clinician input (e.g. ‘My home visitor speaks with me about my personal goals and thoughts about treatment’), positive collaboration (e.g. ‘My home visitor and I share a trusting relationship.’), and non-supportive clinician input (e.g., ‘My home visitor is impatient with me’). For STAR-C, the subscales are positive collaboration (e.g. ‘My client and I share a good rapport.’), positive clinician input (e.g. ‘I am able to take the client’s perspective when working with him/her’) and emotional difficulties (e.g. ‘It is difficult for me to empathize with or relate to the client’s problems.’). For each scale, the sum of the 3 subscales was used, with the negative subscales reverse-coded.

One STAR-P item reduced the internal reliability of the measure at each time point: “My home visitor is stern with me when I speak about things that are important to me and my situation.” We removed the item from the analysis, leaving a total of 11 items. The reliability calculated with omega total for these 11 items was 0.93 (McNeish, 2018). No items on the clinician-rated STAR-C were dropped and the omega total reliability was 0.91.

Demographic Factors

Client self-report questionnaires provided information on demographics and mental health, including marital status, number of children in the home, household income, education, and race/ethnicity. Cumulative measures may be better predictors of multiple outcomes, including retention, than individual factors (Appleyard et al., 2005; Begle et al., 2010; Sameroff et al., 1987), and thus a score was created to index individuals’ cumulative experience of societal factors associated with structural inequities related to economic oppression, racism, and lack of access to quality resources. This demographic factors variable was a sum of the endorsement/presence of (1) education (less than high school), (2) race indicated as African American or American Indian/Alaskan Native and/or Latino/Hispanic ethnicity, (3) marital status (unmarried), (4) low income (< $5000 annually), (5) maternal age (less than 21 when child was born), and (6) household crowding (4 or more children under age 6 in the home). This demographic composite ranged from 0 to 6. In other analysis on the same sample, the demographic factors composite was shown to be related to retention in home visiting (Author paper under review).

Psychological Risk

A cumulative “psychological risk” score included (1) elevated parenting stress (top 20% of sample on the Parenting Stress Index-Short Form [PSI-SF]), (2) clinical range of depression symptoms (10 + on the Patient Health Questionnaire [PHQ-9]), (3) high PTSD symptoms (33 + on PTSD Checklist for DSM-5 [PCL-5]), (4) four or more Adverse Childhood Experiences (ACEs), and (5) clinician rating of mental health as “Poor” or “Very Poor”. Cumulative psychological risk scores ranged from 0–5. Fifty-three percent of the sample had 2 + risk factors, and 22% had 3 + risk factors (for details see AUTHOR REFERENCE).

Clinician Self-Report

Clinicians reported on their emotional relationship to their work with responses to two statements, “I feel burnt out at my job” and “I find my job satisfying” (responses were 5-point Likert scale from “Strongly disagree” to “Strongly agree”). They also reported the amount of time they practiced as IMH clinicians.

Data Analysis

Because recent guidelines advise against describing research results using a “bright line” cut off of p = 0.05 and the words “significant” and “non-significant”, we report all p-values as continuous (e.g. p = 0.017 versus p < 0.05) and, if effect sizes with standard strengths are available, we describe the strength of relationships as “weak” or “strong” (Amrhein et al., 2019; Wasserstein et al., 2019).

Client STAR ratings demonstrated a ceiling effect, in that many of the clients rated their clinician with the highest possible rating. To account for this, we conducted STAR-P analyses with Tobit regression.

For some analyses, we created binary versions of the summed ratings. For each, we used the sum equivalent to answering “often” for every positively-coded item, which was 36 for the clinician-rated STAR-C and 33 for the client-rated STAR-P.

Seventeen clinicians had multiple clients involved in the study. To account for the non-independence of residuals for these cases, analyses were done in Mplus, Version 7.4, (Muthén & Muthén, 2011) using TYPE = COMPLEX, which uses robust maximum likelihood estimation and takes into account the clustering and computes standard errors using a sandwich estimator.

Survival analysis was used to determine the effect of therapeutic alliance on retention in treatment. This analysis was done using PROC SURVEYPHREG in SAS, Version 9.4, with data clustered by clinician and the binary values of therapeutic alliance score as the predictor.

Results

Descriptive Statistics

There were 50 clinicians providing IMH-HV to the 71 clients in the study; 12 clinicians indicated their race as Black and served 19 clients in the study. The 38 clinicians who indicated their race as white served 59 clients. Of the 19 clinician-client pairs where the clinician identified as Black, 12 (63.2%) were matched on race. For the 59 clinician-client pairs where the clinician identified as white, 36 (61.0%) were matched on race.

Table 1 shows characteristics of the therapists. All had a master’s degree in social work (63%) or a related field (37%); however, experience varied considerably. All therapists received some form of reflective supervision; most found their job satisfying and only 8% endorsed feeling burned out.

Table 1 Clinician experience and characteristics (N = 50)
Full size table

The client sample represented demographic factors related to systemic oppression, with 27% of the sample reporting less than $5000/year of income, 27% reporting less than high school education, 78% were not married and 43% reported race as Black.

Clients and clinicians rated TA quite differently (Fig. 1). Most clients rated their clinician very highly at each time point (Fig. 1a). At the 3-month rating, 43% of clients rated their clinician with the highest possible rating of 44; the median response was 43. Only 10% of clients rated their clinician below the score of 33 (corresponding to responding “Often” for every positively-coded item). Similar percentages of clients gave the highest rating possible at 6 months (45%), 9 months (42%) and 12 months (46%). In contrast, the distribution of clinician ratings more closely approached a normal distribution (Fig. 1b). No clinician rated TA at the maximum value of 48. For clinician ratings of TA, the mean score was 38.6 (SD 5.2) and median was 38.

Fig. 1
figure 1

Summary score of a participant ratings (STAR-P) and b clinician ratings (STAR-C) of therapeutic alliance. Note maximum value for clinician ratings was 48 and for client ratings was 44, due to one item being removed

Full size image

Do clients and Clinicians Agree on Therapeutic Alliance?

The relationship between clinician and client rating of TA was examined using Tobit regression, accounting for the ceiling effect of the client ratings, using an upper bound limit of 44, and accounting for the clustering of clients within clinicians. For ratings from 3 months (Tobit regression estimate = 0.13 (SE = 0.17), p = 0.44); at 6 months (Tobit regression estimate = 0.34 SE = 0.29), p = 0.24), 9 months (Tobit regression estimate = 0.34 (SE = 0.32), p = 0.29). At 12 months, clinician ratings were much more positively related to client ratings (Tobit regression estimate = 0.88 (SE = 0.23), p = 0.00).

Predicting Retention with Client and Clinician Ratings

Survival analysis, controlling for cumulative demographic risk, showed that 3-month TA ratings greater than 36 predicted clients remaining in treatment longer. For client alliance ratings, the hazard ratio estimate for the predictor high TA was 0.189 (p = 0.0051), and for clinician alliance ratings the hazard ratio was 0.48, (p = 0.033), each indicating longer time before leaving treatment. Figure 2 shows this effect for client ratings, classified into “High” (> = 33) and “Low” TA at 3 months. Notably, clients who rated alliance as lower at 3 months (9% of the sample) were likely to drop out of treatment very early, such that 62.5% left treatment after 3 or fewer visits (for the total sample the average number of sessions was 30.8 (SD = 17.5)). In contrast, those who rated their alliance with their clinician as high remained in treatment longer, with over 50% remaining in treatment for more than 33 visits.

Fig. 2
figure 2

Survival to time of ending IMH treatment, stratified by high and low ratings of client-rated therapeutic alliance

Full size image

Predictors of Early Therapeutic Alliance

We examined factors that could contribute to TA rated by clients and clinicians early in treatment. Table 2 shows means of TA ratings in relation to characteristics of clients and clinicians. Using Tobit regression as above, we tested marital status (Tobit estimate = 4.9 (2.6), p = 0.06), client race (Tobit estimate = 1.1 (2.1), p = 0.61), low income (Tobit estimate = 2.0 (3.0), p = 0.51), low education (Tobit estimate = 2.1 (2.8), p = 0.44), and psychological risk (Tobit estimate = 0.89 (97), p = 0.36). Of these, marital status was the strongest predictor of client rating, with married clients being more likely to give higher ratings than unmarried clients. In addition, we tested whether client ratings were related to clinician race (Tobit estimate = 1.2 (2.9), p = 0.68).

Table 2 Ratings of therapeutic alliance by clients and clinicians, as related to client and clinician characteristics
Full size table

Clinician characteristics were tested as predictors of 3-month clinician ratings, accounting for multiple clients per clinician (using TYPE = COMPLEX in Mplus). Frequency of supervision (β = − 0.13, p = 0.48), and higher caseload (β = − 0.15, p = 0.30) were weak predictors of differences in ratings, as were clinician self-ratings of burn-out (β = − 0.16 (0.11), p = 0.17) and job satisfaction (β = 0.17 (0.13), p = 0.19). Having 2 or more years of experience as an IMH clinician had a somewhat larger effect (β = − 0.28 (0.13), p = 0.028). We found that Black clinicians had higher levels of reflective supervision (85% of Black clinicians had weekly supervision, whereas only 57% of white clinicians had weekly supervision (logistic regression, odds ratio = 4.3, p = 0.031). After controlling for clinician experience and frequency of reflective supervision, Black clinicians overall had higher alliance ratings than white clinicians (b = 3.1 (1.6), p = 0.049, Fig. 3). We tested matching on race (b = 1.8 (1.2), p = 0.14) and client marital status (b = 0.21 (1.4), p = 0.88) as predictors of clinician ratings.

Fig. 3
figure 3

Clinician-rated therapeutic alliance by race of clinician and client (means + SEM, beta = .40, *p = .045) and (M = 37.3, SD = 3.2, **p = .049)

Full size image

When clients and clinicians were examined separately by race (Fig. 3), Black clinicians (M = 41.3, SD = 4.8) had higher ratings for therapeutic alliance than white clinicians (M = 37.3, SD = 3.2, p = 0.049) regardless of the race of the client. Figure 3 shows results comparing client racial identity across clinician racial identity, while controlling for therapist experience, supervision, and clustering of clients within clinicians. White clinicians rated their alliances lower than Black clinicians rated their alliances (β = 0.40, p = 0.045).

Discussion

We examined TA in the IMH-HV model from client and clinician perspectives as predictors of retention and examined the role of race. The self-report measures of alliance, STAR-C and STAR-P (McGuire-Snieckus et al., 2007), are widely used (Chang et al., 2019; McGuire-Snieckus et al., 2007; Priebe et al., 2011) but with one recent exception have not been studied within home visiting (Mundorf et al., 2017). The STAR-C/P demonstrated good reliability and predictive validity within the current study, with one key caveat related to the item using the term “stern”; future application of the STAR-C with similar populations may benefit from alternative phrasing and/or elimination of this item.

Consistent with prior research, clients in general rated TA with their clinicians to be strong, and both clinician- and client-reported TA at 3 months into care were predictors of retention (Murphy et al., 2004), pointing to the critical need to monitor and address alliance early in home visiting. Similarly, prior studies have indicated that demographic factors (e.g., non-Hispanic Black, unmarried status, and frequent residential changes) were associated with premature termination (Roggman et al., 2008; Wintersteen et al., 2005), and our results suggest that unmarried clients had weaker TA.

A key interest was whether client and clinician race contribute to variations in TA. We found that white clinicians rated TA as weaker if clients were Black compared to their white clients. On the other hand, Black clinicians rated their TA stronger (relative to white clinicians) for both white and Black clients. These findings can speak to a higher potential seen by the Black clinicians in their client families and suggest they may be better able to align with more marginalized clients. Recent work suggests non-Hispanic white adults did not prioritize having a provider that shares their culture, but were likely to see a clinician who shares their background because of limited workforce diversity (Terlizzi et al. 2019). On the contrary, a majority of non-Hispanic Black and Latino adults endorsed a preference for a healthcare provider who understood or shared their culture, and a much lower percentage of families in this population are able to see a provider who shares their culture (Terlizzi et al., 2019).

The relevance of these findings is underscored by research indicating that Black mothers have higher rates and greater severity and chronicity of depression and other mental illness relative to other racial ethnic groups (Siefert et al., 2007), and thus stand to benefit the most from home visiting programs when retention is achieved. Similarly, in general, children in home visiting programs have decreased child maltreatment, reduced doctor visits for accidental injuries, increased immunization and nutrition, fewer behavioral problems, and more secure attachment relationships (Kendrick et al., 2000; Lyons-Ruth & Melnick, 2004; Peacock et al., 2013; Roggman et al., 2008). Our data on disparities in TA, coupled with prior findings that young, single, low-income Black mothers have the highest risk for premature dropout, suggests one pathway for understanding racially-associated health inequities.

Future Directions

The study’s strengths include utilizing a diverse sample of low-income families to examine questions related to TA and client and clinician race in a home visiting program. We discovered race-related differences within the clinician/client relationship which, in turn, related to client retention. As efforts continue to improve the ability to engage and intervene with at-risk, diverse populations, attending to effects of TA is critical. Limitations of this study include our small sample size limiting our power to detect effects. Our measurement of TA relied on self-report, which may limit the generalizability of our findings. Eligible clients were invited by their clinicians to participate in the study, a process that may have introduced selection bias. In addition, it is possible that the association between TA and program retention was confounded by client attitudes and interest in home visiting prior to starting the study, which we did not measure. Home visiting programs are widely implemented as a strategy for supporting families with young children. However, unlike many other national home visiting models, IMH-HV has a core component of infant-parent psychotherapy and home visitors are licensed mental health professionals with specialized training in infant mental health, which may impact capacity to build TA with clients and retain families. These differences may constrain generalizability of these findings to paraprofessionally-delivered programs. Using a cumulative measure of demographics does not address intersectionality. Our sample lacked ethnic diversity with only three clinicians and three clients endorsing Hispanic ethnicity. Future work may utilize more in-depth qualitative measures to gather nuanced information on the dynamics of TA across populations, specifically racially marginalized groups.

Implications

This study has three main implications. First, results confirm there is great value in assessing TA to detect clients at risk for premature termination. Second, this study and prior work convey the need for a more diverse workforce and training to address potential bias among providers. Non-Hispanic white clinicians in particular may benefit from bias and cultural sensitivity training to mitigate the effects of bias and unrecognized privilege (Greene & Blitz, 2012; Murray-García et al., 2014). Diversifying the workforce is important to better reflect the population, specifically for Black clients who have expressed a preference for racial matching. Third, there is a need to develop and deliver effective training for providers, in particular those from non-marginalized groups, regarding potential implicit bias and its effects on the developing alliance (Noon, 2018). As TA—whether clinician- or client-report—is a predictor of treatment retention, addressing effects of race on alliance is a critical step to addressing inequitable access to care and health disparities for high-risk racial minority families.