Cancer Screening Disparities Before and After the COVID-19 Pandemic

This cross-sectional study analyzes changes in breast, cervical, and colorectal cancer screening before and after the COVID-19 pandemic among adults in Ontario, Canada.


Introduction
7][8][9] In the Canadian province of Ontario, this pause and gradual resumption led to a sizeable backlog in cancer screening.There were 41% fewer screening tests performed overall in Ontario in 2020 compared with 2019, 6,10 and by December 2020, there was a backlog of more than 300 000 screening mammograms. 11The screening backlog was expected to be associated with delays in cancer diagnosis: between March and August 2020, there were, on average, 51% fewer high-grade cytological abnormalities detected through screening each month compared with 2019, 12 and Tinmouth et al 13 estimated that it would take 41 months to recover from the provincial backlog of colonoscopies, the recommended test after abnormal colorectal screening results with a fecal immunochemical test.
5][16][17][18][19][20][21] For example, a 2018 report from the provincial advisor on the quality of health care found that only 55% of women living in extremely income-limited, urban neighborhoods were up to date on cervical screening compared with 66% of women in the wealthiest, urban neighborhoods and that 45.6% of people living in extremely income-limited, urban neighborhoods were overdue for colorectal cancer screening compared with only 32.5% in the wealthiest, urban neighborhoods. 22Similarly, Vahabi et al 23 found that 57% of immigrants were up to date on breast cancer screening compared with 66% of nonimmigrants.
It is unclear how the COVID-19 pandemic has affected preexisting screening disparities.Primary care physicians in Ontario were encouraged to consider prioritizing people who were underscreened or never screened for cancers when screening resumed, 24 but it is also feasible that barriers to screening were heightened in the context of COVID-19 for people experiencing disparities.In this population-based, retrospective, cross-sectional study, we aimed to assess whether changes in screening from before the pandemic to March 2022 varied by income or immigration status.
status under Ontario's health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Setting
Ontario is Canada's most populous province with more than 14 million people, of whom almost 30% are immigrants. 25Ontario has a universal health care system in which all Canadian citizens and permanent residents who live in the province can obtain medically necessary hospital and physician services, including cancer screening, at no cost through the Ontario Health Insurance Plan.There are more than 14 000 family physicians in Ontario, many of whom practice in patient enrollment models, in which patients formally enroll with a particular family physician who provides them with primary care. 26,27There are several types of patient enrollment models, including those in which physician payment is primarily fee for service, but there are also additional payments based on capitation and quality performance including cancer screening (referred to in this study as enhanced fee for service).The models with additional payments include those in which payment is primarily capitation based but with a small fee-for-service component and additional payments based on quality performance, including cancer screening (referred to in this study as capitation based, nonteam).Among primarily capitation-based models, those that include government-funded, interprofessional teams are known as family health teams.More than 77% of Ontario residents are formally enrolled with a family physician practicing in a patient enrollment model. 27,28There are also 101 community health centers in Ontario, which are nonprofit organizations that provide primary care and health-promotion services to patients and communities, with a focus on people experiencing marginalization, serving less than 2% of the population. 29These centers also are structured using interprofessional teams.Physicians working in community health centers are paid by salary. 30Family physicians who do not work in a patient enrollment model or in a community health center work under a traditional fee-for-service pay structure.

Data Sources
We accessed several data sets available at ICES, including the Immigration, Refugees and Citizenship Canada Permanent Residents database, which consists of detailed demographic information on Ontario's immigrants and refugees who arrived in the country beginning in 1985.Data include their country of birth, date of arrival, and immigrant category.The Postal Code Conversion File was used to obtain Rurality Index for Ontario (RIO) classification based on postal code of residence.The Registered Persons Database was used to determine residents' health insurance coverage.We also accessed the Primary Care Population Database, which includes every Ontario resident who is eligible for primary care with active provincial health insurance coverage and a health care contact in the previous 7 to 11 years at a given point in time. 31This data set includes patient demographic characteristics, family physician enrollment status, health care usage information, and cancer screening status.The Ontario Health Insurance Plan claims database and Discharge Abstract Database were used for billing code and diagnostic information and for exclusions.The ICES Physician Database and Corporate Provider Database were used to obtain information on family physicians.The Canadian 2016 Census of Population was used to determine neighborhood income quintile (given as Q1-Q5) based on postal code of residence.These data sets were linked using unique encoded identifiers and analyzed at ICES using SAS, version 8.3 (SAS Institute Inc).

Study Outcome
For each cancer screening type, we assessed whether the screening-eligible population was up to date on screening (a binary outcome) at

Study Variables
We described each study cohort based on age category, sex, neighborhood income quintile, immigrant status (immigrant or not), primary care model, rurality of residence based on postal code using the RIO classification, 33 and health care use over the prior 2 years using resource utilization bands (RUBs) (ranges from 0 to 5, with higher values indicating a very high user of health care services) from the case-mix Johns Hopkins ACG System, version 10.0 (Johns Hopkins Medicine), 34 which uses outpatient billing and inpatient hospital records.For immigrants, we also determined their world region of origin based on country of birth and a previously published modification of the World Bank classification system, as screening status has been shown to vary by region of origin. 35,36

Statistical Analysis
We used counts and frequencies to describe the 3 screening-eligible cohorts based on the aforementioned variables as of March 31, 2019.We assessed up-to-date status for each screening type at 6-month intervals, from April 1, 2012, to April 1, 2022, for the overall cohorts, as well as stratified by income quintile and immigrant status.We also compared breast, cervical, and colorectal screening at our 2 time points for the overall cohorts as well as stratified by our study variables.
For each screening type, we conducted multivariable regression analyses using generalized linear models, with the rate difference between the postpandemic and prepandemic time points as the outcome.To assess the rate difference after and before the pandemic, we aggregated individuallevel data to unique combinations of each category of age group, sex (for colorectal cancer only), income quintile, immigration status, patient enrollment model, RIO categories, and RUBs, all of which were included in the model.The mean number of eligible people from before and after the pandemic in each cluster was used as weight to account for different size of denominators.To avoid small denominators and to prevent unstable rates in the regression models, we combined family health teams and community health centers into 1 category, as both use interprofessional teams.We combined those in RUBs 0 and 1 and 4 and 5, and we excluded people with no identifiable family physician.To reduce the outcome of unstable results associated with small clusters, we excluded clusters having fewer than 50 eligible people either before or after the pandemic in the generalized linear models.This excluded no more than 0.36% (for breast screening) of the 2019 population and no more than 0.46% (for breast screening) of the 2022 population.The threshold for statistical significance was 2-sided P < .05.

Discussion
In this population-based cross-sectional study comparing cancer screening prior to the COVID-19 pandemic vs in the postpandemic period, we found that the proportion of people up to date on screening in Ontario decreased for breast, cervical, and colorectal cancers, with the largest decrease for breast screening (9.4 percentage points) and the smallest decrease for colorectal screening (3.9 percentage points).Preexisting disparities in screening for people living in low-income neighborhoods and for immigrants widened for both breast and colorectal screening.The lowest screening rates both before and after the pandemic were for people who had no identifiable family a Generalized linear models were used for multivariable regression analyses.
b The β estimate represents the percentage change in the rate difference vs the reference group, with negative values representing a larger decrease in screening after the pandemic and positive values representing a smaller decrease in screening after the pandemic.
c Resource utilization bands range from 0 to 5, with higher values indicating a very high user of health care services.In the present study, a nonuser is 0, and the highest user is 5. physician, in which rates were no higher than 17.6% (colorectal screening in 2019) and were as low as 9.6% (breast screening in 2022).We also found that patients of interprofessional models, such as family health teams and community health centers, had significantly smaller reductions in postpandemic screening and higher screening rates at each time point compared with other primary care patients.

JAMA Network Open | Equity, Diversity, and Inclusion
Our findings suggest that access to primary care, as well as the type of primary care model to which one has access, played a crucial role in cancer-screening recovery after the COVID-19 pandemic.These results may make a compelling argument for expanding access to interprofessional, team-based primary care as a method of increasing cancer-screening uptake (and other quality of care 37 ) province-wide.Among interprofessional teams, nonphysician practitioners may play a role in identifying patients who are overdue for screening and who need screening outreach or screening education and in performing and/or ordering screening tests.8][39] These efforts will require significant investment: although community health centers are designed for people experiencing marginalization, they currently serve less than 2% of the province's population, and family health teams have traditionally been least available in geographic areas with the greatest need. 40r results are in line with other literature.2][43] Fedewa et al 44 found that US residents who became unemployed during the pandemic were 10% to 30% less likely to be screened for cancers than employed adults.In their analysis of electronic medical records from over 40 000 primary care patients in Michigan, Gorin et al 45 observed an abrupt decrease in cancer screening between March and June 2020 but with a more modest decrease for fecal immunochemical tests.

Limitations
This study has several limitations.First, we limited our population to Ontario residents who had at least 5 years of health insurance plan coverage on our index dates, but a screening colonoscopy is only required once every 10 years if results are normal; thus, we may have categorized people who had screening colonoscopies in other jurisdictions before moving to Ontario as unscreened.
However, a requirement of 10 years of health plan coverage would have eliminated a large number of people from the study.Second, available data did not allow us to identify those who may have been at increased risk of cancer (eg, due to family history or previous abnormal test results).Provincial screening guidelines are different for individuals at increased risk. 32Third, people who immigrated to Ontario prior to 1985 or lived in another province before moving to Ontario would not have been included in the Immigration, Refugees and Citizenship Canada Permanent Residents database and would have been inadvertently classified as nonimmigrants.However, this would likely bias our results to the null.Fourth, neighborhood income is not an individual-level variable and may have subjected our results to ecological fallacy.This likely would also have biased our results to the null.
Finally, to avoid small denominators and to prevent unstable rates in our regression models, we combined some groups and excluded people with no family physician.This reduced the information that the model provided, but the regrouping was based on the relevance and nature of the data, and we deemed it the most appropriate approach.
of AprilIncome quintile 1 represents the lowest; quintile 5, the highest.

Figure 2 . 40 JAMA
Figure 2. Percentage of Ontario Residents Eligible and Up to Date on Breast, Cervical, and Colorectal Cancer Screening at 6-Month Intervals, From 2012 to 2022, Stratified by Immigrant Status

JAMA Network Open | Equity, Diversity, and Inclusion
2 time points: March 31, 2019 (prior to the COVID-19 pandemic), and March 31, 2022.Being up to date on screening was defined as having had a

Table 2
Between 26.1% and 28.4% of people were enrolled in a family health team, less than 1% with community health centers, and between 3.7% and 4.9% with no identifiable family physician.East Asia and the Pacific, Europe and Central Asia, and South Asia were the most common source regions for immigrant Ontario residents.
5.1] years), 3 918 225 women eligible for cervical screening (mean [SD] age, 45.5 [13.2] years), and 3 886 345 people eligible for colorectal screening (51.4% female and 48.6% male; mean [SD] age, JAMA Network Open | Equity, Diversity, and Inclusion Cancer Screening Disparities Before and After the COVID-19 Pandemic JAMA Network Open.2023;6(11):e2343796. doi:10.1001/jamanetworkopen.2023.43796(Reprinted) November 20, 2023 4/15 Downloaded from jamanetwork.comby guest on 11/27/2023 61.8 [6.4] years) in Ontario (Table 1).).For breast screening, women in Q5 had a smaller decrease (8.6 percentage points) than women in other income quintiles (9.2 percentage points to 10.0 percentage points), as did women enrolled in family health teams (8.1 percentage points) vs those in other primary care model types (8.6 percentage points to 10.5 percentage points), excluding those with no family physician.Both groups also had the highest screening uptake prior to the pandemic.For cervical screening, women enrolled in family health teams had the smallest reduction (6.4 percentage points) between the 2 periods, excluding those with no family physician, although women using community health centers had the highest screening uptake among primary care model types in both time periods (69.7% and 62.7%).The highest users of the health care system only had a 2.5 percentage point decrease in cervical screening.For colorectal screening, among primary care model types, both the smallest decrease and highest screening rate after the pandemic were seen in those enrolled in family health teams (2.3 percentage points [70.1% to 67.7%]).Those with the highest health care use had a similar pattern (1.9 percentage point decrease [74.2% to 72.4%]).For all 3 screening types, people with no family physician had markedly lower screening rates and a negligible difference between the 2 time periods, moving from 11.3% in 2019 to 9.6% up to date in 2022 for breast cancer, 10.8% to 10.2% for cervical cancer, and 17.6% to 16.8% for colorectal cancer.Furthermore, for all 3 screening types, immigrants from East Asia and the Pacific had the largest pandemic screening reductions but notably did not have the lowest postpandemic screening rates.Multiple regression analysis results for those with a family physician are shown in

Table 1 .
Characteristics of Ontario Residents Eligible for Breast, Cervical, and Colorectal Cancer Screening on March 31, 2019 with larger reductions in screening after the pandemic.Across all screening types, interprofessional teams (family health teams and community health centers; β estimate for breast screening: 2.14 [95% CI, 1.79 to 2.49]; for cervical screening, 1.72 [95% CI, 1.46 to 1.98], and for Figure 1.Percentage of Ontario Residents Eligible and Up to Date on Breast, Cervical, and Colorectal Cancer Screening at 6-Month Intervals, From 2012 to 2022, Stratified by Neighborhood Income Quintile a Resource utilization bands range from 0 to 5, with higher values indicating a very high user of health care services.In this study, a nonuser is 0, and the highest user is 5. associated

Table 2 .
Individuals Up to Date on Cancer Screening on the Study Dates for Overall Study Cohorts Stratified by Study Variables

Table 2 .
Individuals Up to Date on Cancer Screening on the Study Dates for Overall Study Cohorts Stratified by Study Variables (continued)Resource utilization bands range from 0 to 5, with higher values indicating a very high user of health care services.In this study, a nonuser was 0, and the highest user was 5.colorectal screening, 2.15 [95% CI, 1.95 to 2.36]), followed by nonteam capitation payment models (β estimate for breast screening: 1.34 [95% CI, 1.02 to 1.65]; for cervical screening, 1.40 [95% CI, 1.17 to 1.64], and for colorectal screening, 1.86 [95% CI, 1.67 to 2.05]), were significantly associated with the smallest pandemic reductions in screening after the pandemic. a

Table 3 .
Multivariable Regression Analyses for Cancer Screening Type With the Rate Difference Between the Study Dates as the Outcome, Adjusting for Variables a Abbreviation: NA, not applicable.