key: cord-0968290-irieuw51
authors: McNaughton, C. D.; Augstin, P. C.; Sivaswamy, A.; Fang, J.; Abdel-Qadir, H.; Daneman, N.; Udell, J. A.; Wodchis, W.; Mostarac, I.; Atzema, C. L.
title: Post-acute health care burden after SARS-CoV-2 infection: A retrospective cohort study among 530,892 adults
date: 2022-05-07
journal: nan
DOI: 10.1101/2022.05.06.22274782
sha: dd5ad006d28def120ba49c7ce54f139e95ec0f12
doc_id: 968290
cord_uid: irieuw51

Importance: The SARS-CoV-2 pandemic portends a significant increase in health care use related to post-acute COVID sequelae, but the magnitude is not known. Objective: To assess the burden of post-acute health care use after a positive versus negative polymerase chain reaction (PCR) test for SARS-CoV-2. Design, Setting, and Participants: Retrospective cohort study of community-dwelling adults January 1, 2020 to March 31, 2021 in Ontario, Canada, using linked population-based health data. Follow-up began 56 days after PCR testing. Exposures: Individuals with a positive SARS-CoV-2 PCR test were matched 1:1 to individuals who tested negative based on hospitalization, test date, public health unit, sex, and a propensity score of socio-demographic and clinical characteristics. Main Outcomes and Measures: The health care utilization rate was the number of outpatient clinical encounters, homecare encounters, emergency department visits, days hospitalized, and days in long-term care per person-year. Mean health care utilization for test-positive versus negative individuals was compared using negative binomial regression, and rates at 95th and 99th percentiles were compared. Outcomes were also stratified by sex. Results: Among 530,232 unique, matched individuals, mean age was 44 years (sd 17), 51% were female, and 0.6% had received [≥]1 COVID-19 vaccine dose. The mean rate of health care utilization was 11% higher in test-positive individuals (RR 1.11, 95% confidence interval [CI] 1.10-1.13). At the 95th percentile, test-positive individuals had 2.1 (95% CI 1.5-2.6) more health care encounters per person-year, and at the 99th percentile 71.9 (95% CI 57.6-83.2) more health care encounters per person-year. At the 95th percentile, test-positive women had 3.8 (95% CI 2.8-4.8) more health care encounters per person-year while there was no difference for men. At the 99th percentile, test-positive women had 76.7 (95% CI 56.3-89.6) more encounters per person-year, compared to 37.6 (95% CI 16.7-64.3) per person-year for men. Conclusions and Relevance: Post-acute health care utilization after a positive SARS-CoV-2 PCR test is significantly higher compared to matched test-negative individuals. Given the number of infections worldwide, this translates to a tremendous increase in use of health care resources. Stakeholders can use these findings to prepare for health care demand associated with long COVID.

The public health implications of the coronavirus 2019 (COVID-19) pandemic caused by airborne spread of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) are difficult to overstate. 1 There have been >400 million SARS-CoV-2 infections and 5.9 million deaths reported worldwide, 2 likely gross underestimates as only 3-35% of infections are detected. 3 In addition to acute illness, there is accumulating evidence that SARS-CoV-2 can cause long-term morbidity. 4-9 Of those discharged from the hospital after up to 27% die or are re-hospitalized within 60 days, and as many as 70% of nonhospitalized patients report at least one symptom four months after initial infection. 10, 11 Disease severity assessed by mortality and acute hospitalizations alone underestimates the burden of disease caused by SARS-CoV-2 infection. 12, 13 Health care funders, policy makers, and clinicians need a clear understanding of post-acute health care use following SARS-CoV-2 infection, i.e., long COVID, in order to equitably allocate resources now and plan for future needs. 14 Therefore, we sought to quantify, describe, and compare the post-acute burden of health care use associated with SARS-CoV-2 to negative controls among community-dwelling adults for the entire province of Ontario, Canada.

This retrospective cohort study was performed at ICES, previously the Institute for Clinical Evaluative Sciences, which is an independent, non-profit research institute funded by the Ontario Ministry of Health and Long-Term Care and has legal status under Ontario's Personal Health Information Protection Act (section 45) allowing collection and analysis of health care and demographic data, without consent, for health system evaluation and improvement. 15 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint

We constructed a retrospective cohort of all adults who underwent PCR testing for SARS-CoV-2 between January 1, 2020 and March 31, 2021 in Ontario, Canada. A full list of datasets is in Table E1 . All testing was performed within the health care system of Ontario, which is administered by the Ontario Health Insurance Plan (OHIP) and provides publicly funded physician and hospital services for the 14.8 million residents of Ontario.

The date of the first outpatient PCR test that detected SARS-CoV-2 was used as the index date for exposed individuals. For individuals with multiple negative PCR tests, the index date was last test date. Adults (≥18 years of age) who were alive eight weeks (56 days) after their index date were included. Individuals residing in long-term care facilities on their index date were excluded, as were those without a valid date of birth, sex, or death information.

Individuals were categorized according to results of SARS-CoV-2 PCR testing as test-negative or test-positive. Pending or indeterminate test results (<0.02%) were excluded.

Individuals with a positive SARS-CoV-2 PCR test result were matched to those with only negative test results by sex, hospitalization within two weeks of the index date, test date, public health unit, and a propensity score computed from myriad factors including health care utilization in the previous year, age, baseline socio-demographics and clinical characteristics, neighborhood level socioeconomic indices, and vaccination status (Table E2) . 16 Subjects were matched on the logit of the propensity score using a calliper width equal to 0.05 times the standard deviation of the propensity score. 17 A 7 standardized difference <0.1 in baseline characteristics was considered a good match. 16, 18 Outcome definitions Follow-up to assess post-acute health care utilization began ≥ 8 weeks (≥56 days) after the index SARS-CoV-2 PCR test date. As the definition of post-acute COVID-19 syndrome, or long COVID, continues to evolve, this timeframe was chosen based on the duration of typical SARS-CoV-2 infectivity and acute symptoms. 1, [19] [20] [21] The primary outcome was health care use rate, a composite measure of health care utilization per person-year of follow-up time. Healthcare utilization was the sum of the counts of outpatient encounters (in-person, phone, and virtual), emergency department visits not resulting in hospitalization, days hospitalized, homecare visits (e.g., wound care), and days in long-term care. Each component of health care utilization was examined separately in pre-planned secondary analyses. Follow-up ended on September 30, 2021 or death, whichever occurred first. Sensitivity analyses required hospital discharge to start follow-up, censored at entry into long-term care, censored at six months, and matched by intensive care unit admission.

All analyses were conducted in SAS version 9.04 at the patient level. Baseline characteristics were reported as means with standard deviations (SD), medians and interquartile ranges (IQR), or frequencies, as appropriate. Given the skewed distribution of health care utilization, this was reported as means and SDs, as well as median, 95 th percentile, and 99 th percentile. Outcomes were also stratified by sex (male/female). 22, 23 The mean rates of composite and component health care utilization were compared in the matched sample between test-positive and test-negative individuals using univariate negative binomial regression modelling, accounting for matched pairs. 24

The 95 th and 99 th percentiles of the person-specific rate of health care utilization for . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 

Between January 1, 2020 and March 31, 2021 there were >11 million SARS-CoV-2 PCR tests completed in 3,777,451 unique adults ( Figure E1 ). Of the 3,631,040 individuals who met all inclusion criteria 268,521 (7.4%) had a positive SARS-CoV-2 PCR test. One-to-one matching was successful for 99% of these, thus the matched cohort consists of 530,232 individuals. Demographics, clinical characteristics, and standardized differences between positive and negative individuals for the matched and unmatched cohorts are reported in Table 1 and Table E3 , respectively. Compared to the unmatched cohort, the matched cohort was slightly younger, had fewer women and lower income individuals, was more rural and more ethnically diverse, and a greater proportion underwent PCR testing during late 2020 or early 2021.

In the matched cohort, mean age was 44 (sd 17) years, 51% were female, and 0.6% had received one or more COVID-19 vaccine doses. Within two weeks after PCR testing, 5.3% were hospitalized in both test-positive and test-negative groups, and preexisting conditions such as hypertension, diabetes, asthma, prior pneumonia, and prior venous thromboembolism were common. The median number of outpatient clinic encounters prior to the index date was 4 (IQR 1-8) per person-year, with a median of 0 (IQR 0-0) per person-year of homecare visits, emergency department visits, and days hospitalized (Table E4 ). There was no difference in mortality between groups, including when stratified by sex; mortality was 0.5% for the first six months of follow-up.

Person-specific health care utilization rates and follow-up, overall and stratified by sex, are shown in Table 2 . Median follow-up was 221 days (IQR 187-267) for PCR . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Figure E2 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint Among women (Table 2 and Panel B in Figure 1 ), on average test-positive women had an absolute increase of 2.1 (95% CI 1.8-2.5) additional encounters per person-year compared to test-negative women, or a RR of 1.15 (95% CI 1.13-1.18; Table 3 ). At the 95 th percentile of health care utilization ( Figure E2 In contrast, among men (Table 2 and Panel C in Figure 1 ), mean health care utilization increased by 0.6 (9% CI 0.3-0.9) encounters per person-year for test-positive versus test-negative men, or a RR of 1.06 (95% CI 1.03-1.08; Table 3 

Overall, findings were robust in sensitivity analyses. Requiring hospital discharge to initiate follow-up (Table E5 , Panel A) decreased days hospitalized among individuals . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. at the 99 th percentile from 7.0 to 2.2 additional days per person-year. Censoring on entrance to long-term care produced smaller magnitude additional total health care utilization rates and days hospitalized at the mean and 95 th percentile, and 10.5 additional homecare encounters per person-year at the 99 th percentile health care utilization. Censoring follow-up at six months (Table E5, person-year, respectively in main analyses). Results at the 95 th percentile were similar to main findings, while the 99 th percentile was notable only for fewer homecare encounters.

Finally, matching by intensive care unit admission within two weeks after the index date reduced sample size due to fewer successful matches and smaller magnitude additional health care use compared to main analyses.

Our study is unique in its use of population-wide PCR results, outcomes, and socio-demographic data among nearly all adults in the highly diverse Ontario population, where PCR testing was publicly funded, as well as relatively long follow-up time. We found that among 530,232 community-dwelling adults who underwent PCR testing for SARS-CoV-2 between January 2020 and March 2021, individuals with a positive test had 11% higher mean post-acute health care utilization, after accounting for a multitude of factors including acute hospitalization as a measure of acute disease severity, sociodemographic factors, comorbidities, and pandemic wave. This increase was largely driven by a subset of individuals who experienced large increases in health care utilization: test-positive individuals in the 95 th percentile of health care use had ~2 additional encounters per person-year compared to matched controls, while those in the 99 th percentile had ~72 additional health care encounters per person-year. Together with . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; sensitivity analyses, these findings indicate that eight weeks or more after a positive SARS-CoV-2 PCR test, a subset of individuals are not able to live alone at home without support and they utilize substantial health care resources.

These findings portend a tremendous increase in need for health care related to long COVID. 27 Conservative estimates indicate that 4 million people in Ontario, or 25% of the population, have been recently infected with SARS-CoV-2. 28 Overall, our findings suggest these will lead to 5.6 million additional health care encounters per year, including 2.9 million additional encounters including 280,000 additional days hospitalized per year for individuals in the highest 1% of health care use. In the United States, where an estimated 140 million people have been recently infected with SARS-CoV-2, 29 this translates to an additional 196 million additional health care encounters per year, including 10 million health care encounters per year for individuals in the highest 1% of health care use. Increases of this magnitude will require significant restructuring, innovation, and investment of resources, particularly in the context of existing prolonged wait times to access care, insufficient supply of acute and long-term care beds, and projected loss of health care workers. [30] [31] [32] [33] [34] [35] [36] [37] Patterns of health care use differed by sex, adding to known differences by sex related to COVID-19. 38-45 Among women with a positive SARS-CoV-2 PCR test, increase in total health care utilization was larger than for men (at the 95% percentile, 4 additional health care encounters per person-year versus no difference for men; at the 99 th percentile, 76 additional health care encounters per person-year for women, compared to 37 encounters per person-year for men), and women used more of all types of health care. For men, additional health care use occurred at the high end of utilization, where homecare use also decreased and there was no difference in emergency department use. A full understanding of why health care utilization differs by sex will be necessary to anticipate future long COVID health care resources.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ;

Results of the sensitivity analysis requiring hospital discharge to start follow-up differed only by an expected decrease in the magnitude of additional days hospitalized per person-year among at the 99 th percentile of health care use. Although less than 1% of individuals entered long-term care, the relative impact of SARS-CoV-2 infection on mean days per person-year in long-term care was large, with a RR of 2.55 (95% CI 2.28-2.86; Table 3 ). While censoring follow-up at the date of entry to long-term care (Table   E5 ) attenuated the magnitude of additional health care use for test-positive individuals, results were otherwise similar to main analyses with the exception of homecare among those at the 99 th percentile of utilization, which increased by 10.5 encounters per person-

year. Due to barriers to entering to long-term care, which were exacerbated by the pandemic, 31 our findings may underestimate the true need for long-term care associated with long COVID. Finally, findings were robust to also matching on intensive care unit admission within two weeks after the index date, with smaller magnitude differences but similar patterns in composite and component health care use.

Previous work indicates that 10-40% of individuals report symptoms months after acute COVID-19, including those who were not hospitalized initially, although these estimates vary by population and definitions of cases and outcomes. 46,47 A subset of PCR test-positive individuals in Denmark who were contacted found that 53% reported symptoms 6-12 months after infection, 48 and 5% of respondents in a study conducted in South Korea reported they were receiving treatment for symptoms a median of 454 days after their COVID-19 diagnosis. 49 Several large studies conducted in the United States Veterans Health Administration health care system, ~90% of whom were men and more than 70% were White, found increased outpatient clinic visits after hospitalization for COVID-19 or a positive SARS-CoV-2 PCR test, as well as increased risk of myocardial infarction and stroke in the months after infection; these findings may not be generalizable to other populations. 50, 51 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. Using health care encounters as a measure of post-acute COVID-19 health care needs likely underestimates the true burden of long COVID, particularly as health care decreased during initial phases of the pandemic among some populations and in Ontario, in particular. 30,54-56 There is no generally accepted method for weighing severity of different types of health care encounters 57,58 ; our composite measure gives hospitalization and long-term care more weight due to their severity, although findings were robust in sensitivity analyses.

Because inclusion was conditioned on PCR testing, results may not generalize to populations with significant barriers to testing. Indication for testing and occupation (e.g., health care worker) were not available, so we are unable to determine whether these factors may modify associations between SARS-CoV-2 infection and type of post-acute health care use. To address potential changes in testing indications and capacity over time, test date was included in the propensity score and used for hard matching.

Matching included multiple socio-demographic and clinical factors, although occupation, body mass index, or symptoms and their duration were not available. However, during the study period, testing was widely available for both symptomatic and asymptomatic individuals at no cost, reducing the risk of selection bias.

Results may not generalize to other variants of SARS-CoV-2 or populations with high prevalence of vaccination or re-infection, 59,60 although they may be more applicable . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ;

as new variants evade current vaccine coverage, previous vaccine immunity wanes, and public health protections are removed. 61

Post-acute health care utilization among patients with a positive SARS-CoV-2 PCR test is significantly higher compared to matched test-negative individuals, with higher rates of outpatient encounters, days hospitalized, and days in long-term care.

Women had higher rates of health care use than men, particularly homecare. Given the number of infections worldwide, this translates to a tremendous increase in use of health care resources. Stakeholders can use these findings to prepare for long COVID health care demand.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ;

Jüni P, da Costa BR, Maltsev A, Katz GM, Perkhun A, Yan S, Bodmer NS. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ;

Smith MP. Estimating total morbidity burden of COVID . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 7, 2022. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 7, 2022. ; https://doi.org/10.1101/2022.05.06.22274782 doi: medRxiv preprint

Clinical Case Definition Working Group on Post-COVID-19 Condition. A clinical case definition of post-COVID-19 condition by a Delphi consensus

Readmission and Death After Initial Hospital Discharge Among Patients With COVID-19 in a Large Multihospital System

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Symptom Duration and Risk Factors for Delayed Return to Usual Health Among Outpatients with COVID-19 in a

Multistate Health Care Systems Network -United States

Veterans and COVID-19: Projecting the economic, social, and mental health needs of America's Veterans

Bob Woodruff Foundation

ICES: Data, Discovery, Better Health

An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies