Demographic and co-morbidity characteristics of patients tested for SARS-CoV-2 from March 2020 to January 2022 in a national clinical research network: results from PCORnet^®

PCORnet^® Common Data Model

PCORnet^® operates as a distributed research network, where data are held locally behind institutional firewalls, with queries sent to institutions from a central PCORnet^® Coordinating Center. The participating sites utilize the PCORnet^® Common Data Model, in which each data contributing site maps its clinical and/or administrative claims data into a standardized representation (9). These data are normally updated quarterly by sites, coinciding with data quality curation cycles in which data are examined for conformance with the Common Data Model specifications as well as data plausibility and data completeness.

Prompted by the COVID-19 pandemic, PCORnet^® sites agreed to rapidly implement supplemental data extraction/curation procedures to allow for rapid capture of information on patients with COVID-19. Starting in early April 2020, sites began updating COVID-related data at least monthly (as often as daily) instead of quarterly. Rather than doing this for all patients (which requires considerable processing time and effort), PCORnet^® sites included a subset of all patients, including only those who had at least one International Classification of Diseases, 10^th revision, Clinical Modification (ICD-10-CM) diagnostic code for a potential viral illness, including COVID-19, or had a lab result for a SARS-CoV-2 test recorded beginning January 1, 2020 (Supplemental Figure 1) (10). Through January 31, 2022, this database included 16,317,253 individual patients 20+ years of age.

Queries and Patient Characterization

From this larger population of patients, we used existing modular programs to characterize all patients tested for SARS-CoV-2 (Supplemental Methods Note). These programs return only aggregate data from sites. Submission, and use of these data are covered by a PCORnet-wide Data Sharing Agreement and individual site Institutional Review Board approvals for their PCORnet Common Data Model data. Using designated Logical Observation Identifiers Names and Codes (LOINC) (11), we captured information on patients with results of “positive,” “presumptive positive,” or “detected” on nuclear acid amplification (NAAT)/polymerase chain reaction (PCR) or antigen viral tests for SARS-CoV-2 (referred to as “testing positive” or “positive viral test” hereafter) as well as those with results of “negative (“testing negative” or “negative viral test”). Patients included in the negative viral test strata had only records of negative tests and no positive tests. We stratified patients based on the setting of their viral test: 1) all, 2) ambulatory, 3) emergency department, and 4) inpatient. The latter three categories were mutually exclusive. We assigned a care setting by determining the setting for encounters before and after the test was performed (1 day before and 16 days following). To make care settings exclusive, we removed patients from a lower care setting stratum if they had encounters in a higher care setting (for ambulatory strata, we assessed emergency and inpatient encounters) before and after their positive test (21 days before and 16 days following the date of the test) (Supplemental Note). Patients identified as receiving care in the emergency department included patients who were discharged directly from this setting and not admitted to the hospital within 16 days; those admitted directly from the emergency department or soon thereafter were characterized with the inpatient care setting cohort. One-third of patients did not have a care setting associated with their positive test, perhaps because many patients were tested in drive-thru testing settings that might not have registered an encounter; some of these tests might also have been ordered during telehealth or virtual visits that were not specifically assessed in our queries. We captured all information with respect to the date for an index viral laboratory test, which was the first positive test for patients testing positive and the first negative test for patients testing negative. All codes for case definitions, comorbidities, treatments, and labs were posted on GitHub.

Demographics, Comorbidities, and Disease Severity

Using information available at the time of the test, we assessed demographic characteristics of patients (i.e., age group, sex, race, and ethnicity) using value sets defined in the PCORnet Common Data Model. Sex categories allowed for Male, Female, and Other. Several race categories (i.e., Multiple race, Other race, Native Hawaiian/Other Pacific Islander, and Alaskan Native/American Indian) were less common and typically <1% for all strata; these were grouped together under “Other Race.” Ethnicity was defined as Hispanic or not Hispanic, with a separate missing category. We assessed monthly trends by age group, race, and ethnicity by care setting. We assessed disparities by calculating monthly patient positivity as the total number of patients testing positive separately for each racial, ethnic and age group, divided by the total number of patients tested in that month by care setting. We calculated relative risks (RRs) by month as the percentage of patients testing positive by racial, ethnic, and age group divided by the percentage in a reference group (White; non-Hispanic; age 20-39 years, respectively for each assessment of patient characteristics); these calculations were done separately for race, ethnicity, and age. We noted that disparities were present when RRs were ≥1.5.

We included a detailed assessment of comorbidities, which were selected based on information presented in publications regarding the association of comorbidities with COVID-19 outcomes (12–15). Patients were considered to have a comorbidity if they had at least two ICD-9-CM or ICD-10-CM codes for the disease/condition recorded any time over a three-year period prior to the index date of their viral test. We also captured use of critical care procedure/billing Current Procedural Terminology (CPT) codes (99291, 99292) during this period and mechanical ventilation usage through presence of ICD-10-CM, Healthcare Common Procedure Coding System (HCPCS) and CPT codes any time from the index date of the SARS-CoV-2 test through 21 days after the index date. Lastly, we categorized weight status using the most recent height and weight available within the year prior to the index date. This study was reviewed by the Institutional Review Board of Harvard Pilgrim Health Care, Inc., which deemed it exempt from review because it was a public health surveillance project, conducted under the direction of a public health authority. This activity was reviewed by the CDC and was conducted consistent with applicable federal law and CDC policy.

Characteristics of patients testing positive for SARS-CoV-2

The query captured aggregate data on 1,325,563 patients who tested positive for SARS-CoV-2 and 6,705,868 patients who tested negative from March 1, 2020 through January 31, 2022. Across the 43 sites, percent contributions to the overall sample ranged from <1% to 15% with 15 sites each contributing ≥ 3% of the total; these 15 sites included 73% of all patients testing positive for SARS-CoV-2. Of patients testing positive, 38% were treated in the ambulatory setting, 15% were seen and discharged from the emergency department, and 13% were hospitalized (Table 1). Among hospitalized patients, 10% and 19% were documented as receiving mechanical ventilation and critical care treatment, respectively. Test positivity among patients (not tests) was 17% overall and 15%, 20%, and 9% in the ambulatory, emergency department, and inpatient setting, respectively.

Black or African American patients were 16% of all patients testing positive but were 12%, 28%, and 23% of patients testing positive in the ambulatory, emergency department, and inpatient settings, respectively (Table 1). Other Race patients were 10% of all patients testing positive but were 8%, 16%, and 15% in the ambulatory, emergency department, and inpatient settings. The RR for testing positive for Black or African American patients vs. White patients was 1.0, 1.2, and 1.5 across the care settings, and RR for Other Race patients was 1.1, 1.3, and 1.8 (Table 1). Compared to non-Hispanic patients, the RR for testing positive in the ambulatory, emergency department, and inpatient settings for Hispanic patients was 1.5, 1.5, and 1.8 (Table 1).

Overall, more patients testing positive were in the 20 to 39 years age group (41%) than other age groups, primarily driven by patients testing positive in the ambulatory setting (Table 1). Patients 20-39 years of age comprised the largest proportion of patients testing positive in the ambulatory and emergency department settings, but 55-74 year old patients comprised the largest proportion in the inpatient setting (Table 1). Compared to patients who were 20-39 years, the RR for testing positive in the inpatient setting was 1.7 for patients 40 to <55 years, 1.8 for patients 55 to <65 years, 1.7 for patients 65 to <75 years, 1.9 for patients 75-<85 years, and 2.2 for patients 85+ years. Women comprised a larger proportion of patients testing positive overall (56%), with men more likely to test positive only in the inpatient setting (RR 1.4).

Trends by race, ethnicity, and age

Disparities by race were most apparent during the early phase of the pandemic, especially in the inpatient setting. Among patients hospitalized with SARS-CoV-2 in March and April 2020, Black or African American, Other race, and White patients comprised nearly an equal proportion of patients (Supplemental Figure 2). Black or African American patients still comprised more than 20% of patients hospitalized for 17 of the 23 months examined. When we calculated relative risks by month, we demonstrated that hospitalized Black or African American patients were twice as likely to test positive for SARS-CoV-2 compared to White hospitalized patients from April through August 2020 and 1.5 times as likely to tested positive for 12 of the 23 months, with the lowest RRs from September 2020 to January 2021 and again from June to November 2021 (Figure 1). Patients with “Other Race” had similar patterns when compared to patients of White race, with relative risks of 1.5 or greater for 12 of the 23 months. Patients with Missing or Asian race followed a similar pattern to Black and Other Race groups early in the pandemic and from December 2020 through April 2021. These disparities were similar in the emergency department, with lower relative risks than the inpatient setting (Supplemental Figure 3). Disparities also were noted in the ambulatory setting but only in the early months of the pandemic (Figure 1).

• Download figure
• Open in new tab

• Download figure
• Open in new tab

Figure 1: Monthly relative risks for testing positive for SARS-CoV-2 by race and ethnicity and care setting, March 1, 2020 to January 31, 2022.

These figures show the relative risk for testing positive by month for Black or African American, Asian, Other Race or Missing Race compared to White race as a reference in the (A) ambulatory or (B) inpatient setting and those who were Hispanic, or had Missing Hispanic ethnicity compared to Not Hispanic as a reference in the (C) ambulatory or (D) inpatient setting. Care settings are mutually exclusive. Proportions of patients testing positive by care setting are available in Supplemental Figures 2 and 4; relative risks for testing positive in the emergency department are available in Supplemental Figures 3 and 5. Abbreviations: Hisp-Hispanic; Not Hisp-Not Hispanic; Miss-Missing.

Trends were similar when examined by ethnicity, with a disproportionate number of patients with Hispanic ethnicity testing positive across all care settings; these disparities persisted through mid-2021 in the emergency department in inpatient settings (Figure 1; Supplemental Figures 4-5). Compared to non-Hispanic patients, Hispanic patients had relative risks for testing positive in the inpatient setting of 1.5 or greater for 16 months.

Trends in infections by age demonstrated consistently higher test positivity for 20-39-year-olds in the ambulatory setting (Figure 2, Supplemental Figures 6-7). In the emergency department, younger patients were more likely to test positive than older patients in most months except for March to April 2020 (Supplemental Figure 7). Patients in older age groups (55 years and higher) were more likely to test positive in the inpatient setting than patients who were 20-39 years, with RRs ≥1.5 consistently except in June-July 2020, April-July 2021, and December 2021-January 2022 (Figure 2, Supplemental Figure 6); patients who were 40 to 54 years had RRs ≥1.5 for nearly all months of the pandemic compared to patients 20-39 years, except December 2021-January 2022.

• Download figure
• Open in new tab

Figure 2: Monthly relative risks for testing positive for SARS-CoV-2 by age group and care setting, March 1, 2020 to January 31, 2022.

These figures show the relative risks for testing positive by month and age group in the (A) ambulatory and (B) inpatient setting; reference was 20-39 years of age. Proportions of patients testing positive by care setting are available in Supplemental Figure 6; relative risks for testing positive in the emergency department are available in Supplemental Figure 7. Care settings are mutually exclusive.

Comorbidities

Among patients testing positive for SARS-CoV-2, hypertension was the most common comorbidity (19% for all patients) followed by type 2 diabetes (DM2) (9%), chronic pulmonary disease (7%), and mental health disorders (8%); 7% of patients had hypertension and either Type 1 (T1DM) or T2DM (Figure 3; Supplemental Table 3). These diagnoses were more common among patients with SARS-CoV-2 treated in the inpatient setting, with hypertension present in 38% and DM2 in 22% (19% had hypertension and either DM1 or DM2) of inpatients. The prevalence of obesity also was high at 46% among all patients testing positive but was similar across care settings (45% to 48%). Patients testing negative while inpatient had a lower prevalence of hypertension (29% vs. 38%), diabetes (13% vs. 22%), and obesity (41% vs. 47%) compared with those testing positive (Supplemental Table 4); prevalence was similar for coronary artery disease (10% vs. 13%), mental health disorders (11% vs. 12%), or chronic pulmonary disease (11% vs. 14%).

• Download figure
• Open in new tab

Figure 3: Comorbidities for patients with positive viral tests for SARS-CoV-2 across care settings, March 1, 2020 to January 31, 2022.

These figures show prevalence of various comorbidities for patients with positive viral tests for SARS-CoV-2 across care settings. Each condition includes a broad capture of ICD-9-CM and ICD-10-CM codes, with codes lists available on Github. For example, mental health codes include those for a large number associated with psychotic and mood disorders. Care settings were mutually exclusive. Abbreviations: AV-ambulatory visit; ED-emergency department; IP-inpatient or hospitalized; HTN-hypertension; DM2-diabetes mellitus type 2, DM-diabetes mellitus Types 1 and2, Arrhyth-Arrhythmia; Chron Pulm-Chronic Pulmonary Disease; CAD-coronary artery disease; CKD-chronic kidney disease; MH Dis-mental health disease; CHF-congestive heart failure.

Limitations

Several limitations are inherent to working with structured healthcare data. First, electronic health records and administrative claims data may not have complete data on some patients, especially patients tested and treated entirely outside of a PCORnet health system or outside of the health plan insurance coverage. In such cases, the PCORnet Common Data Model may not have any information about a SARS-CoV-2 infection or inadequate information on chronic diseases and medication use. For example, some institutions had less data on symptomatic ambulatory patients, who may have been referred to outside facilities for testing; other institutions did their own testing for symptomatic ambulatory patients. Most importantly, missing information could have led to misclassification, such as when a patient only had negative tests in the healthcare system but tested positive at an outside institution or on a home test, results that would not be captured in the PCORnet Common Data Model. If missingness was differential by demographic characteristics, proportions reported could be inaccurate.

Second, even for patients who received all care within a PCORnet institution, missing data were still possible, as we found for variables such as race, weight status, and care setting. If missingness was non-random, which could be the case if data were less available on patients of certain race categories, our results could provide some incomplete conclusions. Third, the results we report are descriptive and unadjusted for potential confounders. We discovered higher rates of several chronic conditions among patients with SARS-CoV-2 compared to those without; these differences might be explained by demographic or age differences. We required at least two ICD-10-CM codes for chronic conditions, which may have led to an undercount, especially for those patients who were not seen often in the healthcare system prior to their SARS-CoV-2 test. Fourth, we report differences based on race/ethnicity information available in the EHR. We do not have further information in the Common Data Model, such as socioeconomic status or other information that underlie disparities, such as exposure to racism or other social determinants. We also do not have access to payer status, which could provide some information on disparities by type of insurance, such as governmental or private. Finally, PCORnet does not allow for representative estimates of overall COVID-19 cases across the U.S. Because PCORnet has a fixed number of institutions and is not designed to be representative of the U.S. population, the prevalence of cases across states in PCORnet is not equivalent to the actual state prevalence of patients who have been tested for SARS-CoV-2.