SARS-CoV-2 genomic contextual data harmonization: recommendations from a mixed methods analysis of COVID-19 case report forms across Canada

Abstract

Background: The timely sharing of public health information is critical during a pandemic and is an obstacle that Canada has yet to fully address. During the COVID-19 pandemic, sequencing of the SARS-CoV-2 genome enhanced our understanding of transmission patterns, aided in identifying variants of concern, and supported the development and evaluation of diagnostic tests and vaccines. The Canadian national response faced challenges in aggregating genomic contextual data and carrying out integrated analysis across regions partly due to disparities in COVID-19 case report forms used to capture epidemiological and clinical data that accompanies SARS-CoV-2 sequence data. Such variations delay data integration and make consistent analysis difficult or impossible. The objective of this work was to understand what information was being collected from COVID-19 case report forms used across Canada and identify potential contextual data harmonization issues and solutions.

Methods: Provincial/territorial/national Canadian COVID-19 case report forms were subjected to field-by-field comparisons to identify variations in data categorization, structures, formats, types, granularity, ambiguity, and questions asked. Federal epidemiologists were consulted to substantiate the results.

Results: Data harmonization issues and common data elements were identified. We make recommendations for better national coordination, integrated databases, and data harmonization tools.

Conclusion: This report compares data elements of the various case report forms used across Canada to identify overlaps and differences in the collection method of COVID-19 case information, while also highlighting data harmonization complications and potential solutions. Identifying available data elements will better guide COVID-19 surveillance and research.

Keywords: COVID-19; Canada; Correlation of data; Data collection; Data curation; Metadata; Public health; SARS-CoV-2.

Fig. 1

Differences in how information is collected across case report forms. (a) Excerpts of “Specimen Collection” information from NWT (left) and BC (right). In this example, the different forms use abbreviations and encode specimen information at different levels of granularity. The NWT “Laboratory” section asks for “Specimen Collection Date: YYYY/MMM/DD”, along with checkbox options for “NP swab”, “Throat swab”, “Sputum”, and “Other (e.g. BAL), specify:”. The BC “Laboratory” subsection “Specimen Collected” asks for checkbox entry for “Upper respiratory (e.g., Nasopharyngeal or oropharyngeal swab)” and “Lower respiratory (e.g., sputum, tracheal aspirate, BAL, pleural fluid). (b) Excerpts of “Patient Setting” information from NWT (left) and BC (right). The NWT “Patient Setting” section requests checkbox entry for “Physician office/clinic”, “Home visit”, “ED (not admitted)”, “Facility (LTC, Corrections)”, and then lists checkboxes and YYYY/MMM/DD “Admission date” data for “Inpatient (ward)” and “Inpatient (ICU)”. The BC “Exposures” subsection for exposures that may have occurred 14 days prior to symptom onset request checkbox confirmation for settings of “Acute care facility”, “Long form care facility”, “Group home (community living)”, “Correctional facility”, “School or daycare”, “Workplace not otherwise specified”; along with the role/group relation via checkbox confirmation for “Staff”, “Resident / patient”, “Student”, and “Other, specify”. In this example, it can be observed that different questions are being asked using the same field, e.g., “LTC” and “Long term care facility”. Figure adapted from “Comparison and analysis of Canadian public health SARS-CoV-2 case report forms” [23]

Fig. 2

Examples of common data curation challenges. (a) Input errors. (b) the use of jargon or shorthand that isn’t necessarily known outside the data collectors. (c) Data collected at different granularities, which can cause issues if data systems and curators do not know the hierarchical relationships. (d) semantic ambiguity; the image shows two people using the term “isolation” while one envisions at home (depicted by a house) and the other in a medical facility (depicted by the Caduceus staff). (e) Different date formats. (f) Inconsistent data collection; image shows forms with different fields filled out and/or the same fields filled out differently