Role of norms in variation in cancer centers’ end-of-life quality: qualitative case study protocol

We chose a qualitative case study design at six sites to identify local organizational and provider practice norms that influence variation in EOL treatment intensity, particularly for minority patients. Based on 2016 Medicare claims data analyses, [26] we will recruit 6 of the 11 NCI/NCCN designated cancer centers serving at least 15% African American advanced cancer patients. We based our sample size on recent literature related to sample size sufficiency, recommendations to reach multi-site data saturation, and qualitative research expertise of our study team. First, our study is guided by a theory based on previous research and uses data from multiple sources to test and cross-check for confirming or disconfirming evidence of our theory, a necessary component of ensuring data sufficiency [27, 28]. In addition, conducting qualitative case studies at six sites, each of which will include over 30 interviews plus multiple observations, will produce fine-grained and rich descriptive analysis to generate and compare theoretical insights across sites, as well as across stakeholders (e.g., providers, patients) within sites [29‐31]. Our target numbers for interviews and observations are well within recommendations for reaching data adequacy and saturation given our well defined study aim [32]. Lastly, research suggests that qualitative research expertise, including the quality and depth of interview process, is an important criterion to consider for assessing sample size [33]. Our site visit team has over 40 years of collective experience conducting qualitative research in healthcare topics and settings, ensuring a rigorous and thorough process at each of the six sites.

We will target National Comprehensive Cancer Network (NCCN) and National Cancer Institute (NCI) Comprehensive Cancer Centers for our study because they set national standards for high quality cancer care. We will engage up to six sites serving a high proportion of African American patients, ranging in EOL care intensity, and theoretically sampled, to maximize our ability to compare and contrast organizational and provider practice norms related to EOL care. We defined high proportion of minority patients as > 15%. We measured EOL care intensity based upon risk-adjusted metrics of EOL quality using 2016 Medicare fee-for-service claims data: receipt of chemotherapy in the last 14 days of life (NQF #0210), intensive care unit (ICU) admission in the last 30 days of life (NQF #0213), and non (NQF #0215) or late (NQF #0216) hospice referral. Our approach to calculating these EOL quality metrics has been published elsewhere [26]. Given the multivariable nature of these metrics, we use data visualization to purposively select sites for case study that maximize potential heterogeneity in practice patterns [34]. Following the principles of positive deviance sampling, [35, 36] we will sample higher versus lower EOL quality sites in a ratio of 2:1.

We will employ qualitative case study research methods in this study, including inpatient observation procedures and provider, patient, and leadership semi-structured interviews, that have been previously developed and piloted [23, 24, 37‐39]. We will augment these methods with outpatient and tumor board observation procedures, which we developed and tested at a non-study NCI-designated cancer center serving a white, rural population. We will iteratively revise all data collection procedures based on researcher experiences and thematic insights following each sampled case study site visit.

The study team will collect 3 types of data: field notes from direct observation of inpatient and outpatient cancer care and cancer tumor boards; transcribed audio-recorded semi-structured interviews with cancer center leadership, providers, and patients, family members, and caregivers; and artifacts (Table 1). We will link all data from observations using a unique identification (ID) number. We will use a data collection form for field observations to capture provider and patient demographic information, as well as location, time, and individuals present at the encounter. In addition to relevant clinical data, observers will note socio-linguistic dimensions such as turn taking, tone, affect, body positioning, and eye contact. Artifacts collected during the site visit will include workflows, marketing/informational materials, orientation guidelines, quality reporting, and communication documents used in the cancer center.

Semi-structured interview guides for site leadership focus on institutional norms, including resources, programs, and policies related to EOL care and outcomes, as well as site-specific workflows and scheduling logistics in preparation for site visits. Cognitive mental models semi-structured interview guides for providers, which include EOL case vignettes, explore motivations, decision heuristics, and/or rationalizations [40]. Six EOL vignettes were developed by a medical oncologist, radiation oncologist, and palliative care providers to highlight key decision points common to outpatient or inpatient providers (see Table 2); each vignette has one version with a photo of an African American patient and one with a photo of a white patient. Providers will view one African American case and one white patient case to facilitate mental models debriefing and uncover unconscious beliefs and assumptions related to race that structure advanced cancer decision making. Semi-structured interview guides for patients, family members, and caregivers probe past, current, and future decisions related to their cancer care.

Site visit teams will consist of 2–3 researchers. We will identify a site-specific principal investigator (PI) at each site to help facilitate support from site leadership, and identify and recruit informants for pre site-visit interviews, and providers for site-visit observation and interviews. Up to two months in advance of the site visit, the study team will conduct leadership interviews by phone, including physician and nursing leaders, outpatient oncology practice managers, and key referral service line leaders from palliative/supportive care, hospital medicine, and intensive care. We will approach other site leaders for interviews at the suggestion of the site PI. When necessary, leadership interviews will take place during and after site visits.

One month in advance of the site visit, the study team will recruit providers for observation in the inpatient and outpatient setting. The observation schedule will involve one researcher assigned to each observed provider for a half-day observation in outpatient clinics (morning or afternoon session), the emergency department (by shift), and inpatient setting (by timing of daily service or consult rounds), as well as scheduled tumor boards and family meetings. Selection criteria for providers to observe and interview focus on maximizing our ability to assess provider norms for advanced cancer care within the particular institutional context that we will explore during leadership interviews. Provider selection criteria includes providers who manage patients with: 1) metastatic solid tumors (i.e., we excluded leukemia, lymphoma, and bone marrow transplant providers) and have either 2) high volumes of patients and/or 3) high peer influence, as perceived by the site PI or other key informants. We seek to recruit medical, radiation, and surgical oncology providers as well as palliative/supportive care providers who see cancer patients (see Table 3 for target sampling frame). We also seek to recruit providers from intensive care, hospital medicine, and emergency medicine who care for acutely ill cancer patients. We present an example observation schedule in Fig. 1. We will ask all providers recruited for observation to complete an interview. Interviews with providers will occur in person during or by phone after the site visit. We will digitally record all interviews and compensate all providers for participating in an interview.

At least 2 weeks prior to the site visit, we will send flyers about the study with photos of the study team to the site PI who will facilitate posting of the flier in public settings such as waiting rooms, clinic rooms, infusion suites, and inpatient units. The purpose of this flyer is to alert non-consented individuals to our study purpose and to provide instructions for opting out.

During patient care observation, researchers will directly approach patients and their family/caregivers following introduction by the consented provider. If patients or their family/caregiver verbally consent to be interviewed, the study team member will obtain contact information to arrange for a phone interview at the patient, family member, or caregiver’s convenience at a later date. Selection criteria for patient interviews includes: 1) adults aged 21 years or older; with 2) metastatic solid tumor; 3) whom the provider would not be surprised if they died in the next 12 months; and 4) seen by a consented provider. We will seek to recruit equal numbers of minority and non-minority patients. We will digitally record all interviews and compensate each participant for participating.

We will use a qualitative and mixed methods data analysis platform, Dedoose, to manage and analyze all transcribed field notes, interviews, and artifacts, and link relevant data to contextual information (e.g., patient and provider race; site features) (Sociocultural Research Consultants, LLC). We will develop a codebook first deductively, using our theoretical model, findings in the literature, and prior research and then inductively, through an iterative process of close readings and discussion of the data in order to identify additional codes. Three qualitative researchers will apply the codebook to the data, two who will divide and code all the data, and one who will assess reliability of coding by reviewing a subset of the coded data. All three qualitative researchers will discuss differences in coding and resolve by consensus. We will repeat the analysis process after each site visit, to conduct constant comparative analysis regarding similarities and differences between and within sites in support of study aims 1 and 2. After completion of each site visit, the study team will develop a written summary of preliminary quantitative and qualitative findings specific to the site, which will then be sent to all participants from that site, to assess initial validity of the site-specific findings. After completion of all site visits and analysis of data, we will send final study reports to participating sites.

Finally, we will employ multiple methods of triangulation to assure comprehensiveness and validity of data. Two to three researchers of a multidisciplinary team will participate in each site visit, and an additional three researchers will conduct qualitative analysis, to satisfy investigator triangulation. Method triangulation will include direct observation, semi-structured interviews, and artifact collection. We will achieve data triangulation by observing and interviewing leadership personnel, providers, and patients, family members, and caregivers at each site, of various backgrounds, specialties, and diagnoses, respectively. Qualitative analysis will use both deductive and inductive methods to achieve theory triangulation.

The study has been approved by the Dartmouth College Committee for the Protection of Human Subjects (STUDY00031129) and is considered minimal risk. All participating sites will 1) waive independent IRB approval in favor of acknowledging Dartmouth’s IRB, 2) rely on Dartmouth’s IRB via a SMART IRB reliance, or 3) conduct a local ethical review and approval. We will obtain a waiver of informed consent for participant observation; all providers will provide written electronic consent for observation and interview, and all interviewed leadership, patients, and families will provide oral consent for interview. We have obtained a certificate of confidentiality from the National Institutes of Health (NIH) for this study.

We will not record any identifiable or personal information about providers, patients, family members, caregivers, or staff in field notes, except demographic information. A unique ID number will link data from observations and interviews, including demographic data, to consented participants. The key linking the ID number and identifying information of the consented participants will be maintained on a password-protected server. Only the research team will have access to the linkage file. All data collected on individuals will be linked to their ID number alone. We will audio-record and transcribe all handwritten field notes without any identifiable information. We will store all original field notes in a locked filing cabinet, and all transcripts on a password-protected server, to which only the research team will have access. We will give a discreet lapel pin to all providers, staff, patients, family members, and caregivers who do not wish to be observed, as advertised by the informational flyers posted prior to the study team’s arrival at the site. We will not document any individual wearing such a pin in field notes, nor will we approach them for an interview.

We will give all individuals participating in interviews an information sheet prior to the interview, and we will obtain informed consent verbally at the time of the interview. We will obtain informed consent verbally as many of the interviews will be conducted by phone, either before or after the site visit. The process of obtaining verbal consent has been approved by the Dartmouth College Committee for the Protection of Human Subjects. We will record all interviews, and later professionally transcribe them without any identifiable information. We will store all recordings and transcripts on a password-protected server, to which only the research team will have access. Additionally, we consulted the guidelines for end-of-life research put forth in the “Methods of Researching End of Life Care” (MOREcare) project while designing the protocol for this study [41]. Specifically, we considered the risks (e.g., interviewee distress) and rewards (e.g., potential therapeutic effect) that qualitative interviews may have for patients, family members and care givers while designing these protocols.