https://orcid.org/0009-0007-6018-1367
Figures
Abstract
The Cystic Fibrosis (CF) Impact Questionnaire (CF-IQ) was qualitatively developed to assess the impact of CF in the context of treatment advancements and increased longevity. This study reports the CF-IQ validation. In this noninterventional validation study, people with CF completed the 40-item CF-IQ and validating patient-reported outcome measures (PROMs) via electronic diaries at enrollment (baseline) and at the 4-week follow-up. Validation consisted of modern methods and focus groups to finalize structural validity, and classical methods to assess internal consistency [1–3], test-retest reliability [4,5], concurrent validity [5], and known-groups validity [5] of the CF-IQ. At baseline, 214 adults completed the survey; 193 completed the follow-up survey. Unidimensional item response theory (IRT) models were separately fit to 5 prespecified domains (Control and Burden of CF Treatment Impacts, Physical Activity Impacts, Social Activity Impacts, Emotional Impacts, and Work/School Limitation Impacts). IRT local dependence (LD) statistics identified 17 redundant items. Two independent CF-patient focus groups (14 total patients) confirmed these findings, and the 17 items were dropped. Each domain defined on the final 23 items achieved the criterion of exact model fit as measured by the root mean squared error of approximation (RMSEA, values = 0), Internal consistency (Cronbach’s α) values ranged from 0.81 to 0.89, 4 of 5 domains achieved acceptable test-retest reliability, with intraclass correlation coefficient (ICC) values ≥ 0.7, acceptable concurrent validity was achieved for all domains, and known-groups validity was established. The novel CF-IQ is a psychometrically robust PROM capturing patient-centric impacts of CF in the context of the current standard of care.
Citation: Serrano D, Uzumcu A, Gerstein M, Ayasse ND, Engstrom E, Barnes FB, et al. (2025) Psychometric validation of the Cystic Fibrosis Impact Questionnaire (CF-IQ): A patient-reported outcome assessing impacts of cystic fibrosis. PLoS ONE 20(1): e0317775. https://doi.org/10.1371/journal.pone.0317775
Editor: Emrah Gecili, Cincinnati Children’s Hospital Medical Center, UNITED STATES OF AMERICA
Received: March 8, 2024; Accepted: January 3, 2025; Published: January 24, 2025
Copyright: © 2025 Serrano et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: “The full dataset has not been published due the presence of potentially identifying or sensitive personal information. Vertex Pharmaceuticals is committed to advancing medical science and improving patient health. This commitment includes the responsible sharing of data with qualified researchers. Proposals for the use of the data will be reviewed by a scientific board. Approvals are at the discretion of Vertex Pharmaceuticals and will be dependent on the nature of the request, the merit of the research proposed, and the intended use of the data. Please contact CTDS@vrtx.com if you would like to submit a proposal or need more information.”
Funding: Vertex Pharmaceuticals, Inc, funded this work in collaboration with Pharmerit International – An OPEN Health Company. DS, AU, MG, NDA, EE, FBB, and CI are former employees of and were receiving salary from Pharmerit International – An OPEN Health Company when this work was conducted. The funder had no role in data collection or analysis but contributed to the study design, interpretation of the results, and review of the manuscript as TT and LJM are employees of Vertex Pharmaceuticals, Inc. The specific roles of the authors are articulated in the 'author contributions' section of the manuscript. Vertex Pharmaceuticals, Inc. provided medical writing assistance from a salaried employee, Tracy Bunting-Early, and funded editorial assistance provided by Complete HealthVizion IPG Health Medical Communications and Nucleus Global, an Inizio Company.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: DS, AU, MG, NDA, EE, FBB, and CI were employees of Pharmerit International – An OPEN Health Company, which received funding from Vertex Pharmaceuticals, Inc. CI is currently an employee of Janssen Global Services, LLC. TT and LJM are employees of, receive salary from, and may own stock or stock options of Vertex Pharmaceuticals, Inc. ISG is on the faculty at and receives salary from Institut Necker-Enfants Malades Hospital, Paris, France; she has participated as a principal investigator on clinical studies for Vertex Pharmaceuticals, Inc.
Introduction
Cystic fibrosis (CF) is a life-shortening autosomal recessive disorder that impacts >80,000 people worldwide [6–8]. CF affects multiple organ systems, including the respiratory, gastrointestinal, and reproductive systems, due to mutations in the gene encoding the CF transmembrane conductance regulator (CFTR) ion channel [7,9] Consequent to this genetic mutation, CFTR dysfunction results in abnormal transport of salt and water across epithelial cell membranes throughout the body, causing thickening of mucus secretions [7,9]. This results in mucus retention, bronchial inflammation and infection, and finally, progressive lung damage and premature death [7,9].
Several patient-reported outcome measures (PROMs) have been developed to measure symptoms, impacts, and quality of life (QoL) in CF [10–12]. However, the majority of these PROMs were developed nearly two decades ago [13,14] when people with CF rarely survived to middle age due to more severe disease complications. Recent improvements in standards of care and widespread use of CFTR modulators (CFTRm) have increased life expectancy and QoL in CF [7,9].
This increase in longevity and corresponding expansion of life experiences motivates the development of new, reliable, and sensitive CF-specific instruments. Indeed, patients clearly express that new items are necessary to capture their experience [15]. The Cystic Fibrosis Impact Questionnaire (CF-IQ) was developed following current best practices and PROM development guidelines for use in an increasingly healthy CF population [16–18]. Impacts assessed are those that can only occur in individuals surviving into adulthood; for example, impacts related to advanced education, work and careers, and family. The CF-IQ complements existing CF-specific PROMs by capturing the broader impact of CF on peoples’ lives and characterizing their experiences as well as changes in their QOL in response to therapy [16,18,19]. In the content development of the CF-IQ, concept elicitation and cognitive interview debriefing with CF patients were employed, resulting in a draft 40-item questionnaire with 5- to 7-point verbal rating scales and a specified 7-day retrospective recall period [16]. Readers interested in the qualitative development can refer to McCarrier et al., 2020 for full details [10]. The objective of the present study was to complete the psychometric validation of the CF-IQ.
Methods
Overall study design, participants, and analysis
This noninterventional validation study (NIVS) consisted of 2 parts: part 1) a two-stage, mixed-methods analysis to examine and resolve redundant items in the qualitative development of the CF-IQ; part 2) a classical reliability and validity assessment of the final questionnaire items. The steps taken to validate the CF-IQ are shown in Table 1.
Individuals from the Rare Patient Voice (RPV) CF patient panel [20] were enrolled into this NIVS from 5 February 2020 through 13 March 2020, with the same participants involved in both part 1 and part 2 of the study. Eligible individuals were ≥18 years of age, living in the United States, had a CF diagnosis, were able to read, write, and speak English, were able to provide an estimate of their most recent percent predicted forced expiratory volume in 1 second (ppFEV1), were willing to complete the retest assessment, and were not participating in any interventional studies of CFTRm. The admissible ppFEV1 value range was between 15 to 130. Informed consent was obtained in writing from all individual participants included in the study via an electronic informed consent form prior to study enrollment and any data collection. This study, including all NIVS and focus group data-collection activities, were reviewed and granted an exemption for IRB oversight by the Advarra centralized institutional review board on November 13, 2019.
Between January and April 2020, enrolled participants completed the 40-item CF-IQ and validating PROMs at baseline and at follow-up (4 weeks later) (Table 2). Baseline for this NIVS, as with others, was simply the first data collection. Follow-up was conducted 4-weeks later to comply with the US food and drug administration’s (FDAs) division of clinical outcome assessment (DCOA) recommended retest interval for test-retest reliability. Demographic and clinical characteristics were gathered at enrollment. All data were remotely collected using the QuestionPro Survey environment, which minimized the effects of the SARS-CoV-2 pandemic on the conduct of this study. In addition to the validation analyses (MPMs, internal consistency, test-retest reliability, concurrent validity, and known-group validity), sample demographic and clinical characteristics, rates of participation at baseline and week 4, and the occurrence of any loss to follow-up or missing data were examined. The response distributions for all CF-IQ items were tabulated and examined for floor and ceiling effects and sparse response categories. Note that in all analyses missing data were not imputed and only observed data were analyzed.
Part 1: Resolving item redundancy in the CF-IQ
In summary for part 1, modern psychometric method (MPM)-based local dependence statistics were employed to identify redundant item groups within domains. Any items identified as redundant were eligible for removal from the CF-IQ. Which items to retain from identified redundant item groupings were adjudicated by two qualitative focus-groups. The adjudicated solutions were evaluated in a final round of MPMs.
During qualitative development, the CF-IQ was found to consist of the following domains: Control and Burden of CF Treatment Impacts, Physical Activity Impacts, Social Activity Impacts, Emotional Impacts, and Work/School Limitation Impacts. Using MPMs (i.e., item response theory [IRT] models), the unidimensionality of each domain and any local dependence (LD) [27,28] among items within each domain of the CF-IQ were evaluated. Any items presenting with significant LD (P < .05) were flagged as redundant [27]. For locally dependent item combinations, items proposed for retention were those whose item content was deemed qualitatively to be most relevant to patients and had the numerically largest IRT slopes. Note that LD does not necessarily occur in pairs and in some cases multiple items were found to be jointly locally dependent.
Subsequent to redundant item removal, unidimensional IRT models were refit to each domain composed of the retained items, and model fit [29] and LD were re-evaluated [27,29]. Model fit was evaluated via the C2-based root mean squared error of approximation (RMSEA), for which a value of < 0.05 was deemed an acceptable model fit. If an acceptable model fit and no LD was observed, then that item set was deemed refined. Each domain’s refined item set proceeded to the next stage, where focus groups were asked to assess the refined item set. Two independent virtual focus groups were interviewed and observed by trained members of the research team using a semi-structured interview guide in ~90-minute sessions (S1 Appendix). Focus group members were recruited via email sent by RPV and sampled from participants who had participated in and completed the NIVS.
Participants were presented with each grouping of LD-identified redundant items, one grouping at a time, and asked which item they would retain if they could retain only one item. Participants were instructed to make their decision based on what item they found to be most relevant to their specific experience with impacts from CF. Interviews were audio recorded (with participants’ consent), transcribed verbatim, and analyzed using ATLAS.ti v8.0.
Two independent researchers coded and reviewed both focus groups’ transcripts to ensure inter-coder agreement. Coding outputs were reviewed to ensure that proposed item removal aligned with CF participant perspectives and, when needed, to resolve any inconsistencies between qualitative and psychometric evidence. When focus group data disagreed with removal decisions, the unidimensional IRT model was refit to that domain, alternating the item retained to evaluate the optimal item configuration. ‘Optimal’ was defined by the configuration associated with the best model fit and strongest IRT slope. This, along with consideration of the strength of psychometric findings, original qualitative concept elicitation, cognitive interview results, and focus group analysis, was used to arrive at the final CF-IQ item and domain composition. Please refer to the online supplement for a detailed example of the employed MPM strategy based on the physical activity impacts domain (S1 Summary).
Part 2: Assessing reliability and validity of the refined CF-IQ
After resolving redundancy in Part 1, Part 2 involved assessing the internal consistency, test-retest reliability, concurrent validity, and known-groups validity of the CF-IQ domain scores.
The CF-IQ and validating variables collected for the NIVS are summarized in Table 2 [12,21–26]. Two global impression anchor variables were developed for this study. The Patient Global Impression of Severity (PGIS) is a self-reported, single-item, global scale subjectively assessing the severity of patients’ clinical status. Likewise, the Patient Global Impression of Change (PGIC) is a single item assessing subjective perceived change in clinical status. The FDA’s latest guidance documents, at the time of this study’s conduct, recommended using these anchor variables to support validation activities [19].
CF-IQ reliability was assessed with both internal consistency (McDonald’s ω and Cronbach’s α) [30,31] at baseline and test-retest reliability. Test-retest reliability was computed via the two-way, random, intraclass correlation coefficient (ICC (2,1)) [32] for CF-IQ scores between baseline and week 4 in symptomatically stable subgroups. The two symptomatically stable retest samples consisted of participants reporting no change on the PGIS between baseline and week 4 and participants reporting no change on the PGIC at week 4.
CF-IQ validity was assessed at baseline using both concurrent and known-groups validity. Concurrent validity was estimated using Pearson correlations between CF-IQ domain scores and scores from the Cystic Fibrosis Questionnaire-Revised (CFQ-R), Illness Perception Questionnaire-Revised, Patient-Reported Outcomes Measurement Information System Social Role Short Form-4 Items, PGIS, Treatment Burden Questionnaire, and Work Productivity and Activity Impairment Questionnaire-Specific Health Problem (WPAI-SHP). Known-groups validity was estimated using analysis of variance (ANOVA) models. Known anchor groups were defined on PGIS (“not limited,” “a little limited,” “moderately limited,” “severely limited,” or “extremely limited”), ppFEV1 groups (ppFEV1 ≤ median value or ppFEV1 > median value), and CFTRm therapy duration groups (untreated with CFTRm, treated for ≤4 weeks or treated for >4 weeks). One ANOVA was estimated for each anchor and CF-IQ domain combination. The dependent variable for each ANOVA was the CF-IQ. Known-groups contrasts were structured using reference cell coding; the reference groups for each known group were “not limited,” “ppFEV1 ≤ median ppFEV1 value,” and “untreated” in the PGIS, ppFEV1, and CFTRm therapy groups, respectively.
Statistical software
MPM analyses were conducted using either the multidimensional IRT (mirt) R package [33] or flexMIRT version 3.5 [34]. Descriptive and classical test theory statistics were generated using a combination of Statistical Analysis System ([SAS] Software, version 9.4, SAS Institute Inc, Cary, NC) [35] and R statistical software (R, version 3.4.3, R Development Core Team) [36].
Results
Description of survey sample
At baseline, 296 RPV panelists with CF were identified and sent QuestionPro survey URLs. Of the panelists, 80 did not qualify based on eligibility criteria or did not provide complete survey data. Two of the remaining 216 panelists were removed: one for excessively fast survey completion (<6 minutes) and one for providing a duplicate record, yielding a sample of 214 at baseline, 72% response rate. The baseline sample flow is depicted in Fig 1. All 214 participants were presented with follow-up survey links at week 4 and 193 returned usable data, 90% response rate. For each CF-IQ item, a t test compared the average item response at baseline between participants who completed the week-4 follow-up survey and those who did not. Even without adjustment for multiplicity, no comparison was statistically significant, suggesting that attrition bias was unlikely.
Demographics and clinical characteristics
Participants ranged in age from 18 to 72 years, with a majority (79%) being female (Table 3). Mean (standard deviation) for the most recent ppFEV1 was 70.0 (24.0), taken on average ≈2 months before the baseline survey. Most baseline survey participants (72.9%) experienced ≥1 pulmonary exacerbation (PEx) in the past year; 15.4% had received a lung transplant. Of 157 participants receiving CFTRm, nearly all (98.1%; n = 154) had done so for >4 weeks.
Descriptive assessment of CF-IQ
Almost all CF-IQ items demonstrated some response category sparseness (categories with 10% or less of the sample endorsing); however, this was not a barrier to analysis because sufficient representation was observed across response categories for all CF-IQ items. Overall, most impacts were mildly to moderately severe, with the middle three response categories heavily endorsed.
Part 1: Resolving item redundancy in the CF-IQ
Across the five prespecified domains, 31 (78% of 40 CF-IQ items) CF-IQ items presented with substantial LD across 13 patterns described in Table 4. Observed patterns of LD involved a minimum of two to a maximum of four items. Of these 31 items, 17 (43% of the original 40 CF-IQ items) were found to be statistical drivers of the 13 unique LD patterns. These 17 items were identified as redundant and therefore candidates for removal from the CF-IQ. For example, in the Physical Activity domain, the following items presented with significant LD: CF-IQ item 2 “How difficult was it for you to keep up with others while running” and CF-IQ item 6 “How much of the time did you need to take breaks or rest in order to complete tasks at home, work, or school?”; while both had strong slopes (2.33 and 2.86, respectively), CF-IQ item 6 had the higher slope and was therefore retained. In circumstances where multiple items were found to be jointly locally dependent, the smallest set of items that could be removed and result in LD resolution were identified. For example, in the case of the Emotional Domain items 15,16,17, and 19, with item content of “feeling frustrated”, “feeling irritable”, “feeling moody” and “feeling stressed”, respectively, were found to be locally dependent and the local dependence was resolved by retaining item 15, “feeling frustrated”. When these 17 items were removed and the unidimensional IRT models refit to each domain, model fit was perfect (RMSEA = 0.00) for each domain.
The two independent focus groups were presented the 31 items defining the 13 unique LD patterns and asked to decide which items were redundant and which to retain. The focus groups agreed that 16 of the 17 (94%) items deemed redundant by psychometric evidence, were in fact redundant and should be removed. The one disagreement with the psychometric evidence involved the first unique LD grouping given in Table 4. Both focus groups agreed that item 33 (“Bothered by the amount of time needed to complete CF treatments?”) would evaluate the overall time spent on CF treatment; and that this would provide a more holistic view of treatment burden compared to that of item 35 (“Frustrated about activities you missed because of the time you spent on CF treatment?”), which focused on frustration with activities missed due to CF treatment. With respect to unique LD grouping 2, focus groups (FG) 1 and 2 disagreed on which item to retain: FG 1 agreed with the psychometric evidence and FG2 did not. To resolve which item in each item pair (LD grouping 1: 33 & 35, and LD grouping 2: 34 &36) should be retained, IRT models were refit to the control and burden of treatment domain; and following the criteria described in methods, items 33 and 36 were retained. Therefore, the final item content for the CF-IQ was in agreement with the feedback received from participants in two focus groups.
The final C2-root mean square error approximation (RMSEA)-based model fit and ω-based internal consistency estimates for each of the five CF-IQ domains, defined upon the final 23 CF-IQ items (40 minus 17 items), are presented in Table 5. Per the proofs given by McDonald et al. α underestimates internal consistency and is, therefore, negatively biased relative to the ω. ω is therefore optimal at the stage of domain evaluation [34]. All domains for which model fit was estimable (all but Work/School Limitations) exhibited perfect fit (C2-based RMSEA point estimates of zero and null p values) and were associated with superlative internal consistency, with estimates of ω ranging between 0.87 and 0.92. The domain modifications and IRT item slopes corresponding to the final model specification are given in Table 6. As implied by the magnitude of the observed ω-based internal consistency, IRT slopes for the retained items were uniformly robust, with all exceeding 1.0 and some approaching deterministic slope magnitudes (e.g., CF-IQ item 22 “Difficulty participating in activities you enjoy doing” and item 28 “Hard time keeping up with daily tasks at home/work/school”). Items not retained in the final model are illustrated in grey text in Table 6.
Combined, this evidence concluded the item finalization process. CF-IQ domain scoring was based on unit-weighted sum scores as supported by the ω statistics [28,37], under which higher scores indicate greater disease impact. The refined CF-IQ is available from the Mapi Research Trust at: https://eprovide.mapi-trust.org/instruments/cystic-fibrosis-impact-questionnaire.
Part 2: Assessing reliability and validity of the refined CF-IQ
Reliability and validity for the five domains defined on the final 23-item CF-IQ were robust (Fig 2A–2D). All CF-IQ domains had acceptable internal consistency, as demonstrated by Cronbach’s α values ≥ 0.7. As shown in Fig 2A, Cronbach’s α values ranged from 0.81 to 0.89, with the Physical Activity Impacts and Social Activity Impacts domains having the highest estimates; the Work/School Limitation Impacts domain had the lowest estimate.
Plots of the (A) internal consistency, (B) test-retest reliability, (C) concurrent validity, and (D) known-groups validity of the CF-IQ domains CF-IQ, Cystic Fibrosis Impact Questionnaire.
The Control and Burden of CF Treatment Impacts, Emotional Impacts, Work/School Limitation Impacts, and Physical Activity Impacts domains demonstrated acceptable test-retest reliability for the PGIS anchor, as indicated by ICC values ≥ 0.7 (Fig 2B). The Control and Burden of CF Treatment Impacts, Emotional Impacts, and Social Activity Impacts domains did not reach acceptable test-retest reliability for the PGIC anchor (data not shown).
Concurrent validity was robust, with 96% of estimated correlations between CF-IQ domains and validators meeting or exceeding the criterion of |r| ≥ 0.4, indicating that those instruments expected to measure similar constructs to the CF-IQ domains do appear to measure similar constructs (Fig 2C). Three correlations between the CF-IQ and the WPAI-SHP validator failed to meet the criterion for concurrent validity. Note that Fig 2C presents correlation magnitudes graphically relative to the prespecified cut-offs defining acceptable concurrent validity (|r| ≥ 0.4). While not all numerical values are clearly observable (e.g., when one validator is correlated at 0.48 and another correlated at 0.50, proximity will result in graphical overlap), there are five discriminant validators (the CFQ-R domains as they are reverse-scored relative to CF-IQ) and eight convergent validators. Therefore, readers can evaluate the relative distance of correlations to thresholds and each other, knowing there will be five discriminant correlations for each domain in proximity to -0.4 and eight convergent correlations for each domain in proximity to 0.4.
Known-groups validity evidence estimated in PGIS known groups was strong, though more modest when estimated across ppFEV1 and CFTRm known groups (S1 Table). For all CF-IQ domains, each severity group based on the PGIS had domain scores that were significantly different from the reference group (Fig 2D). In each domain, the effect group score means were higher than the reference group score mean, and the semipartial ω2 estimate of the effect size of the PGIS group predictor exceeded the prespecified criterion of 0.05 (semipartial ω2 range: 0.25–0.47 across domains). This suggests that groups expected to have CF-IQ scores indicating more severe disease do show this trend.
Discussion
The CF-IQ was developed as a CF-specific PROM following current PROM development guidelines [17,18] and standard qualitative practice [16]. In this study, a novel, hybrid study design was utilized to increase patient-centricity to resolve item redundancy and validate the CF-IQ. This hybrid approach allowed the collection of a large validation sample. The evidence derived from this large validation sample was then refined with the assistance of patient input obtained from patient focus group discussions. A total of 17 items were identified as redundant and removed.
Based on psychometric evidence, focus group interviews confirmed all but one item removal decision. This discrepancy between the psychometric evidence and focus group findings was not concerning, as both items in the redundant pair exhibited similar strong item performance; the instrument’s overall performance was excellent with either retained. The hybrid design and resulting evidence demonstrated strengths of linking psychometric and qualitative evidence to clarify guidance from the patient voice. This study illustrates how evidence obtained from correctly implemented MPMs can corroborate the patient perspective and vice versa.
After item pool finalization, 23 of 40 items were retained across five domains: Control and Burden of CF Treatment Impacts, Physical Activity Impacts, Social Activity Impacts, Emotional Impacts, and Work/School Limitation Impacts. Classical validation (reliability and validity) of the 23-item CF-IQ demonstrated that it is a psychometrically robust PROM for use among adults with CF and is fit for purpose in providing a holistic understanding of the impact of CF on QoL. However, results were not uniform. For test-retest reliability, three domains did not meet the prespecified criterion for the PGIC, while the majority of the domains demonstrated acceptable test-retest reliability for the PGIS anchor. Estimates conditioned on the PGIS generally outperform those conditioned on the PGIC, when used for validation. This is routinely explained as a function of recall bias. The PGIC requires the individual to evaluate change over time and is expected to carry recall bias; this attenuates the ICC relative to the PGIS, an absolute measure of perceived severity at the time of survey completion. Validation also confirmed that the CF-IQ domain scores had a high level of agreement with other validated PROMs, including the widely used CFQ-R. The strong correlation observed in concurrent validity between CF-IQ and other validating PROMs allowed us to confirm that CF-IQ accurately measured overlapping concepts across validated PROMs. Note that a relevant distinction between the CFQ-R and CF-IQ is that the latter was developed to characterize impacts reflective of the longer lifespans prevalent in the CFTRm era. To this point, the CF-IQ item responses reflected moderate to severe impacts with polar categories under-endorsed. This is in substantial contrast to patterns observed in the pre-CFTRm era, under which the vast majority of legacy PROM-based impact items possessed non-trivial ceiling effects [26].
This study was conducted during the early months of the SARs-CoV-2 outbreak in January through April 2020. This study was designed (prior to SARs-CoV-2) to employ sampling via a virtual collection deployed in RPV’s CF panel to facilitate efficient data collection. As a result, in-person site-based sampling and laboratory-based confirmation of inclusion/exclusion criteria were infeasible. However, the virtual design allowed us to complete data collection efficiently and unaffected by SARs-CoV-2. The QuestionPro survey environment enabled collection of high-integrity data while preserving methodologic principles that ensured data rigor within a HIPAA-compliant framework. This approach allowed us to achieve an adequate representative sample for validation efforts supporting continued research. Response rates were high (90% at follow-up) and statistical testing provided further evidence against sampling bias in the survey collection.
An additional advantage conferred by the virtual survey sampling design was rapid access to a large sample size in a rare disease population; however, virtual survey design has inherent limitations. Survey respondents were self-selected into the panel, so they may not represent the wider CF population. Consistent with another large online survey of people with CF [38], women were over-represented in the sample; however, lack of evidence of differential item functioning by sex indicates that responses were unbiased. Clinical information (e.g., lung function) was self-reported by study participants and accuracy of these values could not be verified; however, based on the reported data, the study sample appears to be generally similar to the adult population in the US Cystic Fibrosis Foundation Patient Registry [6]. The study did not collect information on CF genotype in participants; therefore, potential response differences by genotype or treatment availability could not be explored. Finally, it is possible that respondents intentionally entered incorrect or false responses; this was mitigated by excluding a small number of participants with excessively fast survey completion or duplicate records.
In conclusion, the CF-IQ is a psychometrically robust PROM capturing patient-centric outcomes to understand the broad life impacts of CF and potential benefits of novel treatments in an adult population with CF. These findings support the further development of the CF-IQ alongside clinical endpoints in future interventional trials and studies with expanded populations. Assessments of the responsiveness of the CF-IQ to change and clinical significance thresholds will be addressed in an ongoing interventional trial.
Supporting information
S1 Appendix. Additional information related to the focus group interviews including interview guide.
https://doi.org/10.1371/journal.pone.0317775.s001
(DOCX)
S1 Summary. Illustrative example of modern psychometric methods.
https://doi.org/10.1371/journal.pone.0317775.s002
(DOCX)
S1 Table. Known-groups validity at baseline across ppFEV1 and CFTRm known groups.
https://doi.org/10.1371/journal.pone.0317775.s003
(DOCX)
Acknowledgments
The authors thank Rare Patient Voice (RPV) for the use of their patient panel to conduct this study. We would also like to thank the patients for their time. We thank Kelly McCarrier, PhD, MPH, of OPEN Health Group, for consultation on qualitative evidence and advising on the CF-IQ domains and item redundancy. Medical writing assistance was provided by Tracy Bunting-Early of Vertex Pharmaceuticals Incorporated. Editorial support was provided by Nucleus Global, an Inizio Company, and Complete HealthVizion IPG Health Medical Communications.
References
- 1. Prinsen CAC, Mokkink LB, Bouter LM, Alonso J, Patrick DL, de Vet HCW, et al. COSMIN guideline for systematic reviews of patient-reported outcome measures. Qual Life Res. 2018;27(5):1147–57. Epub 2018/02/13. pmid:29435801; PubMed Central PMCID: PMC5891568.
- 2. Sloan JA, Dueck A, Qin R, Wu W, Atherton PJ, Novotny P, et al. Quality of Life: The Assessment, Analysis, and Interpretation of Patient‐Reported Outcomes by FAYERS P. M.and MACHIN D. Biometrics. 2008;64(3):996–.
- 3. Mokkink LB, Terwee CB, Patrick DL, Alonso J, Stratford PW, Knol DL, et al. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res. 2010;19(4):539–49. Epub 2010/02/20. pmid:20169472; PubMed Central PMCID: PMC2852520.
- 4. Liljequist D, Elfving B, Skavberg Roaldsen K. Intraclass correlation—A discussion and demonstration of basic features. PLoS One. 2019;14(7):e0219854. Epub 2019/07/23. pmid:31329615; PubMed Central PMCID: PMC6645485.
- 5. Mokkink LB PC, Patrick DL, Alonso J, Bouter LM, de Vet H, Terwee CB User Manual: COSMIN methodology for systematic reviews of Patient‐Reported Outcome Measures (PROMs) 2018 [15 August 2024]. Available from: https://cosmin.nl/wp-content/uploads/COSMIN-syst-review-for-PROMs-manual_version-1_feb-2018.pdf.
- 6. Cystic Fibrosis Foundation. 2019 Patient Registry: Annual Data Report2020 30 Sep 2021. Available from: https://www.cff.org/Research/Researcher-Resources/Patient-Registry/2019-Patient-Registry-Annual-Data-Report.pdf.
- 7. Elborn JS. Cystic fibrosis. Lancet. 2016;388(10059):2519–31. Epub 2016/05/04. pmid:27140670.
- 8. European Cystic Fibrosis Society. ECFS Patient Registry. Annual Data Report. 2018 data2020 24 Jun 2022. Available from: https://www.ecfs.eu/sites/default/files/general-content-files/working-groups/ecfs-patient-registry/ECFSPR_Report_2018_v1.4.pdf.
- 9. Mall MA, Burgel PR, Castellani C, Davies JC, Salathe M, Taylor-Cousar JL. Cystic fibrosis. Nat Rev Dis Primers. 2024;10(1):53. Epub 2024/08/09. pmid:39117676.
- 10. Goldbeck L, Schmitz TG, Henrich G, Herschbach P. Questions on life satisfaction for adolescents and adults with cystic fibrosis. Development of a disease-specific questionnaire. Chest. 2003;123(1):42–8. Epub 2003/01/16. pmid:12527601.
- 11. Gee L, Abbott J, Conway SP, Etherington C, Webb AK. Development of a disease specific health related quality of life measure for adults and adolescents with cystic fibrosis. Thorax. 2000;55(11):946–54. Epub 2000/10/26. pmid:11050265; PubMed Central PMCID: PMC1745639.
- 12. Quittner AL, Buu A, Messer MA, Modi AC, Watrous M. Development and validation of The Cystic Fibrosis Questionnaire in the United States: a health-related quality-of-life measure for cystic fibrosis. Chest. 2005;128(4):2347–54. Epub 2005/10/21. pmid:16236893.
- 13. Henry B, Aussage P, Grosskopf C, Goehrs J-M. Development of the Cystic Fibrosis Questionnaire (CFQ) for assessing quality of life in pediatric and adult patients. Qual Life Res. 2003;12(1):63–76. pmid:12625519
- 14. Quittner AL, Sweeny S, Watrous M, Munzenberger P, Bearss K, Gibson Nitza A, et al. Translation and linguistic validation of a disease-specific quality of life measure for cystic fibrosis. J Pediatr Psychol. 2000;25(6):403–14. pmid:10980045
- 15. Coucke R, Chansard A, Bontemps V, Grenet D, Hubert D, Martin C, et al. "Il faut continuer a poser des questions" patient reported outcome measures in cystic fibrosis: An anthropological perspective. J Cyst Fibros. 2021;20(6):e108–e13. Epub 20210227. pmid:33648900.
- 16. McCarrier KP, Hassan M, Hodgkins P, Suthoff E, McGarry LJ, Martin ML. The Cystic Fibrosis Impact Questionnaire: qualitative development and cognitive evaluation of a new patient-reported outcome instrument to assess the life impacts of cystic fibrosis. J Patient Rep Outcomes. 2020;4(1):36. Epub 2020/05/15. pmid:32405878; PubMed Central PMCID: PMC7221045.
- 17. Patrick DL, Burke LB, Gwaltney CJ, Kline Leidy N, Martin ML, Molsen E, et al. Content validity—establishing and reporting the evidence in newly developed patient-reported outcomes (PRO) instruments for medical product evaluation: ISPOR PRO Good Research Practices Task Force Report: Part 1—eliciting concepts for a new PRO instrument. Value Health. 2011;14(8):967–77. Epub 2011/12/14. pmid:22152165.
- 18. U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER), Center for Devices and Radiological Health (CDRH). Guidance for Industry. Patient-reported outcome measures: use in medical product development to support labeling claims2009 24 Jun 2022. Available from: https://www.fda.gov/media/77832/download.
- 19. U.S. Food & Drug Administration. Patient-focused drug development guidance: methods to identify what is important to patients and select, develop or modify fit-for-purpose clinical outcomes assessments2018 24 Jun 2022. Available from: https://www.fda.gov/drugs/news-events-human-drugs/patient-focused-drug-development-guidance-methods-identify-what-important-patients-and-select.
- 20. Rare Patient Voice. Home page2022 24 Jun 2022. Available from: https://rarepatientvoice.com/.
- 21. Tran V-T, Montori VM, Eton DT, Baruch D, Falissard B, Ravaud P. Development and description of measurement properties of an instrument to assess treatment burden among patients with multiple chronic conditions. BMC Med. 2012;10:68. pmid:22762722
- 22. Reilly MC, Zbrozek AS, Dukes EM. The validity and reproducibility of a work productivity and activity impairment instrument. Pharmacoeconomics. 1993;4(5):353–65. pmid:10146874
- 23. Bevans M, Ross A, Cella D. Patient-Reported Outcomes Measurement Information System (PROMIS): efficient, standardized tools to measure self-reported health and quality of life. Nurs Outlook. 2014;62(5):339–45. pmid:25015409
- 24. Hahn EA, DeWalt DA, Bode RK, Garcia SF, DeVellis RF, Correia H, et al. New English and Spanish social health measures will facilitate evaluating health determinants. Health Psychol. 2014;33(5):490–9. pmid:24447188
- 25. Moss-Morris R, Weinman J, Petrie K, Horne R, Cameron L, Buick D. The revised Illness Perception Questionnaire (IPQ-R). Psychol Health. 2002;17(1):1–16.
- 26. Quittner AL, Sawicki GS, McMullen A, Rasouliyan L, Pasta DJ, Yegin A, et al. Psychometric evaluation of the Cystic Fibrosis Questionnaire-Revised in a national sample. Qual Life Res. 2012;21(7):1267–78. Epub 2011/10/14. pmid:21993695.
- 27. Chen W-H, Thissen D. Local Dependence Indexes for Item Pairs Using Item Response Theory. Journal of Educational and Behavioral Statistics. 1997;22(3):265–89.
- 28. Stout WF. A new item response theory modeling approach with applications to unidimensionality assessment and ability estimation. Psychometrika. 1990;55:293–325.
- 29. Maccallum R, Browne MW, Sugawara HM. Power analysis and determination of sample size for covariance structure modeling. Psychological Methods. 1996;1:130–49.
- 30.
McDonald RP. Test Theory. A Unified Treatment. Mahwah, NJ: Lawrence Erlbaum; 1999.
- 31. Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika. 1951;16(3):297–334.
- 32. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86(2):420–8. Epub 1979/03/01. pmid:18839484.
- 33. Chalmers RP. mirt: a multidimensional item response theory package for the R environment. J Stat Softw. 2012;48:6.
- 34.
Cai L. flexMIRT® version 3.51: Flexible multilevel multidimensional item analysis and test scoring [Computer software]. Chapel Hill, NC: Vector Psychometric Group; 2017.
- 35.
SAS. SAS version 9.4. Cary, NC: SAS Institute Inc.
- 36. The R Foundation. What is R?2022 24 Jun 2022. Available from: https://www.r-project.org/about.html.
- 37. Rodriguez A, Reise SP, Haviland MG. Evaluating bifactor models: calculating and interpreting statistical indices. Psychol Methods. 2016;21(2):137–50. pmid:26523435
- 38. Goss CH, Edwards TC, Ramsey BW, Aitken ML, Patrick DL. Patient-reported respiratory symptoms in cystic fibrosis. J Cyst Fibros. 2009;8(4):245–52. Epub 2009/06/02. pmid:19481983.