Measure Development and Validation of a Self-Efficacy Scale for Advance Care Planning

A disparity exists between the medical intervention people say they want to receive at end-of-life and the care that is typically delivered. Advance care planning (ACP) involves discussing end-of-life care wishes, including relevant values and cultural beliefs, and documenting these preferences for medical providers and loved ones to minimize unwanted suffering and maximize quality of life. Numerous healthcare institutions have emphasized the importance of doing ACP prior to an imminent medical need, prompting researchers to implement awareness campaigns and interventions in earlier stages of healthcare interactions (e.g., primary care). However, motivation to follow through with ACP varies depending on numerous factors including overall readiness, understanding of the risks and benefits, and how one manages the internal experience of facing one’s own mortality. One intrinsic experience that has been shown to be important for health behavior change is situation specific-confidence, or selfefficacy. This work builds on previous research that approaches ACP intervention from the theoretical framework of the Transtheoretical model (TTM) of behavior change, in which self-efficacy is a core component. Study 1 of this dissertation sought to explore the construct of self-efficacy specific to doing ACP with qualitative work including expert interviews and focus groups with older adults in the community about their experiences. The work presented describes the efforts to understand self-efficacy as a barrier to engagement in end of life care planning. Self-efficacy was associated with interpersonal support, access to structured tools to guide discussions, and tolerance of the unpleasantness of negative emotions. Assessment of themes from focus groups and expert interviews was conducted to write items of a scale of self-efficacy to do ACP. Study 2 of this dissertation describes the development and validation of a scale of ACP self-efficacy using a sequential approach to measure development. Qualitative and quantitative methods were utilized for item development/refinement and scale validation. Split-half validation procedures were conducted, with exploratory and confirmatory factor analyses on randomly selected subsamples. The results of several iterations of exploratory factor analyses supported a final set of 12 items loading on one factor, with high internal consistency. The final 12-item ACP self-efficacy scale was found to have good overall model fit in confirmatory analyses, assessed with χ tests of significance and fit indices. Further, the developed scale was validated using previously developed TTM measures of ACP behavior change (Stage of Change, Decisional Balance) and related constructs (General Self-Efficacy, Attitudes Values & Cultural Beliefs). As expected, ACP self-efficacy varied by stage of change, with those in more progressed stages endorsing higher levels of confidence that they could complete ACP behaviors, upholding the relationships hypothesized by the TTM. Together, these two studies address the importance of self-efficacy for engagement in complex behaviors and provide a tool for future use to gain a deeper understanding and increase behavior change in this area.

limited education on what is involved in ACP, fear of thoughts related to death and dying, and avoidance of stressful or emotional conversations with loved ones. 5 With the proportion of older adults in the global population growing rapidly and the increasingly available medical technology allowing for the long-term management of complex medical issues, the costs of end-of-life care on a societal level will also continue to increase. Providing additional treatments that decrease patients' desired quality of life and result in unnecessary expenses and emotional turmoil may be understood as unethical and inefficient. 6,7 Understanding a patient's care preferences, concerns, and the values and beliefs that drive their reasoning are important for healthcare providers to ensure that they are providing patients and their families with the highest quality end-of-life care. It is also important that families and loved ones are made aware of the patient's care wishes and that their experience through the end-of-life care process is acknowledged and respected by the medical community. 8 If barriers exist between patients, their loved ones, and providers in communication about ACP, confusion and tension may impact the quality of the care experience for all involved. 9,10 The primary responsibility for broaching discussions about how people want to be cared for lies with the treatment team, but completion of ACP requires personally evaluating one's own wishes and discussing these with loved ones and medical providers. It is a deeply personal and inherently interpersonal process.
Advance care planning initiatives must acknowledge barriers that may discourage patients from engaging in care planning and seek to bolster confidence that ACP is feasible and worthwhile. Confidence in one's own ability to make effortful, meaningful changeparticularly when confronted with emotional, social, or physical barriers to do sois referred to as self-efficacy. 11 Behavior change researchers across theoretical approaches have identified self-efficacy as a critical component of explaining and intervening upon individuals' motivations and readiness to change a specific behavior. [12][13][14][15] Enhancing self-efficacy has been shown to improve situation-specific performance by increasing the effort put forth toward initiating or persisting in behavior change. 16 Researchers and practitioners have increasingly emphasized the application of health behavior models to ACP initiatives. [17][18][19] Through a series of qualitative and quantitative studies, prominent behavior change and gerontology researchers have conceptualized ACP as a set of separate yet interrelated steps including completing a living will, communicating wishes regarding life-sustaining treatment with a healthcare proxy or agent, and discussing quality versus quantity of life issues with family members and health care providers. 17 Themes that emerged from interviews and focus groups with older adults focused on ACP experiences mapped onto constructs outlined in the Transtheoretical Model of behavior change (TTM), such as one's readiness to engage in ACP. 17 These findings suggested that the TTM might be a good fit for interventions aimed at increasing ACP behaviors. Follow-up measure development studies by Fried and colleagues validated TTM-based scales of Stage of Change, Decisional Balance, and Processes of Change for ACP with older adults in primary care settings. 20,21 Further, ongoing interventions have been designed to utilize individuals' TTM based survey responses to provide tailored feedback over the course of several months, in order to address motivations to engage in ACP. 18 The TTM is a highly researched and well-supported model of intentional behavior change that has been applied to interventions for end-of-life care planning and dozens of behaviors across settings, racial/ethnic populations, and cultures. 21,22 The TTM frames an individual's readiness to change a behavior (Stage of Change) as associated with their perceived confidence to make a change (Self-Efficacy), the perceived benefits and negative consequences of making a change (Decisional Balance), and the use of cognitive/behavioral activities that facilitate change (Processes). Stage of Change is defined conceptually as a continuum through which an individual can move forward or regress in stages; however, it is typically measured as a categorical variable. At a given point in time an individual would be in one of the stages of change for a specified behavior. These stages and typical associated time frames are Precontemplation (not thinking of changing in the next 6 months), Contemplation (thinking about changing in the next 6 months), Preparation (thinking about changing in the next 30 days), Action (made a change in the last 6 months), and Maintenance (made the change more than 6 months ago).
The TTM is an integrative theory of behavior change and the key constructs have been shown to be predictive of behavior change and have been found to be useful for intervening on behaviors that may be particularly resistant to change (e.g., smoking, domestic violence). [23][24][25] Identifying important processes of change for a specific behavior and aligning interventions to address barriers and resistance could help people feel more confident and willing to make changes. 24,26 The integration of constructs makes the TTM a powerful theory for conceptualizing and designing interventions for changing complex behaviors, including planning for end-of-life care. The construct of situational selfefficacy has been shown to be important for behavior change interventions across numerous age groups and cultural contexts, including exercise behaviors in older adults in the U.S. 27 and smoking in Korean adolescents. 28 Self-efficacy has also been shown to be important for interventions that address complex sets of behaviors, such as achievement in schools 29 and integrated service delivery in higher education. 30 To the knowledge of various experts in the field of ACP and behavior change and following a thorough literature review, no scales assessing ACP self-efficacy have been developed and validated. The current project aimed to apply a well-established theory of behavior change to the exploratory phase of a measurement development project for a scale of ACP Self-Efficacy. Using the previously developed measures of engagement in ACP grounded in the TTM as a foundation and the extensive work across behavior change theories which identify self-efficacy as an important construct, this work added to this research area by engaging experts and potential stakeholders (i.e., older adults in the community) in exploration of how confidence to do ACP may impact participation and to assess how a validated measure of ACP self-efficacy may support research and patient care in this area. Exploring the construct of self-efficacy to do ACP within the framework of the TTM addresses a gap in the literature by focusing on the situation-specific confidence that likely impacts engagement in an important public health area. 1

Sample
Expert Interviews. The present study utilized purposeful sampling in order to assemble professionals in the areas of end-of-life care, gerontology, and palliative care that have clinical, research, policy-making, and community-based experience with ACP. 31,32 Several experts who have extensive experience with theory-driven construct exploration and measure development were approached in order to frame discussion guides for interviews and focus groups. Snowball sampling was used such that experts were asked to recommend another expert that would then be approached for further key informant interviews. 33 Focus Groups. Engaging the population of interest in this research area necessitated inviting older adults to share their thoughts, feelings and experiences with advance care planning. 34 Over and above individual interviews, focus group settings allow for the potential for normalization of a common experience and may encourage sharing of personal experiences. 35,36 Older adults who were participating in ACP education initiatives at several senior centers across the state of Rhode Island, provided through the Care Transformation Collaborative (CTC) of Rhode Island, were invited to participate in one-time, 30-60 minute group discussions about ACP following a CTC presentation. The participant pool was limited to individuals over the age of 50 who spoke English.

Procedures
Expert Interviews. Key informants were approached via email with an invitation to participate in a one-time interview at the location of their choice or over the phone. They were sent an informed consent form to review via email before the interview and signed either electronically or in-person. Informed consent included a request to audiotape conversations for future review by the student researcher and research team. Interviews were semi-structured with questions related to experiences with ACP in their work and factors they believed to be important for patients and families' engagement and overall confidence to engage in ACP (Table 1). Participants were asked explicitly about barriers to ACP participation and completion, and how they have engaged patients who were less confident about completing the process.
Focus Groups. Focus group participants were invited to participate in a one-time group discussion on ACP at their local senior center. One group of participants was invited by the student researcher at the conclusion of the presentation by the CTC of Rhode Island on ACP. The second group of participants was assembled with support from an employee at a senior center that had previously hosted the CTC of Rhode Island for a presentation on ACP. All participants reviewed informed consent, were informed of the opportunity to ask questions of the student researcher and supervising professor in person or via phone or email, and consented to have their discussion audiotaped. Group discussions were guided by questions the research team prepared to assess experience with ACP (e.g., thoughts, actions), factors that influence confidence to talk about end-of-life care with loved ones and healthcare providers, and contributors to follow-through with required steps to document their wishes ( Table 2).

Participants
Expert Interviews. Three expert interviews were conducted, two in-person and one via phone. The participants included a physician who led a palliative care team at a local hospital who regularly supported patients in quality of life discussions and end-of-life care decision-making; a physician with a clinical and research focus in gerontology who regularly works with older adults and families to determine decision-making capacity to engage in ACP and also engages in local and national discussions on ACP policies; and a registered nurse with experience treating patients and discussing ACP at various stages of treatment on healthcare teams within palliative, intensive care, and community settings with older adults and families.

Data Analysis
Theoretical thematic analysis was used when listening to key informant and focus group recordings. This type of analysis is concerned with the reported experiences and the meanings of these experiences for the participants and it acknowledges that the data collection and analysis is guided by a theoretical structure. 37 Theory-driven qualitative methods requires staying "open-minded yet mindful of the preconceptions (including theoretical leanings) one has entering a field of study." 32 Two raters with knowledge of literature on TTM based behavior change initiatives and ACP listened to audiorecordings of two expert interviews (one was not audiotaped as it was held via phone) and the two focus groups. Raters both utilized the guiding theoretical framework of the TTM and approached data independently to identify patterns within and across individual interviews and group discussions.
Raters listened to conversations for emergent themes related to self-efficacy to participate in three ACP behaviors: completing a living will, documenting a choice of healthcare agent, and discussing quality versus quantity of life issues with loved ones.
Raters focused on issues that were raised often by multiple participants and whether certain issues appeared to be strongly related to confidence for one or more participants.
After listening to all recordings, raters discussed concepts that they identified from the stakeholders' perspectives, how themes related to the literature on ACP and how participants framed the construct of personal self-efficacy to engage in ACP. Debriefing discussions among the research team designed to aid in writing items that captured elements of the ACP self-efficacy construct.

Overview of Themes
Expert Interviews. The Key Informant interviews yielded several themes related to the general lack of education on the importance of ACP and the fear that most people feel when prompted with ACP questions and discussions by loved ones or providers. Themes related to the goals of care conversations included normalizing the process of discussing quality of life as it relates to medical care decisions. All three experts identified the impact of fear, denial, and procrastination to completing ACPparticularly in engagement in conversations between patients and their loved ones. A central theme was the importance of family members and loved ones in the process of care planning.
Participants discussed the fear of upsetting loved ones as a common barrier for their patients. One participant noted that they regularly tell patients "one of the best gifts you can give to your family is the gift of preparedness." Participants discussed the benefits of recent initiatives (e.g., the CTC of Rhode Island) and published guides such as The Conversation Project that assist in structuring these conversations. 38 As end-of-life care discussions appear to be universally difficult and fear-inducing, conversation guides and education initiatives benefit patients and providers by normalizing the discomfort they may feel and offering a step-by-step process to follow.
Key informants also universally discussed the need for ACP to include discussions with loved ones about quality versus quantity of life, stressing that an advance directive/living will is not sufficient. Although certain medical forms may outline interventions that patients would or would not want, complex scenarios often arise and surrogate decision-makers must be well versed in a patient's values and wishes in order to make an informed decision. For example, if long-term quality of life may be salvaged with temporary usage of a respirator, a patients' healthcare agent must weigh this with their loved one's stated wish to not be kept alive with the assistance of machines. One participant reported that, in her work as a palliative care physician, she has found the discussions with loved ones about quality versus quantity of life to be the most important component of ACP.
Focus Groups. Across the two focus groups, themes emerged related to the lack of understanding and education about what is involved in ACP (e.g., mistaking ACP for having completed a will or trust). Participants endorsed denial of mortality and the desire to avoid unpleasantness as barriers to doing ACP (e.g., "you think you're going to live forever…"). Participants also discussed the importance of being able to communicate wishes to their loved ones and the benefits of having people to hold you accountable to complete ACP. If they felt like they could not talk about their wishes or their experiences with anyone, either because they received negative feedback or they did not have anyone they felt this close to, they were less likely to follow through in doing so.
Overwhelmingly participants noted that witnessing or personally experiencing a dire medical situation that necessitated discussions of end-of-life care wishes underscored the importance of ACP. Until confronted with what might happen if ACP is not completed (i.e., witnessing suffering not consistent with someone's beliefs on quality versus quantity of life), participants noted denial of the necessity to do ACP sooner rather than later. Numerous participants talked about wanting to ensure that their wishes were adhered to so as to maximize quality of life and minimize suffering. Several participants discussed having multiple conversations with their adult children to reinforce these wishes. Confidence to do ACP appeared to be related to many things including feeling supported by others (e.g., family, friends, doctors) in the process, recognizing the necessity to follow through regardless of emotional discomfort, and having experience discussing end-of-life for themselves or for others.

Discussion
The information gathered and analyzed in this study formed the foundation for a measure development project to identify key components of the construct of self-efficacy to do ACP. In order to compose a list of items to be tested using sequential measure development, exploration of a measured construct must be grounded in the voices and experiences of potential stakeholders. 39,34 The utilization of a well-established theoretical model of behavior change as a framework for this line of research allows for guided analysis of themes identified in interviews and focus group discussions about the topic. 37 A theory-driven and thoughtfully developed and validated measure of confidence to engage in care planning behaviors, including completing a living will, documenting a healthcare proxy, and discussing quality versus quantity of life issues, could add to the clinical and research aims of addressing this public health issue.
Older adults who participated in the current study discussed how naïve they felt about the need for and the steps involved in doing ACP before having been confronted with a situation that prompted them to consider it. Several participants wished that there were wider education initiatives to communicate the importance of ACP to younger generations. ACP interventions have historically focused on ensuring that an advance directive has been completed, including completion of a living will and appointment of a durable power of attorney for health care (i.e., an individual who will be responsible for medical decisions; surrogate decision maker). 2,40 However, these documents are not sufficient in outlining how an individual may feel about complex situations that were not covered on check-box forms. More recently, additional interpersonal behaviors such as having discussions with medical providers and family members have been identified as important components. 17,41 Discussing quality versus quantity of life was found to be vital. 17 These discussions include consideration of situations in which a person would prefer not to receive medical intervention because they believe it would diminish their ability to live a life that brings them joy and comfort. Engaging patients and loved ones in discussions about advance care planning can be challenging due to the discomfort in discussing death or 'states worse than death.' 42 Patients may feel less confident broaching these topics with loved ones, including those they may name as surrogates or health care agents, if they believe these conversations may upset them. 20 Patients may consider situations that would be worse than death and times when living with a diminished quality of life would be more important than living as long as possible. 21 When patients and families are able to discuss options and learn the downsides that life-sustaining measures entail (including potential unnecessary suffering), many endorse that the burdens outweigh the benefits and would choose to discontinue or decline life support. 43 Adhering to a dying patient's wishes regarding their desired quality of life may vary depending on whether they would, for example, prefer to be kept alert even if they experience pain or prefer to forego life-sustaining treatments in the form of feeding tubes or intubation. Individuals may choose to decline all life-sustaining treatments or interventions if they feel they are ready to die when it is their time, if they believe the burdens would be too great, or if they believe that the use of these treatments would cause extreme stress for their loved ones. 43,44 Having ACP discussions, including situations where an individual would prefer to forego life-sustaining treatment, could be an emotional process for patients and families. Some patients may avoid these ACP discussions with family members in order to minimize the stress they predict their loved ones will experience. 17 However, it has been shown that not having these discussions could actually increase stress for family members if they become involved in treatment discussions. 8,10 Attitudes toward quality versus quantity of life and care preferences vary based on life stage and health context. Comfort for end of life care discussions and decisions likely varies depending on experiences with death and consideration of options. Although older adults may be more willing to discuss these topics, they have not been shown to have higher rates of actually making end-of-life care plans and may prioritize living in the moment (i.e., 'cross that bridge when necessary') over engaging in uncomfortable and anxiety-provoking conversations with loved ones or HCPs. 5 It would be unethical for health care providers to avoid conversations about ACP and assume then that patients will discuss ACP when they are ready or wait until they feel completely comfortable with these topics. Providers may also feel uncomfortable, unprepared or not have enough time for ACP during patients' regular appointments. Community initiatives may be helpful in these instances to bring ACP materials to older adults and their loved ones sooner rather than later.
Outlining care preferences consistent with patients' values in order to maximize quality of life throughout health care decision-making should be the goal of health care providers who seek to do no harm for their patients and health care organizations who provide support for their consumers. Acknowledgement and assessment of one's confidence to carry through with ACP once it is made a goal may be helpful for providers and family members alike. The Consumer's Toolkit for Health Care Advance Planning 45 and the Starter Kit by The Conversation Project 38 provide advice for patients who might encounter resistance or defensiveness from their loved ones when they initiate ACP conversations. Although these may be helpful suggestions for many patients, the existing tools do not acknowledge that patients and family members may be at varying levels of readiness, degrees of confidence, and may perceive fewer benefits to having these conversations sooner rather than later. The application of the TTM to ACP allows for a well-rounded approach to behavior change, framed by a model that does take into account these factors and has the flexibility to include other constructs related to the research area (e.g., how acceptable one finds a health state, 46   What is your experience with advance care planning in your work?
• What is the setting and stage of care that you typically work with patients and families? • What is your comfort level and competence in this area?
From your clinical perspective, what are the things that seem to get in the way of people doing ACP?
• Specifically I'm interested in what appears to get in the way of patients completing a living will, identifying and documenting a healthcare proxy/agent, and having conversations with loved ones (and providers) regarding quantity versus quality of life issues.
What are the issues that you've seen people really grapple with?
If you were in charge of ACP broadly in society, what would you do differently in order to address this? Setting: Participants were recruited online using a targeted survey sampling company.
Participants: Scale development was based on responses of 310 people aged 50 or older.
Recruitment quotas were placed to ensure a census-based, nationally representative sample across categories of gender, age, region and education.

Measurements:
The full sample was randomly split into two subsamples used for exploratory and confirmatory factor analyses respectively. Item selection and refinement was conducted quantitatively by examining factor loadings and qualitatively by reviewing item content across seven iterations with the exploratory sample. Scale structure was confirmed using confirmatory factor analysis, examining chi-square tests of significance, fit indices and mean square error approximations. External validation of the scale was conducted using ANOVAs to assess the relationship of self-efficacy to constructs outlined in the TTM (stage of change, decisional balance).

Introduction
The Engagement depends on intrinsic (e.g., sense of personal control, dignity beliefs) and extrinsic (e.g., influence from society or one's family) factors. 4 Advance care planning has also been conceptualized as a set of health behaviors that can be addressed with initiatives grounded in well-researched, theoretical frameworks of behavior change. 5 Taken together, past research on engagement in ACP and motivation of behavior change suggests that future programs must focus on identification of the relevant barriers and attempt to meet the individual where they are in terms of personal self-efficacy to complete ACP. This study builds on previous work outlining ACP as a set of health behaviors and reviewing the importance of addressing self-efficacy in regards to motivation to complete them. [5][6][7] The objective of this work is to develop a validated scale of ACP self-efficacy.

Advance Care Planning
Completion of end-of-life care planning requires a process of evaluating one's own wishes and discussing these with loved ones and medical providers. Effective interventions must acknowledge barriers that often discourage patients and their loved ones from engaging in care planning and seek to bolster education about why ACP is feasible and worthwhile. 8 Historically, ACP interventions have mainly focused on ensuring that an advance directive (AD) has been completed, including completion of a living will and appointment of a durable power of attorney for health care (i.e., an individual who will be responsible for medical decisions; surrogate decision maker).9,10 More recently, additional interpersonal behaviors such as having discussions with medical providers and family members have been identified as important for ACP. 5,11 Discussing quality versus quantity of life has been found to be a vital component of ACP 5  Some patients may avoid these ACP discussions with family members in order to minimize the stress they predict their loved ones will experience. 5

Application of a Theory of Behavior Change
Through a series of qualitative and quantitative studies, prominent researchers have conceptualized ACP as a set of separate yet interrelated components included completing a living will, communicating wishes regarding life-sustaining treatment, discussing quality versus quantity of life issues with family members and health care providers, and identifying a healthcare agent or surrogate decision maker. 5 In a series of studies utilizing interviews and focus groups, experiences that older adults shared with researchers involved in ACP mapped onto constructs outlined in the TTM; for example, processes of change, readiness and decisional balance. [5][6][7] These findings suggested that the TTM might be a good fit for interventions aimed at increasing ACP behaviors.
The TTM is a highly researched and well-supported model of intentional behavior change that has been applied to interventions for dozens of behaviors, including ACP, across settings, racial/ethnic populations, and cultures. 20,6 The TTM frames an individual's readiness to change a behavior (Stage of Change) as associated with their perceived confidence to make a change (Self-Efficacy), the perceived benefits and negative consequences of making a change (Decisional Balance), and the use of cognitive/behavioral activities that facilitate change (Processes). 21  Maintenance (made the change more than 6 months ago).
The TTM is an integrative theory of important behavior change constructs which have been shown to be predictive of behavior change and have been found to be useful for intervening on a series of health-related behaviors, especially those that may be particularly resistant to change (e.g., smoking, domestic violence). [21][22][23] Identifying important processes of change for a specific behavior and aligning interventions to address barriers/resistance could help people feel more confident and willing to make changes. 23,24 The integration of constructs makes the TTM a powerful theory for conceptualizing and designing interventions for behavior change. The construct of situational self-efficacy, specifically, has been shown to be important for behavior change interventions across numerous age groups and cultural contexts, including exercise behaviors in older adults in the U.S. 25 and smoking in Korean adolescents. 26 Self-efficacy has also been shown to be important for interventions that address complex sets of behaviors, such as achievement in schools 27 and integrated service delivery in higher education. 28

Self-Efficacy
Confidence in one's own ability to make effortful, meaningful changeparticularly when confronted with emotional, social, or physical barriers to do sois referred to as self-efficacy. 29 Also understood as situation-specific mastery, personal selfefficacy is a cognitive experience that is related to one's past experiences, abilities and skills, as well as expectations about future success or failure. 26,30 Researchers across theoretical approaches to behavior change have identified self-efficacy as a critical component of explaining and intervening upon individuals' motivations and readiness to change a specific behavior. 21,30,31 Enhancing self-efficacy has been shown to improve situation-specific performance by increasing the effort put forth toward initiating or persisting in behavior change. 30 Interventions that aim to motivate individuals to start, stop, or continue to engage in any specific behavior could be improved by acknowledging the potential barriers that may discourage progress and confidence to persist despite the barriers. that measures the extent to which values such as religious beliefs or the desire to live mindfully might impede engagement in ACP. 7 The AVCB scale is not a proxy for selfefficacy, as it does not measure an individual's confidence that they can participate in ACP; rather, it highlights specific attitudinal barriers that may be important for researchers and clinicians to address before an individual may acknowledge that ACP may be right for them.

Collaborations between experts in ACP and
Self-efficacy has been highlighted as a key variable in population-based initiatives and individual interventions aimed at changing complex behaviors. 27,28,32 When applied to a specific situation, the construct of self-efficacy refers to how confident an individual feels they can engage in a particular behavior even when they are faced with certain barriers. Developing a validated measure of self-efficacy for ACP will address a gap in the literature that conceptualizes ACP as a set of health behaviors (i.e., completing a living will, identifying a healthcare agent, discussing quality versus quantity of life issues) by focusing on the situation-specific factors that threaten people's confidence to engage. The purpose of the current study was to build on previous work to develop and validate a scale of ACP self-efficacy and to examine the associations between the measured construct and other variables within the framework of the TTM (e.g., stage of change, decisional balance, processes of change).

Design
A sequential method of measure development was utilized to develop a scale of ACP self-efficacy. 33 Item development and scale structure was supported by input from experts and community members (described in detail elsewhere, Murray 2019), followed by recruitment of an independent sample for testing of the scale structure via exploratory, confirmatory and external validation analyses.

Item Development and Refinement
Initial item development was grounded in a comprehensive review of TTM measures of self-efficacy for other behaviors (e.g., smoking, 26 school achievement, 27 integrated service delivery in higher education 28 ) and from literature on the process of ACP across settings. 5,[9][10][11]14 Experts in the areas of scale development and behavior change interventions were consulted throughout all stages of measure development.
Detailed methods of formative work used to write scale items are outlined elsewhere (Murray, unpublished manuscript, 2019). This process included interviewing experts in clinical, research and policy making efforts to support ACP; conducting focus groups with older adults in the community; and reviewing themes and revising written items with feedback from experts on scale development.

Recruitment & Survey Administration
All materials, surveys and procedures were approved by the Institutional Review Individuals were first asked to review informed consent and were directed to survey questions upon agreement. Data were extracted from Qualtrics into SPSS for exploratory analyses and to EQS for confirmatory analyses. To protect confidentiality, no identifying information was distributed to the researcher from Cint and none of the participants' responses to survey questions were accessible to Cint.

Sample
Participants (N = 310) for the measure development survey were invited to participate through an online survey and panel recruitment company. Eligible participants were adults 50 years old or older who could read and write in English. Quotas were utilized to ensure a nationally representative sample (25% each from Northeast, South, West, and Midwest regions), even representation across men, and a balanced age range for adults over the age of 50 (50-64yo, 40%; 65-75, 40%; 76-99, 20%).

ACP Self-Efficacy.
A total of 35 items were written to represent self-efficacy to do ACP; defined as completing a living will, documenting a healthcare agent, and discussing quality versus quantity of life issues with loved ones. Participants rated how confident they felt that they could do ACP even if the proposed statement were true for them.
If an item did not feel relevant (e.g., "You have not discussed advance care planning with religious or spiritual leaders"), participants were asked to rate this item with a score of 1 ("Not at all confident") to standardize this response pattern across the sample.

TTM Measures. Previously developed TTM-based measures of Stage, Decisional
Balance and Processes of Change for ACP were included in the final survey. 6,7 Stage of Change was measured with two to three questions for each behavior to assess readiness to complete a living will, document a choice of healthcare agent, and discuss quality versus quantity of life issues with loved ones . 7 Participants were asked if they have completed each behavior and, if completed, whether it was done in the past six months (Action) or before (Maintenance); if not completed, whether they were thinking about doing so in the next 30 days (Preparation) or next six months (Contemplation). If they were not planning to complete the behavior in the next six months, they were staged in Precontemplation.
Decisional Balance (i.e., Pros and Cons) for ACP was measured with a 12-item scale in which participants were asked to rate the importance of each item (e.g., Doing advance care planning would simplify how decisions would be made if I were very ill) in their decision whether or not to participate in ACP, from 1, "Not at all important," to 5, "Extremely important." 7 Use of social and cognitive-behavioral strategies that support engagement in ACP, called Processes of Change within the TTM, were measured using a 9-item measure (e.g., I can count on my loved ones to help me with advance care planning). 7 Participants were asked to rate how frequently they used each of the processes in the past month on a five-point scale from 1, "Almost never" to 5, "Almost always."

Related Constructs.
Along with the developed TTM based Self-Efficacy Scale for ACP, theoretically related constructs were also measured for validity testing and further field development. The Attitudes, Values and Cultural Beliefs (AVCB) scale is an 8-item measure of the strength to which an individual endorses various medical misperceptions about participating in ACP, such as beliefs that "I will always be able to make my own treatment decisions." 7 A measure of the Acceptability of Health States (AHS) was included to assess attitudes towards quality of life and states worse than death. 8,12 The three conditions in the AHS measure represent mental and physical states (being bedbound, severe cognitive impairment, enduring severe pain) with reliably diminished quality of life for many people. In previous work, these items were measured on a binary scale of whether these states would be "acceptable" or "unacceptable." The scale for these items was changed to a continuous scale to increase the possibility of variability in responding and to allow for more formal statistical analyses to assess their relationship with the other measures.

Data Analysis
The sample was randomly split and the first half (N=168) was used for the exploratory phase using factor analysis with Varimax rotation on item correlation matrices. Assumptions of normality were tested on the exploratory sample. The number of retained components was based on the minimum average partial procedure (MAP), parallel analysis, and examination of item content. 37 Item selection and retention was an iterative process that involved quantitative and qualitative analysis. Factor loadings were examined and those with <.40 were removed; no item loadings were greater than .90. Inter-item correlations were assessed for multicollinearity. Item pairs with correlations >.70 were assessed for qualitative breadth of construct and to avoid redundancy. Seven iterations of exploratory analyses were performed and involved evaluations of factor loadings and item content as well as discussion of theoretical subcomponents of self-efficacy and literature review.
Cronbach's alpha was examined to determine the internal consistency of the scale in the exploratory sample.
Confirmatory factor analysis (CFA) was performed using the second half of the measure development sample (N=142) to evaluate the fit between the hypothesized model and the sample data. 38,39 Confirmatory factor analysis was utilized as this test accounts for the presence of measurement error, in contrast to principal components analysis that assumes no measurement error, which provides a better estimate of the true variance accounted for by the scale items in the measurement of the underlying variable. 38 Model fit and factor loadings were evaluated for final item selection, which included determination of item clarity, breadth of concept and minimization of redundancy. Along with χ 2 significance tests, the comparative fit index (CFI) and root mean-square error of approximation (RMSEA) were used to assess overall model fit. 40,41 Good overall fit is typically indicated by CFI values greater than .90, with values greater than .95 being ideal. 42,43 RMSEA values less than .10 are seen as acceptable or evidence of mediocre fit, with values < .08 indicating good fit and <.05 indicating very good fit.
The scale's internal consistency was assessed further with Cronbach's alpha.   Table 1. The number of participants within stage groups varied greatly with half of the sample being in the maintenance phase across the three behaviors, having already completed ACP more than six months prior to the survey ( Table 2).

Item development and refinement
The full sample was split randomly using SPSS to yield independent samples for exploratory (N=168) and confirmatory (N=142) factor analyses. Demographic variables such as gender, age, and education did not vary significantly between the two subsamples.

Exploratory analyses
The original 35 self-efficacy items were included in an exploratory factor analysis using SPSS with the first half of the total sample. Minimum average partial procedures and parallel analysis originally suggested six and three factors respectively, one item was removed after the first step. Through an additional 6 factor analyses, the remaining 34 items were reduced to 12 with MAP and parallel analyses each suggesting one factor in the seventh iteration. The final analysis demonstrated that all item loadings were >.5, ranging from .515-.902. The internal consistency for the self-efficacy scale was high, with a Cronbach's alpha value of .95.

Confirmatory analyses
The final 12 self-efficacy items identified in the exploratory phase were included in confirmatory factor analyses using the second half of the split subsample (Figure 1).

External Validation
To test the validity of the ACP self-efficacy measure, ANOVAs were conducted to test how self-efficacy varied across stages of change for each of the three ACP behaviors, all were significant. Confidence to complete a living will (F(4, 305) = 2.520, p = .041, η 2 =.032), document a healthcare agent (F(4)=3.236, p=.013, η 2 =.041), and discuss quality versus quantity of life issues with loved ones (F(4) = 3.261, p = .012, η 2 =.041) significantly differed by stage of change (Table 3). Effect sizes were small to medium.
Post-hoc analyses indicated that self-efficacy to complete a living will was higher for those in maintenance than those in precontemplation; those in action for documenting a healthcare agent were more confident than those in precontemplation; and those in action and maintenance were significantly more confident regarding discussions of quality versus quantity of life issues than for those in precontemplation. There were not enough participants in the preparation stage of change to identify differences in self-efficacy between these individuals and those in pre-action or post-action stages.
For further validation of the construct of ACP self-efficacy, the developed scale was compared with constructs that are theoretically similar for convergent validity.
General self-efficacy was found to be significantly correlated with ACP self-efficacy (r=.26, p<.001). Social desirability was not found to be associated with ACP selfefficacy.

Further Validation of Associated Constructs
Self-efficacy was significantly associated with related constructs, including previously developed TTM-based measures of decisional balance and attitudes, values and cultural beliefs (AVCB) associated with ACP. The developed self-efficacy scale was compared across stages to decisional balance ( Figure 2). Overall, greater ACP selfefficacy was associated with greater endorsement of Pros (r = .454, p <.0001).
Interestingly, greater self-efficacy was also associated with greater endorsement of Cons  20 This project adds to the literature base that addresses ACP as a set of health behaviors by emphasizing the role of self-efficacy to complete these emotionally and logistically complex behaviors and provides a scale that can be used in assessment and intervention efforts designed to increase ACP.
Bandura noted that it is imperative to specify the context/behavior in question when constructing scales of self-efficacy to optimize the explanatory and predictive utility of the measure. 44 Given the complexity of the behaviors addressed within this line of research, isolating the construct of self-efficacy to do several behaviors and persist through barriers proved challenging when writing and revising scale items. However, results of this work suggest that the developed scale did indeed explain a portion of the overall variance for completing ACP, which underscores the importance of including self-efficacy in TTM-based interventions to increase engagement in these behaviors. In this study, participants were asked to respond on a 5-point confidence scale based on their perceived capability to complete ACP even if faced with specific barriers.
Recognizing self-efficacy as a contributor to ACP motivation and engagement can assist future researchers more effectively tailor interventions. Within this study, self-efficacy was found to account for a small-to-moderate amount of variance (based on partial eta squared) across the three ACP behaviors, partially indicative of the complicated nature of this issue and the number of other factors involved in completing these behaviors. As part of a larger program, assessment of self-efficacy and addressing low confidence in this area could increase overall effectiveness of interventions.
The model fit for the developed self-efficacy scale was good, particularly given the complexity of the construct of situational self-efficacy as it relates to three behaviorally and emotionally multifaceted behaviors. 40,42 Notably, the number of participants within stages varied greatly, limiting the power to evaluate differences found between stage groups during validation procedures. Results presented here must also be taken within the context of strong theoretical foundation for self-efficacy as a central component of TTM based behavior change research. 45,46 The conceptual nature of the underlying factor, backed by strong theory, and the formative work of discussing selfefficacy to do ACP with stakeholders in healthcare and community settings, lend additional credence to the value of the constructed scale for future use. Further, this project also served as follow-up validation of the TTM-based measures of ACP decisional-balance, stage of change, and AVCB and introduced a continuous measure of acceptability of health states for use in future studies.
Much of the previous research on advance care planning has been conducted with older adults, primarily those with terminal illnesses such as advanced cancer. However, as population-based initiatives to increase ACP engagement expand, younger adults will be asked to consider their views on death and planning for end-of-life care earlier in their interactions with medical professionals or in community programs. Depending on past experiences with death, young adults (e.g., undergraduate students) have been found to vary in their wishes related to medical intervention. 47 Also, unlike care provision for patients managing terminal diagnoses, primary care providers and those in acute care settings are typically unable to spend extended time with patients to answer questions and assess motivation that may be multifaceted. Recently, brief/tailored ACP interventions have been developed and applied in primary care 48 and emergency departments 49 with promising feasibility. Healthcare providers and institutions will need to continually adapt interventions to meet the individual needs of the patients they serve while being mindful of population-based healthcare costs. This work serves to address a gap in the literature on motivating the public to engage in care planning by evaluating the impact of situationspecific confidence and presenting a new validated measure of ACP self-efficacy.
(a.) completing a living will, (b.) documenting a healthcare agent, and (c.) discussing quality versus quantity of life issues with loved ones.