Facilitating Communication for Critically Ill Patients and their Family Members: Study Protocol for Two Randomized Trials Implemented in the U.S. and France

Abstract

Background:

Critically-ill patients and their families suffer a high burden of psychological symptoms due, in part, to many transitions among clinicians and settings during and after critical illness, resulting in fragmented care. Communication facilitators may help.

Design and Intervention:

We are conducting two cluster-randomized trials, one in the U.S. and one in France, with the goal of evaluating a nurse facilitator trained to support, model, and teach communication strategies enabling patients and families to secure care consistent with patients’ goals, beginning in ICU and continuing for 3 months.

Participants:

We will randomize 376 critically-ill patients in the US and 400 in France to intervention or usual care. Eligible patients have a risk of hospital mortality of greater than15% or a chronic illness with a median survival of approximately 2 years or less.

Outcomes:

We assess effectiveness with patient- and family-centered outcomes, including symptoms of depression, anxiety, and post-traumatic stress, as well as assessments of goal-concordant care, at 1-, 3-, and 6-months post-randomization. The primary outcome is family symptoms of depression over 6 months. We also evaluate whether the intervention improves value by reducing utilization while improving outcomes. Finally, we use mixed methods to explore implementation factors associated with implementation outcomes (acceptability, fidelity, acceptability, penetration) to inform dissemination. Conducting the trial in U.S. and France will provide insights into differences and similarities between countries.

Conclusions:

We describe the design of two randomized trials of a communication facilitator for improving outcomes for critically ill patients and their families in two countries.

Introduction

The impact of critical illness is increasing due to our aging population and advances in effectiveness and availability of critical care.^1,2 Critically ill patients and their families suffer a high burden of symptoms of depression, anxiety, and post-traumatic stress due, in part, to fragmented medical care that is often poorly aligned with patients’ goals.^3–6 Fragmented care may arise from the numerous transitions patients and families experience across clinicians and settings, starting in the ICU and extending to acute care, often including returning to the ICU as well as transitioning to inpatient rehabilitation, skilled nursing facilities, or home.^7–9 Through these transitions, patients and families often struggle to navigate the spectrum of goals of care to match their goals with treatments, communicate goals to their clinicians, and make difficult medical decisions. Unfortunately, poor communication compounds an already stressful experience.^{1,2,7,8,10–14} and can lead to high intensity and unwanted care.^15,16 Ineffective communication has been documented in many countries, including U.S. and France, and differences between these countries have been described.¹⁷ We are testing whether an intervention to improve communication for critically ill patients and their families can reduce family distress in the U.S. and France.

Building on social cognitive theory^18–21 and our prior work,^3,22 we designed an intervention to improve outcomes for patients’ family using nurse facilitators to support, model, and teach communication strategies that enable patients and families to secure care aligned with patients’ goals over an acute episode of illness, beginning in the ICU. In a prior version of this intervention, facilitators were trained to apply communication, attachment, and mediation strategies to facilitate social cognitive components to support families. The social cognitive components were designed to: 1) build self-efficacy for communicating with clinicians; 2) improve outcome expectations that this communication can improve care; and 3) enhance behavioral capability through skill building to resolve barriers to effective communication and mediate conflict.²² In our prior trial, the intervention reduced symptoms of depression among family members of critically ill patients with a predicted mortality of at least 30% at 6 months and reduced hospital length of stay and costs, even after accounting for intervention costs.^22–24

Feedback from families identified limitations of our prior intervention. This feedback contributed to four specific improvements: (1) include critically ill patients with a lower severity of illness to increase proportion of eligible patients; (2) extend the intervention over 3-months, rather than restricting to the hospital stay, to enable support across transitions; (3) focus on eliciting and implementing patients’ goals of care; and (4) examine factors that can impede or facilitate future implementation and dissemination using a type 1 hybrid effectiveness-implementation trial.²⁵

These two trials, conducted simultaneously in the U.S. and France, have three specific aims. First, we will evaluate the effectiveness of the intervention on patient- and family-centered outcomes after critical illness. The primary outcome is family members’ burden of symptoms of depression over the 6 months post-randomization. Second, we evaluate the intervention’s effectiveness for increasing healthcare value. Finally, we use mixed-methods to understand factors and outcomes related to the implementation of this intervention. We chose to conduct two separate trials because of the dramatic differences in the healthcare cultures, the organization of healthcare systems, and the research funding mechanisms.

Methods

Overview:

We are conducting two cluster-randomized trials of a single intervention designed to improve outcomes for critically ill patients and their family. The trials are conducted in four hospitals in Seattle, Washington and four hospitals in France, including 2 hospitals in Paris ad one hospital in Brest and one in Nantes. In the US, we aim to randomize 376 patients to intervention (n=188) or control (n=188) and we estimate 564 family members (approximately 1.5 family member per patient). In France, we aim to randomize 400 patients to intervention (n=200) or control (n=200) and estimate 400 family members. US and French investigators choose slightly different numbers of patients and family members based on loss to follow-up and numbers of family members per patient from prior studies from each country.^3,22,26–30 The facilitators are experienced ICU nurses with specialized training to: 1) define and discuss goals of care with patients and families, with attention to their emotional needs; 2) support and model successful approaches for discussing goals with clinicians; 3) set expectations for success for these discussions; and 4) reduce barriers by identifying and mediating conflict. Training included at least 2 days of combined didactic and practice sessions including role play with simulated family members. A training manual will be available after study completion. Patient- and family-centered outcomes are assessed at randomization, and 1, 3, and 6 months after randomization (Figure 1).

Figure 1:

Overview of the study timeline

Study population and participant eligibility

A. Patients:

Using the EHR and daily ICU rounds (Monday to Friday), study staff identify consecutive patients in the ICU who meet all four of the following eligibility criteria: 1) age ≥18 years, 2) English-speaking (in the U.S.) or French-speaking (in France), 3) a family member available to participate, and 4) a chronic life-limiting illness suggesting a median survival of 2 years or a severe acute illness with a risk of hospital mortality of >15%. Chronic life-limiting illnesses, similar to our prior trial,²² target a median survival (without critical illness) of two years (see Appendix). Acute illness criteria include an APACHE-II (in the U.S.) or SOFA (in France) predicting a ≥15% risk of hospital mortality (the different severity of illness scores was based on tradition of prior research in each country). After initial screen and a “opt out” text page is sent to the patient’s attending physician to confirm appropriateness for the study, study staff contact the patient or, if the patient doesn’t have decisional capacity, a legal next of kin in-person or by phone. Decisional capacity is determined by the attending physician (or designee) to align with clinical practice.³¹ Patients without decisional capacity who had no eligible family are excluded. We anticipate that the majority of eligible patients will not have decisional capacity and the intervention will focus on family participants.

C. Family:

Family participants are identified by the patient if the patient has capacity. If the patient does not have decisional capacity, family are identified via the EHR or by a legal surrogate decision-maker. “Family” is not confined to immediate family. Any family or friend is eligible if they are 18 years or older and identified by the patient or legal surrogate as involved in care for the patient, provided they have primary language proficiency and no cognitive deficits limiting their ability to complete surveys. Family members are recruited by phone or in person. We do not limit the number of family members who can participate but anticipate 1–3 family members per patient.

Theoretical basis of the intervention:

This intervention draws on social cognitive theory,^18–21 which posits that the impetus for behavior change arises from interaction between people, their behaviors, and their environments. Self-efficacy, outcome expectations, and behavioral capabilities are key targets that support behavior change. Facilitators address each in our intervention:³² 1) self-efficacy by discussing ways that patients and families have had difficult conversations in the past and what worked for them; 2) outcome expectations by discussing how having effective conversations with clinicians can help assure treatment is aligned with goals; and 3) behavioral capabilities by teaching communication skills, addressing barriers, and mediating conflict. Figure 2 shows the role of social cognitive theory within a commonly-used model for designing behavioral interventions to improve quality of life, as well as how our Aims fit in this model.^33–37

Figure 2:

Conceptual Model for study

Operationalization of intervention:

Facilitators follow each intervention patient and family for 3 months following randomization and implement the key components of the intervention summarized in Table 1. Since most eligible patients won’t have decisional capacity, intervention components focus on family members.

Table 1:

Overview of the Key Components of the Intervention

GOALS	FACILITATOR SKILLS OR TASKS	TOOLS
A. Increase self-efficacy for identifying and discussing goalsof-care	Support discussions of prognosis and of values, goals and preferences	Facilitated values history51
B. Improve outcome expectations by identifying and resolving barriers	Discuss how effective communication can alter and improve treatment plans	Attachment theory to adapt communication177,178
C. Increase behavioral capacity with skills in communication & mediation	Address barriers; identify and mediate conflict; teach mediation skill	Mediation training for facilitators73,74

A. Increase self-efficacy:

Based on our prior research,^38–45 and the work of others,^46–51 we train facilitators to support, model, and teach family members to identify and integrate patients’ values into goals of care, preferences, and decisions. Facilitators discuss times family members have navigated difficult conversations, praising successful behaviors. Activities to identify and integrate goals of care include: 1) facilitating and supporting discussions of prognosis among patients, families and clinicians;^52–56 2) facilitating and supporting discussions of treatment preferences;^15,46,57,58 and 3) obtaining a facilitated values history for incapacitated patients from family and translate this into treatment preferences.⁵¹ Obtaining a facilitated values history includes a tool to enhance patient-centered discussions of prognosis, goals of care, and treatment preferences: the Facilitated Values History⁵¹ for family-clinician communication that focuses family on the patient’s values and goals of care. Facilitators participate in these discussions and provide guidance with this content. To increase patients’ and families’ capacity to define, understand and convey their goals, facilitators also elicit strong emotion and respond with empathy to allow participants to feel heard and supported. The decision psychology literature suggests that strong emotions can impair information processing and reasoning.⁵¹ Therefore, having a facilitator elicit strong emotion with unhurried listening and empathy can help a person transition to cognitive processing and decision making.

B. Improve outcome expectations for achieving goals of care:

The facilitators model and teach strategies to family members to support their expectations for having successful and effective conversations and how communicating well can improve their care. The focus on goals of care is a new addition based on feedback from the prior trial. These strategies include exploring prior experiences and the outcomes of those experiences. (Did respondents feel encouraged or discouraged? What would have helped them manage the discussion successfully? How can the facilitator help?) Additionally, using attachment theory,⁵⁹ the facilitators identify family members’ interpersonal communication style to ensure that outcome expectations are appropriately tailored to each patient and family. Attachment theory provides a framework to address individuals’ capacities to communicate with others and influences communication with healthcare teams.^23,60–63 Three of the 4 attachment styles pertain to about half the general population^64–66 who may derive particular benefit from the use of targeted communication skills that support difficult decision-making. For example, individuals who are predominantly “self-reliant” in stressful situations (25% of the population) may be reluctant to ask questions or participate in shared decision-making; they may benefit from being given options and feeling in charge. Those who are predominantly “support-seeking” in stressful situations (10% of the population) often have high emotional needs that may be inadequately addressed in harried healthcare settings; setting a structured schedule to have questions answered may be reassuring. Individuals with a “cautious” attachment style (10% of the population) may exhibit approach-avoidance behavior under stress so that attempts at collaborating may be abandoned out of fear of relying on others; communication that is non-judgmental and responsive can enhance trust.⁶³ Facilitators use a brief, valid measure of attachment styles (the Relationship Scales Questionnaire)⁶⁵ that we used successfully in our prior trial.²² Importantly, facilitators use these styles to adapt communication approaches to meet the needs of each patient and family member.

C. Increase behavioral capability through skill acquisition and conflict mediation:

Facilitators model and teach skills that support and advance successful discussions. For example, facilitators help patients and families prepare for discussions by reviewing their goals of care, writing down unanswered questions, and anticipating and planning responses to likely barriers (e.g., “clinicians are too busy”, “I don’t like asking questions because I don’t want to be one of those annoying people who asks too many questions”). Facilitators also model and teach ways to address conflict, a common occurrence in healthcare that frequently involves goals of care and is a key barrier to communication.^67–70 Incorporating mediation into healthcare can identify and resolve conflicts, thereby improving communication, as well as patient and family outcomes.^71–74 If patients and families feel supported and experience less conflict, they may be more able to process information and make decisions. We train facilitators in mediation principles including: summarizing the issues contributing to conflict, avoiding unwarranted assumptions that are underlying the conflict, framing issues neutrally, focusing on issues rather than personalities, and making proposals for resolution.^75–77 Facilitators model these skills for family members during the discussions about goals of care and around transitions in care settings that occur during the study period. We target all of the following types of conflict: clinician-patient, clinician-family, and intra-family.

Description of the conduct of the intervention

A. Implementation of intervention during and after the ICU:

During the ICU stay, facilitators interact in person, and by phone and videoconference, with family members and many types of clinicians (physicians, nurses, social workers, etc.). Following the ICU stay, facilitators interact with patients, family and clinicians in person and by phone for 3 months from randomization (when most first readmissions occur^78–80) or for 1 month after a patient’s death occurring in the first 3 months. In-person contacts include visits to patients’ homes and/or care facilities; phone contacts include calls to family members. Because prior studies suggest frequent contact is important, the schedule for contact is a minimum of every 48 hours in the ICU, every 72 hours in the acute care setting,^79,81–85 within 72 hours of change in care setting, weekly for a month after hospital discharge, and then twice monthly. The facilitators use clinical judgment if they feel more frequent contact is warranted, and family members have access to facilitators through phone and email 5 days per week. In addition to checking directly with patients/families during regular contacts (calls, visits), facilitators also access the medical record to ensure they have accurate information about appointments and treatment plans. Facilitators encourage referral to inpatient or outpatient palliative care services when they identify palliative care needs (i.e., communication, decision-making, and symptom needs). We record all facilitator contacts, including the number and type of contacts for each patient and family to assess intervention fidelity, “dose”, and costs. Facilitators complete a checklist of study activities after each contact (see Appendix.)

B. Training facilitators:

Training is provided by the investigators and consultants with expertise in the following areas: clinical communication skills, use of attachment theory, and mediation. Training addresses the use of these skills across care settings (e.g., inpatient, outpatient, home, care facility). Communication training includes identification of goals of care, incorporating principles of advance care planning and the facilitated values history. Attachment theory training includes understanding the four attachment styles, the consequences of these styles for communication, and the approaches most appropriate for each style. Mediation training covers skills associated with assessment and preparation: rapport building; information gathering and exchange; development and evaluation of options; shuttle diplomacy; and resolution. Facilitators participate in role-playing exercises during training with standardized family members, and they are required to demonstrate mastery of intervention skills before engagement and during fidelity checks. We will enhance and expand our current training manual as the trial goes on, enhancing dissemination of the intervention to other institutions.

C. Intervention fidelity:

To ensure scientific rigor, intervention fidelity is assessed with methods outlined by the NIH Behavioral Treatment Fidelity Workgroup on consistency in dose, providers, delivery, receipt, and enactment of interventions.^86,87 Facilitators meet weekly with investigators to discuss intervention patients and ensure intervention fidelity overall and between facilitators. Formal fidelity checks will be completed for 10% of patients/families, in which encounters are audio recorded and then compared with an intervention checklist for completion of intervention components (Appendix).

Control group:

Participants randomized to control complete the same study measures at all data collection points, but the facilitator is not involved. Contacts with study staff at all data collection points are also designed to enhance survey return rates.

Randomization:

The unit of randomization is the patient. The potential for contamination due to individual clinicians caring for patients in both groups is minimized because the focus of the intervention is specific to the individual patient and family and tailored to their needs. Randomization occurs in variable-sized blocks stratified by hospital, separately for each trial. Family members are clustered within patients. In the US and France, we allow enrollment of multiple family members per patient with no maximum. Based on prior studies, we expect 1.5 family members per patient in the US and 1.0 per patient in France.^3,22,26–30

Outcomes and mediators:

Figure 2 shows a conceptual model for assessment of outcomes, baseline individual and environment factors, and potential mediators influencing the intervention. The outcomes and potential mediators are summarized in Table 2.

Table 2:

Main study measures and data collection protocol

MAIN OUTCOME MEASURES	CONCEPT	DATA COLLECTION: SOURCE & TIME
Aim 1: Outcomes
Primary Outcome: HADS depression score	Depressive and anxiety symptoms	Patients and/or Families: Enrollment (if able),1, 3, 6, months post-randomization for all; 1 month after death for family members
SUPPORT question109	Goal-concordant care
Impact of event scale (IES)88,105,106	Post-traumatic stress symptoms
QUAL-E and QUAL-E (Fam)114–116	Quality of life
Aim 1: Potential Mediators of the Intervention
Perceived Competence Scale (PCS)117,118	Efficacy/capability/expectations for health behavior	Patients and/or Families: Enrollment (if able),1, 3, 6 months post-randomization
Aim 2: Outcomes
Hospitalizations, ICU admits,, palliative care consultations	Healthcare utilization; use of palliative care	Medical record: After-death or 6 months
Hospital and ICU costs	Inpatient costs of care from index hospitalization (direct and indirect)	Medical record/financial databases(US only)
Aim 3: Implementation Factors and Outcomes
Domains: intervention, settings, individual, process	Key factors identified from CFIR125	Patients/families/clinicians: Qualitative interviews
Outcomes: acceptability, fidelity, penetration	Key outcomes identified by Proctor126

Aim 1:

For Aim 1, we will evaluate the effectiveness of the intervention on one primary and several secondary outcomes. Although we will assess psychological symptoms (depression, anxiety, and PTSD) in patients and family, most critically ill patients are unable to participate. Therefore, the intervention and the outcomes focus on the family. The primary outcome is the depression subscale of the HADS over 6 months (including 1-, 3-, and 6-month assessments). These outcomes were selected based on evidence supporting their importance to patients and families.⁸⁸

A. Depression and anxiety (with depression as the primary outcome):

We assess patient and family member symptoms of depression and anxiety with the Hospital Anxiety and Depression Scale (HADS), which has become standard for ICU and post-ICU studies.^4,88–90 The HADS is a reliable and valid 14-item, 2-domain (anxiety and depression) tool used to assess symptoms of psychological distress.^91,92 Each item is scored on a 4-point scale (ranging from 0–3) with scores for each 7 item subscale (anxiety and depression) ranging from 0–21. HADS has been used in over 700 studies with evidence of reliability, validity and responsiveness among critically ill patients and their family.^3,93–104

B. Post-traumatic stress:

We assess patient’s and family’s symptoms of PTSD with the Impact of Events Scale-6 (IES-6). It is derived from IES and IES-R, PTSD measures that are recommended for use with ICU survivors to evaluate PTSD^88,105,106 (Cronbach’s Alpha, .80–.96 for total scores, .46–.56 for subscales; IES-R (0.96).^107,108 Each item is scored on a 4-point scale that addresses symptom severity ranging from “not at all” to “extremely”, and it may be scored either as a summary score or as a mean of the 6 items.

C. Goal-concordant care:

We measure concordance between the care patients want and the care they are receiving with two questions.¹⁰⁹ The first defines patients’ goals: “If (you/the patient) had to make a choice at this time, would (you/the patient) prefer a course of treatment focused on extending life as much as possible, even if it means having more pain and discomfort, or would (you/the patient) want a plan of care focused on relieving pain and discomfort as much as possible, even if that means not living as long?” The second question assesses perceptions of current treatment using the same two options. The outcome is a dichotomous variable of whether the preference matches the report of care received. Although this creates a “false dichotomy” in that many patients want both; this “forced choice” helps identify patients’ top priority.^110–112 This approach mirrors clinical practice in which goals of care are determined by the legal surrogate decision-maker when patients are unable to respond for themselves. Based on prior studies, we expect 50–60% of controls will report goal-concordant care.^109,113

D. Quality of life:

We use the QUAL-E and the QUAL-E (Fam) to assess quality of life. These companion questionnaires, one completed by patients and the other by family members, were developed systematically using focus groups, interviews and surveys to identify items, and then factor analyzed to identify quality of life domains. The QUAL-E includes 20 items in 4 domains (life completion, relationship with the healthcare system, preparation, symptom impact) with acceptable levels of fit (CFI= 0.89, GFI= 0.88, RMSEA =.06) and high levels of internal consistency (Cronbach’s alpha >= 0.70 for 3 of the 4 factors). It demonstrated convergent and discriminant validity analyses against the FACIT-SP and other measures. The QUAL-E (Fam) includes 17 items in two domains (relationship with healthcare providers, completion) and additional items assessing aspects of preparedness. The QUAL-E (Fam)’s convergent and discriminant validity were supported with findings that were in line with predicted associations with the PCAS, FACIT-Sp and other measures.¹¹⁴ Like other quality of life measures in which subscales assess domains that are conceptually distinct and therefore are not easily summed,¹¹⁴ we selected domains and items with relevance for the current study. For the QUAL-E, these include all domains except symptom impact; for the QUAL-E (Fam), we included all 17 items.^114–116

E. Patient and family report of self-efficacy, outcome expectations, and behavioral capacity:

Participants complete the Perceived Competence Scale (PCS) as a measure of these three components derived from social cognitive theory. It is a 4-item questionnaire that has been used to measure respondent’s perceptions of competency and agency at enacting the specific behaviors being assessed.^117,118 Its validity has been supported by significant associations between improved PCS scores and changes in a variety of healthcare behaviors, including short and long term tobacco cessation,^119–121 glycemic control,^118,122 and oral health.¹²³ Reliabilities are reported consistently at alpha > 0.80.^118,123

Aim 2:

We will examine the effectiveness of the intervention for reducing ICU and hospital readmissions and reducing costs of care. The intervention is designed to affect these outcomes by improving overall communication and, more specifically, clarifying and communicating goals of care. In France, because of universal healthcare and limited availability of cost data, this aim will focus on utilization only.

We will measure hospital readmission after initial hospital discharge through the available electronic health records (EHR), institutional billing systems, and patient/family self-reports. By using all three sources for data, we expect to capture hospitalizations that may occur outside of the healthcare system from which the patient was enrolled. Our primary focus will be readmission within 30 days of hospital discharge as this is a national standard,¹²⁴ but we will also collect hospital and ICU length of stay, emergency department visits and outpatient clinic visits from the EHR or surveys.

Aim 3:

Aim 3 enables us to evaluate the factors and outcomes that influence implementation of the intervention. We will use semi-structured interviews with a purposive sample of family members, clinicians, and administrators, as well as facilitators, to identify diverse interview participants with experience with the intervention and continue recruitment until thematic saturation is achieved.

A. Key factors for implementation:

Key factors that influence implementation are explored through semi-structured interviews with family members, clinicians, and administrators. Using the Consolidated Framework for Implementation Research (CFIR), questions are modified to provide a tailored assessment of the intervention and the context in which it is implemented; discussions focus on the effect of inner and outer settings, individual characteristics, and processes of care.¹²⁵

B. Key implementation outcomes:

Outcomes will be guided by the Outcomes for Implementation Research¹²⁶ and use both qualitative and quantitative measures. Qualitatively, semi-structured interviews will explore three key implementation outcomes (acceptability, fidelity, appropriateness).¹²⁶ Quantitative measures will include the proportion of eligible patients enrolled, the proportion of patients randomized to intervention who receive the intervention (facilitators talk with family members), and assessments of intervention dose (examining time spent by facilitators and intervention components delivered).

Additional measures and data

A. Description of participants:

For all participants, we collect age, gender, race/ethnicity, education, employment, and social support. For patients, we collect comorbidities and severity of illness with APACHE-II (in the U.S.)^127–129 or SOFA (in France.)¹³⁰ For family, we collect relationship with the patient and whether they reside with the patient.

B. Attachment style:

To identify attachment styles, we administer the Relationship Scales Questionnaire (RSQ)⁶⁵ which we successfully used as part of the prior facilitator intervention.²²

Data collection:

Study staff involved in outcome data collection are blinded to treatment assignment.

A. Surveys:

To enhance response rates, surveys may be completed in person, by phone, by mail, or online, depending on participant preference. Strategies to enhance response rates include: 1) use of study staff known to participants to contact participants at each data collection point; 2) inclusion of small monetary gratuity ($10) with each follow-up questionnaire (in the U.S. only); and 3) reminder contacts prior to, and following each distribution time point.^131,132 Families complete the same measures at the same intervals as patients with two exceptions: 1) families complete the baseline measures at randomization, while for patients there may be a delay for them to regain decisional capacity; and 2) families of patients who die receive modified surveys 4–6 weeks after a patient’s death, which include the HADS, IES-6, and QUAL-E (Fam).

B. Electronic health record (EHR):

We will use the EHR to supplement family report of healthcare utilization. We will also use the EHR to collect disease characteristics and specific processes of care during and after the ICU stay including treatment intensity (e.g., CPR, mechanical ventilation), transitions in care, and palliative care consults.

C. Hospital financial databases:

For hospitals in the U.S., we will obtain costs associated with inpatient utilization from hospital financial databases and will link this with our EHR data.

D. Qualitative data collection:

Aim 3 collects data with approximately 30 semi-structured interviews with family members (n=15), clinicians (n=10), and administrators (n=5). We use purposive sampling with the goal of a diverse group based on race/ethnicity, age, gender, and, for clinicians, specialty and year of training.

Overview of Analytic Methods

Aim 1 Analyses:

Our primary outcome is family members’ symptoms of depression over 6 months. We will follow the intention-to-treat principle. Our primary analysis will use a linear mixed effects model with family member symptoms at all time points (1,3, 6 months) as the response, main effects for intervention and time points, and random effects to account for multiple measurements (time points) per family member and multiple family members per patient. We will also adjust for hospital, since randomization is stratified by hospital, and for response at randomization to improve precision. This model allows the average response to be different at 1, 3, and 6 months, but assumes the intervention has the same effect at each of these times. We will also explore whether the effect of intervention is different across time by including an interaction between time and intervention. The advantage of using the data at all 3 time points and a mixed model approach is that we can gain precision; it also allows missing responses, assuming the responses are missing at random. We will use a similar approach for the other continuous outcomes, and a generalized linear mixed effects model for the binary outcome of goal-concordant care.

Aim 2 Analyses:

This aim will evaluate the effect of intervention on costs (in the U.S.) and utilization (in the U.S. and France) during the 6-month study period from the EHR, hospital financial databases and surveys. Utilization measures, including ICU-free days and hospital-free days to day 30, days in ICU, and days in hospital, will be evaluated with linear regression. As is standard for cost analyses, we plan to use generalized linear models with a gamma error distribution and identity link function, to explore the difference in mean healthcare costs for total inpatient hospital costs and disaggregated costs (direct and indirect costs), between intervention and control participants during the initial hospitalization and up to 30 days after randomization.¹³³ To estimate the cost of providing the intervention, we will collate study staff records of the time involved in implementing and sustaining the intervention during the trial, not including research activities such as obtaining informed consent and the evaluation of the interventions. We will then determine the average full time equivalent (FTE) over intervention patients, and convert FTE to costs based on average yearly salaries at the study sites, including benefit load. All costs will be adjusted for inflation through the study period.

Aim 3 analyses:

For quantitative analyses of implementation, we will describe the proportion eligible who enroll and the fidelity with the intervention components. For qualitative analyses exploring intervention implementation, we will perform thematic analysis of transcribed interviews to explore feedback on the intervention and ways to improve the intervention delivery and implementation.^134,135 Analyses will be guided by the Consolidated Framework for Implementation Research (CFIR) in exploration of factors affecting implementation¹²⁵ and by Outcomes for Implementation Research in exploration of implementation outcomes.¹²⁶ For example, we can examine facilitators by “high” and “low” fidelity and sites by “high” and “low” penetration to examine cross-case patterns in CFIR constructs that differentiate these groups. Qualitative data will be imported into analytic software (DeDoose and NVivo), where investigators will perform iterative, inductive coding to identify recurrent themes, categories, and relationships among themes and categories. To ensure trustworthiness (a qualitative concept similar to reliability in quantitative analysis^135–138), we will perform a “member check” of the results with prior participants selected for diversity of participant type.

Missing data:

Our goal is to minimize missing data by minimizing respondent burden, offering multiple methods for survey completion, and having trained staff with repeated contact with participants. However, data could still be missing due to skipping individual items on a survey, omissions in the EHR, lack of follow-up, or death. We will quantify the amount of missing data and apply appropriate methods to account for missing data.^139,140

Sample size estimates

The focus for sample size estimation is the primary outcome: family member depression as assessed by the HADS depression subscale. For all calculations, we assume a two-sided test with a significance level of 0.05. If we assume 300 total family members (1 family member per patient and 150 per arm), a standard deviation of HADS depression scores of 4.2 points¹⁴¹ in both arms, 3 measurements of depression (at 1, 3, and 6 months), and an intraclass correlation (ICC) of 0.2, we will be able to detect a difference in mean depression of at least 0.93 points with 80% power and 1.07 points with 90% power. If we had only 1 measurement instead of 3 per family member or, equivalently, if the ICC were 1, we would have 80% power to detect a difference of at least 1.36 points. If we have more than 1 family member per patient (1.5 is expected), we will be able to detect smaller differences at the same power. Across all of these varying conditions, detectable differences are slightly within or below the estimated minimally clinically important difference for the HADS depression scale of 1.6 to 2.0, suggesting we will have adequate power to identify a minimally clinically important difference. In the US we anticipate ≥80% complete data and in France we anticipate >75% complete data, so we aim to randomize 376 patients in the US and 400 in France to achieve at least 300 patients and at least 300 family members with complete data for each trial (U.S. and France).

Adaptations for the COVID-19 pandemic

These trials were underway when the COVID-19 pandemic started in 2020. Both trials stopped enrollment for the initial surge of COVID-19 in the US and France. Once the caseloads of COVID-19 stabilized, both trials resumed with recruitment and intervention activities occurring virtually – by phone or videoconferencing. Both trials are tracking the amount of contact occurring in person as compared to phone or videoconferencing.

Discussion

We have described two parallel randomized trials that are being conducted, one in the U.S. and one in France. These trials provide innovation and advance the science of palliative care in critical care settings in three important ways: enhancements to the intervention based on prior studies, innovation in the model of care coordination the intervention is promoting, and innovation in the use of a type 1 hybrid effectiveness-implementation trial to examine implementation processes and outcomes.²⁵

This intervention is based on the intervention in a prior ICU communication facilitator study²² and builds on prior work evaluating navigators or facilitators in a variety of settings.^142–147 However, our intervention is novel by its inclusion of nurse facilitators with innovative training that supports, models, and teaches communication skills to patients and family facing critical illness and also continues to follow patients and families beyond the ICU for three months. Since this intervention will be implemented both during and after critical illness, it addresses key gaps identified in recent systematic reviews^4,148–151 by: 1) beginning early in the high stress time of critical illness and following patients after ICU; 2) focusing on communication about goals of care; and 3) targeting a diverse population (e.g. age, diagnoses, SES) to enhance generalizability, scalability, and cost-effectiveness.

One of the major challenges for interventions designed to improve communication about goals of care has been a limited number of validated outcomes.^152–155 We address this with standardized measures with good psychometric characteristics. The effectiveness outcomes— reducing the burden of psychological symptoms and impaired quality of life—will be assessed with well-validated measures. These measures are standard for ICU and post-ICU studies.^4,88–90 Our intervention extends over 3 months and we elected to include outcomes extending beyond the intervention to 6 months to assess whether benefit persists beyond the intervention as seen our prior study showing benefit at 6 months.²² In addition, we assess social cognitive outcomes of behavior change (i.e., patient/family self-efficacy, outcome expectations, behavioral capacity) that are targeted by this intervention with a validated measure that will provide insights into the mechanisms that mediate the study’s outcomes.^156–159

Our primary outcome is family member symptoms of depression over the 6 months after randomization. The importance of improving outcomes for family caregivers and the value this brings to patient outcomes is well-established.^160,161 When family members suffer from the burden of critical illness, patients also suffer. Much of the suffering for family members comes from the difficulties of surrogate decision-making and the fact that most critically ill patients do not have decisional capacity.¹⁶² Interventions that reduce family member distress also improve the care they can provide to patients.^3,22 In addition, stress extends beyond the ICU as family bear a significant burden of caregiving after the ICU.^3,93,163 As many as 20–60% of families and patients suffer a high burden of psychological symptoms during and after the ICU.^{3,93,164–168} Improving communication is key because poor communication worsens distress,¹⁶⁹ and interventions that improve communication reduce distress.^3,22 Further, caregiver stress increases healthcare use and reduces quality of life.^170–174

We have designed a hybrid implementation-effectiveness trial to use the innovation of implementation science to promote implementation and dissemination of effective interventions. Implementation science, or T4 translation research, is the systematic study of methods to promote the uptake and integration of health interventions.^{25,126,175,176} Hybrid effectiveness-implementation trials, such as this, represent an innovative design that can facilitate more rapid translation of research into clinical practice while also evaluating the effectiveness of the intervention.²⁵ Improving communication in complex and diverse settings is an ideal intervention for a hybrid effectiveness-implementation trial because evidence of the value of communication is clear, but the core issue is how to efficiently implement interventions that promote effective care.¹⁵¹ We are collecting data on implementation factors and outcomes in order to facilitate implementation and dissemination in the future and advance implementation science for palliative care and communication interventions.

In this paper, we have described the protocol that is being implemented in two randomized trials designed to evaluate a communication facilitator intervention to improve patient- and family-centered outcomes for critically ill patients and their families during and after a patient’s critical illness. Conducting these parallel studies in the U.S. and France provides a unique opportunity to evaluate effectiveness and implementation in two very different medical and social cultures.