An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
S. Darius Tandon, PhD, Jessica K. Johnson, MPH, Alicia Diebold, MSW, Melissa Segovia, MS, Aria Degillio, BA, Jackie Gollan, PhD, Dana Zakieh, BA, BS, Jesus Solano-Martinez, BA, Chen Yeh, MS, and Jody D. Ciolino, PhD.
Author Information and AffiliationsPostpartum depression (PPD) is a common mental health concern that has well-documented negative effects on maternal and child health and disproportionately affects women with low income. Although efficacious interventions exist for preventing onset and worsening of depression among perinatal women, no studies have examined the efficacy of paraprofessionals in delivering a PPD preventive intervention to women with low income.
This study sought to determine whether pregnant women receiving the Mothers and Babies (MB) group-based intervention exhibited greater reductions in depressive symptoms than women receiving usual community-based services, and to examine whether the MB intervention delivered by paraprofessional home visitors would yield similar reductions in depressive symptoms to the MB intervention provided by mental health professionals (MHPs).
A cluster randomized trial was conducted in which 37 home visiting (HV) programs across 7 states were randomized to usual HV services, MB delivered by an HV paraprofessional (HVP), or MB delivered by an MHP. Clusters were defined and randomized at the HV program level whereby an individual HV program served as the unit of randomization. Depressive symptom score at 24 weeks postpartum was the primary outcome, controlling for baseline symptoms, assessed using the Quick Inventory of Depressive Symptomatology (QIDS). Baseline assessments were conducted during pregnancy after study enrollment, with follow-up assessments extending to 24 weeks postpartum. A total of 1316 women were referred for assessment, of whom 874 enrolled. Eligibility criteria were aged ≥16 years, ≤33 weeks of gestation on referral, and Spanish or English speaking. Participants' mean age was 26.3 ± 5.8 years, 70% belonged to a minority racial/ethnic group, and 71% had incomes <$25 000/year. Semistructured interviews with 46 facilitators and 88 intervention participants assessed intervention feasibility and acceptability. We conducted fidelity analysis to examine facilitator adherence and competency. We excluded any participants from analyses who did not contribute any data after the initial baseline assessment, and we analyzed all participants according to the arm to which their site was randomized.
A total of 824 women (94%) contributed data toward our primary analyses. Among women in the 2 active intervention arms, 53% received ≥4 intervention sessions. QIDS scores dropped from a mean 8.0 ± 4.2 points to 5.7 ± 4.5 points overall (scale range, 0-27 points), aggregated across all study arms. While the HVP arm saw the largest drop in mean QIDS score from baseline (8.6 ± 4.3) to 24 weeks postpartum (5.9 ± 4.5), all arms showed an overall mean decrease in primary outcome, with the control arm ending at 5.8 ± 4.6 and the MHP arm ending at 5.3 ± 4.5 at 24 weeks postpartum. Overall, the 95% CI for primary outcome was 7.69 to 8.28 at baseline, 7.03 to 7.60 immediately postintervention, 5.65 to 6.31 at 12 weeks, and 5.32 to 6.00 at 24 weeks postpartum. Thus, we could not claim that either of the intervention arms showed superiority in decreasing depressive symptom scores compared with the control arm (P = .393 when comparing HVP vs control; P = .406 when comparing MHP vs control) at 24 weeks.
However, we have evidence of noninferiority between MB delivered by an MHP vs an HVP. Our model-estimated mean difference in QIDS scores between intervention arms equaled 0.01 points (95% CI, −0.78 to 0.79), and the lower confidence limit remains above our prespecified margin of noninferiority (2 points). The interview data suggest that clients and facilitators found the MB content and group format acceptable, with no significant differences between intervention arms. Challenges included maintaining group attendance, addressing transportation issues, and managing group discussion. Facilitators found the intervention appropriate for pregnant clients, with some challenges presented for clients in crisis situations, experiencing housing instability, and with literacy and learning challenges. Fidelity analyses found no significant differences between HVPs and MHPs in facilitator adherence or competency.
Although analyses from our superiority aim did not find a statistically significant difference between the 2 intervention arms and our control arm, our noninferiority analyses did find that HVPs using MB generated similar reductions in depressive symptoms as MHPs. Additionally, there were no significant differences in fidelity of implementation between HVPs and MHPs, suggesting that lay health professionals may be a viable approach to delivering PPD preventive interventions in community-based settings like HV.
There may have been a potential flooring effect in which there was less room to demonstrate improvement in symptom reduction given the lower-than-anticipated baseline depressive symptom scores of study participants. Among facilitators who delivered >1 MB cohort, interviews were conducted after the final cohort, potentially creating challenges in remembering experiences from earlier cohorts.
Postpartum depression (PPD) is a serious mental health disorder that poses significant health and mental health risks for mothers and their infants.1,2 The prevalence of PPD has been estimated at 10% to 22% and disproportionately affects women with low income of all racial and ethnic groups, with a 20% prevalence rate consistently found among women with low income.1,3-5 Disparities also exist among women with low income exhibiting clinically relevant depressive symptoms but not meeting Diagnostic and Statistical Manual for Mental Disorders (DSM) criteria for major depression. Studies indicate that rates of depressive symptoms among women with low income are 30% to 45%,6,7 double the rates of women of higher socioeconomic status (SES); these disparities exist irrespective of race, ethnicity, or geography.8-10 A growing body of research has documented that depressive symptoms are associated with many of the same negative maternal and child health outcomes as major depression.11 For example, women with depressive symptoms have diminished capacity for sensitive parenting practices, exhibit poorer quality of life (QOL), have increased use of medical services, and exhibit increased mortality rates.11-13
Systematic reviews have highlighted an array of efficacious interventions that reduce the risk of developing PPD.14,15 Among those interventions that have demonstrated efficacy, the majority use health (eg, nurses, midwives) or mental health (eg, psychologists) professionals to deliver individualized or group-based interventions (see Curry et al15 for a review). One exception is the use of peers to deliver peer support via phone,16 although this study was conducted in Canada with predominately White, upper- and middle-class women. As such, there are no interventions led by non-health or non-mental health professionals that have demonstrated efficacy in preventing the onset of PPD and reduction of depressive symptoms among women with low income.
One intervention that has demonstrated efficacy in both reducing depressive symptoms and preventing onset of new cases of PPD is the Mothers and Babies (MB) course.17 MB is a group-based, manualized, cognitive-behavioral intervention that promotes healthy mood management by teaching individuals ways to increase the frequency of thoughts and behaviors that lead to positive mood states. The group-based MB intervention consists of 6 sessions, each lasting about 90 minutes. A central characteristic of MB is its “toolbox” concept whereby participants learn several different skills that help them cope with stress and depressive symptoms. These skills fall into 3 domains: (1) behavioral activation to increase activities that generate enjoyment, (2) restructuring negative cognitions, and (3) increasing social support. The content of MB is tailored to specific needs and issues related to the pregnancy and postpartum periods. A series of randomized controlled trials (RCTs) has demonstrated that MB reduces the incidence of new cases of major depression, reduces depressive symptoms, and improves mood management among perinatal women receiving MB compared with women receiving usual care when MB is led by mental health professionals (MHPs),17-22 including 2 studies in which perinatal women were recruited from home visiting (HV) programs.
A growing number of studies have shown evidence that nonmental health specialists can deliver maternal mental health interventions, particularly in low- and middle-income countries.23 The World Health Organization's Thinking Healthy Programme is the most widely tested of these interventions, with multiple RCTs indicating that trained community health workers (CHWs) delivering cognitive-behavioral therapy (CBT) effectively treats depression among perinatal women with low income in Pakistan,24 India,25 and Vietnam.26 Studies conducted in higher-income countries have also employed a variety of non-mental health professionals (eg, midwives, nurses, health visitors) to treat perinatal depression, including those who have implemented collaborative care models augmenting MHPs with supports provided by primary care physicians, nurse practitioners, or other non-mental health professionals.27 However, we are unaware of published reports describing the use of non-mental health professionals in programs aimed at the prevention of perinatal depression.
HV programs exist in all 50 states28 and typically enroll women prenatally, with services continuing until a child reaches 2 to 4 years of age, depending on the HV model. Core content of HV models typically address (1) preparation for childbirth and having a young child in the home, (2) provision of emotional and tangible (eg, diapers) support, (3) discussion of infant and young child development, (4) linkages to prenatal and pediatric care, and (5) referrals to community resources for social and health services. Many HV models exist, with a recent report highlighting 21 “evidence-based” models, that have shown favorable impacts on ≥1 maternal and child health outcomes using rigorous research designs.29 Several evidence-based HV models, including those that serve the largest numbers of families (eg, Healthy Families America, Parents as Teachers, and Early Head Start), use home visit paraprofessionals (HVPs). Analogous to CHWs, HVPs typically do not have education beyond a bachelor's degree and have not received advanced training in the area of mental health.
Research conducted by Dr Tandon, as well as other research teams, has found rates of clinically elevated depressive symptoms between 45% and 50%30-33 among women with low income in HV programs; these rates are similar to, or exceed, rates of depressive symptoms among women with low income not enrolled in HV. Elevated depressive symptoms are among the strongest predictors of developing PPD. HV programs also enroll women for services who exhibit many demographic characteristics that place them at risk for developing PPD, including low SES, poor social support, and limited partner support. As such, HV programs are a unique and viable setting for delivering interventions aimed at preventing the worsening of depressive symptoms and onset of PPD. However, HV programs typically do not have resources to use MHPs to deliver enhancements to their core program services, thereby limiting the potential scale by which the MB intervention can be used.
Building on our previous work that established MB's effectiveness when delivered by MHPs, as well as literature demonstrating the effectiveness of CHWs to deliver depression treatment interventions in low- and middle-income countries, this study aimed to examine the use of HVPs in delivering the MB group intervention. This study had 4 specific aims:
This project benefited from active engagement of patients and other stakeholders. We created 3 separate entities to obtain patient and stakeholder feedback throughout the duration of the project. First, we created an operations team that consisted of 1 state government official, 1 MHP, 2 HV program managers, 2 HV clients affected by PPD, 2 group facilitators (1 HVP and 1 MHP), and 1 patient advocate. This group met roughly quarterly via phone throughout the project and provided guidance on various aspects of project design, implementation, and dissemination of findings. We created an executive committee that included 3 of the stakeholders from the operations team (Ms Lesley Schwartz, Illinois Governor's Office of Early Childhood Development; Ms Linda Delimata, Illinois Children's Mental Health Partnership; and Ms Sara Barrera, HV manager at Advocate Illinois Masonic Medical Center), with whom we met roughly every 2 to 3 weeks via phone during the project period. The executive committee worked closely with study investigators and research staff to ensure successful and timely achievement of project milestones. In addition to the regularly scheduled phone meetings with the operations team and executive committee, study investigators and project staff maintained ongoing communication with members of both groups to elicit advice on aspects of project implementation. We also created a national advisory committee consisting of PPD experts, HV experts, policy experts, and patient advocates. The national advisory committee met yearly and focused largely on identifying emerging opportunities to disseminate findings to researchers, policy makers, and patient advocacy groups who can facilitate the implementation and sustainability of MB in other HV programs nationally. Below, we describe a few key areas in which our operations team, executive committee, and national advisory committee influenced our study design, study implementation, and dissemination of research results.
This project was initiated not by the Northwestern University (NU) research team, but by state-level officials from the Illinois Governor's Office of Early Childhood Development—including 1 of the members of our operations team and executive committee, Lesley Schwartz—who contacted Dr Tandon about implementing MB across HV programs based on a webinar presentation on MB conducted by Dr Tandon. The Illinois Governor's Office of Early Childhood Development viewed MB as a crucial resource for HV programs that struggle to address maternal depression among their clients. Before this PCORI project, Dr Tandon worked closely with Ms Schwartz to develop a strategy for building the capacity of Illinois HV programs to implement MB. Many HV programs are required to have a group component to their program services, and programs viewed a group intervention focused on maternal depression as a highly relevant group activity for their clients. Moreover, HV programs and HV clients expressed a desire to have a low-cost intervention that can be delivered by home visitors who are already employed by their agencies, given the numerous barriers (eg, stigma, wait lists, emphasis on pharmacological treatment) to clients accessing mental health services via community mental health or primary care providers.
This application sought and received feedback from an array of patients and stakeholders in its development. Dr Tandon worked extensively with Ms Schwartz and Ms Delimata during the planning of this application. Via regular phone calls, emails, and in-person meetings, Dr Tandon, Ms Schwartz, and Ms Delimata developed a research plan that aligned with the interests of patients and stakeholders. Academic and community partners also held a series of in-person and phone conversations with HV managers, HV staff, and HV clients throughout Illinois to develop several key aspects of this project. We received extensive input from a community advisory board of perinatal patients and stakeholders with whom Dr Tandon and colleagues had been collaborating with on an NIH grant aimed at addressing perinatal depression, substance use, and domestic violence in HV programs. HV clients who were engaged in developing this project had low income and were racially and ethnically diverse, thereby reflecting the demographics of the women we anticipated recruiting for this study. The input from stakeholders in developing this project centered on 4 areas:
Early in the project period, all 3 entities (operations team, executive committee, national advisory committee) provided feedback on survey length to minimize participant burden. Specifically, we were able to shorten the assessment tools developed by our research staff at NU (eg, demographic questions, questions about use of MB skills) while maintaining the original validated proposed assessment tools. Another item discussed with stakeholders was the potential barrier of participants being able to complete their surveys. One suggestion that was put into practice during the study was offering computers and computer support at the HV programs to ensure the participants had access to the study assessments. The executive committee also provided guidance on finalizing our study protocol for addressing participants who are experiencing suicidal ideation when completing study assessments.
As participant recruitment and study implementation commenced, stakeholders provided valuable insights related to strategies for conducting successful trainings for MHPs and HVPs who facilitated MB groups, ongoing supervision for these MHPs and HVPs, and communicating key study information to participating program sites. The operations team and executive committee helped develop “crisis” language to distribute to MHPs and HVPs via the project listserv. Stakeholders suggested providing a document to facilitators that outlined the do's and don'ts of handling crisis issues that may be brought up in group sessions by participants and suggested creating a standardized process for handling crisis issues and reporting on them to the NU research team. This resulted in a collaboration between the NU research team and executive committee to develop a crisis protocol, which was shared with all participating sites.
Stakeholders also engaged with the NU research team in discussing recruitment challenges and potential remedies to these challenges. Several strategies were suggested by members of our operations team, including many HV clients who provided keen insights into how best to encourage study participation. Stakeholders suggested peer-to-peer sharing of participant experiences with MB, enhancing and increasing the use of patient testimonials and participant feedback, and asking current study programs with larger caseloads to complete additional group cohorts. The research team incorporated these suggestions and was able to work with several HV programs that completed more than the original 6 cohorts that were discussed when they were recruiting for the study. We also included patient quotes within newsletters and correspondence with stakeholders and participating program sites via our study listserv, and we uploaded patient testimonial videos to our MB website.
The operations team and executive committee helped design and suggest content for separate newsletters that were sent to (1) project stakeholders (eg, HV programs participating in the study, MHPs, and HVPs delivering the MB intervention) and (2) study participants. Our executive committee and national advisory committee, in particular, have played central roles in suggesting venues for disseminating our research findings to maximize this study's impact on researchers, policy makers, practitioners, and patient advocates. As an example, Dr Tandon presented on our interim findings at the Maternal, Infant, and Early Childhood Home Visiting (MIECHV) All-Grantee meeting in February 2019. Dr Tandon co-presented with a member of our executive committee, Ms Schwartz. The MIECHV All-Grantee meeting is attended by state agency leadership from across the United States who typically have oversight for decisions involving the adoption of an intervention like MB into their network of HV programs. Thus, this venue allowed Dr Tandon and Ms Schwartz to share findings pertaining to the implementation of MB, emerging results related to HVPs' adherence and competency in delivering the intervention, and a framework for integrating MB into HV programs using examples from this PCORI-funded study.
Our operations team and executive committee have also played a significant role in brainstorming strategies to promote sustainment of MB at the HV programs involved in this study, as well as to promote the adoption of MB at new HV programs across the United States. Ideas generated by these groups include (1) adding money to state agency budgets as a line item to support MB implementation, (2) developing an outline of costs associated with implementation of MB that can be posted on our MB website, and (3) connecting with the national program offices of the largest HV programs to seek opportunities to share findings with their networks across the United States. Related to this last point, Dr Tandon, Ms Schwartz, and 1 of the HVPs who delivered MB during this project conducted a pre-conference workshop at the National Parents as Teachers Conference in October 2019; Parents as Teachers is one of the largest national HV program models that uses HVPs. Also based on feedback from our operations team and executive committee, the NU research team hosted a webinar for all project stakeholders to thank them for their study participation, review project accomplishments, and review funding suggestions to promote sustainment of MB at their agencies.
Our operations team, and to a greater extent, our executive committee, have also been involved in the development of scientific abstracts and peer-reviewed publications. We have conducted presentations at local, regional, and national meetings and have integrated feedback from our project stakeholders on content that should be prioritized during those presentations. Our executive committee members have contributed to the development and refinement of the focus areas for the peer-reviewed manuscripts that will be published based on this project. For example, stakeholders have provided useful feedback on ways in which we could frame certain manuscripts to ensure a wider readership that extends to HVPs and policy makers. Our operations team has played a central role in reviewing themes emerging from our qualitative data with facilitators and clients, which has formed the basis of an article by Diebold et al.67 Members of our executive committee and operations team are always included as authors or mentioned in the Acknowledgment sections of all presentations and publications.
We conducted a cluster randomized trial with superiority and noninferiority aims that randomized 45 HV programs to 1 of 3 arms using a 1:3:3 allocation ratio. For every 1 control site, we randomized 3 sites delivering MB via MHPs and 3 sites delivering MB via HVPs, using covariate-constrained randomization techniques34,35 to control imbalance on the following HV program-level variables: (1) yearly client volume, (2) population density of geographic area served by program, (3) and percentage of racially/ethnically diverse clients. HV programs were located throughout 7 states in the Midwest region of the United States: Iowa, Illinois, Michigan, Minnesota, Missouri, Ohio, and West Virginia. HV programs in these states were funded through a combination of federal and state funding streams. Federal funding comes via the Health Resources and Services Administration's MIECHV program, while state funding emanated from various state agencies (eg, the Department of Health). Although eligibility for HV services varies, the vast majority of models and states have criteria specifying that clients must fall at or below 100% of the federal poverty guidelines. Additional eligibility criteria that are common across HV models include adverse childhood experiences, past or current child welfare involvement, and social isolation.
HV programs recruited participants enrolled at their agency, with some programs occasionally recruiting from their surrounding community. Potential participants were given a brief overview of the research project and referred to the research team, who subsequently contacted participants to ensure eligibility criteria were met—specifically that participants were aged ≥16 years, ≤33 weeks of gestation on referral, and Spanish or English speaking. Participants unreachable via phone, email, text, or social media received a recruitment letter.
The principal investigator (PI) led 1- to 1.5-day trainings for 69 MHPs or HVPs. Of these, 53 delivered the MB intervention as part of this project.
The MB group intervention consists of 6 sessions (ie, 1 “cohort”). Sessions 1 and 2 provide an introduction to the importance of managing one's mood and CBT content related to pleasant activities. Sessions 3 and 4 focus on thoughts, while sessions 5 and 6 focus on contact with others and course review. Sessions within a cohort were delivered on a weekly basis, with some exceptions due to holidays or weather. The study involved a total of 132 MB cohorts across intervention sites, randomized to either the MHP or HVP intervention arm between January 2017 and October 2018. Sessions were delivered at HV sites or at community locations (eg, library, school) on a day and time that was convenient for the participants and facilitators. Cohorts had an average of 3 participants, and sessions were, on average, 86 minutes long.
Participants in the MHP-led arm received MB delivered by an MHP trained on MB. MHPs were recruited from HV programs (if an existing staff member or consultant met eligibility criteria) or through state professional organizations when HV programs did not have staff meeting MHP facilitator criteria. The research team provided assistance in finding someone who met the qualifications for an MHP to facilitate groups if HV programs had difficulty finding someone. As a result, some of the MHPs were new to the program, whereas some already worked there and/or had an existing relationship. An MHP was required to have at least 5 years of experience working with families and children, and at least a master's degree in a mental health-related field (eg, psychology, social work). Participants in the HVP-led arm received MB led by an HVP employed at the participating HV program; we defined an HVP as someone with no more than a bachelor's degree in child development, mental health, or a related field. A total of 69 individuals received training on MB, 53 of whom delivered the intervention (32 HVPs, 21 MHPs). Training received by these 53 individuals was typically done via an in-person 1- or 1.5-day training conducted by the PI (n = 37), with some facilitators receiving training via a live webinar (n = 6), review of a recorded webinar along with a follow-up call with the PI (n = 5), or a follow-up call with the PI in addition to having previously received in-person training from the PI before this study (n = 5). The content and length of training were identical irrespective of training modality. To promote intervention fidelity, facilitators received supervision from the PI while implementing their first cohort.
Participants in the usual care arm received usual HV services but no MB intervention. As noted earlier, there are many HV models, with each model using a slightly different curriculum and set of procedures for working with perinatal women. In this study, participants in the usual care arm received HV services from programs that used 3 different models. Despite the differences in model types, all models call for weekly or biweekly visits with clients during pregnancy delivered in the client's home. During visits, home visitors focus on a variety of issues, including preparation for childbirth and having a young child in the home, provision of emotional and tangible support, discussion of infant and young child development, linkages to prenatal and pediatric care (if needed), and referrals to community resources for social and health services (if needed). Most HV programs screen for depression and refer women with elevated symptom levels to external mental health or primary care providers. Home visitors typically do not discuss depressive symptoms in the context of a home visit but, depending on the HV model, they may discuss strategies to help clients cope with acute and chronic stressors.
All study activities were approved by NU's IRB (STU00203761). All participants consented to take part in the research study.
On receiving a participant referral, we created a unique record in Research Electronic Data Capture (REDCap), a secure web application for managing online surveys and research data.36 Eligible and interested participants provided informed consent for study participation via REDCap, or by phone with a research assistant (RA). All participants provided informed consent before baseline survey completion. Participant data were collected using self-report surveys at baseline, 1 week postintervention (or 8 weeks after enrollment for controls), and 12 and 24 weeks postpartum. All data were collected via REDCap or by phone in English or Spanish. To increase retention, RAs gathered as much contact information as possible during the screening process, including phone number, email address, Facebook name, mailing address, and a secondary contact person. RAs had a minimum of monthly contact with study participants to ask for updated contact information and to remind participants when their next survey would be coming up. To minimize the instances of missing responses, RAs contacted participants to complete missing survey questions by phone. A total of 88 intervention clients were randomly selected from the intervention arms (45 in the HVP arm and 43 in the MHP arm) to participate in a semistructured interview that focused on obtaining information pertaining to intervention acceptability and feasibility (see Appendix A). Interviews were conducted within 1 month of a participant completing the MB intervention.
All 53 facilitators who implemented the intervention to at least 1 participant cohort were approached to participate in a semistructured interview focused on intervention acceptability and feasibility (see Appendix B), with 46 facilitators (27 HVPs and 19 MHPs) completing interviews. Two facilitators declined to complete an interview and 5 were unresponsive to outreach. There were no demographic differences between facilitators completing interviews and those who declined participation. Facilitators led anywhere from 1 to 12 cohorts (mean = 2.5). All facilitators consented to take part in the study via an online consent form.
The 16-item Quick Inventory of Depressive Symptomatology-Self-Report (QIDS-SR16)37,38 was used to assess severity of depressive symptoms consistent with DSM symptom criteria. The QIDS-SR16 has excellent internal reliability for outpatient samples and established construct validity.37,38 Total scores range from 0 to 27; higher scores indicate greater symptomatology. Mild depression is interpreted as falling between 6 and 10, with moderate depression ranging from 11 to 15 and severe depression ≥16.
Current and past major depressive episodes (MDEs) were assessed using the Maternal Mood Screener (MMS),39 a 9-item self-report checklist supported by previous research as a valid instrument to assess MDE using DSM criteria.40 Individuals met the criteria for MDE if they endorsed ≥5 of the 9 items including either symptoms 1 (feeling depressed) or 2 (losing interest or pleasure), which are present for at least 2 weeks, and endorsement of the question, “Did these problems interfere with your life or activities a lot?”
The Behavioral Activation Depression Scale (BADS)41,42 assesses changes in behaviors believed to underlie depressive symptoms and has demonstrated strong internal consistency, construct validity, and predictive validity.41,42 The BADS consists of 25 items, each rated on a 7-point scale, with higher scores indicating greater behavioral activation. For our analyses, BADS items were averaged to create a mean score ranging from 0 to 6.
The Medical Outcomes Study Social Support Survey (MOS-SSS)43 is a 19-item self-administered survey that includes an overall functional social support index, with greater scores indicating more perceived social support. The MOS-SSS has strong psychometric properties and sensitivity to change over time among parents with low income and culturally and linguistically diverse parents of young children.43 For the purposes of our analyses, we created a mean MOS-SSS score ranging from 1 to 5.
The 30-item Negative Mood Regulation Scale (NMRS)44 assesses mood regulation. The NMRS has demonstrated excellent internal consistency, good test-retest reliability, and correlational evidence of both convergent and discriminant validity with constructs such as depressive symptoms and locus of control, respectively.44 For each item, respondents indicate what they believe they can do to mitigate disappointment or a negative mood, with higher scores indicating greater ability to regulate one's mood. For our analyses, NMRS responses were averaged to create a mean score ranging from 1 to 5.
The Experiences Questionnaire (EQ)45 is a 20-item instrument designed to measure decentering of thoughts and rumination, which are key constructs taught via cognitive restructuring techniques. Specifically, decentering refers to the realization that thoughts are temporary events that may not be true representations of one's experiences. The EQ has demonstrated strong internal consistency in a number of studies examining effects of interventions that incorporate cognitive restructuring techniques.45 For our analyses, we created a mean EQ score that ranged from 1 to 5.
Semistructured interviews were conducted by phone by an NU research team member. We chose a semistructured interview format to be able to follow up on responses given by a respondent. The questions on the interview guide were created to elicit information that would align with both the study aims and elements of focus within the project (eg, MB training, implementation, supervision, effectiveness and acceptability, self-efficacy, group satisfaction, challenges/improvements). Separate semistructured interview guides were created for the client and facilitator interviews, asking about their experiences with MB. Facilitator interviews were conducted in English. Client interviews were conducted in the participant's preferred language (English or Spanish); Spanish interviews were conducted by a bilingual team member. All interviews were audio-recorded, uploaded to a secure shared drive, and professionally transcribed. Interviews conducted in Spanish were translated into English during the transcription process. Consent to be contacted for and participate in an interview was obtained during the initial consenting process for the study for both facilitators and clients. Each client and facilitator was given an identification number used for saving and storing interview recordings and transcripts. The key for these identification numbers was housed in REDCap.
Clients were contacted for an interview within a week of completing the intervention, and interviews were scheduled within the next month to ensure a more accurate recollection of their experience with the intervention. If a client could not be reached within that time frame, another client was selected to participate in an interview. Clients received a $20 gift card for participating in an interview. Facilitator interviews were scheduled, on average, within 3 months after completion of the implementation of groups. Client and facilitator interviews were conducted between April 2017 and October 2018.
The Revised Cognitive Therapy Rating Scale (CTS-R) is a 12-item scale that measures both adherence and competency.46 The CTS has been used to measure competency in delivering individual CBT interventions, and it remains the main assessment tool for measuring the delivery of CBT interventions by MHPs.46 The scale assesses both adherence to therapy method and the skill of the therapist. Both the dimensions of competence in delivering the CBT techniques and the adherence to using prescribed techniques are assessed with the CTS-R.46 Our research team developed and adapted the CTS-R to the MB group (MB-CTS) intervention to assess both adherence and competency of paraprofessionals. The MB instructor manual provides the following for each topic: (1) a summary of key points, (2) a detailed scripts for delivering intervention content, and (3) an estimate of how long it takes to deliver the topic. The detailed scripts for delivering the intervention content are intended to minimize the potential for omission of key content or the introduction of irrelevant content. Fidelity was defined as maintaining adherence to both the appropriate execution of the specific session material in the MB instructor manual and competency in implementing and delivering the curriculum as intended. A total of 132 MB intervention cohorts were delivered, generating a total of 792 intervention sessions (132 cohorts × 6 sessions). A random sample (approximately 20%) of audiotapes were reviewed for protocol adherence (n = 160). Coders reviewed the sample of audiotapes to assess adherence and competency outcomes, and these reviews are the basis for the analyses presented in this report.
The MB-CTS uses the structure and content of the CTS47 and its revision, the CTS-R.48 However, to accommodate the MB content and group format, it is formatted to map onto the content of the 6 sessions of the MB intervention, thereby generating a tailored measure aligning with the MB strategies. It also has 12 items, along with a summary of the group session, which is not found in the original CTS. Thus, the MB-CTS is akin to the CTS and CTS-R structure but adheres to the descriptions and content relevant to the MB intervention.
The accompanying Statistical Analysis Plan and the previously published design paper by Jensen et al66 provide details of the analytic strategy. The primary outcome of interest was the QIDS depressive symptom score at 24 weeks postpartum, controlling for the baseline score. Analyses for superiority and noninferiority focused on this prespecified primary outcome. Secondary outcomes included incidence of an MDE, mean behavioral activation scale score, mean social support score, mean mood regulation score, and mean decentering score. All outcomes were treated as continuous for analyses with the exception of MDE, and post hoc analyses of social support explored dichotomization of this outcome due to the highly skewed nature (ie, “ceiling” effect) of this variable.
Important baseline variables included race/ethnicity; whether the participant was a first-time mother or was currently at risk for a depressive episode (ie, not experiencing an MDE at baseline interview); primary language of intervention receipt or primary language of assessment completion (for control arm); education; and whether the participant was currently in counseling with a therapist at baseline or currently taking medication for depression at baseline. These variables were selected a priori for any adjusted models, as they were deemed clinically relevant or potentially associated with outcome(s). We also collected other demographic information, such as age, marital status, and employment status.
While unit of randomization was at the HV program level (ie, 1 HV program = 1 cluster), analyses proceeded at the participant level. Primary outcome analyses evaluating for superiority of HVP vs control as well as the noninferiority analyses comparing HVP vs MHP at 24 weeks postpartum involved a linear mixed-effects model with fixed baseline QIDS and study arm effects and a random site effect to allow for estimation of an intracluster correlation coefficient (ICC) and distinction between within- and between-site effects. We performed both a fully adjusted analysis, whereby we adjusted for the aforementioned prespecified covariates, and an analysis adjusting for baseline outcome score only, and in each case, we included a random site effect. Analyses for other outcomes mirrored that of the primary outcome. Further, we used data from all study time points in more exploratory longitudinal analyses, where we employed a mixed model with repeated measures to explore outcome behavior over time and time-by-arm interactions. The analysis plan called for exploration of interaction between intervention arm and the aforementioned covariates. These analyses explored the addition of covariate-by-arm interaction terms in the aforementioned model or a 3-way covariate-by-arm-by-time interaction term in the longitudinal modeling in order to explore whether the relative effectiveness of the compared interventions varied by patient characteristics.
We used SAS version 9.4 (SAS Institute, Inc) for all analyses, and assumed a 2-sided, 5% familywise type I error rate, with a Tukey adjustment for pairwise comparisons across arms. Exploratory analyses of interaction terms allowed for a relaxed 2-sided, 10% type I error rate. We excluded participants from analyses if they did not contribute any data after baseline, and we analyzed them according to the arm to which they were randomized if they provided any data after the baseline visit (ie, a modified intention-to-treat analysis).
Power calculations assume an SD in primary outcome of approximately 6 points on the QIDS (range, 0-27), with a meaningful clinical difference corresponding to 5 points, on average, across arms. This meaningful clinical difference of 5 points was chosen to align with the level of symptom changes associated with movement between severity categories on the QIDS. Without much knowledge on ICC, we explored a range from 0.001 to 0.05 and assumed 0.01 to 0.02 as realistic estimates, and we used a conservative type I error rate of 0.05/3 for pairwise comparisons = 0.017. We further allowed for substantial participant-level and whole site-level dropout. Superiority analyses involving a minimum of 5 sites per arm with an average of 23 participants per site for analyses allowed for >95% power to detect a prespecified, meaningful 5-point difference in QIDS score under these assumptions. The noninferiority analyses required the ability to detect a smaller prespecified margin of noninferiority (assumed to be 2 points on the QIDS in our study). A margin of 2 points was deemed to be clinically similar because it reflected a difference that was less than half of the range in any QIDS category. Under the same assumptions on SD and ICC = 0.01, and further assuming realistic site-level dropout to allow for 15 sites per intervention (in each MHP and HVP) arm, an analytic sample size of 20 to 22 participants, on average, per site allowed for 90% power for noninferiority analyses. Thus, 20 to 22 participants × (15 MHP + 15 HV + 5 control sites) would yield between 700 and 770 total participants with analyzable data at follow-up that allowed for at least 90% power for analyses. To account for attrition, our initial recruitment goal was N = 933 participants; however, based on low caseloads at some participating sites and the conservative nature of assumed attrition and follow-up rates, we modified this goal to N = 874 participants for recruitment. From a statistical standpoint, the 1:3:3 randomization allocation was chosen because the noninferiority aim required a larger sample size and more HV programs to be randomized to each of the active intervention arms. From a logistical standpoint, we envisioned site recruitment would be easier if sites had higher probability of randomization to one of the active intervention arms.
The research team developed 1 codebook for the facilitator interviews and 1 for client interviews with codes and subcodes related to the interview questions and key areas of interest for the researchers before analysis. The codebooks were regularly reviewed and updated throughout the coding process as new codes emerged across transcripts. Four members of the research team coded interview transcripts. Before individual coding, 10% of the transcripts were coded by all 4 team members and reviewed to ensure consistency and reliability among coders. Any discrepancies were discussed and resulted in a consensus among all coders.
For the coding and analysis, we used NVIVO Pro, version 11 (QSR International). A thematic analysis approach was used in this study.49 We chose this approach for its flexibility, ease of use, and ability to work with a large data set. Once coding was completed, data were cleaned and organized into framework matrices. Analysis was conducted by 4 staff members and took place from July 2018 to April 2019. Two of these staff members separately analyzed the framework matrices to identify fundamental themes, and the other 2 individuals reviewed results to ensure validity.
Each MB intervention session ranges in the number of topics covered. Raters were instructed to rate the extent to which the facilitator delivered the content of the MB intervention as intended. For example, for each topic, the rater issued scores ranging from (10 = completely covered, 5 = partially covered, 0 = not covered at all). Adherence was determined by presence or absence of the specific topics in each MB session and the appropriateness of the content and delivery. To obtain the overall adherence score for each session, we added the scores for each topic in the intervention session and divided by the total number of topics within that session multiplied by a maximum score of 10. For example, session 1 is composed of 9 topics (9 × 10 = 90). Dividing the facilitator's total score on all 9 topics by 90 yields the percentage adherence. A score <90% adherence triggered follow-up with the facilitator. Ratings were averaged across each individual session to generate 1 score to represent the extent to which the facilitator remained adherent to the session content. Adherence scores could range from 0 to 100, with 0 indicating no adherence to the MB curriculum and 100 indicating complete adherence.
The structure of the general competencies on the CTS were retained. However, technical competencies were adapted to reflect the MB program (eg, item 9, “Emotional expression”; item 10, “Application of MB techniques”; and item 11, “Personal projects”). We added a 12th item, termed “Engagement,” to determine the extent to which the facilitator could engage all participants. Competency items were scored on a scale from 0 to 6. The original CTS scoring structure was retained. For each item, the facilitator's performance was assessed on a scale from 0 to 6. Descriptions were provided for even-numbered scale points (0, 2, 4, 6). If the facilitator's performance was between 2 of the descriptors, an intervening odd number (1, 3, 5) was selected.
Primary outcomes for analyses included a mean percentage adherence score (the average adherence divided by the total possible score) and a mean competency rating (across all 12 items) for each session. Both were treated as continuous measures for analyses. We further explored individual competency item scores. To accomplish primary objectives, we examined several potential predictors/covariates. They included the following:
We calculated descriptive statistics for all variables of interest. Categorical variables were summarized with frequencies and percentages, and continuous variables were summarized with means and SDs or medians and interquartile ranges, as appropriate. Boxplots and histograms, as appropriate, were used to illustrate outcome distributions across arms, sessions, education level, etc.
Primary analyses used separate linear mixed models (LMMs) to examine average competency score and average adherence by study arm and session number. Specifically, models for each outcome included a fixed effect for either arm or session (depending on hypothesis of interest) and a random facilitator effect to account for the same facilitators within arm and sessions. We also considered inclusion of a random site effect to account for differences between sites. However, including a random site effect and a random facilitator effect resulted in an overly saturated model in which both random effects were not estimable. Thus, for analyses in general, we deemed it sufficient to include a single random facilitator effect.
Secondary analysis further used separate LMMs to evaluate the influence of site, participant, and facilitator characteristics on outcomes of average competency score and average adherence. As described above, the models contained a fixed effect for site, participant, or facilitator characteristics (1 at a time in individual models) and a random facilitator effect. Exploratory analysis used the same analytic strategy to evaluate the influence of exploratory variables such as the size of each cohort, as well as facilitator demographics and experience, on these outcomes. Exploratory analyses also examined individual competency score items (dichotomized into an indicator variable based on a score of at least 3, which reflected “satisfactory” competency) via a series of generalized LMMs with logit link and binomial distributional assumption. We evaluated the predictive ability of study arm, session, and race concordance in these models. The models had a fixed effect on the exploratory variable and a random effect on either facilitator or site, as appropriate. Because the individual items are scored on a Likert scale, often with “ceiling” or “flooring” effects, we a priori chose to dichotomize the individual item scores for analyses. This allowed us to explore the behavior of individual items in relation to these potential predictors rather than overall score.
As the fidelity analyses are ancillary to a larger cluster randomized trial with a prespecified analysis plan, they are more exploratory in nature and, thus, there were no formal power calculations in evaluating this aim. All analyses assumed a 2-sided 5% level of significance. If any individual, multilevel variable was significant at the 5% level, we used a Tukey adjustment for pairwise comparisons to further explore any additional contrasts. All analyses were conducted in SAS version 9.4.
This project was originally scoped to be conducted in 42 HV programs across the State of Illinois. With PCORI's approval, we opened up recruitment to other states in the Midwest and West Virginia, as many HV programs in Illinois were experiencing budget cuts that led to decreased numbers of active participants and/or increased turnover of program staff. The covariate-constrained randomization algorithm simulated possible allocations, assessed imbalance across arms with respect to key cluster-level variables (eg, racial breakdown and geographic location) for each of the simulated schemes, and identified a subset of schemes with the greatest degree of balance across arms according to prespecified thresholds for each variable. We ended up recruiting 45 HV programs across 7 states in the Midwest and West Virginia.
During the project period, we had multiple conversations with our PCORI program officer regarding our sample size. We had initially proposed that we would recruit 933 participants and that we would implement the MB intervention in 2 waves of randomization. Based on low caseloads at some participating programs—typically due to budget cuts—and revising attrition rates to be more aligned with our early study findings, we requested a scope reduction in the number of enrolled participants to 843. We also felt that our research team would be better able to oversee logistics related to study recruitment and intervention rollout if we implemented the intervention via 3 waves of randomization. Both of these scope changes were approved by PCORI, and we were able to exceed the updated enrollment goal, with a final enrollment count of 874 participants.
In our original proposal, we stated that online group discussion boards would be created and used for the purposes of supervision. The discussion boards were intended to connect supervisors and group facilitators across participating HV programs to view and reply to each other's comments in a threaded discussion. However, after speaking with our project's executive committee early in our study, we received feedback that online discussion boards would not be the most effective modality of communication with HV supervisors and group facilitators. Members of the executive committee reported not seeing high levels of traffic on other, similar, discussion boards and suggested we develop an alternative modality of communication for supervision support. The NU research team, based on feedback from project stakeholders, instead created a listserv that allowed (1) all group facilitators and HVPs to post questions and experiences throughout the study/intervention, and (2) the NU research team to communicate study updates and preliminary findings during the project. The listserv was launched in March 2017. Participating sites and facilitators used the listserv for multiple topics throughout the project, such as reporting and discussing (1) successes and challenges related to the logistics of delivering MB groups; (2) intervention content that group participants found helpful and how facilitators modified the language of MB content to effectively communicate core concepts; and (3) innovative approaches used by facilitators to engage participants during group sessions. The research team used the listserv for study communication and dissemination.
During the initial phases of our project, the NU IRB made a recommendation to amend the crisis protocol used when a participant denotes suicidal ideation on the study assessments. The IRB noted that we needed to be able to provide the appropriate resources to address the worst-case scenario in nearly real time. We modified our crisis protocol to reflect the feedback from the IRB to include the immediate provision of suicidal support resources to the participant if they indicated suicidal ideation on study assessments.
There were no significant budget modifications associated with this project.
Of the 45 randomized sites, 37 contributed data for analyses (6 control, 16 MHP, 15 HV). Of the 1316 prenatal women referred to the research team, 874 (405 in the HVP-led arm, 310 in the MHP-led arm, and 159 in the control arm) were enrolled in the study. Eight programs either dropped out after randomization or were unable to recruit enough women to form a group. Additionally, 50 participants did not contribute data for analyses as they either completed baseline assessments only or did not complete any assessments after enrollment. Thus, our analytic sample size is N = 824 (149 control, 293 MHP, 382 HVP; see Figure 1). At the primary end point of 24 weeks postpartum, a total of 672 participants (77%) completed primary outcome assessments for analyses at that time point. For these initial preplanned analyses, we did not assess the missingness mechanism or evaluate potential biases. However, future secondary analyses may explore the missing data assumptions and impact on findings.
CONSORT Flow Diagram.
Participants' mean age was 26.3 ± 5.8 years, and median weeks' gestation at baseline was 23 weeks (range, 4-39 weeks). The majority of study participants (70% overall; 36% control, 74% MHP, 81% HVP) belonged to a minority racial/ethnic group and 40% had at least some college education, with the control arm exhibiting the highest percentage (48%) of participants with at least some college education. Income was largely <$25 000 per year (71% overall; 54% control, 65% MH, 81% HV), and 63% of pregnancies were unplanned (64% control, 59% MH, 67% HV; Table 1). Table 2 presents demographic summaries of MHPs and HVPs delivering the intervention. Briefly, all were female, and the large majority (91%) had at least bachelor's degree; 52% self-identified as White, 25% as Black, and 23% as Hispanic/Latina.
Demographics of Study Participants—Overall and by Study Arm.
Descriptive Statistics for Facilitators.
Sites contributed an average of 22 ± 17 participants per site with a coefficient of variation (CV; SD divided by mean, CV) of 0.77. Of the participants randomly assigned to 1 of the active intervention arms (N = 675, MH or HV), 80% received any intervention (ie, attended at least 1 of the 6 sessions of the MB course), and 53% were considered “adherent” according to the definition in the analysis plan, attending at least 4 out of the 6 sessions. A nearly identical percentage of women (52%) attended at least 1 session in each of the 3 CBT modules, indicating there were no differences in adherence based on whether a participant received at least 1 session in each module.
QIDS scores dropped from a mean 8.0 ± 4.2 points to 5.7 ± 4.5 points overall (scale range, 0-27 points), aggregated across all study arms. Although the HVP arm saw the largest drop in mean QIDS score from baseline (8.6 ± 4.3) to 24 weeks postpartum (5.9 ± 4.5), all arms showed an overall mean decrease in primary outcome, with the control arm ending at 5.8 ± 4.6 and the MHP arm at 5.3± 4.5 at 24 weeks postpartum. Overall the 95% CI for primary outcome was 7.69 to 8.28 at baseline, 7.03 to 7.60 immediately postintervention, 5.65 to 6.31 at 12 weeks, and 5.32 to 6.00 at 24 weeks postpartum. Table 3 shows outcome summaries across arms. Related to aim 1, we could not claim either of the intervention arms showed superiority in decreasing depressive symptom scores compared with the control arm (P = .393 when comparing HVP vs control; P = .406 when comparing MHP vs control) at 24 weeks; of note, we observed lower-than-anticipated mean depressive symptom scores overall at baseline.
Primary and Secondary Outcomes by Study Arms.
Related to aim 2, which focused on examining whether there were differences in the magnitude of depressive symptom change in participants receiving the MB intervention delivered by MHPs vs HVPs, we do have evidence of noninferiority when comparing the MHP and HVP intervention arms at 24 weeks postpartum. Our model-estimated mean difference equaled 0.01 points (95% CI, −0.78 to 0.79) points. Our prespecified margin of noninferiority was 2 points, and the upper bound of our model-estimated confidence limit does not surpass this margin. Results are robust, as we have similar findings whether we adjust analyses for our prespecified baseline covariates as above, do not adjust for these covariates (model-estimated mean difference between MHP and HVP of 0.04 (95% CI, −0.96 to 1.03), or conduct a longitudinal data analysis accounting for the arm-by-time interaction (P < .001; model-estimated 95% CI for the comparison between MH and HV arms at 24 weeks, −1.39 to 1.52). Figure 2 shows the mean primary outcome score trajectories over time, by arm. Overall findings are consistent when we conduct a sensitivity analysis on the “as treated” data set, controlling for participant adherence or overall group attendance. We conducted multiple sensitivity analyses, further exploring a dosage variable (no attendance, partial attendance [1-3 sessions], and full dose [4-6 sessions]) and also by grouping those who attended zero sessions into the control arm. Overall findings remain the same in that we do not see evidence of significant differences across study arms.
Mean Depressive Symptom Trajectories Over Time, by Study Arm.
The ICC was not estimable in the fully adjusted model for outcome, as the participant-level covariate information accounted for the majority of site-level variation. However, when baseline covariates were excluded from analyses, the ICC estimate was approximately 0.02, but we do not have evidence against the hypothesis that ICC = 0 even when failing to adjust for influential participant-level covariates (P = .160).
The results for secondary outcome analyses examining incidence of MDE, behavioral activation, decentering, social support, and mood regulation were not significant (refer to Table 3). There was minimal change overall in each of these outcomes over time. Exploratory analyses evaluating potential heterogeneity of intervention effect on primary outcome suggest a significant 3-way interaction between time, arm, and race (P < .001), whereby trajectory of mean QIDS scores over time varies according to whether participants belong to minority racial/ethnic groups and the intervention they received. Specifically, participants receiving HV who belong to a minority racial group showed the largest overall mean decline in depressive symptoms. We further have evidence of a similar 3-way interaction effect for new mothers (P < .001), whereby those receiving either the HVP- or MHP-led intervention showed the largest decline in mean overall depressive symptoms compared with other categories.
Results presented here focus on the following 3 thematic areas: (1) feasibility and acceptability of the MB content, (2) feasibility and acceptability of the MB group format, and (3) challenges related to MB appropriateness.
Overall, facilitators and clients described the content of MB as easy to understand, useful, relatable, and universal:
Yeah, it wasn't hard to read at all … there was other examples and there was activities that you can do as well, so it's not like you're just left there like, ‘I don't know how to do this,' or, ‘I really didn't understand the activity for today or the session for today or the activity that you would have to do at home,' and that was very helpful as well, actually paying attention to your days and how everything goes. [Client]
Clients reported consistent understanding and satisfaction with the content of all 3 MB modules (pleasant activities, thoughts, and social support). The vast majority of the clients discussed using skills from ≥1 of the modules in their daily lives:
I try to look back on the pleasant activities and then also feel like moods and outer reality, I try to keep it not so negative inside. I mean I don't know when I first started out; I had like 13 negative thoughts. Then I got it down to like 6 or 5, I remember. [Client]
Clients noted the valuable role that their facilitator played in the acceptability of the content. Although the content was described as straightforward by clients and facilitators alike, clients repeatedly highlighted how their facilitator provided both meaning and context as needed during the group sessions:
I think the reason why—I know me in particular—received the information so well is because the facilitator was very engaging. You can have facilitators who make the content dry, drab—you know, it feels like our facilitator actually kept it interesting … We read, just it didn't feel like we were just going through the motions. There was always an opportunity for feedback and insight and things of that nature that made us able to apply the content to our own lives a little bit more. We always were able to place things into perspective. [Client]
Clients brought up specific tools from the curriculum they found most helpful, specifically the Quick Mood Scale (QMS) that is used to help track one's mood on a daily basis, as well as vignettes used to introduce each CBT module. Clients noted that the experience of tracking their moods daily helped them draw connections between their thoughts, activities, and/or social supports and their moods. A client described the impact the QMS had on her mood:
It helped me monitor my moods and my activities. I noticed when I wasn't doing activities that my mood was a little low. If I kept busy, my mood was a little better. [Client]
The vignettes were referred to repeatedly in a very favorable light by both clients and facilitators. The vignettes helped clients identify thought patterns and behaviors, and also helped clients draw connections between actions and their moods:
There are examples on some of them [vignettes], where they showed an example of one person and the other person. One person was negative about their day and just didn't want to do anything and the other one started off negative and then they try to get their day better by doing things. And well, sometimes I feel like that. Sometimes I just don't want to get up. It just takes me a while to figure out how to get myself to feel better. [Client]
A significant portion of facilitators felt that the majority of MB content was appropriate, irrespective of client demographics, with some facilitators and clients noting MB was relevant to other family and community members as well as to their own personal lives:
So the fact that it can relate to anybody—it doesn't matter what your race, gender, age, sexual orientation is—it pertains to everybody that's going through this new chapter of their life. Whether it's having their first child, or having a second or a third child, everybody goes through the change in life and I just thought it helped us come together. Just because we're all different doesn't mean that we don't have some common similarities and Mothers and Babies I think really pulled out the similarities in the groups. [Facilitator]
Clients and facilitators reported several challenges related to attendance. Some facilitators shared frustrations with cohorts or individual sessions that were unable to get started due to low attendance. Facilitators and clients noted that the very characteristics of the population that was recruited for groups created challenges in attendance. For example, challenges during pregnancy, such as fatigue and medical complications, impacted the women's ability to attend groups:
I just was not feeling well. Sometimes I would come to class and not feel my best, but I would still stick it out. But it was kind of like a struggle to drag myself there only because I was not feeling well. [Client]
Despite efforts to schedule cohorts at the most convenient times for clients, multiple factors affected their availability to attend groups. The most prevalent factors that impacted attendance, mentioned by both clients and facilitators, were family crises and scheduling conflicts with work and school. Clients also noted having to skip sessions to take care of family members or sick children:
I also feel like that's another deterrent for other people because of the time, because if you're working full-time and things like that, and you don't have a flexible schedule, it's kind of hard to get that 2 to 2½ hours once a week in the middle of the day. [Client]
Additionally, challenges related to transportation to and from groups were identified, including lacking access to reliable transportation, having to travel long distances, and having to travel during rush-hour traffic in urban areas:
You know, the challenges, transportation. You know, getting there. Because sometimes I do—I live far from the office [HV Program], but that part is challenging living far out from the office. And y'all have it the people come get us back and forth. That's—I know it's challenging for me and them because they gotta worry about gas and we gotta worry about our safety. [Client]
Assistance and access to transportation and childcare services played a key role in client attendance. Free childcare, which was provided at the majority of sites, was helpful even for first-time mothers who were taking care of younger siblings or other young family members. This was deemed valuable by all clients, yet facilitators noted the increased time needed to coordinate this service and transition the children to childcare, in order for groups to start on time:
I thought the childcare was perfect because sometimes we have 2 or 3 children that we can't leave unsupervised or that we don't have someone who can take care of them. I thought that was perfect. [Client]
The group environment played a key role in encouraging women to engage in intervention activities. Clients noted that an intimate, welcoming, and confidential space impacted their decisions to stay and share during group sessions:
At first, I was kind of skeptical because I am kind of a quieter person, so I was kind of skeptical about opening up and just revealing personal information about myself, but I knew that the group was confidential, and it was told to us that the group was confidential and nothing that we said was going to be used outside or anything, so that kind of made me open up. [Client]
The majority of the clients felt their cohort size fit in a “Goldilocks band,” where the group had been large enough to learn from peers yet small enough to share freely without feeling rushed. Some clients voiced a desire to have additional women in groups, but others stated that they would not have felt as comfortable if there had been more women in the session. No clients expressed concern with the group size being too large:
It was good at 4 because we could all share a little bit here and there. If it was more than 6, then it would take too long because it's 6 people trying to say something. And you don't know how long you would have. And you don't want to rush certain things. [Client]
Clients emphasized the important role that their peers played in the group experience. Having a space to meet other women who were pregnant and experiencing similar challenges was very meaningful. Additionally, meeting women outside of their home helped address feelings of isolation and loneliness. The experiences shared in group sessions helped clients form friendships and valuable social support networks with peers. Clients and facilitators noted that the creation of social support among peers was one of the most significant outcomes of the groups:
I think a lot of it was just being an intimate group of women…. Even though we all come from completely different backgrounds, me being a first-time mom who works full-time and one having 4 kids and then the other one being a stay-at-home mom, but yet we all shared a lot of the same day-to-day struggles. [Client]
Even though we didn't know each other, you could—they showed care. And that was the biggest thing for me … It's just nice to have—that you still have people out there that is concerned of your well-being…. They would just say, “Happy you came. Good seeing you.” That was another thing that kept me coming by, getting the little support from them, of keeping yourself encouraged to keep coming. [Client]
Several women noted that their main motivation in attending MB groups was to become “better mothers.” They emphasized the value of pregnancy and parenting experiences shared by peers during group sessions. The conversations during sessions helped normalize thoughts and information for clients. Facilitators noted that the normalization of certain feelings also helped decrease feelings of isolation, assuring mothers that they were not alone. It was noted that the normalization of specific feelings was a byproduct of a group environment and would have been difficult to address in a one-on-one setting:
And one of the moms, her eyes lit up, and she was like, “I thought I was the only one, and I felt like I was a horrible mother because I wasn't initially just ecstatic.” So, then we went into those kinds of conversations, and just having them realize that they're not the only ones experiencing whatever it is they're experiencing, I think that was just wonderful for them. [Facilitator]
Many clients voiced that the cohort went by “fast,” and noted feeling sad with the completion of their cohort. The majority of clients stated that the length of the sessions was ideal, with some mothers stating that the sessions could have gone on a bit longer. Only 1 client noted that a cohort could have been shorter than 6 sessions:
And I know that a lot of people were upset when the classes came to an end because we had all felt such a bond together—that we didn't know what to do with our Wednesdays after the class ended. [Client]
Facilitators voiced challenges with the implementation of the group modality with clients who were experiencing crises, housing instability (including homelessness, couch surfing, and precarious living situations), and/or trauma. Several facilitators suggested the program incorporate examples that discussed these challenges into the examples and vignettes found in the manuals to help facilitators address them. A couple of facilitators also suggested incorporating more resources for facilitators to give to clients in the program:
I found a lot of the moms were homeless, and I didn't realize before this group…. I didn't realize that so many moms were homeless. For some reason, they opened up more in the Mothers and Babies group and told me that they were homeless, a lot of them living in shelters and pregnant, or living in a car, or they were squatters living in homes, and just to be able to address those issues…. Well, I think each group facilitator should at least be aware that those stressors are real and that they should have a resource to refer to because it will come up. It came up in every group I had…. So, basically, there is a need that's a great stressor because if a mom has the stress she's homeless, it's hard for her to focus on a group if she has the other stressors. [Facilitator]
A second challenge commonly identified by group facilitators was implementing materials with clients who were either illiterate, had literacy challenges, or had learning disabilities. Some facilitators noted the need to accommodate for different reading levels by completing activity sheets out loud instead of having clients write out their responses. Facilitators suggested that future groups should provide more opportunities for clients to complete materials out loud to avoid singling out clients who are uncomfortable with writing:
So, with the manual we did have 1 participant in 1 cohort that did have difficulty with reading and writing…. I didn't want that to become something that made her uncomfortable in the group and that she felt other people were aware of that…. I just basically tried to read as much as I possibly could from each of the, you know, from the participant's manual, specifically, because of that reason. And I also, of course, the personal projects [QMS] I went over that. [Facilitator]
One of the main challenges highlighted by facilitators was managing discussions during group sessions. Facilitators aimed to create an environment where clients felt comfortable sharing, listening to their peers, and reviewing the information presented in the group. Some facilitators felt this balance was harder to achieve their first time facilitating because they were becoming acquainted with the curriculum. Challenges that impacted finding a healthy equilibrium during group implementation included side conversations between clients, family crises shared during groups, and clients who required additional time due to disabilities, learning, or English as a second language challenges:
If someone is wanting to share something that is really personal or really intense or difficult, that if you were delivering this one on one, you could really spend some time focusing and really talking about that. But in a group setting, you might be able to spend some time talking about that to acknowledge and validate and to recognize just the importance of what someone shared, but you just can't spend the whole time talking about that. And so, I think that inherently can feel really difficult as a facilitator to move on from that, and also being respectful to the other group members, and also knowing that we have material to cover. [Facilitator]
Analyses included a total of 160 rated sessions. Table 4 shows mean outcome scores overall for average adherence and competency and by individual competency item. The mean adherence score was 76.32% and overall mean competency score was 3.76 (slightly above “satisfactory”). Neither arm nor session was significantly associated with mean competency (P = .512 and P = .533, respectively). Although study arm was not significantly associated with adherence score (P = .090; see Figure 3), there was a significant session effect on adherence (P = .009; see Figure 4). After adjusting for multiple hypothesis tests, only one of the pairwise comparisons remained significant: session 3 (model-estimated mean ± SE, 81.2 ± 2.79) vs session 4 (71.2 ± 2.79; Tukey-adjusted mean difference, 10.0; 95% CI, 0.3-19.7; adjusted P = .040), meaning there was greater adherence to session 3 compared with session 4.
Descriptive Statistics for Overall Mean Adherence Scores, Competency Scores, and Individual Competency Items: MB-CTS.
Mean Adherence by Arm.
Mean Adherence by Session.
On examining site-, facilitator-, and client-specific effects, we found that results were largely nonsignificant with a few exceptions. Facilitator education level (2-level factor) was significantly associated with adherence (P = .038; see Figure 5). Those with less than a master's degree tended to show better adherence than those with a master's degree or higher (model-estimated difference, 7.5; 95% CI, 0.5-14.5; P = .038). As noted earlier, facilitator training was conducted in slightly different ways (ie, training modality) across HV programs. Training modality was also significantly associated with adherence (overall P = .047; see Figure 6). After adjustment for multiple comparisons, the pairwise comparison that remained significant was between facilitators previously trained on the MB one-on-one modality who received additional phone-based training on the group modality vs facilitators who received a recording of an in-person training on the MB group modality (model-estimated difference, 18.7; 95% CI, 0.2-37.2; P = .046). Specifically, facilitators trained on the MB one-on-one modality followed by additional phone-based training on the group modality had significantly greater adherence than facilitators who were trained by listening to a recording of an in-person training.
Mean Adherence by Facilitator Education Level.
Mean Adherence by Training Modality.
Race concordance between the facilitator and the group she facilitated was not significantly associated with the outcomes in this data set. Exploratory analyses examining arm and session on individual competency items were nonsignificant except for 1 item. Competency item 1: “Agenda setting and adherence” was significant (P = .033) between session 1 and session 2, with facilitators in session 1 having significantly greater competency on item 1 than facilitators in session 2.
The present findings indicate that a group-based cognitive-behavioral intervention—MB—did not achieve greater reductions in depressive symptoms compared with usual HV services. These null findings related to our superiority analyses are surprising in light of the consistent findings from previous trials that have demonstrated that MB leads to greater reductions in depressive symptoms than usual care,17,18,21,22,55 including 2 RCTs conducted in HV.21,22,55 Our client outcome data did suggest noninferiority of HVPs when comparing them with MHPs, however. These noninferiority findings related to client outcomes complement our analysis of implementation fidelity from this cluster randomized trial, which found no significant differences between HVPs and MHPs in delivering MB according to protocol (adherence) or how skillfully it was delivered (competence).68 Our results also suggest strong acceptability and feasibility of MB based on our qualitative data.
Preventing the onset of PPD and the worsening of depressive symptoms among perinatal women has considerable public health relevance. Along with PPD's immediate impacts on parenting and young child development, research has implicated PPD with longer-term negative sequela on offspring of depressed mothers, including increased incidence of adolescent depression and anxiety and lower academic attainment.50-52 Significant barriers exist in providing mental health services to women at risk for developing PPD. Most notably, health and mental health providers have limited capacity to serve women at risk for PPD given limited availability of services for women experiencing subclinical symptoms, with most efforts focused on treating women already exhibiting the disorder. This is true despite a growing body of research suggesting that depressive symptoms during pregnancy are the strongest predictor of PPD and that depressive symptoms are associated with many of the same negative maternal and child health outcomes as PPD.11 Additionally, many perinatal women—particularly those with low income and those who are racial/ethnic minorities—may feel stigmatized if they receive care from an MHP or may not know where to seek care.53,54
Because HV programs serve women across all 50 states, they are a viable setting to provide mental health services to perinatal women, thereby making MB a potentially impactful intervention for them to consider implementing. Results from our qualitative data show that overall both clients and facilitators found the content and group format to be acceptable and appropriate. Clients not only stated how much they enjoyed the content and group setting, but also spoke to the effectiveness of the content and its delivery by providing examples of how they used core concepts outside the group setting. Clients emphasized how well the content was delivered, noting that facilitators explained concepts in a way that made them easy to understand. Facilitators generally found the content to be universal to most populations, highlighting the appropriateness of MB for pregnant women of varying backgrounds. One exception was that facilitators found the material difficult to implement with women experiencing crises, housing instability, and/or trauma. Given the frequency with which these challenges related to delivering MB to women with certain life circumstances were highlighted by interventionists, a logical next step may be to modify the MB instructor manual and training to include examples and suggestions that guide facilitators in working with women facing these life situations.
Financial support was provided for transportation to sites and childcare to minimize participant absences at group sessions. Participants stated that this support was an important enabler of group attendance, consistent with findings from other studies examining group interventions delivered to women prenatally.56,57 Interestingly, childcare was deemed necessary by participants regardless of whether they already had children of their own. Clients explained they often had childcare responsibilities for other people's children (including family members), and if childcare was not provided they would have been unable to attend the group. Although transportation support was described as beneficial in encouraging attendance, the reliability of transportation and long distances to travel to intervention sessions were also noted as barriers to attendance for some clients. The support that was provided, while helpful, may not have been enough. These findings suggest that HV programs and other agencies implementing group-based perinatal interventions such as MB should allocate adequate resources for transportation and childcare to increase attendance despite the cost associated with these supports.
Clients and facilitators noted it was sometimes difficult for clients to get out of the house or actively participate due to the discomforts and medical issues associated with pregnancy. Ensuring that group sessions have comfortable seating, several breaks, and flexibility in scheduling could assist in achieving higher attendance. Group size did not seem to impact participants' overall enjoyment of sessions. The composition of groups, including the age of participants and whether participants were first-time mothers, also did not seem to have a meaningful impact. Largely, participants were able to find common ground in pregnancy and shared lived experiences. The overwhelmingly positive feedback regarding the relationships built in groups exemplifies the importance of social support for pregnant women.56,58-60 This was also evidenced by the desire voiced by some participants for cohorts to go on longer.
Fidelity analyses suggested no significant differences between HVPs and MHPs for either their overall adherence to delivering the MB intervention or their competency in delivering it. Moreover, we do not have evidence of significant differences between HVPs and MHPs on any individual competency item. There was only 1 significant session effect on adherence, with greater adherence displayed when delivering the first session (session 3) in the curriculum's “Thoughts” module compared with the second session (session 4) in that module. Because the MB curriculum intentionally has content in the second “Thoughts” module session that builds on content from the first “Thoughts” module session, it is possible this difference could be due to facilitators not feeling they needed to cover materials they believed were adequately covered in the initial “Thoughts” module session. Decreased redundancy used in this session, along with an increased focus on explaining the rationale for the inclusion of content found in session 4 during facilitator training and supervision, may increase adherence in future MB implementation.
Facilitators with a master's degree or above had overall lower average adherence than those with less than a master's degree. Facilitators with higher education, particularly related to mental health, may have had more confidence delivering the curriculum's CBT-based content, giving them increased comfort in going off script more often than facilitators who may have had less familiarity with the CBT concepts. The conceptual framework for examining intervention implementation introduced by Aarons et al,61 known as Exploration, Preparation, Implementation, and Sustainment (EPIS), suggests that varying clinician attitudes toward evidence-based interventions may be another possible explanation for our finding that MHPs had lower adherence when delivering MB. For example, some studies have found that MHPs may find evidence-based interventions too restrictive in meeting clients' varied needs irrespective of how long the MHP has worked in the field as a clinician.62,63 Although most previous research has examined attitudes of MHPs toward evidence-based interventions, Jensen-Doss et al64 examined attitudes of both MHPs and providers with less formal education. Their study found that MHPs with more mental health training had less positive attitudes toward evidence-based treatments.
Similarly, lower-than-average adherence was found for facilitators who were trained via listening to a recording of a live training compared with facilitators who had previously been trained on the MB one-on-one modality and had their training supplemented by a call with the PI on the differences associated with delivering the MB group modality. Bryan et al65 note that 1 key adult-learning principle that may affect success in training public health practitioners is that learners are actively involved in the learning process. Facilitators who listened to the recording were given an opportunity to ask questions and review the material with the PI; however, this did not provide the same degree of interactions with the PI and peers as the in-person trainings, which may have impacted the learning process. This finding highlights an important consideration for future MB intervention trials as well as other intervention research when considering ways in which facilitator training is conducted.
Across all competency items, “Engagement” had the highest mean score among both types of facilitators. This exemplifies facilitators' abilities to involve clients in the discussion and encourage them to participate in activities, which are hallmarks of the MB group intervention. This high rating could be a reflection of the previous experience that facilitators had leading groups, as only 6% had no prior experience leading any type of group. Many facilitators who were on staff at the HV programs, particularly HVPs, had previous relationships with clients at the programs who participated in groups. These existing relationships could also explain the high “Engagement” scores.
Findings from our primary outcome analyses suggest that the MB intervention may have greater benefits for women who belong to a minority racial/ethnic group and/or are first-time mothers. Specifically, such women receiving MB from an HVP exhibited the largest overall mean score declines in depressive symptoms, and women receiving MB from an HVP or MHP showed larger declines in depressive symptoms than usual care participants. These are important findings, particularly as they relate to ongoing implementation of MB in HV programs. The National Home Visiting Yearbook28 overview of the demographics of HV clients across the United States indicates that 39% of HV clients are from a minority racial group, with 30% of women reporting Latina ethnicity. Moreover, some HV models serve only first-time mothers, while others place increased emphasis on recruiting first-time mothers. While MB will continue to be presented as a viable option for reducing depressive symptoms among women in HV programs irrespective of race, ethnicity, or parity, our findings suggest that HV programs can have confidence that MB is an effective option for their clients who identify as being a minority race or ethnicity, or who are having their first child.
Several limitations merit consideration when interpreting study findings, particularly as they relate to the null findings related to our superiority aim. Most notably, our overall sample's baseline depressive symptom scores fell into the mild symptom range and were lower than in other MB trials. For example, in a previous MB trial conducted with 4 HV programs that served primarily Black women, baseline depressive symptom scores were in the moderate severity range, with postintervention scores falling into the normal symptom range.55 A similar decrease from a moderate to normal depressive symptom range was found in a separate MB trial conducted with Asian American and Pacific Islander women enrolled in HV.22 Thus, because of the low baseline scores in our study, there may have been a potential flooring effect in which there was less room to demonstrate improvement in symptom reduction. Due to pragmatic considerations associated with assembling large enough MB groups at each participating HV program and the potential for underreporting depressive symptoms, we did not use a depression symptom cutoff for study inclusion whereby participants needed to exhibit a certain level of symptomatology. Moreover, many of the predictors of developing PPD (eg, poor social support, low SES, being a teenage mother) are the same as eligibility criteria for HV program enrollment, which suggested that the study would still recruit women at risk for developing PPD without including a depression symptom cutoff. Another reason for the low baseline depressive symptom scores across the sample may be related to the percentage of women who indicated at baseline that they were using medication or seeing a counselor, which was slightly higher than in previous MB trials.
Limitations related to implementation of the MB intervention may have also contributed to our study's null findings related to MB superiority over usual HV services. Across active intervention arms, just over half the sample received ≥4 intervention sessions, which may have prevented participants from obtaining the full set of skills found in the MB curriculum and benefiting from engagement with other perinatal women in an MB group. We found that facilitators—both HVPs and MHPs—adhered to our MB instructor manual roughly three-quarters of the time when delivering the intervention. It is possible that greater adherence could have resulted in greater comprehension of MB materials and/or skill practice among participants, leading to greater reductions in their depressive symptoms. Similarly, the combining of MB sessions may have resulted in fewer opportunities for intervention participants to practice using MB skills between intervention sessions, potentially limiting their depressive symptom reduction.
Other limitations related to the study should also be noted. The use of a self-report tool (the MMS) to assess MDEs may have resulted in an underreporting of depression episode incidence despite prior research demonstrating good convergence between the MMS and a structured clinical interview. Related to our qualitative interviews, the 3-month gap between completing intervention delivery and participating in an interview may have affected recall among our facilitators, particularly among those who implemented >1 cohort. Interviews with clients were only conducted with women who attended at least 1 of the group sessions. While outside the scope of this study, in the future it could be beneficial to interview participants who did not attend group sessions to better understand barriers to attendance.
Limitations of our fidelity analysis included not capturing unique sessions. Some facilitators had to combine sessions (most commonly sessions 5 and 6) due to agency and client needs. Because this would have created complications for the coders, these sessions were eliminated from being scored. In the future, it may be helpful to score these sessions to determine whether there is any variation in fidelity when sessions are combined. A second limitation was the number of statistical hypothesis tests. Although we used a Tukey adjustment for multiple pairwise post hoc comparisons, the number of tests performed increased the risk of making a type I error. Foreseeing this issue, we developed an analysis plan a priori to document and outline the statistical hypothesis tests and to prevent too many ad hoc analyses.
Our findings suggest several important future directions. First, additional post hoc analyses should be conducted to generate a more comprehensive understanding of our client self-report data. In particular, additional analyses could be conducted to compare intervention recipients with controls who entered the study with more significant levels of depression. This is particularly relevant given the flooring effect previously described. It is also important to understand the extent to which the intervention led to a clinically significant decline in depressive symptom scores, defined as a drop of at least 5 points on the QIDS. Second, attention should be placed on determining strategies to maximize attendance at MB group sessions in light of our findings that only 53% of participants were adherent to the intervention. Although we addressed 3 of the areas that previous trials had identified as attendance barriers—transportation, childcare, and food during sessions—it is possible that shortening intervention sessions to reduce a participant's weekly time commitment might ensure greater attendance.
For programs experiencing low caseloads there are also pragmatic considerations associated with implementing MB groups. We believe that one of the greatest challenges experienced by HV programs in recruiting for this project was our inclusion/exclusion criteria that limited participation to women <33 weeks' gestation at the time of study enrollment. Because HV programs continue to serve women after childbirth, expanding MB delivery to include postpartum women would allow for the creation of larger MB groups.
Finally, given the frequency with which challenges associated with client trauma, unstable housing, and crisis situations were mentioned, a logical next step may be to modify the MB instructor manual and training to include examples and suggestions that guide facilitators in working with women facing these life situations. Another area for improvement of the MB materials is to include more suggestions for facilitators delivering the intervention with participants who have learning and/or literacy challenges. This is currently covered during the MB training, but the frequency of facilitator comments on this topic suggests a need to add more information during both training and supervision.
Our study findings suggest the need for greater selectivity when considering implementation of the MB group intervention. Although previous studies have generated evidence of MB's effectiveness, the current study found no significant differences in depressive symptom reduction between women receiving MB and women receiving usual HV services. These results may be a function of the lower-than-anticipated baseline depressive symptom scores among study participants, which hindered our ability to detect symptom improvement. Our superiority findings yielded no differences between intervention and usual care participants, and we found that HVPs generated similar outcomes to MHPs and also showed comparable adherence and competency in delivering the intervention. These findings occurred in settings in which HVPs and MHPs were provided with appropriate training and support while implementing MB. Whether the findings apply to other circumstances, including women with higher depressive symptom scores, is unknown.
Both clients receiving MB and facilitators delivering it found the intervention to be acceptable and feasible. Given the scope of HV nationally, coupled with HV's use of paraprofessional staff, using HVPs to deliver PPD preventive interventions to perinatal women provides an enormous opportunity to reduce the burden of depression across the United States. HV programs should consider the balance of evidence from this study and previous MB trials in considering whether and how to implement MB. Moreover, HV programs can use results from this trial as emerging evidence that, should they choose to implement MB, the data suggest that HVPs can deliver the intervention with fidelity.
We would like to acknowledge our advisory board members and stakeholders who significantly contributed to our study.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (AD-1507-31473). Further information available at: https://www.pcori.org/research-results/2016/testing-effectiveness-adding-group-therapy-home-visiting-services-reducing
Mothers and Babies Group Program: Semistructured Interview for Intervention Participants (PDF, 115K)
Mothers and Babies Group Program: Semistructured Interview for Group Facilitators (PDF, 121K)
Tandon SD, Johnson JK, Diebold A, et al. (2021). Testing the Effectiveness of Adding Group Therapy to Home Visiting Services on Reducing Postpartum Depression in Women with Low Incomes. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/03.2021.AD-1507-31473
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
