Treatment and Outcomes Among Hospitalized COVID-19 Patients With Immune-mediated Inflammatory Diseases: An Electronic Health Record Database Cohort Study

J Clin Pathways. 2021;7(6):32-43. doi:10.25270/jcp.2021.0708.1
Received April 7, 2021; accepted July 19, 2021.

Coronavirus disease 2019 (COVID-19) is a disease caused by infection of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). It has quickly spread and overwhelmed health care resources in the United States after the first known case was confirmed in Washington state on January 20, 2020. By the end of 2020, the United States surpassed 20 million SARS-CoV-2 infections with >340,000 deaths.¹ Most cases result in mild symptoms, such as fever, cough, fatigue, and shortness of breath; however, some progress to a potentially lethal acute respiratory distress syndrome (ARDS), which is believed to be caused at least in part by a cytokine storm.² A cytokine storm is a hyperinflammatory syndrome that results from excessive release of proinflammatory cytokines, which can be triggered by a variety of infectious and noninfectious diseases.³ Many cytokines appear to be involved (eg, interleukins [IL], interferons [IFN], tumor necrosis factors [TNF]), including IL-1, IL-6, IL-12, IFN-β, IFN-γ, and TNF-α.² In particular, there is evidence that elevated IL-6 levels are associated with adverse clinical outcomes related to ARDS among patients with COVID-19.⁴ There are many targeted immunomodulatory therapies (TIMs) that impact cytokine pathways and have been shown to be safe and effective for immune-mediated inflammatory diseases; therefore, several studies have investigated the potential role of these medications for patients with COVID-19, but a clear consensus has yet to be reached.^5-9

Given the role of hyperinflammation in COVID-19 pathogenesis, there exists a need to better understand patient characteristics, treatment journeys, and clinical outcomes among patients with severe COVID-19 and comorbid immune-mediated inflammatory diseases such as rheumatoid arthritis (RA), psoriatic arthritis (PsA), psoriasis (PsO), and/or inflammatory bowel disease (IBD, including Crohn's disease [CD] and ulcerative colitis [UC]). Recent studies have used various data to evaluate risk factors and outcomes for COVID-19 related outcomes, such as hospitalization, intensive care unit (ICU) admission, intubation, and death among patients with specific conditions that include IBD,^10-15 rheumatic diseases,^16-18 or PsO.^17,19,20 

The use of biologics-based therapies to treat these conditions generally has not been associated with a greater risk of severe outcomes among COVID-19 patients with comorbid IBD,²¹ rheumatic diseases,^22-24 or PsO.^17,19,20 In fact, consistent with their antiinflammatory mechanism of action, some studies have observed a reduced risk of hospitalization or a composite outcome of ICU admission, ventilator use, and death among patients with IBD and rheumatic diseases who use biologics.^25-27 However, additional evidence for patients with comorbid immune-mediated inflammatory diseases is needed to inform choices to initiate or continue treatments in the context of the COVID-19 pandemic. Therefore, the primary objective of this study was to evaluate patient characteristics, treatment, and outcomes during hospitalization for patients who were hospitalized due to COVID-19 and had comorbid RA, PsA, PsO, CD, and/or UC. In addition, as there is evidence of improved outcomes during later phases of the COVID-19 pandemic,²⁸ this study compared outcomes among patients admitted to the hospital before May 1, 2020 with those who were admitted later.

This study was conducted using the Optum De-identified COVID-19 Electronic Health Records (EHR) database, which captures point of care diagnostics specific to COVID-19 patients during initial presentation, acute illness, and convalescence with >500 mapped labs and bedside observations, including COVID specific testing. This is a multi-dimensional database containing information on outpatient visits, diagnostic procedures, medications, laboratory results, hospitalizations, clinical notes, and patient outcomes primarily from integrated delivery networks (IDNs) that include 80 million patients in the United States (≥10 million patients from each Census region). As of October 2020, 2.4 million patients met the inclusion criteria imposed by Optum in selecting the patients for the COVID-19 EHR database. 

All study data were accessed with protocols compliant with US patient confidentiality requirements, including the Health Insurance Portability and Accountability Act of 1996 (HIPAA). Because this study consisted of secondary data analyses of deidentified patient records, the study did not constitute human subjects research and was exempt from Institutional Review Board registration and review requirements under the US Federal Policy for the Protection of Human Subjects or “Common Rule.” 

Study Design and Study Population

This retrospective, observational cohort study analyzed EHR data of patients who had a COVID-19-related hospitalization between January 21, 2020 and October 16, 2020 and had a history of RA, PsO, PsA, CD, or UC (Figure 1). 

Patients who had a hospital admission and fit one of the following three scenarios on or after January 21, 2020 were included in the study: (1) a primary diagnosis of coronavirus-related disease (ie, an International Classification of Diseases, 10th version Clinical Modification [ICD 10 CM] code of U07.1, B97.29, B97.21, B34.2); OR (2) a secondary diagnosis code for coronavirus-related disease (ICD-10: U07.1, B97.29, B97.21, B34.2) within 14 days before or after COVID-19-related symptoms (ICD-10: J12.81, J12.89, J20.8, J22, J40, J80, J98.8, A41.89, R05, R06.02, R50.9); OR (3) a positive COVID-19 specific non-antibody lab test within 14 days before or after COVID-19-related symptoms (J12.81, J12.89, J20.8, J22, J40, J80, J98.8, A41.89, R05, R06.02, R50.9). These criteria were designed to capture COVID-19 patients that were admitted with nonspecific coronavirus-related codes during the early stage of the pandemic,²⁹ per interim guidance from the Centers for Disease Control and Prevention (CDC),³⁰ as well as patients admitted with the COVID-19-specific code, U07.1, created in late March 2020.³¹ Codes intended for screening or potential exposure to viral diseases (eg, Z03.818, Z11.59, Z20.828) were not included as part of our COVID definition because they were to be used only as “rule out” codes, per CDC guidance.³¹ Index date was designated as the admission date of the first COVID-19-related hospitalization (ie, having one of the above diagnoses or lab tests observed within 14 days before or after the admission date of the hospitalization, on or after January 21, 2020). Additional inclusion criteria were aged ≥18 years as of the index date; presence of EHR activity in the database related to labs, diagnoses, procedures, or prescriptions at least 12 months prior to the index date (baseline period); and presence of at least one diagnosis code of RA, PsA, PsO, CD, or UC any time within the 36-month preindex period. Women with pregnancy during the hospitalization, and patients with tuberculosis infection during baseline or the hospitalization (ICD-10-CM: A15) were excluded from the study.

Study cohorts were created for patients with a history of RA, PsO, PsA, CD, or UC diagnosis, respectively. This study was overall descriptive in nature and was not intended to detect treatment effect; thus, power and sample size calculations were not performed. Given a patient might have multiple immunologic conditions within 36 months prior to the index date, study cohorts were not required to be mutually exclusive. Subgroup analyses were performed for patients who were treated with approved TIMs for their corresponding immunologic condition during hospitalization. As the number of patients with PsO, PsA, UC, or CD treated with these medications of interest during the COVID-19 hospitalization was small, sub-groups were classified as (1) on-label treated RA, (2) on-label treated PsA/PsO, or (3) on-label treated CD/UC. 

Outcomes

Outcomes were measured during hospitalization, from the index date to the date of discharge or end of study period, whichever happens first; or otherwise specified. The outcomes included occurrence of ICU admission during hospitalization; hospital and ICU length of stay (LOS); need for mechanical ventilator support and treatment with TIMs (specifically, monoclonal antibodies or recombinant proteins that bind IL-1 receptor, IL-6 receptor, IL-12, IL-17, IL-17 receptor, IL-23, IL-12/23, CD20, CD80/86, or α4β7 integrin and small molecule inhibitors of Janus kinase [JAK] or Phosphodiesterase-4) for RA, PsO, PsA, CD, and UC during hospitalization (irrespective of whether they were treated prior to hospitalization). We also assessed number of days on mechanical ventilation, proportion of patients requiring and number of days on supplemental oxygen use, and extracorporeal membrane oxygenation (ECMO), discharge status, and mortality.

Demographic and clinical characteristics including age, sex, race/ethnicity, geographic region, insurance information, calendar month of the index date in 2020 (before or on/after May 1), Quan-Charlson Comorbidity Index score, general comorbidities, type, and frequency of immunologic treatments during the baseline period were assessed. Consistent with previous studies,^28,32 May 1, 2020 was chosen as a cutoff date since the daily number of US deaths peaked in late April³³ and states with stay-at-home orders had begun lifting restrictions by late May.³⁴ 

Data Analysis

Descriptive analysis was conducted using univariate statistics for all study cohorts (ie, COVID-19 patients with RA, PsO, PsA, CD, or UC) and subgroup of patients (ie, RA, PsA/PsO, or CD/UC patients treated with TIMs). Mean and standard deviation (SD) were reported for continuous variables, and frequencies and proportions were reported for categorical variables. To test statistical difference in outcomes among patients admitted to the hospital before vs on/after May 1, 2020, t-tests and chi-squared tests were performed. A P value of <.05 was considered statistically significant. All data analyses were conducted using SAS Enterprise Guide 7 (SAS Institute, Cary, NC).

Of 1084 patients who met all the study criteria (Table 1); 605 patients were included in the RA cohort, 153 in PsA, 229 in PsO, 125 in CD, and 153 in the UC cohort. Overall, 62.9% of patients were female and the mean age (SD) was 65.7 (14.5) years; 69.9% were White, and 17.6% were Black (Table 1).

Among the study patients, the most frequently occurred comorbid conditions were hypertension (65.8%), hyperlipidemia (52.6%), heart disease (45.3%), diabetes mellitus (35.8%) , obesity (34.1%), and kidney disease (33.2%). Across the five cohorts, 8.8% of patients used TIMs within 1 month prior to the index date; the most frequently observed treatments were TNF-α blockers (3.0%), followed by IL-1 inhibitors (2.5%) and JAK inhibitors (0.8%). Similarly, among the 14.9% of patients who used TIMs within 3 months prior to the index date, the most frequently observed treatments were TNF-α blockers (4.7%), IL-1 inhibitors (5.1%), and JAK inhibitors (1.6%) (Table 2).

Outcomes

The average LOS for the study patients was 8.6 (SD, 8.8) days. We observed that 257 (23.7%) patients were admitted to the ICU with an average ICU stay of 10.8 (SD, 11.1) days, 147 (13.6%) patients required mechanical ventilation with a mean of 7.9 (SD, 8.5) days of use, and 203 (18.7%) required supplemental oxygen with a mean of 4.3 (SD, 5.2) days of use. Use of ECMO was reported for 1 patient (0.1%) (Table 3). There were 494 patients (45.6%) discharged to self-care, 171 (15.8%) were discharged to the care of another health service organization, 108 (10.0%) were discharged to a skilled nursing facility, and 26 (2.4%) were discharged to hospice care. There were 203 (18.7%) deaths during the study period; 26.1% of these patients died after they were discharged to home, another health service organization, a skilled nursing facility, or hospice care (Table 3).

Overall, there were 243 (22.4%) patients who used at least one TIM during hospitalization (RA, 18.3%; PsA, 28.8%; PsO, 26.2%; CD, 32.0%; UC, 26.1%). Of those using TIMs during hospitalization, nearly 70% did not have a record of TIM use in the 3 months prior to hospitalization (Figure 2A). Only 10.2% of patients not using TIMs during hospitalization had used a TIM in the 3 months prior to hospitalization (Figure 2B). The top 3 most frequently observed TIMs were IL-1 inhibitors (14.3%), IL-6 inhibitors (4.6%), and TNF-α blockers (2.6%). This pattern was generally consistent across the five cohorts (Table 3). These medications were used by 55 (5.1%), 2 (0.2%), and 51 (4.7%) patients, respectively, in the 3 months prior to hospitalization. Specifically, among patients with RA (n=605), 10.6% used an IL-1 inhibitor, 4.1% IL-6 inhibitors and 2.8% used TNF-α blockers during hospitalization. Among patients with PsA/PsO, 13.7% of PsA patients and 14.8% of PsO patients used an IL-1 inhibitor, respectively, during hospitalization; 6.5% and 6.1% used IL-6 inhibitors whereas 4.6% and 3.1% used TNF-α blockers. Among patients with IBD, 24% of CD patients and 19.6% of UC patients used an IL-1 inhibitor, respectively, during hospitalization; 4.8% and 2.6% used IL-6 inhibitors whereas 4.0% and 1.3% used TNF-α blockers. 

Patients admitted to a hospital prior to May 1, 2020 had a longer mean hospital LOS (9.6 vs 8.0 days, P=.004), a higher proportion using mechanical ventilation (16.8% vs 11.5%, P=.012), a higher proportion of deaths (25.8% vs 14.2%, P<.001), and a higher proportion using TIMs during the hospitalization (28.7% vs 18.4%, P<.001), in particular, IL 6 (8.1% vs 2.4%, P<0.001) and IL-1 (19.0% vs 11.3%, P=.001), when compared with those admitted to a hospital on/after May 1, 2020. When compared with patients admitted on/after May 1, 2020, those admitted prior to May 1, 2020 were less likely to be admitted to the ICU (18.5% vs 27.0%, P=.001) (Table 4).

Subgroup Analyses

Demographics and clinical characteristics of the 243 patients treated with TIMs during (but necessarily prior to) hospitalization were similar when compared with the overall study population. Overall, 31.3% of patients treated with TIMs during hospitalization were treated with TIMs in the 3 months prior to admission. Overall, 155 (63.8%), 50 (20.6%), and 28 (11.5%) patients used IL-1 inhibitors, IL-6 inhibitors, and TNF-α blockers, respectively, during hospitalization. These medications were used by 28 (11.5%), 1 (0.4%), and 20 (8.2%) patients, respectively, in the 3 months prior to hospitalization (Table 5). The mean hospital LOS for patients who received a TIM during hospitalization was 12.2 (SD, 11.6) days; 24.7% of these patients were admitted to the ICU, with an average ICU stay of 14.3 (SD, 13.2) days. Mechanical ventilation was provided to 31.7% of patients with an average 8.8 days of use, and 21.0% of patients were given supplemental oxygen with a mean 4.0 (SD, 6.7) days of use. Among the same subgroup, the majority were discharged to self-care (35.0%) or the care of another health service organization (17.3%). Among all patients treated with TIMs during hospitalization, 74 (30.5%) died during the study period; 12.2% of these patients died after they were discharged to home, skilled nursing facility, or hospice care (Table 5).

This study used representative data from a large EHR database in the United States to gain insight into the outcomes and treatment journey for hospitalized COVID-19 patients with comorbid immunologic conditions. A moderate number of ICU admissions and use of ventilation and TIM was observed, with improved outcomes noted after May 1, 2020. Our findings add to the growing body of evidence obtained from real-world settings that may inform disease management for various patient populations during the challenging and dynamic heath care crisis posed by the COVID-19 pandemic. 

The mean age and most commonly observed comorbidities in study patients align with common chronic illnesses in the US population as well as previously reported risk factors for hospitalization due to COVID-19 infection, such as chronic kidney disease, diabetes mellitus, obesity, hypertension, and asthma, after adjusting for age, sex, and race/ethnicity.³⁵ Similar to other studies of COVID-19, the female:male patient ratio in our study was about equal for all cohorts, except for patients with RA, among whom 71% were female. 

The rates of ICU admission, mechanical ventilation, and death among our study population were similar when compared with those reported by the general population of hospitalized COVID-19 patients. In a recent chart-review study of 2491 US adults hospitalized with COVID-19, 32% required ICU admission, 19% required invasive mechanical ventilation, and 17% died in hospital.³⁶ Similarly, an analysis of EHRs of 173,000 adults hospitalized with COVID-19 reported an in-hospital mortality of 13.6%.³⁷ A systematic literature review by Rees and colleagues³⁸ reported median hospital LOS for COVID-19 patients ranged between 4 and 21 days (excluding studies from China, where some hospitals were used initially as isolation settings) and ICU stay ranged between 5 and 19 days (including studies from China), and our findings fall within these ranges. Only one patient in our study population received ECMO. Although there is evidence that ECMO can improve outcomes in patients with COVID-19-related ARDS,39 this technique was not employed widely during the COVID-19 pandemic. For example, a US study of health care providers hospitalized with COVID-19 between March and May 2020 found that only 5 of 438 patients received ECMO.³⁹ The lack of ECMO use in our study population is presumably due to scarce ECMO equipment/resources and current guidelines to reserve this approach for patients with more favorable prognoses (aged <65 years, body mass index<40, no cancer).⁴⁰ 

Our study found a 50% increase in the number of patients using TIMs during hospitalization (over a mean hospital stay of 12 days) when compared with the prior 3 months. Among those patients using TIMs during hospitalization, the majority received IL-1 and/or IL-6 inhibitors and a substantially greater proportion of this subgroup used these medications during hospitalization (63.8% and 20.6%, respectively) when compared with the 3 months prior (11.5% and 0.4%, respectively). 

We observed evidence of improved outcomes among cases occurring on or after May 1 when compared with the early part of the pandemic (17% shorter mean hospital stay, 32% fewer patients using ventilators, and 45% fewer deaths). This is consistent with several studies that reported decreases in poor outcomes such as ICU admission, mechanical ventilation, or death among hospitalized COVID patients in the later phases of the pandemic.^28,41 Similarly, Jorge amd colleagues reported lower risks of ICU admission, mechanical ventilation, and death among hospitalized COVID-19 patients with rheumatic and musculoskeletal diseases admitted on or after April 20 when compared with earlier cases.³² 

These trends toward improved outcomes may reflect various factors related to the dynamic nature of the early phase of the COVID-19 pandemic, such as resource constraints, changes in patient behavior (eg, greater urgency in seeking care), or improvements in disease management. Indeed, recent studies have found mortality rates for hospitalized COVID-19 patients were associated with higher admission rates, ICU occupancy, or community prevalence of COVID-19, suggesting that resource strains on hospitals in the early phase of the pandemic were a contributor to poor outcomes.^41-43 In addition, the genome of the SARS-CoV-2 virus has changed rapidly, with variant strains (in particular those containing the D614G mutation) becoming predominant within the first few months of the pandemic.⁴⁴ However, although there is evidence that mutations such as D614G resulted in increased transmission, the impact of these early variants on mortality appears to have been minimal.⁴⁵

Regarding changes in disease management, we observed less frequent use of TIMs, in particular IL-1 and IL-6 inhibitors, and less frequent use of mechanical ventilation among those hospitalized after May 1. We also observed a higher proportion of hospitalized patients being admitted to the ICU in the later phase of the pandemic, which may indicate a change in disease management. These and other changes in treatment practices reported in the literature, such as prone positioning and pharmacologic treatments such as steroids, neuromuscular blockade, anticoagulants, and remdesivir may have contributed to decreased mortality in the later stage of the pandemic.^46,47 A recent study of COVID-19 patients admitted to ICUs at 65 US hospitals reported a wide range in risk-adjusted 28-day mortality rates (6.6%-80.8%) as well as significant variation in treatments, consistent with the important role of clinical management in patient outcomes.⁴⁸ Further studies of clinical pathways using real-world data are warranted to generate additional insights to improve management of COVID-19, especially for immune-compromised patients who may not have fully protective immunity even after immunization or a previous SARS-COV-2 infection.

Limitations

Several limitations should be considered when interpreting the results of this study. Due to the observational nature of this descriptive study, we cannot infer causality between the treatment and the outcomes. Also, the study was unable to account for assessments that might have been made outside of the clinics/hospitals that contribute data to the Optum EHR database. The primary analysis required only one diagnosis code of RA, PsA, PsO, UC, or CD at any time within the 36-month index period to define the patient cohorts, so it is possible that patients assigned a single “rule out” diagnosis code were inappropriately included but based on the published literature this risk appears to be low. Similarly, these five cohorts were not mutually exclusive (ie, there were patients with multiple immunologic conditions who were included in multiple cohorts). 

In addition, to protect patient confidentiality, only the month and year of death is available in the Optum database, so we are not able to determine whether patients died in the hospital or after their discharge from the hospital. Information on discharge status and death is incompletely captured in the Optum COVID-19 EHR database, so we are unable to accurately calculate the location of death or time from hospitalization to death. Finally, although the COVID-19 coding algorithm was developed after fully considering CDC guidelines and consulting with multiple clinicians and coders, it has not been formally validated.

Our study showed that patients with a history of immunologic conditions (RA, PsO, PsA, CD, or UC) who are hospitalized for COVID-19 have a similar patient journey as other hospitalized COVID-19 patients. We also observed an increased use of TIMs during hospitalization in all five cohorts, as compared with the 3 months prior to hospitalization. Lastly, we observed shorter mean hospital stay, decreased use of TIMs or mechanical ventilation, and fewer deaths among patients hospitalized on or after May 1, 2020, possibly indicating rapid changes to the standard of care for hospitalized COVID-19 patients within the first few months of the pandemic. Further studies focusing on admission assessment/treatment journey in patients with immunologic conditions are warranted to inform optimal management of COVID-19 disease in this population.  

Authors: Alex Z Fu, PhD^1,2; Qian Cai, MSPH, MS¹; Emily S Brouwer, MPH, PharmD PhD³; Christopher D Pericone, PhD¹; Kimberly Woodruff, PharmD, PhD¹ 

Affiliations: ¹Janssen Scientific Affairs, LLC, Titusville, NJ
²Georgetown University Medical Center, Washington, DC
³Janssen Research and Development, LLC, Titusville, NJ

Correspondence: Kimberly Woodruff, PharmD, PhD
Phone: (484) 816-7698
Email: kwoodru1@its.jnj.com

Disclosures: Alex Fu, Qian Cai, Christopher Pericone, and Kimberly Woodruff are employees of Janssen Scientific Affairs, LLC (a Johnson & Johnson company) and hold stock in Johnson & Johnson. Emily Brouwer was an employee of Janssen Research and Development, LLC (a Johnson & Johnson company) during study conduct and drafting of this manuscript and holds stock in Johnson & Johnson.

Acknowledgements: This research and preparation of this manuscript was sponsored by Janssen Scientific Affairs, LLC. Mohit Sharma, (Mu Sigma Business Solutions Pvt. Ltd) provided data programming support. Leo J. Philip Tharappel and Vijay Rayasam (SIRO Clinpharm Pvt Ltd.) provided medical writing support and editorial assistance for the manuscript.