Article Text
PDF
Abstract
Objective To identify potential risk factors for adverse long-term outcomes (LTOs) associated with COVID-19, using a large electronic health record (EHR) database.
Design Retrospective cohort study. Patients with COVID-19 were assigned into subcohorts according to most intensive treatment setting experienced. Newly diagnosed conditions were classified as respiratory, cardiovascular or mental health LTOs at >30–≤90 or >90–≤180 days after COVID-19 diagnosis or hospital discharge. Multivariate regression analysis was performed to identify any association of treatment setting (as a proxy for disease severity) with LTO incidence.
Setting Optum deidentified COVID-19 EHR dataset drawn from hospitals and clinics across the USA.
Participants Individuals diagnosed with COVID-19 (N=57 748) from 20 February to 4 July 2020.
Main outcomes Incidence of new clinical conditions after COVID-19 diagnosis or hospital discharge and the association of treatment setting (as a proxy for disease severity) with their risk of occurrence.
Results Patients were assigned into one of six subcohorts: outpatient (n=22 788), emergency room (ER) with same-day COVID-19 diagnosis (n=11 633), ER with COVID-19 diagnosis≤21 days before ER visit (n=2877), hospitalisation without intensive care unit (ICU; n=16 653), ICU without ventilation (n=1837) and ICU with ventilation (n=1960). Respiratory LTOs were more common than cardiovascular or mental health LTOs across subcohorts and LTO incidence was higher in hospitalised versus non-hospitalised subcohorts. Patients with the most severe disease were at increased risk of respiratory (risk ratio (RR) 1.86, 95% CI 1.56 to 2.21), cardiovascular (RR 2.65, 95% CI 1.49 to 4.43) and mental health outcomes (RR 1.52, 95% CI 1.20 to 1.91) up to 6 months after hospital discharge compared with outpatients.
Conclusions Patients with severe COVID-19 had increased risk of new clinical conditions up to 6 months after hospital discharge. The extent that treatment setting (eg, ICU) contributed to these conditions is unknown, but strategies to prevent COVID-19 progression may nonetheless minimise their occurrence.
- COVID-19
- epidemiology
- public health
- respiratory infections
Data availability statement
Data may be obtained from a third party and are not publicly available. Data were licensed from Optum and interested researchers may contact Optum for data access requests. All interested researchers can access the data in the same manner as the authors. The authors had no special access privileges.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
This study used a large electronic health record database containing a rich source of patient-level medical and administrative records from hospitals, emergency departments and outpatient centres across the USA.
Multivariate logistic regression analysis was used to adjust for measured confounders and assess the association of increasing COVID-19 severity (proxied by treatment setting) with the risk of new clinical conditions being diagnosed up to 6 months after COVID-19 diagnosis or hospital discharge.
A sensitivity analysis assessing the association of increasing COVID-19 severity (proxied by treatment setting) with the risk of a new cancer diagnosis served as a negative control.
The main limitation of this retrospective study is that we use treatment setting as a proxy for COVID-19 severity and therefore it is difficult to tease out associations specific to the treatment setting (eg, invasive ventilation) from the underlying COVID-19 severity; any differences that exist between cohorts could bias the results and as all potential confounders may not be controlled for, the results do not indicate causality.
Additional limitations include missing information on smoking status, the lack of a COVID-19-negative control group, the possibility of missing data, being restricted to examining conditions captured by International Classification of Disease-10 codes, the lack of information on COVID-19 treatments received and the lack of laboratory values or other biomarkers to better characterise disease.
Introduction
The COVID-19 pandemic caused by the novel SARS-CoV-2 has imposed an immense burden of morbidity and mortality worldwide.1 Although the majority of patients experience mild or moderate symptoms that resolve within a few weeks of initial infection, increasing evidence suggests that a subset of patients continue to display symptoms beyond 4 weeks after infection.2 3 These symptoms are wide ranging and often extend beyond the typical initial symptoms of COVID-19 to include respiratory (eg, dyspnoea, decreased exercise capacity), cardiovascular (eg, heart palpitations, chest pain) and mental health (eg, confusion, disorientation) disorders.4 5 Notably, such outcomes have been observed even in patients with mild acute COVID-19 symptoms.6 These prolonged symptoms have collectively been referred to by several names including postacute COVID-19, post-COVID-19 syndrome (PCS), postacute sequelae of SARS-CoV-2 infection and possibly more commonly ‘long COVID-19’.7 8 However, due to the overlapping and non-specific range of symptoms experienced, the medical community has not yet converged on precise definitions and it is possible that distinct subsets of patients with long COVID-19 exist. It has also been suggested that long COVID-19 can be further subdivided into subacute COVID-19 (4–12 weeks after initial onset of COVID-19 symptoms) and PCS (beyond 12 weeks).4 9 The underlying pathogenic mechanisms of long COVID-19 are not well understood, but multiple causes have been proposed, including immune dysregulation and viral persistence.10 Additionally, in patients with severe disease requiring treatment in the intensive care unit (ICU), non-specific secondary effects cannot be ruled out, similar to those observed in ‘postintensive care syndrome’.11
High-quality clinical data on respiratory, cardiovascular and neurologic sequelae of SARS-CoV-2 infection are beginning to emerge12–14 and several observational studies and patient registries have been established to better understand the long-term outcomes (LTOs) of COVID-19.15 16 However, little is known about the potential baseline factors that may predict the development of long COVID-19.
Retrospective cohort studies using electronic health records (EHRs) are uniquely positioned due to their size and convenience to provide insights into factors underlying long COVID-19 development and the range of long COVID-19 conditions that exist. The Optum deidentified COVID-19 EHR dataset contains patient-level medical and administrative records from hospitals, emergency departments, outpatient centres and laboratories across the USA. This dataset has previously been used to describe key epidemiological features of a large cohort of hospitalised patients with COVID-1917 and to develop a prognostic model of in-hospital mortality.18
The current study used the Optum deidentified COVID-19 EHR dataset to better understand the types of LTOs encountered by patients with long COVID-19, to define the factors that predict their diagnosis and to understand the role that treatment setting (as a proxy for COVID-19 severity) plays in the manifestation of these outcomes.
Methods
Database
Individuals with COVID-19 diagnosed between 20 February 2020 and 4 July 2020 were extracted from the Optum deidentified COVID-19 EHR dataset (569 149 individuals from 3 832 315 in the entire dataset). This dataset contains patient-level medical and administrative records from hospitals, emergency departments, outpatient centres and laboratories across the USA. All data were deidentified according to the Health Insurance Portability and Accountability Act Expert Method and managed according to Optum customer data use agreements. The COVID-19 EHR dataset comprises clinical information sourced from hospital networks that provided data meeting Optum’s internal data quality criteria. Data cleaning methods used were as described previously.17
Patients and study design
Eligible patients (overall COVID-19 cohort) had ≥1 of the following: a COVID-19 diagnosis code (U07.1, U07.2), a positive diagnostic test for SARS-CoV-2 infection (eg, molecular or antigen test) or a B97.29 diagnosis code (other coronavirus as the cause of diseases classified elsewhere) without a negative SARS-CoV-2 molecular test within 14 days. The index date was defined as the date of COVID-19 diagnosis or COVID-19-related hospitalisation (as defined below), whichever occurred first. The baseline period was defined as the 12 months prior to the index date and a minimum of 180 days follow-up was required for all patients. The overall study design is shown in figure 1.
Overall study design.
Eligible patients were assigned into the following six subcohorts according to treatment setting: (1) Outpatient, patients with a COVID-19 diagnosis and no record of hospitalisation or an emergency room (ER) visit within 21 days of diagnosis; (2) ER on diagnosis, COVID-19 diagnosis on the same day as ER visit; (3) ER, COVID-19 diagnosis prior to ER visit, that is, patients with an ER visit within 21 days after COVID-19 diagnosis (excluding diagnosis date); (4) Hospitalisation without ICU, patients hospitalised with no record of ICU admission; (5) Hospitalised with ICU but no ventilation, patients hospitalised with record of ICU admission but no record of ventilator or extracorporeal membrane oxygenation (ECMO) use during ICU stay and (6) Hospitalised with ICU and ventilation, patients hospitalised with record of ICU admission and ventilator or ECMO use during ICU stay.
Hospitalisation was defined as an inpatient or ER overnight visit with an initial COVID-19 diagnosis made during hospitalisation and within 7 days of admission or an inpatient or ER overnight visit within 21 days of the initial COVID-19 diagnosis, where the hospital had a record of this diagnosis. Contiguous ER and inpatient visits with a gap of up to 1 day were considered a single hospitalisation. If a patient had multiple eligible hospitalisations, only data from the first hospitalisation were considered, as described previously.17
Modeling and statistical analysis
LTOs occurring >30–≤180 days after hospital discharge or COVID-19 diagnosis were categorised into one of the two time windows (>30–≤90 days or >90–≤180 days) and were further classified as respiratory, cardiovascular or mental health conditions (online supplemental table 1).19 LTOs were selected to capture a broad range of potential sequelae, even if there was no strong clinical or pathological rationale for their choice, given the absence of sufficient clinical data regarding established complications associated with COVID-19. Multivariate logistic regression analyses were performed to determine the association of disease severity (proxied by treatment setting) with the three LTO classifications. Covariates were intended to encompass the main known risk factors for developing severe COVID-1920 and included demographic information (ie, age, gender, race, ethnicity, diagnosis month, insurance type, obesity status) and baseline health conditions (ie, those included in the Charlson Comorbidity Index (CCI) (online supplemental table 2). CCI was treated as a numeric variable, while all other variables were treated as categorical. Age was binned into <18, 18–29, 30–39, 40–49, 50–65, 65–74, 75–84 and ≥85 years. Date of diagnosis was also binned into months in 2020 (pre April, April, May, June, July; allowing for ≥180 days follow-up until 31 December 2020 at the latest). Patients were excluded from the regression model examining a specific LTO category if they had a diagnosis in that category in the 12 months prior to the index date (eg, if a patient had an asthma diagnosis 12 months prior to the index date, they would be excluded from the model for respiratory LTOs).
Supplemental material
All statistical analysis was performed using R V.3.6.3.21 Using the sjstats package, regression was performed using the function ‘glm’ and the risk ratio (RR) was calculated by converting the OR using the function ‘OR to RR’.22 Increased risk of diagnosis of a health condition was implied when the RR and both the low and high 95% CI limits were >1 and decreased risk was implied when the RR and low and high 95% CIs were <1.
Sensitivity analysis
A sensitivity analysis was performed to investigate the potential association of disease severity (proxied by treatment setting) with risk of a new cancer diagnosis, to serve as a negative control. The same set of covariates was used as per the main analysis, but cancer diagnosis was the only LTO examined. Currently, no evidence exists to suggest that COVID-19 severity increases the risk of a new cancer diagnosis. Thus, an association here may indicate that the associations from the main analysis may be driven by other differences between patients across treatment settings.
Patient and public involvement
No patient involved.
Results
Patient population
In total, 57 748 patients were eligible for the overall COVID-19 cohort. Table 1 presents descriptive statistics of the patients by subcohort. Mean age tended to be higher in patients in hospitalised subcohorts (53.2–57.7 years) than in those in non-hospitalised subcohorts (41.0–46.8 years). Overall, 53.3% of patients were female. Across all patients, 50.3% were Caucasian, 22.8% were African American, 3.2% were Asian and the remaining 23.6% were missing information on race. Additionally, 67.5% were of non-Hispanic ethnicity, while data on ethnicity were missing for 11.8% of patients. Overall, 19% of patients were obese and the mean weighted CCI Score was 1.20. Information on smoking status was missing for 93.1% of patients (table 1). Full details of demographics and baseline characteristics are provided in online supplemental table 3.
Baseline characteristics of patients with COVID-19, overall and by subcohort
The proportions of patients with incipient respiratory, cardiovascular and/or mental health conditions that were diagnosed either >30–≤90 days or >90–≤180 days after COVID-19 diagnosis or hospital discharge are provided in table 2. The proportions of patients with new LTOs were generally higher in the subcohorts with more severe disease (ie, the ER subcohort and all hospitalised subcohorts) compared with the outpatient subcohort. In addition, the proportion of patients with respiratory LTOs was higher than the proportions with cardiovascular or mental health LTOs. New respiratory LTOs were diagnosed more frequently during the earlier time window across subcohorts, except in the outpatient subcohort where the proportion of patients diagnosed was the same in both time windows (both 8.1%; table 2). No clear temporal trends were noted for diagnosis of cardiovascular or mental health LTOs, with similar proportions of patients with new cardiovascular and mental health LTOs observed in the >30–≤90-day and >90–≤180-day windows for each subcohort (table 2). The proportions of patients with LTOs in more than one category (ie, ‘respiratory and cardiovascular’, ‘respiratory and mental health’, ‘mental health and cardiovascular’ or ‘respiratory, cardiovascular and mental health’) were lower than the proportions of patients with LTOs in a single category, suggesting that a diagnosis in one category did not necessarily lead to a diagnosis in another.
Long-term outcomes that were diagnosed >30–≤90 days or >90–≤180 days post COVID-19 by subcohort
Regarding individual conditions, the prevalence of newly diagnosed pneumonia, dyspnoea and respiratory failure in the >90–≤180-day window closely followed the pattern of initial COVID-19 severity (as proxied by treatment setting), with most cases being diagnosed in the ‘ICU with ventilation’ subcohort (online supplemental table 4). Similarly, although encephalopathy, confusion or disorientation, cardiac arrhythmia and myocardial infarction were less common, the prevalence of these conditions also increased with increasing COVID-19 severity. Full details of conditions that were diagnosed in the >30–≤90-day and>90–≤180-day windows following COVID-19 diagnosis or hospital discharge are provided in online supplemental table 4.
Modeling
The most striking potential covariate associated with increased risk of newly diagnosed respiratory conditions at >30–≤90 days and >90–≤180 days post COVID-19 diagnosis or hospital discharge was increasing severity of illness according to increasing hospitalisation severity, using the outpatient subcohort as the reference group (figure 2 and online supplemental table 5). ICU with ventilation was associated with increased risk of a novel respiratory condition diagnosis compared with the outpatient subcohort at >30–≤90 days (RR 2.64, 95% CI 2.27 to 3.04) and >90–≤180 days post COVID-19 diagnosis or hospital discharge (RR 1.86, 95% CI 1.55 to 2.21); in addition, ICU without ventilation was associated with increased risk during the >30–≤90-day time window (RR 1.69, 95% CI 1.39 to 2.03), while ER was associated with increased risk at both >30–≤90 days (RR 1.39, 95% CI 1.17 to 1.65) and >90–≤180 days (RR 1.33, 95% CI 1.10 to 1.58) post COVID-19 diagnosis or hospital discharge. By contrast, patients with an ER visit on the COVID-19 diagnosis date were less likely than those in the outpatient subcohort to be diagnosed with a new respiratory condition at >30–≤90 days (RR 0.64, 95% CI 0.56 to 0.74) and 90–180 days post COVID-19 diagnosis or hospital discharge (RR 0.56, 95% CI 0.48 to 0.65). Additional covariates associated with increased risk of new respiratory conditions were older patient age and obesity. A COVID-19 diagnosis during or prior to April 2020 exhibited a non-significant trend towards increased risk of new respiratory condition occurrence compared with later diagnosis, which may reflect changes in treatment algorithms over time. Full results are presented in online supplemental table 5.
Relative risk of new respiratory conditions occurring from >30 days to ≤180 days after COVID-19 diagnosis or hospital discharge. Relative risk of new respiratory conditions occurring at (A) >30–≤90 days and (B) >90–≤180 days after COVID-19 diagnosis or hospital discharge. Graphs represent relative risk and 95% CIs. Reference groups are: <18 years (age), female (sex), African American (race), Hispanic (ethnicity), non-obese (obesity), diagnosis in February and March 2020 (diagnosis month), commercial (insurance) and cohort 1: outpatient (cohort). CCI, Charlson Comorbidity Index; ER, emergency room; ICU, intensive care unit.
Increasing hospitalisation severity was also found to be associated with increased risk of a new cardiovascular condition occurring post COVID-19 diagnosis or hospital discharge (figure 3 and online supplemental table 5). Notably, ICU with ventilation was associated with increased risk of the occurrence of novel cardiovascular conditions compared with the outpatient subcohort at >30–≤90 days (RR 3.16, 95% CI 1.83 to 5.18) and >90–≤180 days post COVID-19 diagnosis or hospital discharge (RR 2.65, 95% CI 1.49 to 4.43), while ICU without ventilation was associated with increased risk during the >90–≤180-day time window (RR 2.41, 95% CI 1.25 to 4.23). Similar to the findings regarding respiratory conditions, patients with an ER visit on the COVID-19 diagnosis date were less likely than outpatients to be diagnosed with novel cardiovascular conditions in both the >30–≤90-day (RR 0.45, 95% CI 0.27 to 0.71) and >90–≤180-day windows (RR 0.59, 95% CI 0.38 to 0.89). Additional covariates associated with an increased risk of new cardiovascular conditions occurring included older patient age and non-Hispanic ethnicity. Full results are presented in online supplemental table 5.
Relative risk of new cardiovascular conditions occurring from >30 days to ≤180 days after COVID-19 diagnosis or hospital discharge. Relative risk of new cardiovascular conditions occurring at (A) >30–≤90 days and (B) >90–≤180 days after COVID-19 diagnosis or hospital discharge. Graphs represent relative risk and 95% CIs. Reference groups are: <18 years (age), female (sex), African American (race), Hispanic (ethnicity), non-obese (obesity), diagnosis in February and March 2020 (diagnosis month), commercial (insurance) and cohort 1: outpatient (cohort). Relative risk in the >30–≤90-day time window was not calculated as no new diagnoses were made in the reference group (<18 years) during this time. CCI, Charlson Comorbidity Index; ER, emergency room; ICU, intensive care unit.
The risk of a new mental health condition occurring post COVID-19 diagnosis or hospital discharge also increased according to increasing hospitalisation severity (figure 4 and online supplemental table 5). ICU with ventilation was associated with increased risk of a new mental health condition occurring compared with the outpatient subcohort at >30–≤90 days (RR 1.89, 95% CI 1.51 to 2.35) and >90–≤180 days post COVID-19 diagnosis or hospital discharge (RR 1.52, 95% CI 1.20 to 1.91) and ICU without ventilation was similarly associated with increased risk of a new mental health condition diagnosis during the >90–≤180-day window (RR 1.34, 95% CI 1.02 to 1.73). Of note, compared with those <18 years, all age groups examined appeared to be at higher risk of the occurrence of new mental health conditions at >30–≤90 days post COVID-19 diagnosis or hospital discharge. In the >90–≤180-day window, only the 65–74 and 75–84 years age groups were not at higher risk. Additional covariates associated with increased risk of a new mental health condition occurring included obesity, Caucasian race and non-Hispanic ethnicity. See online supplemental table 5 for full results.
Relative risk of new mental health conditions occurring from >30 days to ≤180 days after COVID-19 diagnosis or hospital discharge. Relative risk of new mental health conditions occurring at (A) >30–≤90 days and (B) >90–≤180 days after COVID-19 diagnosis or hospital discharge. Graphs represent relative risk and 95% CIs. Reference groups are: <18 years (age), female (sex), African American (race), Hispanic (ethnicity), non-obese (obesity), diagnosis in February and March 2020 (diagnosis month), commercial (insurance) and cohort 1: outpatient (cohort). CCI, Charlson Comorbidity Index; ER, emergency room; ICU, intensive care unit.
Sensitivity analysis
With the exception of older age, COVID-19 severity (proxied by treatment setting) did not predict a new cancer diagnosis up to 180 days after COVID-19 diagnosis or hospital discharge (online supplemental figure 1 and supplemental table 6), giving confidence in the results of the original analysis.
Discussion
By using EHRs of over 55 000 patients from hospitals and clinics across the USA, this study set out to examine the types of new LTOs (ie, only those that were identified after COVID-19 diagnosis or hospital discharge) associated with long COVID-19 and to identify potential underlying factors that may contribute to their occurrence. Severe disease was found to predict an increased likelihood of a new LTO diagnosis, whereby increasing hospitalisation severity was associated with increased risk of new respiratory (eg, pneumonia), cardiovascular (eg, myocardial infarction) and mental health conditions (eg, confusion or disorientation). In severely affected patients with COVID-19, some LTOs were diagnosed between three and 6 months after hospital discharge, suggesting that the overall COVID-19 burden extends far beyond the acute infection phase. In addition, although patients with severe disease were most at risk of presenting with new LTOs, non-hospitalised patients also experienced a relatively high incidence of LTOs, suggesting that even patients with mild disease are at risk of adverse long-term effects associated with COVID-19.
Although the data show a clear general trend of increased LTOs that correlated with COVID-19 severity (proxied by treatment setting), the specificity of this effect to COVID-19 is unclear, as ICU survivors commonly develop a range of new conditions on discharge collectively referred to as ‘postintensive care syndrome’, regardless of their underlying diagnosis.11 Nonetheless, preventing the development of more severe disease, where possible, may decrease the likelihood of health problems post infection and would be expected to simultaneously increase the probability of survival. Together, these effects would have a cumulative positive impact on both patients and healthcare systems.
Interestingly, the ‘ER on diagnosis’ subcohort exhibited a reduced incidence of LTOs compared with the outpatient subcohort. The reasons for this are not clear but are likely due in part to the lower mean age and reduced incidence of comorbidities in this subcohort relative to the other subcohorts. In addition, it is possible that in the context of the pandemic, when primary care physicians had more limited personal protective equipment and other resources, these patients were directed to the ER to be tested for COVID-19, despite not having severe enough disease to warrant an ER visit. Finally, depending on the hospital setting and processes in place, asymptomatic patients who attended the ER for non-COVID-19 reasons may have tested positive while there, which may have led to the inclusion of milder COVID-19 cases in this subcohort.
Previous studies have examined the link between COVID-19 severity and LTOs. A study of 2469 hospitalised patients with COVID-19 in Wuhan, China, showed that more severe disease correlated with increased risk of LTOs up to 6 months after infection, including fatigue, sleep difficulties and anxiety or depression.23 Anxiety or depression was observed in 23% of patients in that study compared with ~10% in our study; this difference is likely because our study was limited to newly diagnosed disorders in both inpatients and outpatients, while the previous study included new or worsening symptoms in hospitalised patients only. A separate, large study of patients with COVID-19 that used a US EHR database (N=2 36 379) to examine 6-month outcomes (inpatients and outpatients) reported that ~7% of patients had a first anxiety disorder compared with ~17% that had any anxiety disorder and that increased incidence was correlated with increased disease severity.13 A further study compared 73 435 non-hospitalised patients with COVID-19 who were users of the Veterans Health Administration with 4 990 835 control patients and reported an increased risk of incident sequelae including, but not limited to, respiratory, cardiovascular and mental health disorders after a median follow-up duration of 126 and 130 days, respectively.19 Smaller, single-site hospital studies in the UK have reported similar trends between disease severity and shorter-term outcomes, with breathlessness commonly reported up to 12 weeks post COVID-19.24 25 In addition, self-reported data in patients with COVID-19 (N=4182) showed that upper respiratory complaints (eg, shortness of breath) and cardiac symptoms (eg, palpitations, tachycardia) were commonly reported in patients with long COVID-19 (symptoms lasting ≥28 days)26 and data from a separate study using wearable devices provided further evidence of prolonged tachycardia in symptomatic patients with COVID-19.27 The current study builds on these previous reports and provides additional evidence of a link between COVID-19 severity (proxied by treatment setting) and increased risk of developing LTOs, using a large dataset from both hospitalised and non-hospitalised patients. In addition, our study provides a detailed summary of the incidence of a wide range of specific health conditions that occurred up to 6 months after COVID-19 diagnosis or hospital discharge, providing a useful resource to better understand and characterise the range of conditions that constitute long COVID-19.
Our study categorised three major classes of LTOs that occur in patients with long COVID-19: respiratory, cardiovascular and mental health. This is broadly in keeping with a previous retrospective cohort study in England that followed 48 780 patients hospitalised with COVID-19, who had significantly higher rates of respiratory and cardiovascular disease after a mean follow-up of 140 days.28 In addition, a retrospective study that used a large administrative all-payer database including 27 589 inpatients and 46 857 outpatients demonstrated that post COVID-19, patients were more likely to experience a range of conditions, including respiratory, nervous and circulatory system conditions, than outpatient control patients.29 A greater understanding of the conditions that characterise long COVID-19 is needed to better anticipate the future healthcare burden of COVID-19 and to optimise strategies to minimise long COVID-19 development. In this regard, signals detected in the current study such as lung fibrosis, as well as other factors including paediatric long COVID-19, vaccination effects and healthcare utilisation, are topics that may warrant future analysis. In particular, a greater understanding of the long-term economic consequences of COVID-19 and the impact of long COVID-19 on patient quality of life is needed.
A major limitation of this analysis is that treatment setting is used as a proxy for COVID-19 severity; therefore, it is difficult to tease out the effect of treatment setting procedures (eg, invasive ventilation) from the underlying COVID-19 severity. Furthermore, our analysis did not distinguish short-term outcomes from chronic health conditions. Additional limitations include missing information on smoking status, the restriction of follow-up to only 6 months, the lack of a COVID-19-negative control group, the possibility of missing data (eg, patients may have sought care for an LTO not captured in the Optum deidentified COVID-19 EHR dataset), the lack of information on COVID-19 treatments received and the lack of laboratory values or other biomarkers to better characterise disease. Finally, capture of health conditions relies on International Classification of Disease-10 (ICD-10) codes, whereas some conditions of interest (eg, anosmia, ageusia and brain fog) lack specific ICD-10 codes and other conditions are known to be undercaptured. The B97.29 diagnosis code includes other coronaviruses in addition to SARS-CoV-2 and may therefore be a potential limitation of our study; however, the majority of our COVID-19 cohort (>85%) was diagnosed from April to July using the official U07.1 diagnosis code that is specific to COVID-19, meaning it is unlikely that a substantial number of infections, if any, were from other coronaviruses.
Conclusions
Although LTOs were reported in patients across all subcohorts, increased risk of new respiratory, cardiovascular and mental health conditions was observed with increasing COVID-19 severity, using treatment setting as a proxy. Strikingly, the risk of new conditions being diagnosed remained high up to 6 months post COVID-19 diagnosis or hospital discharge, suggesting that the burden of COVID-19 extends far beyond the acute infection phase. Future research is warranted to understand specific factors that lead to the occurrence of new LTOs in patients with COVID-19 and to distinguish between the relative effect of COVID-19 severity versus any general effects that may occur after acute critical illness.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data were licensed from Optum and interested researchers may contact Optum for data access requests. All interested researchers can access the data in the same manner as the authors. The authors had no special access privileges.
Ethics statements
Patient consent for publication
Ethics approval
The use of the Optum deidentified COVID-19 EHR dataset was reviewed by the New England Institutional Review Board (IRB) and was determined to be exempt from broad IRB approval, as this study did not involve human subject research.
Acknowledgments
The authors thank Shemra Rizzo for valuable contributions. Third-party medical writing assistance, under the direction of the authors, was provided by John Bett, PhD, from Ashfield MedComms, an Ashfield Health company.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors All authors were involved in drafting and revising the manuscript, approved the final version and agree to being accountable for all aspects of the work. NJ contributed to the conception of the research question, study design, analysis and data interpretation. XC contributed to study design, analysis and data interpretation. UB contributed to the conception of the research question, study design, analysis and data interpretation. KZ contributed to the conception of the research question, design of the analysis, selection of outcomes and data interpretation. DC contributed to the conception of the research question, design of the analysis and selection of outcomes. LT contributed to study design and data interpretation. MB contributed to data interpretation. MN contributed to selection and categorisation of key complications for study design. VY contributed to the study design, acquisition, analysis and data interpretation.
Funding This work was supported by F. Hoffmann-La Roche Ltd.
Competing interests NJ and UB are employees of F. Hoffmann-La Roche Ltd. MB is an employee of Roche Nederland BV. UB and MB hold shares in F. Hoffmann-La Roche Ltd. XC, DC, LT, MN and VY are employees of Genentech, Inc. and hold shares in F. Hoffmann-La Roche Ltd. KZ is a former employee of Genentech, Inc. and holds shares in F. Hoffmann-La Roche Ltd.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.