key: cord-0970382-4o315y8l authors: Singh, Shailendra; Khan, Ahmad; Chowdhry, Monica; Bilal, Mohammad; Kochhar, Gursimran S.; Clarke, Kofi title: Risk of Severe Coronavirus Disease 2019 in Patients With Inflammatory Bowel Disease in the United States: A Multicenter Research Network Study date: 2020-06-06 journal: Gastroenterology DOI: 10.1053/j.gastro.2020.06.003 sha: d2da3ae40332334bc81bab9fcde5234a4cec4450 doc_id: 970382 cord_uid: 4o315y8l nan Crohn's disease (CD) and ulcerative colitis (UC), may be at an increased risk for severe coronavirus disease 2019 owing to their immunosuppressant medications or the chronic inflammatory disease state. 1 Recently, a worldwide registry Surveillance Epidemiology of Coronavirus Under Research Exclusion (SECURE-IBD) consisting of physician-reported patients with IBD with COVID-19 reported the clinical course of COVID-19 among patients with IBD and the factors associated with severe COVID-19. 2 However, there are limited data regarding the comparison of clinical characteristics and outcomes among patients with IBD with COVID-19 and other patients. Moreover, the outcomes of patients with IBD with COVID-19 predominantly in the United States remain unexplored. Our study aimed to evaluate the characteristics and outcomes of patients with IBD with COVID-19 in the United States and compare them to a large cohort of patients without IBD with COVID-19. This was a population-based retrospective cohort study conducted using TriNetX (Cambridge, MA), a federated health research network data set. We performed a real-time search and analysis of electronic health records of more than 40 million patients from multiple health care organizations (HCOs) globally to identify patients with IBD diagnosed with COVID-19 between January 20, 2020, and May 26, 2020, based on a positive laboratory test result or assignment of COVID-19specific ICD code. During the same time period, patients diagnosed with COVID-19 and who had no history of or documentation of a diagnosis of IBD ever were included in the non-IBD control group. The outcome of interest was the risk of severe COVID-19 disease, defined as a composite outcome of hospitalization and/or 30-day mortality postdiagnosis of COVID-19. Outcomes were compared in patients with IBD with COVID-19 and patients without IBD with COVID-19 after 1:1 propensity score matching for demographics and comorbid conditions (listed in Table 1 ) using logistic regression and greedy nearest-neighbor matching algorithm with a caliper of 0.1 pooled standard deviations. Details of data source, quality checks, codes used for patient selection and medications, and statistical analysis have been described previously 3 and are discussed in the Supplementary Materials. Of 196,403 patients with IBD from 31 HCOs, 1901 patients underwent testing for COVID-19, and a total of 232 patients with IBD (CD, 101; UC, 93; indeterminate, 38) were diagnosed with COVID-19. During the same time period, 19,776 patients without IBD were also diagnosed with COVID-19 from the same HCOs. The mean age was similar between the groups, and there were more female patients and more prevalent comorbidities in the IBD group (Table 1) . A higher proportion of patients in the IBD group presented with nausea and vomiting (10.77% vs 4.31%, P < .01), diarrhea (8.19% vs 5.14% , P < .01), and abdominal pain (7.75% vs 2.70%, P < .01) ( Table 1) . In a crude, unadjusted analysis, there was no difference in the risk of severe COVID-19 between the IBD and non-IBD groups (risk ratio [RR], 1.15; 95% confidence interval [CI], 0.92-1.45; P ¼ .23). After propensity score matching, both groups were well balanced, and the risk of severe COVID-19 was similar (RR, 0.93; 95% CI, 0.68-1.27; P ¼ .66) ( Table 1) . Overall, patients with IBD with severe COVID-19 were older and had a higher proportion of multiple comorbidities (Supplementary Table 1) . Medication data were collected up to 1 year preceding the diagnosis of COVID-19 and were available for 166 patients in the IBD group. Sixty-two patients were on immunemediated therapy (biologics, 37 and/or immunomodulators, Table 2 ). The composite outcome of hospitalization or mortality after COVID-19 in patients with IBD is similar to patients without IBD. In addition, patients with IBD with COVID-19 on long-term biologics or nonsteroid immunomodulatory therapies did not have a higher risk of poor COVID-19 outcomes. However, recent corticosteroid use that may as well imply poor disease control may be related to worse outcomes. The risk for severe COVID-19 in patients with IBD is also similar to the widely recognized risk factors for COVID-19 outcomes, such as advanced age and comorbidities, 4 and such patients should be closely monitored. There are concerns that patients with IBD may be at increased risk for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-induced infection and poor outcomes. SARS-CoV-2 has been detected in stool samples of patients with COVID-19, 5 and high concentrations of angiotensin-converting enzyme 2 (ACE2), the binding site for SARS-CoV-2, are found in the terminal ileum and colon 6 and can increase in the inflamed gut of patients with IBD. 7 COPD, chronic obstructive pulmonary disease; SD, standard deviation. a Demographics and comorbidities are compared before and after propensity matching of cohorts. b Numbers rounded off to 10 to protect Protected Health Information (PHI). However, there is no evidence that these factors can influence the course, infectivity, or severity of COVID-19. Another concern in patients with IBD with COVID-19 relates to the use of immune-mediated therapies. Generally, these therapies can be associated with an increased risk of infections. However, these medications are key in inducing and maintaining remission of IBD with subsequent prevention of disease flareup that may require hospitalizations and corticosteroids, which can increase the risk of severe COVID-19. Furthermore, the use of these therapies could be advantageous in suppressing the inflammatory response or cytokine storm described in patients with severe COVID-19. 8 Our study is limited by the inherent limitations of an electronic health records based database. A composite primary outcome of hospitalization or death was chosen because the number of individual events was small to evaluate separate endpoints. Despite limitations, this is the first attempt to compare characteristics and estimate the risk of severe COVID-19 in patients with IBD compared to other patient populations while adjusting for confounding variables. IBD patients in remission and on immunomodulators and biologics should stay on their medications and should exercise social distancing principles like the general population. Patients with IBD with advanced age, multiple comorbidities, or with a poorly controlled disease requiring corticosteroids who develop COVID-19 infection should be aggressively managed, given the increased risk of worse outcomes. Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at https://doi.org/10.1053/ j.gastro.2020.06.003. TriNetX (Cambridge, MA) uses electronic health record data collected from member HCOs. A typical HCO is a large academic health center with data coming from the majority of its affiliates. A single HCO frequently has more than 1 facility, including main and satellite hospitals and outpatient clinics. In the majority of cases, the data originate from the primary electronic health record system. A typical organization has a complex enterprise architecture where the data flow through several different databases, such as a data warehouse and a research data repository, on its way to TriNetX. In addition to electronic health record data, which are usually available in a structured fashion (eg, demographics, diagnoses, procedures, medications, laboratory test results, and vital signs), TriNetX has also the ability to extract facts of interest from the narrative text of clinical documents using natural language processing. TriNetX maps the data to a standard and controlled set of clinical terminologies. The data are then transformed into a proprietary data schema. This transformation process includes an extensive data quality assessment that includes data cleaning, which rejects records that do not meet the TriNetX quality standards. The TriNetX software checks the basic formatting to ensure, for example, that dates are properly represented. It enforces a list of fields that are required (eg, patient identifier) and rejects those records where the required information is missing. Referential integrity checking is done to ensure that data spanning multiple database tables can be successfully joined together. As the data are refreshed, the software monitors changes in volumes of data over time to ensure data validity. TriNetX requires at least 1 nondemographic fact for a patient to be counted in a given data set. Patient records with only demographic information are not included in data sets. Coding system If an HCO provides data in ICD-9-CM, a 9-to-10-CM mapping based on general equivalence mappings (GEM) plus custom algorithms and curation to transform data from ICD-9-CM to ICD-10-CM. All statistical analyses were performed in real time using TriNetX. The TriNetX uses a custom-built platform developed from Java 1.8.0_171, R 3.4.4 (R Core Team, Vienna, Austria) , and Python 3.6.5 with their software language packages to ensure the accuracy and validity of results. The means, standard deviations, and proportions were used to describe and compare patient characteristics. Categorical variables were compared by using the Pearson chi-square test and continuous variables by using an independentsamples t test. Logistic regression on our input covariates was used to obtain propensity scores for each patient in both cohorts. Logistic regression was performed in Python using standard libraries numpy and sklearn. The same analyses were also performed in R software to ensure that the outputs match. After the calculation of propensity scores, matching was performed using a greedy nearest-neighbor matching algorithm with a caliper of 0.1 pooled standard deviations. The order of the rows in the covariate matrix can affect the nearest neighbor matching; therefore, the order of the rows in the matrix was randomized to eliminate this bias. For each outcome, the risk ratio (RR) with a 95% CI was calculated to compare the association of obesity with the outcome. An a priori defined 2-sided alpha of less than .05 was used for statistical significance. TriNetX obfuscates patient counts to safeguard protected health information by rounding patient counts in analyses up to the nearest 10. Risk ratio cannot be estimated because of outcomes of 10 in the nonsteroid group. need-extra-precautions/people-with-medical-conditions. html?CDC_AA_refVal¼https%3A%2F%2Fwww.cdc.gov %2Fcoronavirus%2F2019-ncov%2Fneed-extra-precautions %2Fgroups-at-higher-risk Conceptualization: Lead; Data curation: Lead; Formal analysis: Lead; Investigation: Lead Conceptualization: Lead; Data curation: Lead; Formal analysis: Lead; Investigation: Lead Conceptualization: Supporting; Writing -review & editing: Equal) Conceptualization: Supporting; Writing -review & editing: Equal) Conceptualization: Equal; Methodology: Supporting Supervision: Supporting; Writing -review & editing: Equal) AGAF (Conceptualization: Equal; Investigation: Lead Conflicts of interest This author discloses the following: Kofi Clarke has served as a research grant reviewer/consultant/speakers bureau for Pfizer We acknowledge Charleston Area Medical Center Health System and West Virginia Clinical and Translational Science Institute, which provided us access to and training on the TriNetX global health care network. We also acknowledge the TriNetX (Cambridge, MA) health care network for design assistance to complete this project.