What does it mean for a population health study to truly reflect the population?

It’s a question researchers have wrestled with for decades, and one that became more urgent during the COVID-19 pandemic, as rapid-response studies were launched across the country to monitor infection trends and inform public health action. But when it comes to understanding who was included in that data, and who may have been left out, new research is helping fill in the gaps.

A recent study published in BMC Public Health examined the sociodemographic composition of six major SARS-CoV-2 serosurveillance efforts conducted in Canada between April 2020 and November 2023, including one using national CanPath data. Led by Matthew Knight, MSc, and senior author Dr. W. Alton Russell of McGill University, the study reveals how recruitment strategies influenced the demographic makeup of study participants, which can shape the conclusions drawn from them.

“During the COVID-19 pandemic, serological surveillance studies provided key data to help researchers understand trends in SARS-CoV-2 infection and immunity,” said Knight. “However, pandemic restrictions required these studies to use creative recruitment strategies to obtain blood samples for testing, leading to a diversity of recruitment strategies, each with their own strengths and limitations.”

That diversity presented a unique opportunity. “Using data from six Canadian SARS-CoV-2 serosurveillance studies, we wanted to assess how study representativeness differed across various recruitment strategies and study designs. This can help inform future approaches for infectious disease monitoring generally, not just for SARS-CoV-2.” Knight explained.

The team compared the age, sex, urban/rural location, material deprivation, and racialized status of participants across six studies. These included pre-existing longitudinal cohorts like CanPath, convenience samples using residual blood, and a newly recruited probabilistic survey cohort.

Surprising findings, persistent gaps

While no single study achieved full representativeness across all groups, some recruitment methods performed better than others, and some findings defied initial expectations.

“We were surprised to see that some demographic subgroups were equally or better represented in studies using simple convenience samples among blood donors or patients submitting samples for laboratory testing, compared to prospectively sampled health surveys using more complex recruitment strategies,” said Dr. Russell.

Convenience samples, they noted, may avoid one major challenge of many traditional recruitment efforts: non-response bias. But they usually come with the caveat that researchers cannot ask participants for contextual information that can be used to characterize the population and adjust for selection bias.

“Because convenience sampling is relatively cheap compared to other sampling techniques, this suggests that using routinely collected blood donations or samples may be a cost-effective method to obtain representative study populations in resource-constrained situations,” Dr. Russell added. “But these convenience samples have an important limitation. While they are less expensive and sometimes more representative on certain dimensions, they do not have a sampling frame that can be used to generate sampling weights. Because of this … it may be more difficult to quantify and correct for representation bias.”

The study also confirmed concerns about consistent underrepresentation of racialized groups, an equity issue that public health researchers have flagged for decades.

“We were also surprised to observe that some racialized minority subgroups were underrepresented in each study in which they were assessed,” said Knight. “However, we lacked the data to analyze the race and ethnicity of the convenience sample in patients, which may have had better representation … as compared to the blood donors and participants in health studies like CanPath.”

The evolving role of population cohorts

CanPath — the Canadian Partnership for Tomorrow’s Health — was among the six data sources included in the study. As Canada’s largest population health study, CanPath collects a wide range of health, lifestyle, environmental, and biological data from over 330,000 participants across the country.

In the context of pandemic surveillance, cohorts like CanPath offered something particularly valuable: infrastructure and participant populations that were already in place.

“Longitudinal cohorts played a critical role in enabling responsive, real-time research during the pandemic,” said Dr. Jennifer Brooks, Executive Director of CanPath. “At a time when health systems were stretched thin, these studies made it possible to act quickly, and as we saw here, to evaluate how representative those actions were.”

The findings also point to the need for more consistent and timely collection of sociodemographic data, particularly race and ethnicity, to help uncover structural inequities.

“Identifying participation barriers could improve demographic representation and detection of health trends within subgroups,” said Knight. “We hope our work also underscores the need for more consistent measurement of racial and ethnic identity in Canada to improve the assessment of health inequities.”

What’s next: building a smarter surveillance system

Looking ahead, the researchers hope the findings inspire change, both in study design and national infrastructure.

“There are many compelling reasons for Canada to maintain and expand this infrastructure, infostructure, and expertise by developing an integrated, pan-Canadian serosurveillance network,” said Dr. Russell, referencing their recent commentary in the Canadian Journal of Public Health. “Ideally, such a network could quickly obtain samples and data from diverse sources for integrated analyses, including low-cost residual blood… alongside well-characterized cohorts like CanPath.”

They also point to a need for stronger methodological tools and faster data linkages to support representative, equity-informed decision making in future crises.

“Ideally, these methods would take advantage of the rich information available about populations from administrative data,” Dr. Russell added. “Currently, linkage of administrative data is typically possible within a province, but the processes are far too slow to aid in real-time decision making.”

CanPath remains committed to supporting research like this by providing open access to harmonized data and biosamples, enabling linkages to administrative and environmental data, and offering tools to explore population-level patterns in health and disease.

For more information, please contact:
Megan Fleming
Communications & Knowledge Translation Officer
Canadian Partnership for Tomorrow’s Health (CanPath)
info@canpath.ca