Postpartum depression affects approximately 20% to 40% of mothers with young children (Chaudron et al., 2004; Field, 2010), with rates highest among women with low socioeconomic status (Goyal et al., 2009; Lorant et al., 2007). The impact of postpartum depression on children and families can be far-reaching, and treating postpartum depression can improve both maternal and child health outcomes (Field, 2011; Slomian et al., 2019). Reach Out and Stay Strong Essentials for mothers (ROSE), is an evidence-based intervention (EBI) that aims to prevent postpartum depression among low-income, pregnant women (Zlotnick et al., 2016). ROSE was initially tested in clinical settings among racially and ethnically diverse women, and previous research found 31% of control participants had an onset of postpartum depression at 6-months post-delivery, compared to 16% in the intervention group (this effect was maintained at 12 months post-delivery: 40% of controls vs. 26% of intervention participants) (Zlotnick et al., 2016).

Postpartum depression has been associated with trauma and life stressors (Mukherjee et al., 2017; Stone, et al., 2015), and consequently rates are believed to be high among homeless postpartum women. In the U.S., almost 60,000 families and over 100,000 children are homeless (National Alliance to End Homelessness, 2018; Peer TA Network, 2018), and homelessness is a particular challenge in urban settings. Addressing postpartum depression (PPD) among women in a large urban homeless shelter system is a critical step towards ensuring the health of families in shelter. To our knowledge, to date no evidence-based postpartum depression prevention intervention has been adapted for implementation among prenatal and postnatal women in homeless shelters. Although EBIs are considered the gold standard to be replicated in varied settings, it is unclear whether such validated interventions are as effective or sustainable in different settings or populations (Alvidrez et al., 2019; Hailemariam et al., 2019).

Adaptation Framework

There has been discussion in the literature about the tension between adhering to fidelity of EBIs and adapting EBIs to specific stakeholder groups (Castro et al., 2004), but many scholars have concluded that it is feasible and imperative to do both (Castro et al., 2004; Kemp, 2016). Adapting interventions to different populations and settings enables EBIs to address the local context, and using an implementation framework helps ensure that feedback from participants is continually incorporated into the intervention in a way that is intentional and maintains fidelity to the model (Castro et al., 2004; Chambers et al., 2013). Further, current literature suggests that initial planning and ongoing development, evaluation and refinement in real-world, diverse populations and systems are critical to an intervention’s success, and, perhaps even more importantly, to the ongoing sustainability of the intervention (Chambers et al., 2013).

Several implementation science models have been developed to adapt and implement EBIs. The Dynamic Adaptation Process (DAP) includes identifying core elements and adaptable characteristics of an EBI, and supporting implementation through training and fidelity monitoring (Aarons et al., 2012). The Exploration, Preparation, Implementation and Sustainment (EPIS) model consists of four stages (Aarons et al., 2011; Moullin et al., 2019): Exploration (assessing areas in need of attention and the best strategies/programs to meet the needs); Preparation (identifying facilitators and barriers to implementation, assessing needs for program adaptation, and developing a detailed implementation plan); Implementation (training, coaching, and active facilitating of practices); and Sustainment (maintaining the use of new programs/strategies). We chose the DAP and implemented the EPIS model within each DAP cycle because this approach allowed for a sequential yet iterative approach that integrated feedback from stakeholders at multiple points in time (Aarons, et al., 2012). However, most previous studies utilizing this approach focus mainly on the organizational stakeholders (e.g., Gunderson et al., 2021), whereas the present study aimed to center both community (participant) and organizational (facilitator) input.

In our combined process, each DAP embeds three of the four EPIS stages, and feedback loops allow the Implementation stage from one DAP to inform the Preparation stage in the next DAP; this continuous cycle allows organizational stakeholders, caregivers (community), and researchers to refine and improve programs on an ongoing basis. The fourth stage, Sustainment, was conducted after our 4 DAP cycles (Fig. 1). This approach was particularly appealing considering we aimed to adapt an EBI to a new population and setting and wanted to ensure that we included input from multiple stakeholders, maintained fidelity, and collected engagement, satisfaction and outcome data as key indicators of success (Aarons et al., 2012).

Fig. 1
figure 1

EPIS-DAP model as applied to the adaptation of ROSE into SIS

Full size image

Present Study

ROSE is a 4-8 session, manualized intervention that aims to prevent postpartum depression by increasing social support, decreasing stress and increasing effective communication. ROSE can be implemented in an individual or group format, and although it was originally developed for implementation during pregnancy, it has been implemented in the postnatal period as well. The core elements of ROSE include psychoeducation and skill-building. The psychoeducation focuses on postpartum depression, the importance of managing the stress that may come with the transition to motherhood, and accessing social support as a buffer against PPD. Skill-building elements consist of effective communication, stress management, and social support-building skills (Women and Infants, 2023). Given the high risk for PPD in homeless shelters, our specific research aim was to examine whether the EPIS framework and DAP process could be used together to include both organizational and community input in the adaptation of an EBI for implementation in a large public U.S. system while maintaining fidelity to the EBI’s core components.

Methods

Study Design and Participants

This study used a mixed methods approach, including pre- and post-intervention surveys and post-intervention qualitative feedback sessions. Staff in a large, urban, U.S. shelter system recruited participants through flyers, invitations, and information sessions among residents. Eligible women were English speaking residents of a family shelter who were 18 years of age or older, did not have an untreated mental illness (including substance abuse), and were pregnant at the time of the study (during the study, eligibility criteria were expanded to include mothers of children one year of age or younger). This research was reviewed and approved by our institution’s Institutional Review Board. Participants provided written informed consent before participating in the research. In this consent, they were informed that this research was completely voluntary, they could leave at any time, lack of participation would not impact receipt of other services, and their participation and data would be strictly confidential, with all standard protections in place. Participants were compensated 30 USD for data collection and 30 USD for program participation.

Procedures and Measures

The Exploration, Preparation and Implementation stages of EPIS were conducted within each of 4 DAP cycles (April, 2018-June, 2019; July, 2019-December, 2019; January, 2020-March, 2020; May, 2020-December, 2020). After these cycles, we used what we learned to create scalable plans in the Sustainment Stage. Each stage is described below. Figure 1 displays the EPIS/DAP process.

Exploration

Engaging partners early and continually and understanding the potential participant population were essential components to our Exploration Stage. Specifically, this stage included (1) engaging senior leadership and social work staff and (2) understanding the needs of shelter staff and shelter residents, as well as the acceptability of a PPD intervention like ROSE.

Preparation

In the Preparation Stage, adaptations were made based on findings from the exploration stage (e.g., in DAP 1) and the implementation stages from the previous DAPs (e.g., in DAPs 2, 3 and 4). These changes included adaptation to logistics and workflows as well as to the content and facilitation.

Implementation

The modifications made in the Preparation Stage were then examined in the Implementation Stage to inform the next cycle. For example, fidelity tracking is essential when adapting evidence-based interventions; fidelity to the core elements of ROSE was tracked through three primary measures: session fidelity checklists, audio recordings of sessions and weekly consultation calls. The session fidelity checklists asked facilitators to check off all key elements of the session that they were able to complete during group time. Checklist links were e-mailed to staff using an automated system (OpenREDCap) directly after group times. Sessions were recorded if all participants consented. These recordings allowed research staff to better understand how shelter staff were using the SIS manual scripts and how content was both being delivered and being received by participants. The project’s consulting clinical psychologist reviewed these tapes weekly to record fidelity notes and subsequently provide feedback on adherence to the model to group facilitators during weekly consultation calls. These calls were also an opportunity for shelter staff to provide qualitative feedback on the intervention, noting highlights and challenges they encountered.

At the end of each intervention cycle participants were asked to meet with research staff for a program debrief session in which they were asked about their experiences in the group. After each intervention cycle, staff and participants were also given a quantitative satisfaction survey. Together, the data collected in each implementation phase informed the adaptations in the Preparation phase of the subsequent DAP.

Sustainment

Based on what we learned in our first 4 cycles of implementation, we developed a plan for scalability and sustainability of our intervention in the family shelter system.

Analytic Approach

Thematic analysis was used as a method for identifying, analyzing and reporting patterns within qualitative data (Braun & Clarke, 2006). Field notes included summaries of group meetings and were reviewed and analyzed by multiple study staff with previous qualitative experience. Themes were identified through a process that allowed for consensus to be built from learnings across research staff. After several group discussions, the reviewers reached consensus and common themes emerged. Descriptive analyses were conducted to analyze the satisfaction data.

Results

Setting and Population

In this large, urban setting, the family shelter system serves adults with partners and parents with children. Overall, about one-third of participants were between the ages of 18 and 24, and half were between 25 and 34. They mostly identified as Black (47%) or Hispanic/Latina (41%), and two-thirds reported having 1 (33%) or 2 (33%) children. The majority were single (69%) and had earned a High School education (54%). Three-quarters of participants used Medicaid insurance and 60% were unemployed.

DAP 1 (April, 2018–June, 2019)

Exploration (April, 2018–July, 2018)

Shelter staff engagement began with senior leaders at three shelter sites to discuss the value of the intervention, review existing workflows and plan for the proposed implementation. Shelter social work teams were then invited to discuss logistics, modality and incentives. Shelter staff expressed preference for more tools related to management of PPD symptoms, the prevalence of a history of domestic violence among shelter residents, and an acknowledgement of challenges related to implementation (e.g. recruitment, staff time, the structure of the shelter system).

Research staff then held focus groups at each of the three sites (three groups, 18 participants total) to better understand participants’ acceptance of, and readiness for, a PPD prevention intervention. Overall, participants were very interested in the proposed intervention; they highlighted the extreme isolation they faced in shelter settings with a new baby and the important connection between maternal health and child outcomes as the main drivers of their motivation.

Preparation (August, 2018–December, 2018)

Provider and community engagement in the Exploration Stage led to adaptations that were specific to our target community. Based on staff feedback, we decided to adapt a version of ROSE that had previously been used for postnatal women within the first year of the baby’s life and to adapt the intervention name, changing it from ROSE to Strong in Shelter (SIS) to more specifically reflect the experience of the target population.

It was important to ensure that the language and corresponding visuals in both the manual and participant workbooks were at the right literacy level and would resonate with participants living in shelter. For example, we anticipated that the title of self-care activities, “Pleasant Activities,” might not resonate with women. Since these activities are a critical component in elevating mood among people at risk for depression, it was important to engage participants in this component of the intervention. We therefore renamed the activities, “Me Time.” Due to the prevalence of domestic violence experienced at some point among many women currently living in shelter, we additionally shifted the focus of the effective communication module to include any individual in a resident’s network, with the goal of reducing risk for those who remained in unsafe relationships. To address mothers’ feedback on high levels of social isolation, we incorporated an extra 30-minute window of socializing prior to group sessions, that included a meal and play-mats for the babies. Logistics discussed in the Exploration phase were solidified, including identifying staffing, developing workflows, trouble-shooting barriers, and planning for upcoming staff training and implementation.

Implementation (January, 2019–June, 2019)

We completed two intervention cycles across three shelter sites during the first Implementation Stage.

Training, recruitment and sample

In the first DAP, we trained 12 shelter social work staff and supervisors, 7 of whom became facilitators on the initial adaptation of the intervention. The trainings were 15 hours each, and included didactics regarding the risks and protective factors for postpartum depression, the aims of the research study, step-by-step teaching of the manualized intervention, and live role plays to practice administration of the intervention. The SIS training was a part of the iterative process, with shelter staff continuing to provide feedback on the adapted intervention throughout the trainings, and research staff incorporating that feedback into the intervention manual.

Research staff worked with shelter staff to engage as many eligible residents as possible, describing the program, consenting interested residents and distributing a set of baseline questionnaires. Participants were given gift cards as incentives for survey completion, and complimentary food was provided at all group gatherings. If barriers came up that prevented participants from attending group, shelter staff scheduled one-on-one make-up sessions to allow them to receive the material at a more convenient time and continue with the group the following week.

A total of 53 shelter residents were eligible for SIS. Of those, 21 (40%) consented and provided baseline data. Of the 21, 16 (76%) completed all sessions, and 17 (81%) completed at least half of the sessions across the 2 cycles of implementation.

Fidelity

Fidelity data were not available for the first cycle of this DAP, but of the 18 sessions completed in the second cycle, checklists were available for 17. Almost all (16, or 94%) of those checklists indicated that all session content was completed.

Satisfaction/Feedback

As reported in the staff satisfaction surveys, of the 7 facilitators who completed post measures all reported being quite or very satisfied with SIS on a Likert scale from 1 to 4, and indicated that SIS was helpful in enhancing their ability to help clients manage or reduce their stress. In the consultation calls, facilitators shared that they completed many individual make-up sessions to help participants complete all sessions, suggesting that this dual model was effective. Despite having completed most of the items on the fidelity session checklists, facilitators indicated that one session was too long and was difficult to implement in the allotted time. They also mentioned how challenging the relaxation exercise was with older children in the room. Further, facilitators provided feedback that many participants were not comfortable doing self-care activities without their babies, and that they had a hard time grasping the concept of “Me Time.”

Participants also completed a satisfaction survey after each group. We report on participant satisfaction data from the second cycle only, as the first cycle utilized slightly different questions and thus the two cycles could not be combined. Participants were highly satisfied with the intervention, with 7 out of 7 participants indicating that SIS helped them with the stress of a new baby, with effectively asking for help, and with identifying a support system a little or a great deal; participants responded positively to all other satisfaction questions as well.

In the debrief sessions following the intervention cycle, shelter residents consistently expressed their appreciation for the format and content of the groups. A participant recounted her experience in this way, “[The group] helped me see that I could make a different plan that included my baby.” They also reported actively using many of the skills they learned in the group. At the same time, participants noted language and format issues that did not resonate well with them.

Participants expressed a deep appreciation for the ability to meet and socialize with other new mothers. One mother relayed, “The group helped me get up out of bed and socialize.” They valued the connections they made and described further social plans outside of group, often with fellow residents that lived on their floors or even across the hall but whom they had never met. Participants expressed deep sadness when the groups ended and requested ongoing alumni/parenting groups to sustain a similar group-meeting format.

DAP 2 (July, 2019–December, 2019)

Exploration (July, 2019)

In DAP 2, we were able to leverage the exploration work done in DAP 1, given that the 2 new sites were from the same shelter system as the previous sites. In this DAP, the Exploration Stage focused on the logistics necessary to implement SIS in the new sites, with less discussion on the fit of the intervention for these particular sites.

Preparation (July, 2019–August, 2019)

As a result of feedback in the Implementation Stage of DAP 1, we made adaptations to the SIS model. We broadened our examples of self-care and relaxation exercises to include activities that could be done both “with baby” and “without baby,” both in the session and as “homework.” The section titled “Me Time” was changed to “Refill Activities,” and was accompanied by an analogy of a pitcher of water being refilled when empty. We broke one session into two sessions to be more manageable and effective, and language and format changes were made in the manual and workbook. Feedback from the consultation calls also led to the addition of session-specific “Notes to the Interventionist” in the manual to alert facilitators to important themes that might emerge during group discussions.

Implementation (September, 2019–December, 2019)

We completed two intervention cycles across 5 sites during the Implementation Stage.

Training, recruitment and sample

In DAP 2, we trained an additional 5 new shelter staff and retained 3 facilitators from DAP 1, all of whom were facilitators on the newly adapted version. The training and recruitment processes were the same as described in DAP 1. Of 70 eligible shelter residents, 31 (44%) consented and provided baseline data; of those, 18 (58%) completed all sessions and 21 (68%) completed at least half.

Fidelity

Of the 30 sessions completed in this DAP, checklists were available for 24. All of those checklists indicated that all session content was completed.

Satisfaction/Feedback

Similar to the previous DAP, all 6 staff who completed the post measures reported being very satisfied with SIS and indicated that SIS was helpful in enhancing their ability to help clients manage or reduce their stress. During consultation calls, facilitators shared that the new Refill Activities and relaxation exercise options were well-received by the participants. Facilitators additionally shared that many participants formed friendships, offered to assist one another with child-care, and arranged activities outside of the group. However, facilitators also shared that some of the instructions in the manual were confusing, and they described challenges with recruitment and retention.

Similar to DAP 1, ratings on the 15 participant satisfaction surveys were high. In the debrief sessions participants expressed valuing the format and content of the groups, and having the opportunity to practice newly learned techniques. One woman stated, “This was the first time I felt comfortable enough to disclose I had PPD with my first child.” Another recounted, “Learning to ask for help is the part that really helped lower my stress.”

Participants expressed valuing the opportunity to connect with other mothers in similar circumstances and expressed a desire to participate in more groups similar to SIS. Participants also indicated issues that they would like to see changed. One mother suggested, “When we start each session, we should refresh the previous session to let everyone get a sense of what’s going on before we move on to the next topic.”

DAP 3 (January, 2020–March, 2020)

Exploration (January, 2020)

In DAP 3, the Exploration Stage included meetings with staff from 3 additional sites, and, similar to DAP 2, leveraged the work done in previous DAPs and focused on necessary logistics. During this Stage, we learned that the new sites had larger numbers of pregnant women than previous sites.

Preparation (January, 2020)

The DAP 3 Preparation Stage was short, but important. In this Stage, we broadened the eligibility criteria to include pregnant women, and made appropriate adaptations to the intervention manual specifically for the prenatal population. Based on feedback in the previous DAP, we worked with shelter staff on ways to increase recruitment among residents, such as delivering personal invitations, and restructured the format of the manual to make it easier for facilitators to follow by adding more narrative directions, icons to indicate scripts, and a separate outline page for each session. We also added a “refresher” section to the beginning of each session to review the previous session content.

Implementation (February, 2020–March, 2020)

This Implementation Stage was disrupted in March due to the onset of the COVID-19 pandemic. We started one intervention cycle across three sites during that time period.

Training, recruitment and sample

In DAP 3, we trained an additional 3 new shelter staff and retained 5 facilitators from previous DAPs, all of whom were facilitators with the newly adapted version. Of 37 eligible shelter residents, 10 (27%) consented and provided baseline data. Not surprisingly, given the onset of the COVID-19 pandemic, only two completed all sessions and six completed at least half of the six sessions. Due to the small numbers and unusual and challenging circumstances, we are not reporting here on the implementation data from DAP 3.

DAP 4 (April, 2020–December, 2020)

In March of 2020, we understood that shelter residents were increasingly isolated and unable to access many of their usual sources of support (although they did continue to see shelter staff both virtually and in person). Offering SIS to buffer some of the environmental stressors seemed critical at this time, so we adapted the SIS intervention to a virtual format.

Exploration (April, 2020)

In this Exploration Stage, we met with senior leadership and social work staff from the 3 sites where we were previously working. Meetings centered on the logistics necessary to implement SIS in a virtual format. It was determined that participants would need to be provided with technological devices and hotspots to ensure WIFI access.

Preparation (May, 2020)

Our team quickly pivoted to be responsive to the changing social environment with adaptations that included virtual administration as well as new content that directly addressed the challenges of the pandemic. Research staff were scheduled to be available to walk participants through the technology prior to the group, and staff planned to sit in on the groups to trouble shoot any technological issues that arose. Modifications were also applied to the manual, workbook and session content, as well as staff training. For example, COVID-19 and the accompanying social isolation, was specifically discussed as a stressor, and “Refill Activities” were updated to include virtual activities.

Implementation (June, 2020–December, 2020)

We completed two intervention cycles across 7 sites in this Implementation Stage.

Training, recruitment and sample

In DAP 4, we trained 13 new shelter staff and retained 4 facilitators from DAP 2 and DAP 3, 13 of whom were facilitators of the SIS virtual intervention via the WebEx platform. During this time, 69 shelter residents were eligible for SIS; about one-third (26 or 37%) of those residents provided baseline data, and of those, 21 (81%) completed all sessions and 22 (85%) completed at least half .

The virtual platform allowed us to further innovate, in some cases having facilitators from two different shelters co-lead a group consisting of participants from both shelters. Research staff consented participants and collected baseline and follow-up data virtually, and helped participants with the technology.

Fidelity

Data were available for all 42 sessions completed during this DAP; checklists for 34 sessions (81%) indicated that all session content was completed. Through these session checklists, we determined that some facilitators did not have time to complete all session content due to participants’ technical issues with devices.

Satisfaction/Feedback

Staff and participant satisfaction in the virtual groups was mostly consistent with the in-person versions of SIS. All 10 staff surveyed indicated that they were somewhat, quite or very satisfied with SIS, and 9 of 10 respondents indicated that SIS was helpful in enhancing their ability to help clients manage or reduce their stress. In consultation calls, facilitators reported challenges in assisting participants with technology issues and ensuring participants were in a private space.

Ratings on the 21 participant surveys were high, although slightly lower than in previous DAPs. In the participant debrief sessions, some participants expressed a preference for meeting in person, but others appreciated the convenience of the virtual format. Many indicated that the group helped decrease the isolation they increasingly felt during the pandemic. One mother indicated that the best part of the group was “getting to talk with other moms and feeling less alone.”

Importantly, the virtual format worked considerably better for women who were in shelter living only with their one infant. For women who were living with partners or older children, it was extremely challenging to find private space within their individual unit to engage in the intervention. That said, several participants in the virtual cycles requested further forms of virtual communication with one another once the group ended. In one mother’s words, “It helped me a whole lot, I wish we could keep going with the group. It distracted my mind from a lot of stuff, from corona stuff, everything that’s going on because of corona.”

Sustainment

Based on knowledge gained over the 3 DAP cycles, we have worked with the shelter system to develop a plan to scale and sustain SIS throughout the shelters. Our current plan is to train staff from across the system 3 times per year, and to have our team psychologist be available upon request; the prenatal and postnatal manuals and workbooks have been translated into Spanish to expand access for those with limited English proficiency living in shelters. This plan is in early stages, and we will monitor our progress and make modifications as necessary.

Discussion

In this study, we adapted an evidence-based intervention to a new setting and population, focusing on the inclusion of both organization and community voices, the use of an iterative process, and maintenance to model fidelity. Combining the EPIS and DAP models provided a pathway to focus intensively on planning, engage in intentional, continuous and iterative implementation, and learn from data collection and systematic feedback. We applied this model to both the organizational context, including staff and facilitators in the process, and the community context, seeking and incorporating participant feedback into the adaptation.

Organizational Context

Previous literature suggests that contextual factors of the organization are important to determine the optimal fit of an intervention in a particular setting (Chambers et al., 2013). Specifically, organizational stakeholder engagement has been shown to be important to the adoption, implementation and sustainability of interventions (Holcomb et al., 2022). While several frameworks (including the EPIS and DAP) have been developed for this purpose, there is limited literature detailing all stages of the adaptation and sustainment processes in the real world (Moullin et al., 2019; Hailemariam et al., 2019). Our study confirmed the importance of including organizational stakeholders in identifying appropriate implementation strategies, and outlined the steps to do so.

Research suggests that organizational readiness is critical to the adoption and implementation of EBIs and can vary among different people and teams (Weiner, 2020). Our process allowed research staff to assess readiness among different organizational stakeholders, address concerns as they arose, and design an intervention that resonated with shelter staff and administrators. Relationship-building with all levels of staff also sent a message that staff voices were valued. The resulting early buy-in led to committed and engaged stakeholders who worked with research staff to continually improve the intervention and further engage participants, supporting fidelity to the core components of the model.

Managing how adaptations fit into the local context without jeopardizing the core components is essential to the effectiveness of an evidence-based model (von Thiele Schwarz et al., 2021). Importantly, our framework monitored implementation closely and our measures indicated high fidelity, suggesting that our adaptations did not stray from the core components of ROSE. Notably, fidelity to the model was slightly lower in the virtual format, due to technology constraints.

Community Context

Most adaptation frameworks in the extant literature do not focus on the inclusion of participants or community members, and there has been a call to broaden engagement efforts to be more inclusive (Triplett et al., 2022). The importance of involving community members in the adaptation process has been well described in the cultural adaptation literature (Burlew et al., 2020). Some benefits of this approach include reaching and engaging the audience, reinforcing messages, increasing fit, and responding to contextual barriers (Carvalho et al., 2013). Similar to our approach, Chen et al. (2013) suggest that the entire intervention, not just discreet parts, must be implemented before seeking community feedback. However, most of the literature does not describe this approach as an iterative process (e.g. Chelbowski et al., 2020). Our study adds to the research base by utilizing an iterative approach, EPIS and DAP, and extending that to include deep community input.

This iterative approach is essential because relying on a linear approach assumes that an intervention that is “one size fits all” will perform equally in all settings and populations, which does not account for the complexities of real-world contexts (Chambers et al., 2013). The iterative feedback loop utilized here enabled us to continually adapt the intervention in real-time in ways that reflected the reality of the population being served. In each DAP, additional adaptations were made based data from the previous DAP, including feedback on the ease of implementation, the appropriateness of the language, and the relevance of the format. Without these data, and the opportunity to revise the intervention based on these data, staff and participants may not have engaged so highly with SIS.

As an example, self-care is an important part of the SIS model – engaging in fun, meaningful activities can help boost mood and is a significant protective factor against depression (Lin et al., 2018). However, the terminology we initially used, and the examples we provided, did not resonate with participants. The feedback loop in the EPIS/DAP model enabled us to learn about these disconnects (Implementation Stage DAP 1) and modify the intervention (Preparation Stage DAP 2) to better reflect the population and setting while maintaining fidelity to this core intervention element. Data collected in the Implementation Stage of DAP 2 indicated that our adaptation resonated more with our participants. Without this process, an important part of the intervention would have been implemented on paper, but would not have had the desired effect in this population.

Recruitment to SIS was challenging. That said, once shelter residents did engage in SIS, their participation throughout the full intervention was fairly consistent (except for the DAP interrupted by COVID-19) and they consistently reported that the components of the SIS intervention were valuable and relevant to issues they were facing, suggesting that our focus on community input supported engagement. Analysis of satisfaction data had similar findings, with nearly of all women who participated in SIS noting the intervention was helpful in some way. As with fidelity, satisfaction scores were slightly lower in the virtual sessions, suggesting that more work is needed to optimize that format.

Sustainment

The literature suggests that adaptation to local context is essential to ensuring that an intervention can be sustained over time (Chambers et al., 2013). However, the iterative adaptation work described above was time-intensive and generously supported by a private philanthropic institution. Going forward, we have minimized the implementation support required by research staff so that SIS can be scaled broadly throughout the family shelter system, enabling all women in shelter to benefit from the intervention. In order for this effort to be truly sustainable, though, financial support for facilitator training and consultation must come from the agency where implementation is occurring. The system continues to explore opportunities to take on that financial role going forward.

Notably, we found consistent trends in acceptability in the virtual groups, despite slightly lower satisfaction ratings among staff and participants, suggesting that the intervention is nimble and able to be modified to different situations. This may be important to future rollouts of SIS, as conducting virtual groups allows more flexibility in the implementation of groups across shelters. A cross-site model, with facilitators from more than one site, may allow both an increase in staff efficiency and a broader reach to more shelter residents, even post-pandemic. This model may have broader implications for other interventions among shelter residents and in other hard-to-reach populations.

Limitations and Future Research

The SIS adaptation was limited by a relatively small number of sites in a very large system. However, we identified consistent themes across most of the shelters, which enhanced our confidence in the findings. Given the high number of Spanish-speakers in the shelter system, the fact that SIS was only available in English limited our reach. SIS has since been translated into Spanish; the original ROSE has been successfully implemented with the Latinx population (Zlotnick et al., 2016), and we are looking forward to rolling out the SIS Spanish-language versions throughout the system. Future research should examine how well this version is able to be implemented with fidelity to the model, and whether further adaptations are necessary in the Latinx population in shelters.

While SIS adaptation and implementation have been successful, this study did not illuminate how well SIS can be scaled across the system. We are currently working with the shelter system to test out a process of training and supporting staff throughout the system to deliver SIS to all eligible women living in shelter. Further examination of the implementation of this model at full scale, including how fidelity to the model is maintained, how cost-shifting to the local agency works, and how valued virtual sessions are, will be needed. Research should also examine mechanisms to enhance technology literacy and thus participants’ abilities to take part in virtual group interventions throughout the system.

Finally, this implementation did have challenges with recruitment. The field would benefit from additional research on strategies to better recruit and retain participants in group formats.

Conclusions

The specific processes and methodologies used to adapt an evidence-based intervention for new populations or settings are critical to the ultimate success of that EBI in the new context. In this paper, we illustrated how the combined EPIS-DAP iterative framework can be used to not only address the fit of the intervention in the organizational context, but also extend this approach to involve deep community participants’ input. Our data suggest that by engaging staff and the population to be served early and continually, and utilizing an iterative and flexible process of adaptation, evidence-based interventions such as ROSE can implemented with fidelity and accepted by both organizations and participants in new settings.