The Perspectives of Women and Their Health-Care Providers Regarding Using an ECA to Support Mode of Birth Decisions

Abstract

This study used focus groups to assess the feasibility and acceptability of adapting an Embodied Conversational Agent (ECA) to support decision-making about mode of birth after previous cesarean. Twelve women with previous cesareans, and eight prenatal providers at an academic, tertiary-care medical center, viewed a prototype ECA and were asked to share feedback on the potential role in helping women prepare for decision-making. Both groups felt that although it was somewhat “robot-like,” the ECA could provide easy access to information for patients and could augment the visit with providers. Future work is needed to improve ECA visual appeal and clarify the role and timing for utilization of decision aids using ECA technology to enhance the shared decision-making process.

INTRODUCTION

In 2017, the cesarean rate in the United States was 32.0%, more than double that recommended by the World Health Organization (Betran et al., 2016). In response to these high and rising rates, the National Institutes of Health recommended shared decision-making to support women choosing mode of birth after cesarean options (Cunningham et al., 2011). Given the complexity of the choice between planning vaginal birth after cesarean (VBAC) and elective repeat cesarean birth (ERCB), women need effective ways to access evidence-based decision support and prepare for partnership with their health-care providers to make shared decisions based on values and preferences (Chinkam et al., 2020; Shorten et al., 2014).

Interactive Internet-based decision aids have been shown to support women with previous cesarean in reviewing their birth options and preparing for shared decision-making with their provider (Shorten et al., 2015; Shorten et al., 2019). This paper explores an innovative effort to push technology beyond static decision support tools for pregnancy, toward greater interactivity, using an Embodied Conversational Agent (ECA). Health information technology, using ECAs, offers new possibilities for women to access easy to use, virtual, responsive health information, as well as decision-making support tailored to their personal preferences (Hoffman et al., 2013).

LITERATURE REVIEW

ECAs are computer animated characters designed to simulate face-to-face conversation using both verbal and nonverbal behaviors, integrating best practices from provider–patient communication theory (Cassell et al., 2000; Gardiner et al., 2013). They represent an advancing area of research about applying health information technology to support patient education and decision-making. ECAs provide an avenue for patient education that is tailored to each person's health literacy level and personal values as well as cultural and logistical preferences (Gardiner et al., 2013).

An ECA is an animated anthropomorphic interface agent that can engage in real-time, multimodal dialogue. ECAs talk using synthesized speech and can display a wide range of nonverbal behaviors synchronized with their speech, including facial expressions, head movements, eye gaze, body postures, and various hand gestures. The user interacts with the ECA primarily by selecting a response from a multiple-choice touch screen menu of options updated at each turn of the dialogue (Bickmore et al., 2005). Although ECA dialogues are typically scripted, they can be tailored based on information from the unfolding discussion with the patient.

Several ECA systems have been developed and evaluated for effectiveness in providing education on various women's health topics, with positive results. The “Gabby” system was created to promote preconception care among young African American women. The system was programmed to screen for over 100 general and reproductive health risks and address these in a series of conversations with a virtual counselor named “Gabby” (Gardiner et al., 2013).

Jack et al. (2015) conducted a randomized controlled trial comparing the “Gabby” system to usual care. They enrolled 100 nonpregnant African American women aged 18–34 years who were screened for over 100 preconception health (PCH) risks and randomized to either the “Gabby” or control group. The women in the “Gabby” group interacted with the system for up to six months; the control group received a letter indicating their PCH risks with a recommendation to talk with their clinician. The numbers, proportions, and types of risks were compared between groups. They found that women who interacted with “Gabby” had greater reductions in the number of PCH risks (8.3 vs. 5.5, P < .05) and the proportion of risks compared to controls (27.8% vs. 20.5%, P < .01). The majority of the women said that it “was easy to talk to Gabby” and 64% used information from Gabby to improve their health. The authors concluded that Gabby was significantly associated with preconception risk reduction.

A virtual lactation consultant, “Tanya,” was also developed to promote breastfeeding behaviors among first-time mothers (Edwards et al., 2013; Zhang et al., 2014). Mothers in the intervention group demonstrated significantly greater intent to exclusively breastfeed and greater knowledge following the interaction with the ECA compared to the control group (Edwards et al., 2013). Zhang and Bickmore (2018) developed a virtual decision coach to help women participate in shared decision-making regarding prenatal testing for Down syndrome. The women who participated showed significant increases in knowledge, high levels of satisfaction with their final decisions, and low levels of decisional regret. The ECA was found to be effective in the role of decision coach and to facilitate the implementation of shared decision-making (Zhang & Bickmore, 2018).

Shared decision-making, in the context of a previous cesarean, requires providers and women to have bidirectional discussions about planned mode of birth. This process is particularly vital for mode of birth after cesarean counseling because the risks and benefits are complex, outcomes are not guaranteed, and choices are personal and value-laden (Munro et al., 2017; Shorten et al., 2014). Ideally, shared decision-making is an informed, interactive, supportive, and individualized method of decision counseling that draws on each woman's experience of birth, as she works with her provider to establish her preferred mode of birth (Charles et al., 1999; Cox, 2014; Shorten et al., 2005).

Although numerous studies have demonstrated the benefits of decision aids in the context of shared decision-making, integrating decision support consistently into prenatal care remains challenging (Barry & Edgman-Levitan, 2012; Shorten et al., 2015; Shorten et al., 2019; Stacey et al., 2017). Practical solutions for helping to facilitate shared decision-making in pregnancy care are long overdue. Building from the success of Zhang's work regarding Down syndrome screening, it is timely to explore whether ECAs can play a role in facilitated prenatal shared decision-making for birth after cesarean (Zhang & Bickmore, 2018).

The purpose of this study was to engage the opinions of women and their health-care providers about using an ECA to support the process of shared decision-making about mode of birth after cesarean.

METHODS

Designs

This is a qualitative stage of a two-part pilot study using focus groups to assess the feasibility and acceptability of using an ECA, from the perspective of women with a previous cesarean and their prenatal providers. The study protocol was approved by the Institutional Review Boards of the study site (BUMC IRB approval number H-34725). The study was conducted from May to November 2016.

Study Setting

The study site was a busy academic, tertiary-care medical center that serves a large city safety net hospital. The patient population of the medical center is racially and ethnically diverse: 40% non-Hispanic Black, 35% Hispanic, 10% non-Hispanic White, 5% Asian, and 10% mixed/other. Fifty percent of patients do not speak English as their first language. Eighty-five percent of the 2,800 annual births are publicly insured. Eighty-two percent of women who gave birth in the setting in 2016 with a previous cesarean birth were candidates for labor after cesarean (LAC). Of those women only 40% selected to have a vaginal birth and overall, 23% had a successful VBAC (S. Chinkam. personal communication). Pregnant women (PW) who receive care at the Obstetrics and Gynecology Ambulatory Care Clinic, the outpatient clinic located within the study site, are seen by Certified Nurse-Midwives (CNM), nurse practitioners (NP), obstetric attendings, and residents.

Study Participants

Study participants included two groups of women and one group of prenatal providers. The first group of women were pregnant with a singleton pregnancy, having experienced no more than two previous low transverse cesarean surgeries (LTCS), were medically eligible to LAC, and spoke and read English. The second group had experienced previous LTCS between January and June 2015, were of reproductive age (15–45) and planning to be pregnant within the next 5 years. All were medically eligible to consider LAC for their future birth and spoke and read English. The prenatal care providers group was made of CNMs, NP, obstetric attendings, and residents who provided prenatal care at the Ambulatory Care Clinic.

Women were excluded if they had a contraindication for LAC such as: uterine rupture, classical or T-uterine incision, extensive transfundal uterine surgery, or placenta previa.

Study Procedure

The prenatal clinic schedules were checked weekly by a study team member to identify women with an upcoming first prenatal visit whom, based on chart review, might be eligible for study participation. The women were first sent a letter of invitation, and then contacted by phone to personally invite them to participate in the study. At their first prenatal visit, women who expressed interest over the phone met a team member to learn more about the focus group. There were no women who were initially approached but later turned out to be ineligible for study participation.

Enrollment of nonpregnant women (NPW) was performed by identifying eligible women from the hospital birth data, then mailing them an invitation about the study and study participation, with a follow-up personal phone call by one of the research staff. Women were selected according to their date of giving birth, starting from January 2015 until June 2015.

Recruitment of women for each group (pregnant and nonpregnant) was ceased once 20 women had agreed to participate, with an anticipated focus group attendance rate of 75%. All women who expressed interest in participating were called by a study team member to confirm the location, date, and time of the focus group beforehand.

Enrollment of providers was performed by the principal investigator emailing an invitation to participate in the study to all prenatal providers at the study site. Providers were asked to reply and indicate interest (or not) in participating. Providers who did not respond to the email were also invited individually, in person.

Written consent was obtained from all participants. Participants were each given a $75 gift card at the completion of the focus group to thank them for their time.

Focus Group Process

At the start of each focus group, the moderator reviewed the purpose of the focus group, reminded participants about confidentiality, reviewed the consent, and each participant chose a pseudonym. Each woman was asked if she had seen or heard of an ECA. The moderator then briefly explained the ECA and showed a video of a sample ECA on a television screen (Figure 1). The virtual agent shown in the focus group was developed as a virtual lactation educator with speech output produced on the server with a commercial speech synthesizer. Dialogues are scripted using a custom hierarchical transition network-based scripting language (Bickmore et al., 2005). Nonverbal conversational behavior is generated using Behavior Expression Animation Toolkit (BEAT), and includes hand gestures and eyebrow raises for emphasis, gaze away behavior for signaling turn-taking, and posture shifts to mark topic boundaries, synchronized with speech (Cassell, 2004; Zhang & Bickmore, 2018).

Figure 1. Screen shots of Embodied Conversational Agent (ECA) shown on a television screen.

After seeing the ECA, participants were asked their thoughts about the ECA and if they would consider using an ECA developed specifically to help patients make the decision about mode of birth after cesarean.

ANALYSIS

All focus group discussions were audio recorded. Each participant was given a code number for their identification, data collection, and analysis. The coding was done numerically starting with PW for pregnant woman, NPW for nonpregnant woman, and CNM for Certified Nurse-Midwife, NP for nurse practitioner, OBA for obstetric attending, and OBR for obstetric resident. All transcripts were transcribed and analyzed using NVivo 10.

Data were analyzed by two members of the study team (IH & AS) using inductive thematic analysis (Guest et al., 2012; Miles & Huberman, 1994). This involved an iterative and interpretative process allowing for repeated exploration of participant perspectives. An inter-coder agreement was established through researchers' discussion of codes and emergent themes. We applied a constant comparison method, which allowed for systematic generation of codes and themes (Creswell & Creswell, 2017; Strauss & Corbin, 1998). Researchers' independent interpretations of the data were discussed in several stages during coding.

RESULTS

A total of seven PW, five NPW, and eight providers participated in the focus groups. Women who attended focus groups were aged between 22 and 41, most were single (n = 7) and had experienced only one previous cesarean (n = 11). Most self-identified as Black African American (n = 8) with the remaining women White/Hispanic (n = 2), Black/Haitian (n = 1), and Native Hawaiian (n = 1). None of the women had any previous experience with an ECA. In the provider group there were a range of provider types including CNMs (n = 3), obstetric attending (n = 3), and obstetric residents (n = 2). Most providers were female (n = 6) and had been practicing at the study site for between 2 and 12 years. No providers had any previous experience with ECAs.

Three themes about the ECA (two positive and one negative) emerged from the analysis of focus group data, with the PW and NPW combined: (a) ECA provides easy access to information about mode of birth after cesarean, (b) ECA could augment the visit with providers regarding mode of birth shared decision-making, and (c) ECA is robotic and not the same as face-to-face discussions. Provider discussions revealed three similar themes.

Women's Perspectives

1. ECA provides easy access to information about modes of birth after cesarean.

   Women from both groups expressed that an ECA could be potentially useful as a convenient and easy to access source of information about birth options. The convenient, 24 hours access to the ECA at home, on their computer, or phone was appealing to participants because it meant that they could access mode of birth information and interact with the ECA about their questions/concerns regardless of their location.

   “You know sometimes you have a question and by the time you actually get there, you know in the midst of doing everything you kind of forget to ask. So, I think it's convenient cause you kind of have access to it.” (PW)

   “Sometimes they're not available to talk to you, so it'd be easier online, on the computer.” (PW)

2. ECA could augment the visit with providers in regard to mode of birth shared decision-making.

   Women felt the ECA could be a source of important information about their mode of birth choices and possibly prepare them for discussions with their provider. Some believed that it could serve as a bridge between prenatal appointments and shared decision-making discussions. It could also assist them to spend time thinking about their birth options while waiting to see their clinic provider and identify important topics to talk to their provider about, or questions to ask, in regard to LAC versus ERCB.

   “I think in the waiting area is the right place so that if I want to get more information, or it might prompt me to think about more questions I can ask when I'm in the appointment.” (NPW)

   “That's good, especially if they give it to her- the people in the waiting room, cause if they are looking at it and- I mean they ask the questions and “oh! I wanted to ask my doctor about this word, that word.” It's like refreshing your memory while you're at the appointment…. So while you're in the waiting room, looking at it, the question we had two weeks ago might pop in my mind, like “oh, yeah this is what I'm going to ask my doctor.” (PW).

3. ECA is robotic and not the same as face-to-face discussions.

   Despite women thinking about ways the ECA could be used to give them basic information about their mode of birth options, and prompt them to ask their provider questions, some of the first impressions about the prototype format were negative.

   “That's a little creepy. You can ask it a question and it will answer? That's crazy! I'm sorry, this is crazy.” (PW)

   Women saw the ECA as an adjunct to care and did not think that the ECA could replace face-to-face discussions with their provider. Women emphasized the importance of a personal approach to consultations around birth after cesarean.

   “I understand that computers are very intelligent, and I understand that. But I think childbirth and pregnancy is a very emotional time for women in a way that you can't control, and for a person not to be making that decision with you - a person whom you trust with your life and your child's life is in this person's hands, and then I can't leave that up to a computer.” (NPW)

   “I'd just rather have someone in front of me, talking. I can talk back. Though I think this would be helpful for some people, but I like to talk to people…as a 40-week pregnant woman I would throw something at that computer. “Are you seriously trying to tell me this is what I'm doing here?” You know I would walk out and go to a different facility. There's like- I mean no. That's not delivering my baby, so if the information is important, they're going to tell me how important it is, not the computer.” (NPW)

Provider Perspectives

Three similar themes regarding the ECA emerged from the analysis of focus groups with providers: (a) ECA could be used as a source of information about mode of birth after cesarean; (b) ECA could be used to supplement routine shared decision making about mode of birth after cesarean during prenatal visit; and (c) ECA was very impersonal.

1. ECA could be used as a source of information for patients about mode of birth after cesarean.

   Provider comments were consistent with comments from the women. Providers felt the ECA could be another important avenue for patient education about mode of birth decisions. In particular, they felt it would be useful in part due to its being able to be accessed at any time and any location.

   “I kind of agree with Z [pseudonym of another provider]… Like if it was part of our flow after our (trial of labor after cesarean) counseling to send them a link to their email or something of that nature. And they could go on their phone, and click it, and be there, and do it. I think it would be more highly utilized than like if they have to wait till they get home, and log into a computer or log onto a computer here.” (OBA)

   “It might confirm their decision. I don't know if it would help sway them.” (CNM)

2. ECA could be used to supplement and prenatal visit and shared decision-making about mode of birth after cesarean during.

   Providers considered potential options for integrating the ECA within the clinic workflow for patients needing to make decisions about their mode of birth. Most felt it would be a useful adjunct to personal discussions, either before or after their visits.

   “I would probably want to use this before the discussion, rather than after. Just like preface it and then let them use it as you're finishing something else. And again, not instead of time with you, but in addition to. So, hopefully then the discussion afterwards would be a little more focused and have more questions that are relevant to themselves.” (OBR)

   “You know if the partner can't make it to the prenatal visit. It can be something that they can show them.” (OBA)

3. ECA was very impersonal.

   First impressions of providers were that the current format was impersonal and robotic. They felt that a more realistic voice and human form in combination, is necessary for future prototypes.

   “It feels very impersonal. And my ears shut off pretty quickly.” (OBA)

   “For me, I just feel like it's impersonal. I know if I go to the doctor, I want the doctor to talk to me. And I feel like my patients, or like – I feel like they want our sound. I'm not saying it's like it's not a helpful tool, but I don't think that would replace” (CNM)

   “I think it has a lot to do with the audio part of it. I think the visual part was fine. But if it was a different voice, it might be a different experience…” (OBR)

DISCUSSION

A number of Internet-based health information technology decision tools have been created to address the changing needs and expectations of patients and providers (Zhang & Bickmore, 2018). Women seek access to personalized health education outside clinic visits at a time and place that is convenient to them. Providers recognize that they are not always able to optimize preparation for and the process of shared decision-making discussions with their patients during busy clinic schedules. Although not a replacement for face-to-face interaction, ECAs may serve to supplement and complement personal interactions. The challenge of meeting both patient and provider needs is clearly important in mode of birth counseling for women with a history of previous cesareans.

This study adds to the literature that generally supports the potential usefulness of ECA as an adjunct to reinforce shared decision-making regarding mode of birth counseling after a cesarean. Some of the women and providers had negative first impressions of the ECA due to the somewhat robotic nature of the prototype. Recent work has refined ECAs, so they are accepted as both friendly and trustworthy. However, the effectiveness of an ECA also depends on the individual attributes and needs of the user (Astrid et al., 2010; Isbister & Nass, 2000). In the future, opportunities for tailoring ECA personalities to match the individual preferences and characteristics of the user can be explored. Additionally, current state-of-the-art speech synthesizers are much more natural and empathetic than the one presented to the focus groups, which should lead to greater acceptance in the future.

Despite skeptical first impressions, it was clear that as discussions continued, both groups were able to appreciate the potential benefit of more sophisticated ECAs. They saw the ECA as an additional source of information that could be accessed anytime/anywhere and acknowledged that it could better prepare them and/or augment shared decision-making discussions with their provider. Particularly with a topic as personal and value laden as LAC versus ERCB, all the focus group participants seemed to recognize the importance of opportunities for women to have access to information that would allow them to make the best mode of birth decision for themselves and their families. This general utility of ECAs is consistent with previous studies showing they are capable of supporting health counseling, including delivery of complex health information to patients with varying literacy levels (Edwards et al., 2013; Gardiner et al., 2013; Jack et al., 2015). Improved outcomes are likely due to ECAs ability to adapt their messages to the particular needs of patients and the immediate context of the conversation, and by providing health information in a consistent manner within a low-pressure environment, where patients are free to take as much time as they need to thoroughly understand the information that supports decision-making (Bickmore et al., 2010).

Additionally, both groups discussed values of the ECA that acknowledge the diversity of both patients and providers. They recognized the benefits of using the ECA for non-English speaking women. Studies support the usefulness in such cases and show that ECAs with the same ethnic background and language as the user are well accepted when providing health information or counseling (Sue et al., 1996; Zhou et al., 2017; King et al., 2011; Yin et al., 2010). This is particularly important in settings where there are large numbers of women with diverse ethnicities and languages. Having an ECA available to support mode of birth decisions in various languages could be invaluable.

The development of innovative technological solutions to address time constraints during consultation and to improve patient access to tailored information, in the form of an ECA, came as a surprise to most of the study participants. Both women and the providers emphasized the importance of maintaining face-to-face consultations during their prenatal visits, with technological solutions viewed primarily as an adjunct to personal communication. Concern about personal consultations with providers being substituted, or completely replaced, by robotic communication was expressed by participants in all groups. Despite those concerns it seems clear that the ECA could allow women to obtain needed information without having to worry about the limited time in their prenatal visit or feeling hesitant to ask questions. Moreover, providers do not always consistently follow practice guidelines, creating variations in the delivery of health care information. The ECA could support provider and patient discussions and offer programmed consistency in preparatory information.

There are several limitations of this study. First, although the recruitment process for women met the initial target of 20 women expressing interest, attendance was less than the 75% anticipated. As a result, the sample may not have been large enough to reach data saturation. Additionally, due to the nature of the prototype available for viewing, the audiovisual presentation of the ECA was suboptimal and at times difficult for participants to hear. The ECA prototype video was also short and did not fully demonstrate the ECAs capacity for interaction with the patient she was talking with. Moreover, the video was of a lactation consultant, and not related to shared decision-making after cesarean simply because there are currently no existing ECAs that address the topic. These factors likely influenced how both women and providers perceived the ECA's potential role as a decision support tool for their decision-making process. As previously noted, the ECA demonstration represented an early prototype, and more recent ECAs are more sophisticated and have better graphics and audio quality.

Implementation for Practice

This study supports and adds to the literature surrounding the feasibility and benefit of ECAs to help women and their providers better prepare to decide on the type of birth each woman wants to have. The need to develop and use this type of technology has become more urgent during pandemic situations such as COVID-19. When routine prenatal care as well as childbirth education classes are moved to phone or video telehealth formats to minimize the risk of spreading of infection, gaps in information exchange can occur. Developing the ECA for wider use with easy access could help fill a critical care gap and enhance shared decision-making about the mode of birth after cesarean. This is particularly important when counseling women from ethnic minorities and/or women with low health literacy. Having an ECA developed in different languages, tailored to meet the cultural values and needs of individual women, could potentially help fill the gap for better shared decision-making on mode of birth after cesarean.

CONCLUSION

It is critically important to ensure that the implementation of new technology to support shared decision-making meets the needs of users and is valued as a means of enhancing the quality of the decision-making experience. Overall, the ECA was regarded as a potentially useful source of information for women to supplement face-to-face counseling with providers about mode of birth after cesarean. However, future work is still needed to improve the user experience, as well as to clarify the role and timing for access and utilization of decision aids with integrated ECA technology.

Biographies

SOMPHIT CHINKAM is a staff nurse-midwife at Boston Medical Center and a clinical assistant professor of obstetrics and gynecology, Boston University School of Medicine, Boston, Massachusetts.

COURTNEY STEER-MASSARO is a staff nurse-midwife at Boston Medical Center and an assistant professor of obstetrics & gynecology, Boston University School of Medicine, Boston, Massachusetts.

IVAN HERBEY is an adjunct faculty at UAB School of Medicine, Alabama.

ZHE ZHANG is a PhD candidate in the Khoury College of Computer Sciences at Northeastern University.

TIMOTHY BICKMORE is a professor in the Khoury College of Computer Sciences at Northeastern University #202, West Village Residence Complex H, 440 Huntington Ave, Boston, MA 02115.

ALLISON SHORTEN is a professor and department chair in the School of Nursing at the University of Alabama at Birmingham, and Director of the UAB Office of Interprofessional Curriculum.

DISCLOSURE

The authors have no relevant financial interest or affiliations with any commercial interests related to the subjects discussed within this article.

FUNDING

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Obstetrics and Gynecologist department seed grant, Boston University School of Medicine, Boston University, awarded to Somphit Chinkam.