The study involved participant recruitment for semi-structured interviews that were audiotaped and then assessed using deductive content analysis. Study protocols were approved by the Children’s Hospital of Philadelphia’s Institutional Review Board (IRB protocols #16–013072 and #16–012804). Informed consent was obtained from parents and assent was obtained from children. We recruited participants into the sample until saturation was achieved for each domain—i.e., when no new facets had been elicited for a minimum of 5 consecutive participants.
Participant recruitment
Participants with Crohn’s disease were recruited from an institutional Center for Pediatric Inflammatory Bowel Disease, which maintains a curated list of patients to ensure accuracy of diagnosis. The sub-group of children who participated in the ImproveCareNow quality improvement network (improvecarenow.org) [8], were 8–17 years-old, and, by parent-report, could provide self-reports in English were invited to join the study via email. Parents wishing to learn more about the study after reading the email completed an electronic questionnaire that confirmed eligibility and asked for contact information. Parents of enrolled children were contacted by telephone to schedule an interview for the child.
Participants with chronic kidney disease were recruited after a nephrology clinic visit. Eligibility criteria were age 8–17 years-old, able to self-report in English according to their parents, and had an estimated Glomerular Filtration Rate (eGFR) consistent with chronic kidney disease (i.e., between 15 and 89 mL/min/1.73 m2). Parents of eligible children who expressed interest in the study completed a consent form and provided contact information. They were contacted after the visit to schedule an interview for the child.
Semi-structured interviews
We chose to implement semi-structured interviews rather than focus groups because of our long-standing experience that children, particularly those in middle childhood, provide more information in discussion directly with an interviewer rather than in group settings. The semi-structured interviews were done to elucidate children’s experiences of the PROMIS patient-reported outcome domains of pain interference (i.e., the impact of pain on everyday life), fatigue (i.e., experiences of feeling tired and the impact of those experiences on everyday life), sleep disturbance (i.e., difficulties falling and staying asleep), and sleep-related impairment (i.e., sleepiness and its impact on daytime functioning). Interviews of children with Crohn’s disease probed on pain interference and fatigue, and those for children with chronic kidney disease addressed fatigue, sleep disturbance, and sleep-related impairment.
Two interviewers (JC, KDF) conducted the interviews. Both were trained in the semi-structured interview method by an experienced qualitative researcher (CBF). Each conducted simulated interviews with adult actors to ensure that all procedures were followed closely, including not using closed-ended questions to prompt children to think about their experiences. Open-ended questions addressed the experiences of pain, fatigue, and sleep and how they affected participants’ everyday lives. Each interview began with an assent process followed by questions intended to build rapport (e.g., “What are some of your favorite things to do after school?”). Participants were asked what “[they] think about when [they] hear the word(s) [pain, fatigue, or problems sleeping].” This first question was intended to elicit experiences of the domain. Then interviewers asked about impact on daily life: “Now let’s think about a time when you had [pain, fatigue, or problems sleeping]. What are some things that you couldn’t do in your life?” Once no new information was obtained, the participant was asked to think about a second experience, and the same process was used. Interviews were audio-recorded and transcribed verbatim.
Deductive content analysis
We tested the hypothesis that the types and breadth of experiences reported by children with chronic conditions regarding the domains of pain interference, fatigue, sleep disturbance, and sleep-related impairment would be no different from counterparts from the general population who were part of the initial instrument development samples. Thus, we took a deductive approach to content analysis [9]. Deductive content analysis starts with an existing conceptual framework and evaluates the degree to which it is applicable to newly generated data. The deductive approach to test an extant framework’s applicability to patients with chronic disease contrasts with the inductive or grounded theory approach which makes no a priori assumptions about the categories represented in the data.
To build the conceptual framework for each of the four PROMIS measures, we identified conceptually distinct categories, which we term facets. The facets for the two sleep domains were those reported by Bevans and colleagues that described the qualitative development of the item pools [10]. Facets were not reported in the manuscripts describing the development of the pain interference and fatigue item banks, so we developed new groupings for this study based on all the items that underwent psychometric testing (i.e., the item pools). Facets could have one or more item-level concepts. Some facets represented in the item pool were dropped from the item banks during psychometric evaluation and item response theory calibration. It is important to note that the four PROMIS item pools were generated after concept elicitation from children drawn from the general population, their parents, and experts as well as a literature review of instruments that measure the same domain.
Two investigators independently reviewed each audio-transcript and extracted meaning units, which we defined as conceptually distinct statements in the words of the participant regarding the experiences of a particular outcome (e.g., fatigue) or how it impacted their lives (i.e., how fatigue affected their daily functioning). Interviewers met to compare meaning unit extraction and adjudicated differences in order to produce a single set for each interview. The meaning units were independently assigned to a facet category and item-level concept within that category with differences adjudicated by one of the authors (CBF). New facets were added when none existed in the domain conceptual frameworks.
Data analysis
We described each participant sample in terms of demographic and clinical characteristics. The duration of each interview was recorded in minutes, and we computed the median and interquartile range for the two samples. The number of item-level concepts elicited was recorded for each measure, and we identified examples of meaning units that illustrated our approach for coding audiotext. For each domain we determined the cumulative number of new facets elicited from participants in sequential order to identify saturation—i.e., when no new facets had been elicited. Finally, we counted the number of participants from whom each conceptual facet was elicited.