1. Introduction

Cystic fibrosis (CF) is a genetic multisystem disease with an incidence of approximately one per 3848 live births in Canada and an estimated 80 new diagnoses annually [1, 2]. Impressively, in Canada in 2022, the median survival in people with CF (pwCF) has reached 60 years in contrast to two decades ago when it was 37 years of age based on the Canadian CF Registry (CCFR) data and speaks of the advancements in research and care in this field [1].

Initially, the CCFR was an extension of the United States’ CF Foundation (CFF) Patient Data Registry whose cornerstone was laid by Dr. Warwick in 1966. By 1973, Canadian CF clinics (N = 20) were contributing Canadian data to the United States’ CFF Registry and the first formal joint United States–Canadian CF report summarizing data captured from 1966 to 1976 that was published in 1978 [3]. This report detailed information on national CF birth and death rates, survival curves, and age of diagnosis. In the late 1970s, clinical data such as lung function and markers of nutrition (e.g., height and weight) and microbiology were added to the registry. In 1984, CF Canada established a standalone CCFR that included all historical data dating back to the 1970s and continues to capture patient demographic and clinical and vital statistics longitudinally across the lifetime of the individual.

Currently, all accredited CF clinics are supported by clinic incentive grants contingent on data submission to the CCFR, and each clinic is uniquely able to access aggregate data relating to national epidemiologic trends as well as patient-level data pertaining to their respective patient population. Incentives for pwCF to receive care from Canadian CF centers and hence have their data captured in the registry include being able to obtain medications from a pharmacy associated with a CF clinic where many of their CF medications are paid for by provincial drug plans. It is estimated that less than 1% of individuals who are approached to participate in the CCFR decline consent to participate (personal communication, CF Canada), which enables powerful population-based analyses as well as access by pwCF to research trials and quality improvement initiatives [4].

Given the fact that the CCFR is a valuable tool in clinical care, surveillance of epidemiologic trends, and research, much effort has gone into iterative updates and improvements in data collection and monitoring. The CCFR data can be used by care teams in their clinical interactions, planning for needed resources, and understanding their practice in the national context. Each year, an Annual Data Report is published by CF Canada which documents key demographic and clinical metrics to paint a picture of the evolving landscape [1]. With the large therapeutic pipelines in CF, the CCFR plays an important role in the provision of pharmaceutical postapproval data for regulators as well as obtaining an understanding of real-world effectiveness [5]. To obtain CCFR data, researchers are able to submit an application to CF Canada, which is then reviewed by a panel of CF researchers for approval [6], and these embedded record-level case-report data may be used for retrospective or prospective analyses.

Leveraging population-based registries for research and clinical purposes requires that the data integrity is of the highest quality. In the United States, Knapp et al. conducted an audit of the CFF Patient Registry (CFFPR) and demonstrated that the data reported in the registry had a high agreement with the patient’s medical charts [7]. Although CF Canada strives to ensure the same with the CCFR, no formal validation of the data has been previously conducted. Therefore, the aim of this study was to conduct an audit of the accuracy of the data captured within the CCFR. The hypothesis was that the data would be highly accurate and complete for key demographic and clinical metrics.

2. Methods

3. Results

4. Discussion

As a recognized example of a successful population-based registry, the CCFR comprises the vast majority of pwCF attending accredited CF clinics. In our validation of the CCFR data, we identified that the registry data are complete and accurate for meaningful clinical variables with little missingness. Of the Canadian population included in the registry, we identified that clinical measurements and hospitalizations were accurate in approximately 96% or more of cases. Biological sex (at birth) and date of birth were concordant between the CCFR and medical records in all cases. Medications and CF-related complications were also entered with high accuracy in more than 90% of cases. The lowest accuracy in the medications category was for pancreatic enzymes use, which may be due to how data are entered, since the variable indicating pancreatic status (pancreatic insufficiency vs. sufficiency) is entered as a separate field and matched in 99.0% of cases. This speaks about the need to consider challenges with accuracy when variables with overlap or redundancy are collected. Notably, newer therapies such as CFTR modulators inclusive of start dates were entered with complete accuracy. This is of particular importance as the registry can be used to identify those eligible for clinical trials of novel therapeutics and requires that these data be accurate.

With the evolution of registries and their interfaces (i.e., online portals) such as the CCFR, quality improvement initiatives have become an increasingly important focus and more feasible at the population level. The successful use of the United States-based CFFPR for individual, clinic, and national level initiatives as well as serving as a repository of quality improvement resources has led to the adoption of this model across multiple countries. As an example, Germany developed a quality improvement program enabling the identification of approaches that can lead to improved CF outcomes [8, 9]. The comparison of nutritional outcomes in the Toronto and Boston CF centers is yet another example of the power of a proactive quality improvement program [10], as these results led to significant shifts in nutritional interventions for pwCF. The SARS-CoV-2 pandemic also showcased the ability of multicountry registries to harmonize and identify the impacts of the pandemic on pwCF [11]. There is a completed and ongoing work to assess differing care models, treatment approaches, and their resulting outcomes demonstrating the power that can be harnessed from population-based registries [12–15]. Furthermore, efforts are underway to link CCFR data with other sources which may further the utility of the registry and serve as another quality assurance strategy [16, 17].

Other countries with established CF registries report coverage ranging from 41% to 99% of their CF population and may also vary in the breadth and design of data collection such as whether the data are collected at each encounter or only annually [18]. The Knapp et al.’s audit of the United States’ CF registry estimated that the CFFPR includes approximately 81%–84% of pwCF in the country on the basis of their attendance at an accredited center, where the estimate of individuals is not yet diagnosed as well as the consent for the registry [7, 19]. The United States’ CFFPR audit identified that they had a similar high accuracy of clinic visits and hospitalizations. A French CF registry audit undertaken by Pellen et al. demonstrated that there was a high degree of accuracy in clinical measures, but variability was noted in the measurement processes such as for lung function (based on the use of different reference equations) between clinical centers [20]. As there have increasingly been harmonized analyses to examine important outcomes relating to lung health and survival [21, 22], it is important to ensure that the registries are sufficiently similar in their accuracy and documentation on variable definitions to ensure appropriate harmonization of data across registries [23].

Although a number of key clinical variables relating to health encounters and medications were accurately recorded in the CCFR, variables such as date of diagnosis and race were less accurate. Challenges with validating the date of CF diagnosis may reflect the loss of these precise data as individuals transition from pediatric to adult clinics and may be recorded as an early or late diagnosis in adulthood. It may be that the date of diagnosis has a limited utility and predictive value or impact in an audit rather than having a general timeframe of diagnosis (i.e., infancy, early childhood, and adult). In our validation, race had a high level of accuracy (99.4%); however, it also has a large number of records that could not be validated (28). This variable is often determined by the CF clinic staff rather than reported by the patients themselves and may not accurately reflect the true diversity of the CF population. Furthermore, the categories used to define race are not comprehensive and may not reflect multiracial backgrounds accurately. Evaluations in other large observational health datasets such as the United States-based Healthcare Costs and Utilization Project have demonstrated that race or ethnicity was unknown in 57% of patients (of 160 million), and information was discrepant with the electronic health record in two-thirds of cases [24]. Race has long been collected in datasets for administrative, clinical, and research purposes. However, there are a myriad of issues to its collection, not the least of which is that race is a social construct without a clear correlation to ancestry [25, 26], and compounded by issues of cultural insensitivity and lack of understanding relating to its use and risks for discrimination [27–30]. To work toward diminishing health inequities [31], employing multifaceted approaches inclusive of patient engagement and adopting an antiracism culture may be key to better collection of self-identified ethnicity/ancestry data more compassionately and accurately [26].

Our validation had little to no missingness in the CCFR data with the exception of lung function values (110 records, 16% of the total number of records). However, it is important to note that lung function is not typically measured in children < 6 years of age. Young children will have height and weight measured during a clinic visit but their lung function (i.e., ppFEV₁) would be recorded as missing for the purpose of data validation. Although we made efforts to obtain a representative sample of CF clinics by geographic region and clinic size for the audit, there was overrepresentation by pediatrics, smaller clinic sizes, and English-speaking clinics. However, among the individuals with CF who were included in the data validation, they appeared to be representative of the general Canadian CF population, with a slight overrepresentation of those living posttransplant. Moreover, we cannot be certain that the full Canadian picture is represented by our sample although we used a stratified random sampling strategy within clinics and there are national standards for clinical practice which would minimize this bias. We made an assumption about the accuracy of the clinical record, but it is possible that inaccuracies from the medical chart were transcribed into the registry when original documents were not available. Given the cross-sectional nature of our audit, longitudinal trends or loss to follow-up were not systematically evaluated; however, the loss to follow-up rates are low (2.5%) as reported in prior publications [21] and in annual CF registry reports [1]. Our variable selection was broadly consistent with that of Knapp et al. in the 2016 CFFPR audit enabling comparison [7], but future data validation projects could include preventative measures such as vaccines, as well as pregnancy data, as these may be important health metrics to consider. Despite the limitations, our national validation of the CCFR demonstrated a high degree of accuracy and completeness and further reinforces the importance of population-based registries for advancing clinical care.

5. Conclusion

In summary, our validation of the CCFR identified that key demographic and clinical metrics were highly accurate and complete when compared to the medical chart. The inaccuracies in the date of diagnosis are likely related to archiving or loss of medical records as individuals transition from pediatric to adult centers and may ultimately be of less importance as compared to the general timeframe of diagnosis. Race is a social construct with limited correlation to ancestry, and inaccuracies in its collection may be representative of a complex problem that requires thoughtful, multifaceted approaches to improve self-reported collection relating to ethnicity. The study reinforces the role of the CCFR as an important resource for research, clinical care, and quality improvement for pwCF living in Canada.

Disclosure

An abstract version of this manuscript was presented at the North Aiguremerican Cystic Fibrosis Conference in 2024.

Conflicts of Interest

Ranjani Somayaji receives funding from the Cystic Fibrosis Foundation and the Canadian Institutes of Health Research. Anne L. Stephenson receives a stipend from Cystic Fibrosis Canada for her consultant work as Medical Director of the Canadian CF Registry and received grant funding from Cystic Fibrosis Canada and the Cystic Fibrosis Foundation. Christopher H. Goss receives grant funding from the National Institutes of Health and the Cystic Fibrosis Foundation. The remaining authors declare no conflicts of interest.

Funding

This study was funded by the Cystic Fibrosis Foundation (SOMAYA21-002752A121).

Supporting Information

An Appendix with the data validation program database template that was used for the validation by sites is included.