Advancing Symptom Alleviation with Palliative Treatment (ADAPT) trial to improve quality of life: a study protocol for a randomized clinical trial

The ADAPT study utilizes the conceptual framework depicted in Fig. 1 which is based on integrating elements of Lenz’s unpleasant symptom theory [23] into an adaptation of the Wilson and Cleary model of health-related quality of life [24]. Given the relationship between underlying disease severity, symptoms, depression, and impaired quality of life, interventions aimed at improving quality of life must target symptoms and depression concurrently. Therefore, in the ADAPT trial, disease-specific care is provided in conjunction with symptom-focused care, a psychosocial intervention, and advance care planning and communication by employing a multidisciplinary, team-based approach. By addressing individual facets of advanced chronic disease, including symptoms and depression, we hypothesize improvement in participants’ quality of life.

In addition to evaluating effectiveness, the ADAPT study design incorporates a mixed “hybrid effectiveness-implementation” approach [25], evaluating individual intervention components simultaneously with the resources required to implement, maintain, and disseminate the intervention.

The primary hypothesis is that veterans with CHF, COPD, and ILD randomized to the ADAPT intervention will have improved quality of life at 6 months compared to the control arm of usual care. Secondary hypotheses are that veterans randomized to the intervention will have decreased disease-specific symptom burden, improvement in depression symptoms, increased advance care planning and communication, and decreased hospitalizations compared to usual care.

Based on the primary hypothesis, the first aim of the study is to determine the effectiveness of the ADAPT intervention on improving quality of life. As the ADAPT intervention targets key contributors to impaired quality of life—symptoms and depression—evaluating these and other secondary outcomes of the study are both clinically meaningful and important to patients, informal (e.g., family) caregivers, and providers.

The secondary aim of the study is to examine implementation of the ADAPT intervention. Utilizing a mixed-methods approach, key stakeholders, including participants, informal caregivers, study personnel, and providers, will provide feedback on the intervention components to assess facilitators and barriers to implementation. Additionally, given that a primary barrier to implementation is the costs and resources required for start-up and maintenance of an intervention [26], these will be evaluated. Incorporating these two facets of intervention testing—effectiveness and implementation testing—can reduce the time delay from innovation discovery to implementation, advance scientific knowledge, and increase policy relevance of clinical research.

The study is a randomized clinical trial conducted at two Veterans Health Administration (VA) facilities, the VA Eastern Colorado Health Care System (VA ECHCS) and VA Puget Sound. Each health care system encompasses a tertiary care medical center and seven community-based outpatient clinics.

Veterans with CHF, COPD, or ILD with impaired quality of life who are at increased risk for hospitalization or death are eligible for participation in the study. Enrolled subjects are randomized 1:1 to usual care or the intervention arm (team-based care plus usual care; Fig. 2). Randomization occurs at the patient level with computer generated random block sizes, stratified by study site and disease.

The study will enroll 300 veterans with CHF, COPD, or ILD with self-reported poor quality of life and increased risk of hospitalization or death who are able to participate in the intervention. Complete inclusion and exclusion criteria are provided in Table 1.

To identify veterans with increased risk for hospitalizations or death, we use a VA-developed prognostic tool, the Care Assessment Need (CAN) score [27]. The CAN creates a probability estimate of hospital admission or death within one year. Patients with CHF or COPD are potentially eligible to participate in this study if they are in the top 20th percentile of risk for hospitalization or death in the next year. As the CAN score is not validated for use in ILD patients, it will not be used as an eligibility criterion for those with ILD. However, patients with ILD, particularly fibrotic ILD, have high rates of hospitalization [28, 29] with significant associated in-hospital or subsequent mortality and are therefore eligible for participation. ILD was approved for addition to the study protocol on May 25, 2018 to expand eligibility criterion.

After meeting diagnostic and at-risk criteria, potential subjects are screened for quality of life using the Functional Assessment of Chronic Illness Therapy-General (FACT-G) questionnaire [30]. A score ≤ 70 (with lower scores indicating worse quality of life) indicates potential eligibility. This cutoff was chosen because a score ≤ 70 identifies poor quality of life as validated by declining performance status and increasing disease burden [30].

To facilitate recruitment and assess eligibility, a HIPAA waiver was granted at both study sites to allow screening of administrative databases and review of medical records. Eligible subjects at both study sites are identified electronically using validated combinations of diagnostic codes for CHF, COPD, or ILD [31, 32]. This generates a list of potential subjects who meet eligibility criteria based on available administrative data. Study personnel then screen individual medical records to confirm eligibility.

After confirmation, patients’ primary care providers are contacted to confirm the study team can contact their patients and to explain the study. With primary care provider approval, veterans are mailed letters describing the study and providing contact information for study staff if they are interested in participating. If an eligible veteran does not contact the study team, they are contacted by telephone. We anticipate that it will take 25 months to recruit 300 subjects.

This study has been approved by the Colorado Multiple Institutional Review Board protocol #15–1891, the VA Puget Sound Multiple Institutional Review Board protocol #00857 and is registered at ClinicalTrials.gov (NCT02713347). The study is funded by VA HSR&D IIR 14–346 (Bekelman, Principal Investigator).

The intervention is a multidisciplinary, team-based approach to addressing symptoms and psychosocial needs of participants (Table 2). The team-based approach is based on the evidence-based collaborative care model of health care delivery [33, 34]. The intervention personnel include a registered nurse (RN) and Master’s level social worker (MSW). They integrate into a larger collaborative care team (“Team”) that includes a representative primary care provider (PCP) and palliative care specialist. Specialist support with a cardiologist or pulmonologist is available for the Team for additional management recommendations if needed. Each site has a Team that meets weekly for 30–60 min, integrating palliative symptom management with disease-specific care plans.

The Team is responsible for recording recommendations in a progress note in the electronic medical record, implementing non-pharmacological recommendations, and writing orders for medications or tests that the participants’ individual PCPs review and sign at their discretion. This integration creates improved communication between the intervention Team and the PCP and an additional level of safety for participants.

Based on the conceptual model of the study, the intervention includes three pillars: symptom-based care, psychosocial care and advance care planning. Each facet of the ADAPT intervention takes place in collaboration with the Team and participants’ individual PCPs. RNs and MSWs interact directly with the participants across several visits as subsequently described.

Participants randomized to the control group receive usual care at the discretion of their care providers. They participate in the same number of study visits in addition to completing questionnaires and self-assessments at the same intervals as those in the intervention arm. Usual care will be enhanced by providing participant PCPs with the results of baseline depression surveys if a participant screens positive for depression. Individual providers will then assume responsibility for follow-up of the positive depression screen. Participants in the control arm do not have any limitations on care recommendations or referrals, which may include management by subspecialists, mental health providers, or palliative care at the discretion of their PCPs.

The outcome measures and frequency of data collection are summarized in Table 3. Patients self-complete the survey measures and mail surveys to study staff. Surveys are labeled with a study identification number and double-entered by study staff unaware of participant’s treatment arm assignment.

Implementation of the intervention will be evaluated using a mixed-methods approach. Using qualitative methods, participants and key study personnel will provide feedback of the implementation process and intervention components utilizing selected Consolidated Framework for Implementation Research (CFIR) domains [54]. Study participants will complete an interview after the 6-month visit with a focus on the value of different intervention personnel, the content and value of the intervention, communication and care coordination, and their perceptions with respect to sustainability of the intervention.

The intervention team will participate in a structured focus group discussion at the completion of the study. The discussion will elicit feedback about what parts of the intervention worked well, might be streamlined, enhanced, or eliminated and views about patient, informal caregiver, and PCP receptivity and responsiveness.

An intervention database will be used throughout the study to track intervention content and processes. PCPs with patients who complete the intervention will be surveyed to assess their satisfaction and experience with the implementation process. The survey can be completed via email, phone, or in-person to improve total response rates.

We will evaluate the necessary resources to implement and maintain the intervention. We track the time and other resources associated with intervention implementation and maintenance and will subsequently assign cost. The main implementation resource is personnel training. Maintenance costs include personnel time to provide the intervention, including phone calls, patient visits, team meetings, and care coordination with PCPs.

Data from all participants will be included regardless of level of participation using an intent-to-treat approach. The primary outcome is the difference in FACT-G score at 6 months analyzed as a continuous variable. The sample size was determined to detect a clinically significant difference in the primary outcome of mean FACT-G scores between the intervention and control arms. With an analytic sample size of 115 veterans per arm, we will have 85% power to detect a moderate effect size of 0.4 (two-sided test, alpha = 0.05). We plan to enroll 300 veterans and anticipate 5% will die and 15–20% will have missing outcome data. The minimal clinically important difference on the FACT-G score is 4–6 points [55], and with a standard deviation of 15, a Cohen’s d effect size of 0.4 will be on the high end of clinical significance.

Because of the anticipated high correlation of baseline FACT-G with follow-up FACT-G (r > 0.5), we will include the baseline FACT-G as a precision variable in a linear mixed model of the longitudinal outcome measures. To describe the treatment by disease interaction, we will estimate the treatment effect and its confidence interval within each of the disease groups (CHF, ILD, COPD) using disease-specific health status measures (KCCQ, K-BILD, CCQ) at the 6-month endpoint. In exploratory analyses, we will estimate treatment effect within illness subgroups (CHF, ILD, and COPD) on the primary outcome, and within subgroups of illness, including CHF (preserved vs reduced ejection fraction) and COPD (defined by post-bronchodilator FEV₁/FVC < 0.70 on spirometry [56]). Missing data will be reviewed to identify potential patterns and examined to assess how these patterns impact our results. Specifically, we will examine plots of group means over time stratified by the time of the last completed observation to determine if biases are evident due to missing data. When data are missing at random, unbiased results can still be obtained from the maximum likelihood method that will be used in the linear mixed model analysis. To account for the possibility of data missing not at random, sensitivity analyses will be performed using pattern mixture models and results will be presented to assess the impact of missing data on the reported conclusions. Among those who were hospitalized or died, we will examine for differences in intensive care utilization and patterns of care at the end of life. We will also examine for intervention effect on longer-term health care utilization (e.g., hospitalization, intensive care utilization).

The quantitative data on intervention component implementation will be examined using descriptive statistics. For example, for team meetings, the number and type of medical orders written and completed will be summarized. For social worker visits, the median, range, interquartile range, and types of modules completed will be displayed. This type of analysis will show what components of the intervention were actually done. It will characterize the true “dose” and content of the intervention that was provided. This will contribute information about what may have led to intervention success or failure.

The qualitative data on intervention components and processes will be analyzed using a combination of inductive and deductive methods. We will create an evolving set of codes linked to units of text (fragments, sentences, or paragraphs). A qualitative analyst and the research assistant will serve as primary coders for qualitative data, and the PI will review coding and codebooks as they are developed. We will follow a systematic process to enhance coder agreement in assigning codes and a peer debriefing process that requires regular meetings with a qualitative analyst, the PI, and the research assistant to review and refine codes, code definitions, and conceptual boundaries for our analysis. The iterative analysis will begin by using a priori codes based on the CFIR model, supplemented by codes reflecting intervention content and structure and questions used for data collection. Codes will be refined and new codes added as new insights emerge. Through systematic coding we will quickly develop working themes and hypotheses about critical intervention components and processes that will be examined (and inform any minor changes in data collection interview guides/survey). These themes will also describe facilitators and barriers to intervention implementation. We will both audio-record and take detailed notes during all data analysis meetings in order to document proposed codes and code revisions, proposed themes and their descriptions, and other decisions made during these working meetings.

We will use several recommended strategies to enhance the validity and credibility of qualitative findings: 1) structured interview guides administered by well-trained interviewers; 2) coding templates and detailed descriptions of codes, coding decisions, and analysis strategies to document all phases of the data analysis (audit trail); and 3) team approaches (at least two analysts) to develop coding templates and independently code subsets of transcripts/notes to determine their agreement and application of codes and code definitions.

In addition to analyzing and summarizing quantitative (implementation tracking database and provider surveys) and qualitative (patient interviews and intervention team focus groups) findings separately, we will also merge findings to draw overarching lessons learned from multiple methods used. In this process, findings will be summarized in a table placing qualitative themes side by side with the quantitative findings to show the extent to which the data converges. Merging these findings will provide “triangulation”: findings from each data source will be used to validate and confirm findings from the other data sources. With the combination of qualitative and quantitative data, we expect to provide a more complete explanation of why certain intervention components and processes are more critical than others, as well as facilitators and barriers to the implementation of the intervention.

We will determine resources and cost of the intervention by first calculating the resources (personnel hours or FTE, and other costs) to implement and maintain the intervention during the study. Personnel costs associated with the program will be calculated based on actual VA nurse (and other staff) wage and benefit rates. Total intervention costs and costs per intervention participant at each site will be calculated. Actual salaries and benefits will be used when calculating personnel costs. Second, we will estimate the resources and costs to implement and maintain the program in a variety of VA settings.

Several sensitivity analyses will estimate the range of intervention costs using alternative assumptions for costs that may vary in different implementation contexts. Three sensitivity analyses are planned: 1) variable labor costs with more or less experienced nurses or physicians; 2) variable efficiency of the nurse and social worker in part-time vs full-time intervention roles—this will be done using actual data that reflect the range of time spent per patient early in the study vs later in the study; and 3) variable patient case mix using the range of time spent per patient. In exploratory analyses, we will estimate which patient-level predictors (e.g., cardiac ejection fraction, spirometry, quality of life) are associated with time spent per patient (outcome) using linear regression modeling.