Future Directions in Youth and Family Treatment Engagement: Finishing the Bridge Between Science and Service

ABSTRACT

The field has spent more than 50 years investing in the quality of youth mental healthcare, with intervention science yielding roughly 1,300 efficacious treatments. In the latter half of this period, concurrent efforts in implementation science have developed effective methods for supporting front-line service organizations and therapists to begin to bridge the science to service gap. However, many youths and families still do not benefit fully from these strategic investments due to low treatment engagement: nearly half of youths in need of services pursue them, and among those who do, roughly another half terminate prematurely. The negative impact of low engagement is substantial, and is disproportionally and inequitably so for many. We contend that to build a robust and “finished” bridge connecting science and service, the field must go beyond its two historical foci of designing interventions and preparing therapists to deliver them, to include an intentional focus on the youths and families who participate in these interventions and who work with those therapists. In this paper, we highlight the significance of treatment engagement in youth mental healthcare and discuss the current state of the literature related to four priorities: conceptualization, theory, measurement, and interventions. Next, we offer an example from our own program of research as one illustration for advancing these priorities. Finally, we propose recommendations to act on these priorities.

To improve the quality of mental health services for youths and families, the field has pursued strategies to bridge science and service for at least a quarter century (National Institute of Mental Health [NIMH], 1998). Situated on one side of the span is an extensive, scientifically derived knowledge base, enabled by more than five decades of policy (Kefauver-Harris Drug Amendments, 1962; NIMH, 2001), leading to massive public and institutional investments in experimental interventions research (Hoagwood et al., 2018). This evidence base now represents more than 1,500 randomized controlled trials (RCTs), yielding protocols from roughly 1,300 treatment arms that meet a general standard for being evidence-based, addressing a wide variety of clinical concerns from infancy to early adulthood (PracticeWise, 2022).

Situated on the other side of the span are healthcare organizations and their service providers, whose context typically includes large caseloads, administrative and billing demands, workforce turnover, and limited professional development opportunities. Notably, these factors bear little resemblance to the context of research trials. Several major historical reviews showed that in these settings, typical effect sizes were low (Weisz et al., 1992, 1995, 2006) and that uptake or application of evidence-based practice was minimal (Garland et al., 2010; Weersing & Weisz, 2002; Weisz et al., 2006). In light of these observations, the attention of policymakers began to shift to address the known “chasm” between science and service (Institute of Medicine [IOM], 2001), resulting in increased investments in implementation science. Over the past 25 years, these investments have yielded new models that aim to fit efficacious interventions to community service contexts, with attention to several interconnected factors such as service delivery structure; training and consultation; measurement of meaningful outcomes; and organizational structure, resources, and preferences (e.g., Chorpita, 2003; Fixsen et al., 2005; Glasgow et al., 1999; Greenhalgh et al., 2004; Hoagwood et al., 2020; Schoenwald & Hoagwood, 2001). These advances required institutional commitments and investments in new types of research, albeit they remain dramatically lower for dissemination and implementation research than for interventions research (Purtle et al., 2015). This investment has begun to pay off, and an accumulation of trials has shown positive outcomes in service contexts (e.g., Chorpita et al., 2017; Jensen et al., 2014; Schaeffer et al., 2021), as well as sustained application of evidence-informed care within large health systems (e.g., Hoagwood et al., 2020; Nakamura et al., 2014; Southam-Gerow et al., 2014). In light of these positive findings, it is fair to conclude at long last that the field has built a passable structure and that crossing the chasm is now possible, although it is widely acknowledged that improving the quality of mental health services is not as straightforward as “building a bridge and driving across it with a truckload of manuals” (Shirk & Peterson, 2013, p. 107). Shirk and Peterson’s claim thus brings us to the question referenced in our title of how to move from a rudimentary structure to a fully robust and complete connection between science and service.

While addressing the complexities involved in the scientific study of organizations and systems, Nobel laureate Herbert Simon pointed out that, under typical conditions, a bridge will perform as planned – by carrying a path over an obstacle – but that it is the taxing conditions that expose the bridge’s weaknesses (Simon, 1988). In our field, “taxing conditions” are so commonplace that the evidence-based support needed by most youths and families does not make it across this bridge. In other words, our bridge from science to service is replete with metaphorical potholes and hazardous debris; it has an absence of lane lines and is missing stretches of guardrails. Technically, it can be spanned, but pragmatically, at scale, and under taxing conditions, it fails too often and for too many. We contend that one of the most insidious taxing conditions involves treatment engagement. That is, even when evidence-based care is delivered with a high degree of integrity and quality by community therapists, which is a significant achievement, roughly half of families withdraw from treatment and still more are unable to participate meaningfully in their care to achieve the desired outcomes. Thus, we contend that the topic of “future directions of treatment engagement research” is of paramount relevance to finishing the bridge from science to service. This recognition of the need to expand the focus of research beyond the two well-established areas of intervention science and implementation science is not new. Calls for increased family empowerment in mental health services have been made by leaders in the field for more than two decades (e.g., Bickman et al., 1998; Jensen & Hoagwood, 2008; McKay et al., 1996; Olin et al., 2010). There is, however, a new perspective emerging from the small but developing evidence base on treatment engagement, whose synthesis now offers increasingly clear direction for the future.

The Extent of the Challenge

The scope of this problem is extraordinary: low engagement is the most ubiquitous challenge among youths and families with mental health needs. National survey findings suggest that only one-third of youths with mental health needs access services (Merikangas et al., 2011). Of those who do enroll in treatment, between 50% and 70% terminate earlier than recommended by their therapists, meaning that they withdraw from therapy before achieving the expected benefits (de Haan et al., 2013; Guo et al., 2014; Nock & Ferriter, 2005; Pellerin et al., 2010). One study of a U.S. insurance claims database revealed that 45% of youth discontinued treatment within 30 days and those who remained in treatment averaged less than one session per month across a 6-month period (Harpaz-Rotem et al., 2004). Another study revealed that 40% of youths enrolled in mental health services attended only a single appointment (Saloner et al., 2014).

Low engagement is not an artifact of “usual care” services, nor is it confined to the mental health service sector. Similar attrition rates have been observed in studies testing evidence-based treatments (EBTs). For example, studies of Trauma-Focused CBT have reported attrition rates of 30–70% (Cohen et al., 2011; Scheeringa et al., 2011; Wamser-Nanney & Steinzor, 2016, 2017; Yasinski et al., 2018). A summary of 262 studies of behavioral parent training revealed that 25% of those participants who passed eligibility screening declined subsequent participation; another 13% discontinued before the first session; and another 13% withdrew during the study, suggesting that 51% of those appropriate for parent training did not complete treatment (Chacko et al., 2016). Even when treatments are designed to be especially responsive to youth and family needs and outperform high-quality EBTs on clinical and engagement outcomes, discontinuation rates are still unacceptably high (Chorpita & Daleiden, 2018). Attrition is so problematic in research that scholars now routinely summarize attrition rates across studies of specific EBTs (e.g., dialectical behavior therapy; Dixon & Linardon, 2020) and youth populations (e.g., depressed youth; Wright, Mughal et al., 2021) and publish ideas for retention/recruitment strategies specific to improving research studies (e.g., Daykin et al., 2018; Robinson et al., 2015).

It is even more alarming that attrition rates are not evenly distributed; typically, attrition is worse for those with greater mental health impairment (Pellerin et al., 2010; Rich et al., 2014). Moreover, youths and families of color are even less likely to access services than are their white/European American counterparts (Carson et al., 2011; Cook et al., 2013; Georgiades et al., 2018; Merikangas et al., 2011; Saloner et al., 2014). Youths of color attend fewer visits within a specified timeframe and are more likely to discontinue treatment than are white youths (Carson et al., 2011; de Haan et al., 2013; Marrast et al., 2016; Saloner et al., 2014). Other traditionally underserved groups, including those who have immigrated to the U.S. (Georgiades et al., 2018), those with incomes below the poverty line (Merikangas et al., 2010; Saloner et al., 2014), and those living in rural regions (Kelleher & Gardner, 2017) also have well-documented service use disparities.

Beyond the individual client, low engagement impacts mental health therapists, systems, and even other consumers. Missed appointments and therapy discontinuation consume staff time and reduce revenue (Barrett et al., 2008; Gross et al., 2011; Pekarik, 1985). Therapists rate low engagement as a highly stressful client behavior (Farber, 1983) that contributes to therapist distress (Piselli et al., 2011) and diminished confidence (Coutinho et al., 2011), with clear implications for turnover. Moreover, a missed appointment by one client increases the time that others spend on the waiting list. Waitlist time, in turn, is associated with increased likelihood of missed intake appointments when an appointment finally becomes available, thereby perpetuating a cycle of low engagement for the consumer population (Schraeder & Reid, 2015; Sherman et al., 2009).

The diminished return on our investment in the bridge from science to service is even worse than the data on attrition would suggest. This is because dropout has been posited as the final step on an engagement pathway that can begin with other engagement challenges that also have been shown to limit outcomes, including weak therapeutic alliance (e.g., Karver et al., 2018; McLeod & Weisz, 2005; Shirk et al., 2011), poor homework completion (e.g., Simons et al., 2012), insufficient skill rehearsal (e.g., Wu et al., 2020), low perceptions of treatment credibility (Wergeland et al., 2015), and low expectations of therapeutic success (e.g., Norris et al., 2019).

Let us also be clear: this vexing problem of treatment discontinuation and low engagement is not a recent one. Attrition rates of 30–40% were reported in the youth mental health literature 65 years ago (e.g., Levitt, 1957, 1958; Ross & Lacey, 1961). Across youth and adult clinic populations, attrition was examined with such frequency that a review paper was published as early as 1975 (i.e., Baekeland & Lundwall, 1975). Around this time, initial research on engagement interventions took place (e.g., Eyberg & Johnson, 1974; Fleischman, 1979; Holmes & Urie, 1975), laying the foundations for seminal work in the next two decades that (a) included behavioral observation of client and therapist interactions to understand client resistance (e.g., Chamberlain et al., 1984; Patterson & Chamberlain, 1994; Patterson & Forgatch, 1985), (b) identified and examined the impact of common barriers on treatment participation (e.g., Kazdin & Wassell, 1999; Kazdin et al., 1997), and (c) tested engagement interventions designed to reduce the interference of these barriers for families seeking services in front-line settings in urban contexts (e.g., McKay et al., 1998; McKay, Nudelman, et al., 1996; McKay, McCadam, et al., 1996).

Prioritizing Solutions

To produce a robust and finished bridge that performs under taxing real-life conditions, we contend, as others before us have (e.g., Shirk, 2004), that the field must prioritize research on treatment engagement, as it continues to invest in research on interventions and on dissemination and implementation. The disparity between scientific examination of treatment engagement and the evidence base of more than 1,500 RCTs testing children’s mental health interventions is striking: despite its significance in research and practice, there are fewer than 60 randomized controlled trials testing engagement interventions (Becker et al., 2018; Lakind et al., 2022). This disparity in research emphasis is amplified when considering the small literatures on the conceptualization, theory, and measurement of engagement relative to the many thousands of studies on psychopathology. However, the prioritization of treatment engagement cannot be borne by scholars alone, responding only to aspirational commitments documented in federal reports to improve access or reduce disparities in mental health services. Disappointingly, funding has decreased for youth mental health research (Hoagwood et al., 2018) at a time when the field requires strategic commitments and programmatic funding from sponsors that establish new, dedicated, and significant investments in research on treatment engagement. In addition, to incentivize a new generation of engagement researchers, we contend that organizational recognition of treatment engagement as a substantive area of scientific inquiry in its own right (i.e., beyond the contexts of specific mental health conditions, such as substance use or depression) will be required.

It is time for a deliberate effort to engineer a more robust bridge connecting science and service, one that works under the taxing conditions faced by most young people and their families who need mental health services. In the following sections, we discuss the present status of four priorities related to treatment engagement: conceptualization, theory, measurement, and interventions. We then offer an illustration from our own program of research as one strategy for advancing these integrated priorities. Finally, we offer recommendations for future directions related to these priorities.

Current State of the Science of Treatment Engagement

Conceptualization

Conceptualization involves the way the field organizes ideas to represent treatment engagement. As such, conceptualization provides a critical foundation for all other related work in the field, including theory development, measurement, and intervention. The current conceptualization in the field is that engagement is multidimensional, dynamic, and transactional.

Engagement is typically conceptualized as a latent variable with multiple composite or causal indicators. This means that there are multiple identifiable manifestations that reflect an individual’s level of engagement, rather than engagement as a global or unidimensional construct. Research has confirmed the structural validity of engagement as comprising behavioral, cognitive, and social dimensions (Chorpita & Becker, 2022). Thus, it is possible for an individual to be more or less engaged across different dimensions (e.g., have a positive view of the therapeutic alliance but have low expectations that treatment will be helpful). Additionally, high and low engagement can look different for different people (e.g., one person might have low expectations for treatment whereas another might have difficulty attending treatment regularly).

Although there are many conceptual models of engagement as a multidimensional construct (e.g., Becker et al., 2018; Haine-Schlagel & Walsh, 2015; King et al., 2014; Piotrowska et al., 2017; Pullmann et al., 2013; Staudt, 2007), scholars have not yet come to a consensus about the specific dimensions of engagement and their manifestations. There is general agreement that behavioral dimensions include behaviors such as treatment enrollment, attendance, session participation, and homework completion. In contrast, cognitive dimensions significantly vary across models (e.g., attitudes, self-efficacy, locus of control, understanding of treatment, readiness to change). Social dimensions typically represent the dyadic interactions between a therapist and their client (e.g., quality of the therapeutic alliance) and have been included explicitly in some models (e.g., Becker et al., 2018; Staudt, 2007) or conceptualized as primarily cognitive phenomena in others (e.g., perceptions of therapist efficacy, King et al., 2014).

In addition to being multidimensional, engagement is typically conceptualized as dynamic, meaning that it can change across time (Staudt, 2007). Multiple studies have demonstrated that alliance typically increases within a session (Zlotnick et al., 2020), changes across the course of treatment (Chu et al., 2014; Halfon et al., 2019), and improves even for those with the weakest alliance at the outset of treatment (e.g., Zilcha-Mano & Errázuriz, 2017). Engagement is also conceptualized as a transactional phenomenon, meaning that engagement is not a feature of an individual, but rather the result of transactions between the client and their ecology, including people (e.g., therapist, support network, community) and systems (e.g., the service agency, the network of services they receive), as well as the interface between the client and the treatment itself. Certainly, clients have agency to decide whether and how much to participate in treatment, yet the transactional nature of engagement means that therapists can and do influence the engagement of youths and families (Fjermestad et al., 2021; Ovenstad et al., 2020), even repairing alliance ruptures when they occur (Eubanks, Muran, et al., 2018; O’Keeffe et al., 2020).

The conceptualization of engagement as multidimensional, dynamic, and transactional largely diverges from how research on engagement historically has been carried out. In fact, an uncritical review of the evidence base could give one the misleading impression that engagement is unidimensional, static, and a stable attribute of the individual. This is because the research literature is siloed by discrete dimensions (e.g., the evidence base on alliance is separate from the evidence base on therapy expectancies; Becker et al., 2018), is characterized by studies that often measure engagement outcomes at only a single point in time (e.g., attendance or expectancies at the first appointment; Lakind et al., 2022), and features studies that examine client demographic characteristics (e.g., age, race, gender) and their association with one’s engagement in therapy (Karver et al., 2018).

Despite how the research has historically been organized, it has thus far yielded productive insights that have shaped the current conceptualization of engagement. Indeed, discrete lines of research are necessary and have provided a valuable understanding of such dimensions as alliance (e.g., McLeod & Weisz, 2005; O’Keeffe et al., 2020; Shirk et al., 2011) and expectancy (e.g., Horvath, 1990; Kazdin & Krouse, 1983; Schleider & Weisz, 2018). At the same time, the accumulating research now affords consolidation and synthesis that can advance a conceptualization of engagement as multidimensional, dynamic, and transactional, reflecting the complexities with which engagement manifests in clinical practice. An example of such a consolidated view is offered below (see “A Process Illustration” section).

Theory

Whereas conceptualization involves organizing ideas about how to represent a phenomenon, theories aim to explain phenomena by specifying how concepts relate to one another and by providing testable hypotheses (Fried, 2020; Meehl, 1990; Robinaugh et al., 2021). In psychology broadly, the pursuit of identifying, validating, and describing patterns of increasingly complex phenomena has greatly outpaced theory development (Proulx et al., 2021; Robinaugh et al., 2021). The same is true for the phenomenon of treatment engagement. Theories that specify potential mechanisms of change have influenced discrete engagement procedures, as in the example of operant conditioning (Skinner, 1938) that inspired the study of incentives for participation in therapy (e.g., Dumas et al., 2010). However, the field lacks bona fide theories of treatment engagement that attempt to explain: (a) why, when, and how treatment engagement will occur, (b) mechanisms of action, including those that represent different theoretical constructs (e.g., self-efficacy, social learning, operant conditioning), (c) mediators and moderators, and (d) engagement outcomes from a multidimensional perspective. Statistical models of treatment engagement are a bit more common than theories, but these represent patterns of data associations, rather than a priori explanatory theories that can account for current observations and also extrapolate to unforeseen contexts (Haslbeck et al., 2021).

Scholars have long lamented the waning influence of theory in psychological science (e.g., Eronen & Bringmann, 2021; Fried, 2020; Meehl, 1978, 1990; Proulx et al., 2021; Robinaugh et al., 2021), yet the value of a theory regarding engagement as a multidimensional, dynamic, and transactional phenomenon cannot be understated. Development and rigorous testing of theory elevates the knowledge gained from scientific investigation above one’s intuition (Lilienfeld, 2010; Meehl, 1978). Theory offers safeguards against biases, including a priori specified associations (vs. ad hoc “happenstantial” explanations; Beutler, 1991, p. 226) and an emphasis on falsification (vs. confirmation). In contrast to unwavering rationalization, strong theories evolve in an iterative, self-correcting way as the understanding of a phenomenon deepens and additional constructs or causal links lend themselves to testing (Lilienfeld, 2010). Theory is the connector among discrete empirical findings, serving to integrate these findings in a coherent manner thereby contributing to future empirical and intellectual breakthroughs (Prochaska et al., 2020). One caveat, though, is that most theories remain intact despite opposing empirical data. By testing competing theories within the same study (e.g., six theories of parenting practices and adolescent sexual behavior; L. G. Simons et al., 2016), scientists can contest prevailing assumptions and spur paradigm shifts that significantly advance the field (Joireman & Lange, 2015).

We have chosen to briefly discuss the Theory of Planned Behavior (TPB; Ajzen, 1985) as an exemplar; not because it fully explains the phenomena of low engagement, but because it embodies the features of strong theories and underscores the significance of consistent application of theory to psychological research. The TPB clearly specifies a clear causal chain: (a) attitudes, subjective norms, and perceived behavioral control influence (b) intentions to perform a behavior which then influence (c) behavioral performance (Ajzen, 1985). Importantly, the TPB made explicit ideas that might underlie clinical intuition (e.g., surrounding an individual with others who support behavior change will increase the likelihood of the individual’s behavior change) and it integrated concepts that, previously, had discrete lines of research. The strong theoretical foundation of the TPB accelerated the related priorities of measure and intervention development. Measures representing the TPB constructs have proliferated and contributed to a growing body of research clarifying the ways in which different measurement methods affect study findings (e.g., Courneya et al., 2006; Hardeman et al., 2013; Mankarious & Kothe, 2015). The TPB has guided the development of interventions for diverse health behaviors (e.g., diabetes management, Caro-Bautista et al., 2021; smoking cessation, Lareyre, Gourlan, et al., 2021), amassing voluminous research, including studies that diverge from what would be predicted by theory. Challenging studies and published criticisms have promoted intermittent scholarly discourse clarifying the TPB and evaluating its research evidence. Not only have these exchanges identified where the theory is more or less robust but they have also elevated the scientific discourse about expectations of theory and the role of theory in science (see Ajzen, 2011, 2015, 2020; Sniehotta et al., 2014, 2015; Trafimow, 2015). Although widely applied in healthcare research, the TPB has been used infrequently in mental health research, and rarely in youth mental health research (e.g., Chang et al., 2019). The few studies that exist in mental healthcare suggest that TPB is useful for predicting treatment attendance (e.g., Kelly et al., 2016).

The example of the TPB reveals what is possible for theory development related to treatment engagement. Developing theories that represent engagement as a multidimensional construct will require deep reflection about the integration of historically siloed lines of research, yet has the potential to create synergy that will accelerate our understanding of how different dimensions relate to one another. Elaborating theory to also reflect the dynamic and transactional nature of engagement will expand our capacity to represent the factors that influence engagement and identify mechanisms of change. We expect that initial attempts at theory will serve as stepping stones for future revisions of conceptualization and theory informed by theory-driven research. Importantly, prioritizing theory development and theory-driven research will enhance the field’s ability to strategically measure and intervene on treatment engagement.

Measurement

Measurement involves the quantification of a phenomenon (e.g., score on an alliance survey). Given that engagement is multidimensional, dynamic, and transactional, it is important to consider the capability of the field’s measurement methods to characterize the nature of engagement. Lakind et al. (2022) reviewed 112 measures used in 52 randomized controlled trials (RCTs) testing engagement interventions. They found that 61.6% of these measures reflected treatment attendance, with far fewer representing other behavioral (e.g., in-session participation or out-of-session homework completion), cognitive (e.g., expectancy or understanding of therapy), and social (e.g., therapeutic relationship) dimensions (for a list, see Lakind et al., 2022). Most studies collected outcome data for only a single engagement dimension and at a single point in time; when more than one engagement dimension was measured, it was often done so using different methods, thereby introducing method variance into any analyses examining the associations between dimensions. Measurement of youth (19.2% of studies) or caregiver (26.9%) perspectives of their own engagement was infrequent relative to use of clinic records, in part due to the overreliance on attendance as the primary indicator of engagement (Lakind et al., 2022). Certainly, there are measurement constraints and tradeoffs within the context of research studies, but it is important to recognize that the collective way that engagement has been measured within these studies shapes our ability to understand the nature of engagement, emergent challenges, the ebb and flow of engagement over time, mechanisms of action, the perspectives of youths and families, and mediators and moderators of engagement. Thus, considering ways to better align measurement with the multidimensional, dynamic, and transactional nature of engagement will significantly advance science and practice. As an example, we created youth and caregiver self-report questionnaires that measure five engagement dimensions (see “A Process Illustration” section of this paper). Data collected from a large sample of youths and caregivers participating in community mental health services allowed us to establish the structural validity of engagement as a multidimensional construct, with support for a five-factor structure apparent across youths and caregivers and robust to differences in individual characteristics (e.g., youth age and race; Chorpita & Becker, 2022).

Separate from intervention research, more extensive measure development and application has been pursued by independent scholars for the in-depth study of specific engagement dimensions, most notably, alliance. For alliance, measures have been developed, tested, and refined over the course of the past 30 years, beginning with therapist- and youth-report questionnaires (e.g., DiGiuseppe et al., 1996; Shirk & Saiz, 1992). To supplement self-report measures and address some of their potential limitations (e.g., demand characteristics, developmental considerations), scholars have developed observational measures of alliance (e.g., Diamond et al., 1999; Faw et al., 2005; Karver et al., 2008; McLeod & Weisz, 2005; O’Keeffe et al., 2020). Having a suite of measurement strategies has facilitated in-depth study of alliance, yet meta-analytic reviews also raise concerns about how the lack of consensus about the definition and measurement of alliance influences the conclusions that can be drawn about its association with treatment outcomes (Karver et al., 2018; McLeod et al., 2021). Discourse about the imprecise measurement of other dimensions has been ongoing in the literature for more than a decade (for a review of in-session participation and homework, see Haine-Schlagel et al., 2022; for alliance, see Karver et al., 2018; Shirk et al., 2011; Shirk & Karver, 2003; for attendance see O’Keeffe & Midgley, 2022; Warnick et al., 2012).

Pioneering work related to the measurement of barriers that interfere with treatment attendance (Kazdin et al., 1997) laid the foundation for scores of studies exploring consumer perspectives about obstacles that interfere with help-seeking and treatment participation (e.g., Becker et al., 2012; Kazdin & Wassell, 1999; Vázquez et al., 2022). Other scholars have focused on developing measurement methods for detecting signals of low engagement or emerging concerns, such as through therapist report (e.g., Becker, Wu, et al., 2021; Gellatly et al., 2019) and behavioral observation methods (e.g., O’Keeffe et al., 2020; Wright, Brookman-Frazee et al., 2021). This is an important area of research because signs of low engagement can be subtle and covert dimensions of low engagement (e.g., weak alliance, poor expectancies) often precede decisions to withdraw from treatment (Aubuchon-Endsley & Callahan, 2009; Garcia & Weisz, 2002).

Scholars have also pursued measurement efforts to characterize the ways that therapists interact with youths and families (e.g., Gellatly et al., 2019) and the specific clinical procedures therapists use to facilitate engagement (e.g., Martinez & Haine-Schlagel, 2018; Martinez et al., 2017; E. G. Wu et al., 2022). Measurement methods also track behaviors that might interfere with client engagement (e.g., criticizes, misunderstands; Diamond et al., 1999; Karver et al., 2008; O’Keeffe et al., 2020) as well as how therapists repair the relationship following some sort of tension or professional misstep (e.g., O’Keeffe et al., 2020).

In sum, specialization of measurement has yielded important insights about certain dimensions of treatment engagement (e.g., alliance, attendance), yet the tradeoffs of measurement insularity include the relative scarcity of measurement of other dimensions and concerns about the variety of definitions and measures for any one dimension. Aligning measurement to represent engagement as a multidimensional, dynamic, and transactional phenomena and to represent theoretical advances will, in turn, enhance future iterations of conceptualization and theory and also improve the design and impact of engagement interventions.

Interventions

There are approximately 55 published randomized controlled trials testing engagement interventions related to children’s mental health services (Becker et al., 2018; Lakind et al., 2022). Like other aspects of the engagement literature, the interventions literature tends to be fragmented by engagement dimensions (e.g., interventions aimed at improving attendance at the first treatment session do not also target alliance). This fragmentation presents challenges because clients can experience different levels of engagement across different dimensions, and our current interventions thus require therapists to assemble engagement strategies in vivo based on their own judgment and exposure to the evidence base. In a departure from the typical stage model for developing and testing psychosocial interventions that begins with well-controlled efficacy trials, the initial tests of many engagement interventions were conducted within front-line service settings and involved general service populations (e.g., Coleman & Kaplan, 1990; McKay, Garcia et al., 1996; McKay, Nudelman, et al., 1996), thereby increasing the generalizability of the findings across diverse service populations and settings. Along similar lines, there are several valuable examples of independent efforts to address outstanding engagement needs for specific populations (e.g., depressed African American youth; Breland-Noble & The AAKOMA Project Adult Advisory Board, 2012; culturally diverse families participating in behavioral parent training; McCabe et al., 2020).

Research has revealed an array of procedures that work to improve treatment engagement in some way (e.g., appointment reminders, cultural acknowledgment, motivational enhancement, psychoeducation; Lindsey et al., 2014), and accordingly, recent efforts have begun to focus on better understanding what procedures work best for what dimension or aspect of engagement (Becker et al., 2015). For example, a review of 50 RCTs testing engagement interventions found that certain engagement procedures (e.g., psychoeducation) are associated with positive engagement outcomes across cognitive, behavioral, and social dimensions of engagement, whereas other engagement procedures have more focal associations with engagement outcomes along one specific dimension (e.g., appointment reminders might only improve attendance; asking about someone’s background and culture might only improve the relationship; Becker et al., 2018).

Some research has also investigated the timing of engagement interventions. A review of 12 engagement intervention studies by Nock and Ferriter (2005) yielded guidance about ways to organize practices across the course of treatment. They found that most interventions could be classified as preparatory enhancement strategies (i.e., those for which engagement strategies are delivered only at the outset of treatment). This finding is significant because preparatory enhancement strategies (e.g., psychoeducation) presumably set the stage for some threshold of sufficient engagement at the outset of treatment. However, Nock and Ferriter suggested that continuous enhancement strategies that are delivered over the course of treatment (e.g., contingencies, telephone communication in between sessions) better reflect the dynamic nature of engagement, are focused on maintaining engagement over time, and might be useful for reducing or addressing engagement concerns that emerge over the course of treatment. This idea of designing engagement interventions that are matched to an engagement target or dimension and that are continuous across time is also reflected in person-centered approaches that assess youth or family perspectives about engagement or barriers at the outset of treatment and use this information to tailor responsive care (e.g., Becker et al., 2019; McCabe et al., 2020; McCabe & Yeh, 2009; Sanchez et al., 2022).

Many engagement interventions are also program-agnostic, meaning that they were developed independent of a specific therapeutic approach or treatment program, presumably supporting increased generalizability if practicing therapists had a way to know what was in the engagement evidence base. Whereas most EBTs for childhood problems are designed to include clinical procedures that target a specific clinical focus (e.g., anxiety, depression), they also commonly include several preparatory engagement procedures, most likely those that are considered best-suited for broad application across clients (e.g., rapport building, psychoeducation) (Chorpita & Daleiden, 2018; E. G. Wu et al., 2022). Research examining how therapists use manuals revealed that therapists return to those “preparatory” engagement strategies for more than one-third of cases, after they are otherwise out of the preparatory phase of the treatment (Park et al., 2015). Given concerns about treatment attrition, scholars have added focal engagement strategies to further enhance specific EBTs (e.g., Trauma-Focused CBT; Dorsey et al., 2014; Supporting Teens’ Academic Needs Daily (STAND); Sibley et al., 2016). Situating additional engagement strategies within EBTs is likely to increase their application, as research has shown that therapist selection and use of engagement procedures are largely guided by the protocol they are using (Martinez et al., 2017; E. G. Wu et al., 2022). At the same time, this insular approach to enhancing one EBT at a time does not allow treatment developers, or therapists, to leverage the entire evidence base on engagement. In its entirety, the evidence base on engagement interventions has many strengths, but further investment is needed to understand the complexity of how best to arrange these procedures, by studying how they work in the context of the general service landscape (e.g., studying what is in the engagement evidence base to address any of the challenges that arise in practice), rather than only in population-specific contexts (e.g., studying how to increase motivation within a substance use population).

A Process Illustration

We offer an illustration of our initial pursuits of these four priorities and our attempts to integrate our efforts across each.

Conceptualization

In our own work, we have organized our understanding of the engagement literature in terms of a multidimensional conceptualization. Specifically, we proposed the REACH framework (Becker et al., 2018) that corresponds to five dimensions of engagement: Relationship (e.g., therapeutic alliance; Shirk & Karver, 2003), Expectancy (e.g., beliefs that treatment will be helpful and that one can participate successfully in treatment; Nock & Kazdin, 2005), Attendance (e.g., presence at treatment sessions; Nock & Ferriter, 2005), Clarity (e.g., understanding about the treatment approach or the roles of each person involved in treatment; Shuman & Shapiro, 2002), and Homework, which reflects multiple participation behaviors (e.g., homework completion, in-session participation; Nock & Ferriter, 2005).

The REACH dimensions were rationally proposed (Becker & Chorpita, 2016) based on two initial literature reviews (e.g., Becker et al., 2015; Lindsey et al., 2014). We then conducted a systematic review of 50 randomized controlled trials testing engagement interventions in children’s mental health services (i.e., Becker et al., 2018), in which we identified a sample of effective engagement interventions and classified their discrete outcomes (within the REACH framework or otherwise), which demonstrated that the diverse ways in which the field operationalizes treatment engagement are easily consolidated within this framework. We then applied a distillation method to classify engagement procedures (i.e., “practice elements,” Chorpita & Daleiden, 2009; Chorpita et al., 2005), such as appointment reminders, psychoeducation, and instilling hope, across effective engagement interventions, thereby permitting us to summarize the common and unique procedures used across the engagement interventions literature (K. D. Becker et al., 2015, 2018). Distillation offers advantages as a literature synthesis approach beyond the typical method of identifying EBTs based on their discrete lines of research, including promoting a common language for describing the clinical features of each intervention across a taxonomically diverse literature (Chorpita et al., 2005) and facilitating synthesis of a literature in which there exist few formal protocols.

We then analyzed the data to identify practice elements common to interventions that were effective at increasing engagement within each REACH dimension. Practices such as psychoeducation were found to be relevant to all REACH dimensions, whereas other practices had unique associations (e.g., Relationship: cultural acknowledgment; Expectancy: motivational enhancement; see Becker et al., 2018). Thus, across the evidence base for engagement interventions, we extracted a multidimensional framework of treatment engagement as well as a model for matching engagement procedures with the dimensions they might be well-suited to address, which could then inform intervention development.

Theory

In our quest to build an intervention that would make as much of the evidence base as accessible as possible, we used two theories to guide our thinking. First, in their classic text, Malone and Crowston (1994) defined coordination as managing dependencies among resources and activities to accomplish a goal. In mental health, for example, resources might include published research findings, manualized protocols, data collected about a client, an assessment report, a treatment plan, etc. Resources might also include people invested in youth mental health (e.g., therapists, supervisors, other service providers, youths, families). Activities can include writing an assessment report, treatment planning, delivering therapy, supervising, etc. Coordination organizes the dependencies by defining which resources are utilized (e.g., a supervisor and therapist meeting together to discuss an assessment report) for which activities (e.g., to review that report to plan together how to address the identified problem) and in what order (e.g., the plan made in supervision will precede a therapy session in which the plan is implemented). Thus, coordination theory prompted us to use the REACH framework as a consistent organizing vocabulary across all intervention resources and activities.

Second, there is an ideal flow and logic model for clinical decision-making that is represented by the Action Cycle component of Graham et al.’s (2006) Knowledge to Action framework (KTA). The KTA framework articulates how research evidence is synthesized and molded into knowledge resources (e.g., manuals or other evidence summaries) for a therapist and the Action Cycle specifies a behavioral workflow that represents decision points at which therapists can interact with knowledge resources in practice. For example, problem identification is an early decision in which a therapist might interact with assessment information. Intervention selection is a subsequent decision in which a therapist might interact with intervention protocols or evidence summaries. Moreover, in a well-coordinated model, problem identification should precede the choice of a solution. This behavioral workflow of decisions is not guaranteed; rather, it must be managed independently within the service system, such as with policies, training, or management processes. Thus, the KTA framework prompted us to represent these clinical decisions of the Action Cycle in our intervention model, structure a workflow to guide intentional reflection about each decision in sequence, and ensure knowledge resources relevant to each decision were accessible when appropriate.

We were intentionally agnostic in terms of a specific psychological theory of engagement because our reliance on coordination theory and the KTA framework meant that the knowledge architecture of our intervention system should be able to accommodate the entire evidence base of engagement procedures, which each might reflect different theories or hypothesized mechanisms of change. Moreover, the architecture explicitly allows the introduction of engagement theory into the problem-solving sequence under appropriate conditions (e.g., when interventions are not getting the expected outcome; see Chorpita & Daleiden, 2014).

Measurement

Given the challenges with identifying engagement risk across a variety of dimensions, some quite subtle (e.g., Gearing et al., 2012; Hunsley et al., 1999), as well as with siloed instrumentation in the field, we chose to pursue a multidimensional measurement strategy. Specifically, we developed two consumer-focused self-report instruments for youths and caregivers to indicate their own level of engagement in youth mental health services. The initial versions of the My Thoughts about Therapy instruments for youths and caregivers (MTT-Y and MTT-CG, respectively) include 35 items, with 7 items representing each of the 5 REACH dimensions. Respondents rate their agreement with each item using a 4-point Likert-type scale (0 = strongly disagree to 3 = strongly agree). These measures are freely available in English and Spanish, along with a user’s guide and scoring program, at https://www.childfirst.ucla.edu/resources/.

Using these instruments, we recently empirically validated the multidimensional structure of treatment engagement. In a large sample of youths and/or their caregivers (n = 1,807), we demonstrated the empirical superiority of the 5-factor structure of the REACH framework relative to 1-factor and 4-factor models of treatment engagement, thereby providing evidence for the hypothesized multidimensionality of treatment engagement (Chorpita & Becker, 2022). We also demonstrated that the distinct dimensions of engagement were similar across different therapy participants (i.e., youths and caregivers), and that the factor structure was consistent across youth age, youth race, region, and caregiver language (Chorpita & Becker, 2022). Prior to this study, the multidimensional factor structure of treatment engagement had been presumed, but had not yet been confirmed, even in the adult services literature. We expect that as we and the field learn more about engagement, these and similar instruments will require enhancements and extensions, such as by representing more precise conceptualization (e.g., adding or editing items; adding dimensions), by conducting item-response analyses to identify items that predict subsequent engagement outcomes (e.g., attendance or attrition), or by creating briefer measures.

Intervention

We created the Reaching Families Engagement System (RFES; referred to as a coordinated knowledge system in our pilot study – Becker et al., 2019) that coordinates multiple evidence resources into a single system to support decisions consistent with Graham et al.’s (2006) Action Cycle, including identification of engagement problems, selection of an intervention that is well-suited to the focal problem, preparation for delivering the intervention, delivery of the intervention, and evaluation of its effect on engagement (Becker et al., 2019). The RFES includes three knowledge resources to support each of these actions (identified in italics). First, scores from the MTT questionnaires that youths and caregivers complete about their engagement in services (identify, evaluate). Second, a worksheet used by therapists that (a) includes specific prompts representing each REACH dimension to elicit reflection about youth/caregiver engagement from the perspective of the therapist (identify) and (b) provides an explicit mapping of youth engagement problems to specific engagement practices (identify, select, prepare, evaluate) based on the associations we identified between REACH dimensions and engagement procedures in our literature review (Becker et al., 2018). Third, a library of brief guides that provide concise descriptions about how to use specific engagement procedures with a youth or caregiver (prepare, deliver).

The development of the RFES was guided by principles and activities consistent with the architecture of Managing and Adapting Practice (MAP; Chorpita & Daleiden, 2014, 2018). Following the MAP design principles (Chorpita & Daleiden, 2014), the RFES structures a purposeful, goal-directed, and self-correcting clinical reasoning process involving problem identification, selection, preparation, implementation, and evaluation in a recursive loop that continues until the engagement problem is resolved. These decisions and activities are consistent with a Deming cycle (Deming, 1993), a well-established decision-making process that is often referred to as a Plan-Do-Study-Act (PDSA) cycle. Research on the RFES thus far has demonstrated promising findings that the system supports the identification of engagement problems and selection of engagement procedures well-suited to that challenge (Becker et al., 2019), although the investigation of this system in larger trials with a wider variety of outcomes will be important.

Future Directions for the Science of Treatment Engagement

Our illustration is just one working example of a possible process of how integration across the four priorities can create synergy to advance discovery. Beyond this type of process, however, there are many ways to accelerate the science and practice of treatment engagement. Below we offer recommendations that, collectively, attempt to address gaps in the field and establish a set of guiding principles and shared concerns.

Explore Additional Dimensions of Engagement

Research appears to support five core dimensions of engagement: relationship/alliance, expectancies, attendance, understanding of therapy, and homework/session involvement (Chorpita & Becker, 2022). Because conceptualization informs theory, measurement, assessment, and intervention, identifying additional dimensions of engagement is an important next step. Possibilities include an affective dimension (e.g., emotions related to treatment such as relief or frustration) or a social network dimension that expands beyond the typical dyadic interactions reflected in the therapeutic alliance to represent the ways in which social norms, community stigma, or familial support for therapy encourage connections to therapy. Additional engagement dimensions would then require integration with current conceptualizations of engagement, elaboration of theory and measurement, and focus within intervention research. Given the substantive volume of research on consumer perspectives about therapy (e.g., Baker-Ericzén et al., 2013; Buckingham et al., 2016; Sylwestrzak et al., 2015; Wasson Simpson et al., 2022), a synthesis of the literature on consumer perspectives about therapy might be instrumental for elaborating existing engagement dimensions or identifying additional ones.

Develop a Structured Vocabulary to Represent Engagement Dimensions and Interventions

As is common in psychological science, there are a myriad of engagement-related terms, which impede efforts to synthesize concepts and findings across discrete models and lines of research (National Academies of Sciences, Engineering, and Medicine, 2022). Given the independent lines of research inquiry associated with different engagement dimensions and given that these dimensions are likely to be related in meaningful ways (e.g., lack of understanding of treatment rationale could affect homework completion or impact attendance), conceptual models as well as intervention approaches are likely to benefit from the application of one or more unifying terminology sets to the entire domain of engagement research. In the broader field, there are examples of organizing taxonomies and ontologies for classifying mental health problems (e.g., DSM, ICD), indexing published articles (i.e., Medical Subject Headings; MeSH), and characterizing behavior change theories and interventions (see Larsen et al., 2017) that facilitate a shared understanding and advance the field’s ability to synthesize evidence produced by a wide variety of research programs and approaches. Efforts to harmonize terminology could be fruitfully applied to dimensions of engagement (e.g., Haine-Schlagel & Walsh, 2015; King et al., 2014; Piotrowska et al., 2017; Pullmann et al., 2013; Staudt, 2007), specific engagement-related client behaviors (e.g., Chamberlain et al., 1984; Gellatly et al., 2019; McLeod & Weisz, 2005; O’Keeffe et al., 2020), specific engagement-related therapist behaviors (e.g., Karver et al., 2008; Martinez & Haine-Schlagel, 2018; O’Keeffe et al., 2020), and engagement intervention procedures (e.g., Becker et al., 2018; Pellecchia et al., 2018).

Deliberate attempts to create shared specifications of relevant constructs across the entire field of engagement science are likely to expose meaningful similarities and differences across discrete research endeavors. For example, some scholars might consider a client’s understanding about what therapy involves as a facet of alliance (e.g., Karver et al., 2008), whereas other scholars might conceptualize this type of understanding as a discrete cognitive dimension (e.g., K. D. Becker et al., 2018). Thus, one downstream benefit of explicitly attending to terminology is the potential to improve precision within and across each of the four integrated priorities. By revealing overlap, omissions, and underspecification across discrete research efforts, the application of a controlled vocabulary or terminology set can increase conceptual clarity and precision, which, in turn, advance our ability to clearly articulate associations among theoretical constructs, to develop methods that validly represent these constructs, and to specify our interventions clearly to test theory-driven hypotheses (Larsen et al., 2017).

Consider Reporting Standards to Increase Consistency Across Studies

As knowledge accumulates about treatment engagement, the synthesis across discrete studies with distinct methods, measures, and terms, becomes increasingly more complex. In psychology more broadly, efforts to increase the quality of reporting have been successful at making data from primary studies more interpretable in general and more usable for secondary synthesis in particular (e.g., Consolidated Standards of Reporting Trials [CONSORT]: Altman et al., 2001; Turner et al., 2012; Journal Article Reporting Standards [JARS]: APA Publications and Communications Board Working Group on Journal Article Reporting Standards, 2008). Consensus statements have been developed about reporting for specific types of studies, such as child and adolescent anxiety disorder intervention studies (i.e., Creswell et al., 2021). Reporting standards will likely vary based on the research question and methods. Given the ubiquity and ease of measuring attendance, and the need for continued measure development and/or refinement for the other dimensions, we offer three ideas for the application of reporting standards specific to attendance that could be implemented in the immediate future.

First, reporting standards could recommend reporting attrition and/or attendance rates for every study testing a youth mental health intervention. One study revealed that 30–50% of intervention studies of behavioral parent training did not report information about study attrition or attendance (Chacko et al., 2016); thus, this recommendation would increase the interpretability of each discrete study’s outcomes and intervention feasibility and impact. Second, reporting standards could recommend multiple ways for calculating and reporting attrition and attendance rates within the same study, thereby further increasing the interpretability of any one study’s rates as well as the synthesis across studies. This is important because research has shown that rates vary according to how attrition and attendance are operationalized and measured (Warnick et al., 2012). Third, when attrition and attendance are dependent variables, reporting standards could recommend that study authors conduct analyses with different operationalizations of the dependent variables to examine if and how operationalization affects results (e.g., risk factors for attrition might vary according to its operationalization; Warnick et al., 2012). This standard would not only enhance the interpretability of study findings and aid synthesis across studies but also contribute to the understanding of how measurement affects results and what operationalization is useful for what purpose (Warnick et al., 2012).

In addition, although attendance and attrition are pragmatically significant engagement indicators, they are the bare minimum that should be recommended by reporting standards. Similarly, psychology journals are the tip of the iceberg when it comes to applicability of these proposed reporting standards. For example, robust reporting recommendations could include each of the five engagement dimensions and be recommended in intervention studies in all of healthcare. Multidimensional reporting standards on a widespread basis would accelerate our understanding of the impact of interventions on an array of engagement outcomes as well as the extent to which these discrete engagement dimensions covary, not just within individual studies, not just within psychology, but across the 75 RCTs published per day in healthcare fields (Bastian et al., 2010). Moreover, multidimensional measurement could yield valuable information when an intervention did not produce the expected clinical effects. Knowing that an individual had limited understanding of the therapeutic rationale or that they did not complete therapy homework regularly might reveal potentially malleable targets for future intervention. In this vein, multidimensional measurement closely aligns with the logic behind experimental therapeutics initiatives, which attempt to understand why an intervention did or did not work and thus point to where future efforts should be directed to enhance impact and outcomes, that are widespread in healthcare and explicitly supported by the National Institute of Mental Health (Gordon, 2017). In addition, multidimensional measurement of engagement parallels calls made decades ago by leaders in the field for the multidimensional measurement of intervention outcomes (e.g., functioning, consumer perspectives, environmental contexts) that extended beyond clinical symptoms (Hoagwood et al., 1996; P. S. Jensen et al., 1996) and resulted in notable expansions in multidimensional reporting (Hoagwood et al., 2012; Krause et al., 2019).

Develop and Test Theories that Represent the Multidimensional, Dynamic, and Transactional Conceptualization of Engagement

Although all theories are incomplete representations, we can use existing research and theory to develop new theories with sufficient verisimilitude to provide a foundation for testing and amending our ideas (Meehl, 1990). In its simplest form, such a theory could put forth clearly specified constructs, articulate the associations among them (e.g., alliance and expectancies will be positively correlated; C. L. Patterson et al., 2014) and hypothesize the mechanism of association (e.g., as people develop connections with someone they view as caring and competent, they have increased optimism for success). A more elaborate theory might also include clinical outcomes and predict the magnitude of the effect of each dimension on outcomes based on what has been previously observed in research (e.g., alliance will have a stronger effect on outcomes than will expectations; Constantino et al., 2021; Karver et al., 2018). Given that engagement is dynamic, another iteration might specify temporal associations among dimensions (e.g., pre-treatment expectancies will be associated with mid-treatment alliance; Kirsch et al., 2018; change in alliance will predict later participation in therapeutic activities; McLeod et al., 2014), evaluate mediational effects (e.g., in-session rehearsal will mediate the effects of pre-treatment expectancies on outcomes; M. S. Wu et al., 2020), and examine the direction and reciprocity of effects between engagement and clinical symptoms (e.g., early alliance will predict subsequent clinical symptoms but early symptoms will not predict subsequent alliance; Labouliere et al., 2017).

Further elaborating the real-world phenomena that influence engagement dimensions is also an essential step in theory development. This involves extending beyond post hoc analyses of the associations between pre-treatment youth or therapist characteristics and engagement and instead building upon previous correlational and experimental research and articulating causal relationships that are manipulable within experimental designs. Elaborating and testing theories about what interventions impact different engagement dimensions, mechanisms of change, the thresholds at which engagement problems emerge, or the conditions that influence high or low engagement would significantly advance our understanding of engagement. For example, an evolving theory might predict that certain engagement procedures (e.g., psychoeducation, rapport building) would generally be useful at the outset of treatment to establish a minimum but sufficient level of alliance and expectancy, but that if new problems emerge, other, more robust engagement procedures (e.g., exploring a client’s specific identities or motivation) would be required to address these concerns.

Develop and Apply Measurement Methods that Represent the Multidimensional, Dynamic, and Transactional Conceptualization of Engagement

Further advances in the realms of conceptualization, theory, and intervention are predicated on measurement that adequately represents the conceptualization of engagement. Thus, an important step toward enhancing theory and the application of science in practice is to include measures representing each engagement dimension and to measure dimensions across time (cf. Hoagwood et al., 1996, in the context of mental health services). Using measures and methods that reflect the multidimensional and dynamic nature of engagement within the same study will yield insight about some of the questions deemed essential by our theories: what measures are well-suited to assess different engagement dimensions, how early different types of engagement problems can be detected, how different engagement dimensions change over time, what interventions bring about change on different dimensions, how different engagement dimensions relate to one another over time, and how each dimension is related to treatment outcome. In addition, multidimensional measurement across time within a single study would set up the conditions necessary to conduct mediational analyses that might be hypothesized in theory but are largely unexplored in the literature (Becker et al., 2018).

Although measuring multiple engagement dimensions within a single study is currently possible with existing methods and instruments, this approach would typically require using different instruments that each reflect a discrete engagement dimension. Given the issues introduced by method variance confounds (e.g., observation for one dimension, questionnaire another), it will be important to consider approaches that have some capacity for measuring multiple dimensions of engagement within a single measurement instrument for each informant. A uniform measurement approach would provide greater confidence that the results are not merely an artifact of different measurement strategies but instead are attributable to engagement as a construct.

Identification of candidate indicators to build new measures or “indicator libraries” can possibly be accelerated through natural language processing (NLP), which involves techniques to quantify words, phrases, and syntax into meaningful sets to promote comparison and synthesis. As has been done in behavioral medicine (e.g., Larsen & Bong, 2016), for example, NLP permits comparison across instruments to identify similar items and constructs. One could apply NLP to alliance measures, for example, to determine the facets of alliance represented by different measures (e.g., therapist alliance represents bond and task agreement whereas youth alliance represents an affective component; Ormhaug et al., 2015).

Aside from the in-depth work on alliance, much of what is currently known about engagement is based on coarse behavioral observation or global ratings, which do not easily represent engagement as a transactional phenomenon. Thus, an important future direction involves development of not only more granular multidimensional measures but also those that elicit perspectives from different individuals (e.g., youths, caregivers, therapists, case managers, clinical supervisors, etc.) regarding youth and caregiver engagement in services. A recent meta-analysis yielded moderate associations between youth-therapist and parent-therapist reports of dyadic alliance, with youths and parents reporting higher alliance than therapists (Roest et al., 2022). Obtaining information from different informants can reveal each respondent’s unique contexts and perspectives, as well as whether certain perspectives are more relevant for certain engagement outcomes. For example, studies suggest that youth-rated alliance is more strongly associated with treatment outcomes than is therapist-rated alliance (Murphy & Hutton, 2018; Ovenstad et al., 2022). Informant discrepancies provide rich and complementary perspectives that enhance our understanding of a phenomenon and that might be otherwise unnoticeable by eliciting a single perspective (De Los Reyes et al., 2022; Kraemer et al., 2003). Important research questions include: what information do therapists use to assess engagement, on what dimensions do therapist perspectives vary from those of youths or caregivers, and how do therapists adjust their clinical care when given feedback about engagement.

Build Pragmatic Detection and Monitoring Technologies

As measurement approaches mature, it will also be important to consider how to organize assessment for the purpose of ongoing risk identification and associated risk management. Given that therapists often have difficulty assessing engagement (e.g., Becker, Wu, et al., 2021; Hunsley et al., 1999), rapid, timely, and intermittent assessment is essential to detect engagement concerns before they escalate. Detection technology might include a questionnaire administered and scored within a digital platform that is completed by a youth or caregiver prior to or after a treatment session. Scores that indicate engagement concerns could be shared with the therapist to guide intervention, as well as tracked over time to monitor progress. This type of engagement monitoring and feedback could leverage decades of research demonstrating the value of measurement and feedback systems to clinical practice (e.g., Chorpita et al., 2008, 2016; Lambert, 2005; Lambert et al., 2003), including a meta-analysis that revealed that feedback on process indicators (e.g., alliance, in-session participation) produced greater effect sizes in clinical outcomes compared to feedback on progress indicators (e.g., reported symptoms) (Poston & Hanson, 2010). It is critical that any such approaches be efficient and pragmatic, which means future research will benefit from the application of such methods as Item Response Theory and adaptive testing to large sets of empirical indicators of engagement dimensions, as well as consideration of passive monitoring strategies, such as machine-assisted analysis of a wide variety of naturally occurring inputs (e.g., facial expression, attendance and arrival records, online homework completion or quality, client or therapist vocal intonation, chart note documentation, live dialogue in therapy events, etc.).

Build Responsive Intervention Systems

Research has demonstrated that therapists typically rely on a narrow subset of engagement procedures from the evidence base (Becker, Dickerson, et al., 2021) and use ecologically proximal supports to guide their selection and use of engagement strategies (e.g., procedures recommended in a treatment manual being used, as opposed to what the broader engagement literature might suggest; E. G. Wu et al., 2022). Thus, in the absence of a system of supports, it is naïve to expect therapists to maximize opportunities to promote engagement or to substantially reduce the impact of engagement challenges. Moreover, any such system must also be responsive to the substantial array of different possible clinical presentations of low engagement. Selecting only a handful of engagement procedures from the evidence base would be insufficient, instead representing a “tip of the iceberg” approach to responsive intervention (Becker et al., 2018, p. 18). Instead, to leverage the full evidence base of more than 40 years of engagement interventions research, it would be helpful to organize intervention resources into a common system with a unifying interface (Becker et al., 2018), perhaps along with a library of instructional resources for delivering these engagement procedures.

To capitalize on the connection between assessment and intervention, this system would require some harmonization across detection/monitoring technologies and intervention resources, such that there is a clear mapping of engagement targets and challenges to their indicated solutions, supporting therapists to know what to use when (Becker, Dickerson, et al., 2021). A responsive intervention system would also benefit from the capacity to integrate the continually growing and evolving evidence base and support therapist use of the various engagement procedures in that evidence base, so that they are best matched to the context at hand. Moreover, members of the workforce beyond the individual therapist could be leveraged to deliver select engagement procedures or components thereof (e.g., front office staff who provide appointment reminders could use additional engagement strategies to enhance these reminders; Donohue et al., 1998). As another example, informational brochures or videos could be designed to capitalize on engagement strategies as part of an organizational-level approach to enhancing treatment engagement.

Reconceptualize the “Engaging” Therapist

There are implicit premises about what it means to be effective at engaging youths and families in treatment, which are often revealed in a description of the traits or attributes of a therapist (e.g., warm, caring, thoughtful). This implicit notion that engagement may be dependent on the therapist’s traits inhibits a more deliberate inspection of which specific behaviors are important for which youths in which context. Moreover, a corresponding inference is that these traits are naturally-occurring baseline attributes of a given therapist, rather than reflective of deliberate and intentional behaviors used to engage youths and families in treatment and the transactional nature of therapist–client relationships (G. R. Patterson & Chamberlain, 1994). Although there is some natural variation across therapists in certain interpersonal characteristics, engagement represents a set of related competencies that can be developed in anyone through the teaching of specific behaviors (e.g., explaining the rationale for an intervention, expressing optimism, or modeling a new skill for a youth). The precise set of behavioral competencies and transactional or contextual conditions that can effectively engage youths and families will evolve as new research emerges. However, the science of engagement will benefit from adopting an explicit perspective that engaging behaviors and conditions represent an inspectable set of phenomena that can be dismantled, manipulated, or trained, rather than presumed to be merely trait-like or background features of the therapist in context.

Elevate Consumers and their Communities in Research and Efforts Related to Engagement

Although engagement interventions have been tested with success in front-line service contexts and with racially diverse samples (Becker et al., 2018), more attention is needed about including individuals in research who represent diversity across race, ethnicity, language, religion, gender and sexual orientation, socioeconomic status, and geographic region (Polo et al., 2019). Such initiatives should not only increase inclusion of groups of individuals whose numbers are grossly underrepresented in research but also amplify the voices and perspectives of individuals from diverse backgrounds. A substantial body of research has elicited youth and caregiver perspectives about obstacles that make it difficult for them to access and consistently participate in treatment (e.g., therapist turnover, feeling blamed or misunderstood by therapists, competing priorities, stigma, discrimination, scheduling, transportation, etc.; Bornheimer et al., 2018; Buckingham et al., 2016; Kazdin et al., 1997; Lindsey et al., 2013; McKay & Bannon, 2004). Finding solutions to these explicit challenges as experienced by mental health consumers is a significant priority because many existing interventions do not align with family priorities or concerns but instead are guided by a specific engagement target (e.g., rather than reducing stigma, an intervention might focus on increasing expectancies or alliance).

Research has identified community-level characteristics that are associated with reduced help-seeking and treatment engagement (e.g., Cooper et al., 2020; Ijadi-Maghsoodi et al., 2018; Price et al., 2022; Wadsworth et al., 2018); thus, social-ecological models offer promise for our four integrated priorities. Initiatives to theorize, measure, and intervene on community-level support for mental health services could complement present efforts to impact engagement within the dyadic relationship between therapists and clients and might provide avenues for connecting with the many youths who might otherwise never enroll in mental health services (Merikangas et al., 2010). Future research could apply multidimensional measurement of engagement to pre-treatment help-seeking or population-level perspectives about mental health services to identify dimensions that could serve as facilitators (e.g., strong relationships with faith leaders) or barriers (e.g., distrust of medical system) for enrollment in services.

This work will be advanced by connecting with important and untapped lines of research that could produce much synergy and if they could be integrated with engagement research, including research on stigma (e.g., Price & Hollinsaid, 2022), mental health literacy (e.g., A. Wright et al., 2006; Jorm, 2012), and direct to consumer marketing (e.g., Becker et al., 2020). Therapist perspectives are also of great value to the four integrated priorities. Research that examines therapist experiences with engaging youth and families (e.g., Becker, Dickerson, et al., 2021; Becker, Wu, et al., 2021; Eubanks, Burckell, et al., 2018) or with using specific engagement interventions (e.g., W. Chu et al., 2022; Haine-Schlagel et al., 2017) can shed light on the features of their professional judgment and habits that align with theory and consumer perspectives as well as other areas that might be supported through professional development.

Recognize, Prioritize, and Incentivize Treatment Engagement as an Intellectual Focus Independent of EBTs

Twenty years ago, implementation science was not recognized as a dedicated focus of expertise within clinical psychology. Rather, implementation was an important focus, but efforts fell under the purview of treatment developers to design implementation strategies for their specific EBT, which were largely siloed from the growing body of implementation research within our field, and the even larger evidence base on implementation outside of psychology. So too is our field trending in this way when it comes to treatment engagement. Many studies on treatment engagement have been conducted ad hoc as an aspect of intervention outcome studies for specific populations. A search on NIH Reporter revealed that much of the funded engagement research is also within the context of examining engagement for those receiving a specific intervention. Although these discrete, effects-in-context efforts can yield value, developing engagement interventions for discrete EBTs will not sufficiently advance science, let alone impact practice, if the field must wait for each EBT developer to design and test their own set of engagement strategies. An alternative vision is that, as has occurred with implementation science in psychology, the field would uncouple engagement and clinical interventions research and pursue a standalone area of engagement science, focusing broadly on the four priorities outlined above. Importantly, institutions would need to recognize the intellectual independence of treatment engagement, such as by explicitly recruiting engagement science experts, as has become more commonplace for those with implementation or health disparities expertise.

Coordinate Efforts in the Field to Advance these Priorities

Given expertise in the field within discrete lines of research, now is an optimal time for collaboration across parties invested in youth mental health – scientists, funders, professional organizations, publishers, and consumers – to pursue these recommendations and advance these four priorities. For example, terminology development requires coordination across invested parties to formally specify a shared vocabulary and definitions that will be used in science, funding announcements, and reporting standards to further accelerate integration of knowledge about treatment engagement across the field. Theory development requires deemphasizing statistical identification of effects (van Rooij & Baggio, 2021) and instead committing to robust explanatory paradigms (Lilienfeld, 2010; Proulx & Morey, 2021), which could be accelerated and elevated by collaboration to clarify concepts and measurement, hypothesize associations between concepts, prioritize research questions considered important by theory, and recommend methodological strategies for answering prioritized questions (Proulx et al., 2021). Models and recommendations for coordinating efforts are available in the literature (see Michie et al., 2018; National Academies of Sciences, Engineering, and Medicine, 2022).

Heed Diction

It is of course important to be cautious with potentially biased terms. Words such as compliance, adherence, and motivation are widely used in academic writing and federal reports. These terms might reflect a desire for an all-purpose term to represent complex phenomena, but they are problematic because they presume the superiority of an expert-identified course of treatment and situate accountability for treatment participation squarely on the shoulders of the mental health consumer (Nock & Ferriter, 2005; Staudt, 2007). Related labels, such as difficult, non-compliant, unmotivated, resistant, and hard to engage, that have penetrated the professional lexicon are even more problematic because they ignore research that shows that inconsistency or ambivalence about behavior change is a universal feature of normal behavior (Patterson & Chamberlain, 1994). Such language conveys a speaker’s internal and stable attributions about youths and families, exacerbate implicit bias against traditionally underserved populations, and inhibit action on the part of the therapist. Moving away from this language and toward language that reflects engagement as dynamic (e.g., Chu et al., 2014; Halfon et al., 2019) and transactional (Fjermestad et al., 2021; Ovenstad et al., 2020; O’Keeffe et al., 2020) will advance our thinking and hence our science. In our own work, we prefer to use the terms “engagement” to describe the general concept and “low engagement” to refer to circumstances when problems are apparent. Our strong preference, however, is to disambiguate the term “engagement” and refer to more granular and focal indicators.

Conclusion

The field owes a debt to pioneers in the field who, for 65 years, have studied patterns of attrition, analyzed interactions between clients and therapists, pursued deep understanding of therapeutic relationships and processes, examined barriers to treatment, designed interventions to improve engagement, and raised their voices about engagement and empowerment. These scholars knew that for services to help the most vulnerable populations, the field must establish favorable conditions for youths and families to make meaningful connections to therapy. These efforts paved the way for scores of others to pursue engagement research in a scientific context, which continues to lack recognition as a discrete discipline and that is disproportionately under-resourced relative to its relevance to having a robust and finished bridge between science and service.

Our reflection on this important work embodies two paradoxes. First, there have been so many critical efforts related to treatment engagement that a lengthy review to consolidate and make sense of them precluded an in-depth review of any one of them. Second, because there are already many scholars making significant contributions to the science of treatment engagement, we are at a time when we actually need even more people working in this area, and specifically, building a collaborative infrastructure for that community of scholars to work in a more cumulative and coordinated manner. The field cannot remain satisfied with the rudimentary span arising from our good investments in intervention and implementation science. Rather, to honor those investments, the field and its institutions must prioritize engagement science as well. This focus will require the same incentives that have accelerated other research initiatives, making it easier to collaborate and to advance and elevate these four integrated priorities we have outlined above. We look forward to a bridge from science to service that can work for all, even under the most taxing conditions.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported, in part, by the William T. Grant Foundation. Award #187173: Coordinated Knowledge Systems: Connecting Evidence to Action to Engage Students in School Mental Health (PI: Chorpita; Co-PI: Becker). Award # 190304: Consistently Crucial but Invariably Ignored: Testing the Role of Coordination in the Use of Research Evidence (PI: Becker; Co-PI: Chorpita).