The coronavirus disease 2019 (COVID-19) pandemic has resulted in an unprecedented surge in hospitalization and utilization of resources globally. Reference Carr1 To accommodate the influx of patients, healthcare systems have had to modify staff allocation while also optimizing space and available resources. Reference Cole and Barnard2 These rapid changes in the healthcare system have affected overall infection prevention and control activities and the quality of care provided to patients during the pandemic. Reference Stone and Feibel3 However, quantifying these effects remains challenging. Reference Austin and Kachalia4
Policy makers consider unplanned readmissions to reflect the quality of care related to discharge planning, patient engagement, and care transition. 5 The 30-day readmission rate is an important indicator of patient safety and quality, and it is linked to hospital reimbursements in the United States. Reference Horwitz, Partovian and Lin6 The 30-day timeframe is consistent with the other readmission measures approved by the National Quality Forum and publicly reported by the Centers for Medicare and Medicaid Services. Reference Horwitz, Partovian and Lin6 It is defined as an unplanned admission for any cause to an acute-care hospital within 30 days of discharge. 7 Given the scale of the COVID-19 pandemic in the United States and the continued emergence of new severe acute respiratory coronavirus virus 2 (SARS-CoV-2) variants, hospital readmissions increase resource utilization, impose an additional burden on healthcare systems, and increase the risk for hospital-acquired infections. 8 Hence, evaluating 30-day readmissions among COVID-19 patients could help with decision making on when to discharge patients, could help identify populations that require closer clinical follow-up after discharge, and could serve as an essential indicator of healthcare performance. Reference Grimm9
Although previous studies have evaluated readmissions following COVID-19 hospitalization, most have limitations. Reference Lavery, Preston and Ko10–Reference Yeo, Baek and Kim21 Previous studies were restricted to specific patient populations, included only a limited number of comorbidities identified as high-risk, had modest sample sizes, and were conducted in the early stages of the pandemic, when overwhelmed healthcare systems could bias the estimation of readmission rates. Moreover, no study thus far in the United States has evaluated 30-day readmission for COVID-19 as the primary reason for readmission. The present study was conducted to overcome these limitations.
The primary objective of this study was to use a large claims database to evaluate the association between comorbidities present during the index COVID-19 hospitalization and the risk of all-cause readmission to the same hospital within 30 days. The secondary objective was to evaluate the association between comorbidities and risk of readmission with COVID-19 as the primary diagnosis within 30 days of discharge.
Methods
Study design and data source
Using the Premier Healthcare database, we conducted a retrospective observational cohort study of patients discharged from hospitals after first-time admission for COVID-19. The Premier database is an all-payer repository of claims and clinical data, including records from 865 nongovernmental, community, and teaching hospitals that contributed inpatient data during the study period. 22 Hospitals in the Premier database represent geographically diverse areas across the United States and capture ∼1 of every 4 hospital discharges. Premier internally validates all data before their release into the database. 22 For most data elements, <1% of records had missing information, and for key elements, such as demographics and diagnostics, <0.01% of data were missing. 22 The Premier database has been used by traditional academic institutes (including previous research by our group), the National Institutes of Health, and the Centers for Disease Control and Prevention (CDC) to conduct COVID-19–related studies. Reference Pineles, Goodman and Pineles23–Reference Goodman, Magder and Baghdadi27 This study did not include personally identifiable information and was exempt from institutional review board review. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Reference von Elm, Altman and Egger28
Study population, COVID-19 case, 30-day readmission definitions
All inpatient admissions with discharge dates between April 2020 and April 2021 were included in the study. COVID-19 patients were identified through the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) discharge diagnosis code of U07.1 (COVID-19, virus identified). 29 The code had 98% specificity, 99% sensitivity, and a positive predictive value of 92% when internally validated in the Premier database against reports of laboratory-confirmed infection by polymerase chain reaction with SARS-CoV-2. Reference Kadri, Gundrum and Warner30 Our cohort, hereafter referred to as the index COVID-19 cohort, was defined as inpatients with a first-time diagnosis of COVID-19 and discharge between April 2020 and March 2021. Excluded from the index COVID-19 cohort were patients who died and those who left against medical advice. Any subsequent hospitalization occurring within 30 days of the index COVID-19 hospitalization discharge date, whether for COVID-19 or another indication, was considered 30-day readmission. Reference Lavery, Preston and Ko10 The Premier database records readmissions only if a patient returns to the same hospital where the index hospitalization has occurred. 22
Outcomes
The primary outcome was all-cause 30-day readmission following the index COVID-19 hospitalization. The secondary outcome was readmission for COVID-19 as the primary readmission diagnosis within 30 days.
Patient and hospital variables
For each patient admitted in the index COVID-19 cohort, we extracted the following information from the Premier database: a) demographic variables, such as age, sex, race, ethnicity, and marital status; b) hospital characteristics, including teaching status, urban versus rural location, number of beds, and geographic census division; c) year, quarter, and month of discharge for index hospitalization; d) time to readmission (days) calculated as the difference between the date of first readmission within 30 days and the date of discharge for index COVID-19 hospitalization; e) discharge disposition including mortality during hospitalization; and f) ICD-10-CM codes which were mapped to 29 Elixhauser comorbidity categories using standardized Agency for Healthcare Research and Quality methodology and software. 31 An unweighted Elixhauser score was calculated for each patient in the cohort. 31 g) COVID-19 severity indicators identified through literature review and expert clinical consensus (A.H., J.B., B.P., and G.N.) included supplemental oxygen therapy, intensive care unit (ICU) stay, vasopressor therapy, mechanical ventilation (invasive and noninvasive), renal replacement therapy, and extracorporeal membrane oxygenation (ECMO). Reference Bhatt, Jering and Vaduganathan32–Reference Kompaniyets, Goodman and Belay34 In the Premier database, these were mapped using the chargemaster (a comprehensive list of all items billable to a hospital patient or to a patient’s insurance) and from International Classification of Diseases, Tenth Revision, Procedure Codes (ICD-10-PCS) (Supplementary Table 1 online), 22 h) primary discharge diagnosis for readmission was categorized using the Clinical Classification Software Refined (CCSR) categories to approximate the primary reason for hospital readmission using ICD-10-CM codes. 35
Statistical analysis
Unique patients in the index COVID-19 cohort were followed through April 2021 to capture readmission within 30 days (yes or no). The unit of analysis was the unique patient. Descriptive statistics for patient, comorbidities, and hospital characteristics were calculated using the mean (standard deviation [SD]), median (interquartile range [IQR]), or frequency count (percentage). We estimated the odds ratios (ORs) and 95% confidence intervals (CIs) of all-cause readmission and COVID-19–specific readmission within 30 days. We analyzed the association between 29 Elixhauser comorbidity categories and the odds of readmission using a logistic regression model to obtain crude ORs and corresponding 95% CIs. The multivariable model was adjusted for a priori confounders that included age, sex, severity of COVID-19, covariates with a P ≤ .02 in the bivariate analysis, and biologically plausible covariates based on expert consensus. Reference Lavery, Preston and Ko10,Reference Goodman, Magder and Baghdadi27 The covariates included in the multivariable model were demographics (age, sex, race), hospital characteristics (number of beds, census division), coexisting comorbidities, discharge disposition (home self-care, transfer to skilled nursing facility, home health organization, hospice, ongoing care), discharge year and quarter of the index hospitalization, and COVID-19 severity (cf, ICU stay, supplemental oxygen therapy, renal replacement therapy, vasopressor therapy). Multicollinearity between comorbidities included in the model was assessed using the variance inflation factor. Reference Schreiber-Gregory and Jackson36 We used C-statistics to check the model predictions for outcomes. 37 All P values were 2-tailed and analyses were performed using SAS version 9.4 software (SAS Institute, Cary, NC).
Results
Characteristics of index COVID-19 cohort
Among 378,818 unique patients with an index COVID-19 hospitalization between April 2020 and March 2021, 44,277 (11.6%) died and 3,405 (0.9%) left against medical advice on the index admission. Of the remaining 331,136 patients (index COVID-19 cohort), the patient and hospital-level characteristics stratified by all-cause 30-day readmission are shown in Table 1.
Table 1. Characteristics Among 331,166 Index COVID-19 Hospitalized Patients, Stratified by All-Cause 30-Day Readmission and No Readmission
Note. IQR, interquartile range; SD, standard deviation.
a Discharge Quarters. Quarter 2, 2020: April, May, June; Quarter 3, 2020: July, August, September; Quarter 4, 2020: October, November, and December; Quarter 1, 2021: January, February, March.
b Hospice includes discharged to hospice-home or discharged to hospice-medical facility
c Ongoing care categories include discharge or transfer to a cancer center, to federal hospital, swing bed unit, another rehabilitation facility, long-term care hospitals that provide acute inpatient care with an average length of stay ≥25 d, a psychiatric hospital, and a critical-access hospital.
d Other categories includes patients who were discharged to facilities other than those listed in ongoing care.
e Clinical/Severity markers were captured for first time COVID-19 hospitalization
f Vasopressor therapy was defined as receiving any of the following vasopressor agents during hospitalization: epinephrine, nor-epinephrine, phenyl epinephrine, dopamine, dobutamine, and vasopressin.
g ICD-10 CM diagnoses discharge codes were mapped to Elixhauser comorbidities using the standardized Agency for Healthcare Research and Quality methodology and software into 29 comorbidity categories.
Characteristics of patients readmitted within 30 days
All-cause readmission
Among the 331,166 patients in the index COVID-19 cohort, 36,827 (11.1%) were readmitted within 30 days at least once to the same hospital as the index hospitalization. The median length of stay during the index COVID-19 admission was 5 days (IQR, 3–10) and the time to readmission was 5 days (IQR, 2–12). Approximately one-third 11,325 (30.8%) of the readmitted patients were aged ≥80 years. Among the readmitted patients, 18,027 (49%) received supplemental oxygen therapy, 5,539 (15%) required ICU stay, and 2,913 (8%) received renal replacement therapy on the index admission. The most frequent primary diagnoses among the readmitted patients were infectious diseases (n = 16,873, 45.8%), followed by diseases of the circulatory system (n = 3,864, 10.5%) (Table 2). All-cause mortality during hospital readmission occurred in 6,466 patients (17.6%).
Table 2. Patient’s Primary Discharge Diagnosis for Readmission Hospitalization
COVID-19 as primary readmission diagnosis
In the index COVID-19 cohort, 11,382 patients (3.4%) were readmitted within 30 days of the index hospitalization with COVID-19 as the primary readmission diagnosis. The characteristics were similar to those of patients with all-cause 30-day readmission (Supplementary Table 2 online). Among patients readmitted with COVID-19, the median length of stay during the index COVID-19 admission was 4 days (IQR, 2–6), and the time to readmission was 3 day (IQR, 1–7) days. All-cause mortality during hospital readmission for COVID-19 was observed in 1,952 patients (17.2%).
Comorbidities associated with 30-day readmission
In the index COVID-19 cohort, 313,963 patients (95%) had at least 1 comorbidity at the time of discharge. The median unweighted Elixhauser comorbidity score of readmitted patients (for both all-cause and readmissions with COVID-19 as primary readmission diagnosis) was 4 (IQR, 3–6). Using the mean unweighted Elixhauser score patients in the index COVID-19 cohort were grouped into no comorbidities (n = 17,173, 5.2%), 1 comorbidity (n = 41,175, 12.4%), 2 comorbidities (n = 61,949, 18.7%), 3 comorbidities (n = 67,013, 20.2%), 4 comorbidities (n = 55,668, 16.8%), 5 comorbidities (n = 39,005, 11.8%), and >5 comorbidities (n = 49,153, 14.9%). Furthermore, >4 comorbidities during the index COVID-19 hospitalization were present in 143,826 patients (43.4%). Patients with 1 or more comorbidities present during the index hospitalization had an increased adjusted odds of readmission for both all-cause and readmission with COVID-19 as the primary readmission diagnosis compared to patients with no comorbidities (Fig. 1). In the multivariable model, each additional comorbidity category added was associated with an 18% increase in the odds of all-cause readmission (adjusted OR, 1.18; 95% CI, 1.17–1.19) and a 10% increase in COVID-19 as the primary readmission diagnosis (adjusted OR, 1.10; 95% CI, 1.09–1.11).
Fig. 1. Forest plot of adjusted odds of all-cause and COVID-19–specific 30-day readmission.
In another multivariable model including individual comorbidities (Table 3), most of the Elixhauser comorbidity categories were associated with increased odds of 30-day all-cause readmission. Metastatic cancer (adjusted OR, 1.98; 95% CI, 1.80–2.18), lymphoma (adjusted OR, 1.96; 95% CI, 1.76–2.17) and drug abuse (adjusted OR, 1.55; 95% CI:1.43 –1.69) were associated with the highest odds of readmission. Obesity was associated with decreased odds of readmission (adjusted OR, 0.91; 95% CI, 0.88–0.93). The C-statistic for the multivariable model (Table 3) for predicting the outcome of all-cause 30-day readmission was 0.71. For the multivariable model for readmission with COVID-19 as the primary readmission diagnosis, 52% (15 of the 29 categories) of the Elixhauser comorbidity categories were associated with an increased risk of readmission. Lymphoma (adjusted OR, 1.86; 95% CI, 1.58–2.19), renal failure (adjusted OR, 1.32; 95% CI, 1.25–1.40), and chronic lung disease (adjusted OR, 1.29; 95% CI, 1.24–1.34) were the comorbidities most strongly associated with COVID-19 as primary readmission diagnosis. Among patients readmitted with COVID-19, obesity (adjusted OR, 0.92; 95% CI, 0.88–0.97) and weight loss (adjusted OR, 0.68; 95% CI, 0.62–0.75) were associated with decreased risks of 30-day readmission. The C-statistic for the multivariable model (Table 3) to predict COVID-19–specific readmission was 0.67.
Table 3. Multivariable Model of Comorbidity Categories and 30-Day Readmission Following Index COVID-19 Hospitalization
Note. CI, confidence interval. Odds ratio obtained from logistic regression.
a Comorbidities present at the end of first-time COVID-19 hospitalization.
b Adjusted odds ratio for demographics (age, sex, race), hospital characteristics (number of beds, geographic census division), coexisting comorbidities, discharge disposition (home self-care, transfer to skilled nursing facility, home health organization, hospice, ongoing care), discharge quarter of the index hospitalization, and COVID-19 severity (ICU stay, supplemental oxygen therapy, renal replacement therapy, vasopressor therapy).
Temporal trends in 30-day readmission
We evaluated the trends in readmissions among COVID-19 patients over the study period. As illustrated in Table 4, after controlling for demographics (age, sex, race) and hospital characteristics (number of beds and census division), the risk of all-cause readmission was 27% higher in May 2020 (adjusted OR, 1.27; 95% CI, 1.19–1.35) and for readmissions with COVID-19 as the primary readmission diagnosis was 43% higher in July 2020 than in April 2020 (adjusted OR, 1.43; 95% CI, 1.29–1.57).
Table 4. Adjusted Odds Ratios of 30-Day Readmission Over the Study Period From April 2020 to April 2021
Note. CI, confidence interval. Adjusted for demographics (age, sex, race) and hospital characteristics (number of beds, geographic census division).
Discussion
To the best of our knowledge, this is the largest study to date in the United States to investigate 30-day readmission following index hospitalization for COVID-19. In this cohort of 331,136 unique patients hospitalized for COVID-19, the all-cause 30-day readmission rate to the same hospital was 11.2%. The readmission rate within 30 days for COVID-19 as the primary readmission diagnosis was 3.4%. Furthermore, with each additional Elixhauser comorbidity category present during the index COVID-19 hospitalization, an 18% increase in the odds of all-cause readmission and a 10% increase in the odds of COVID-19 readmission were observed.
Comorbidities (eg, chronic lung disease, diabetes with complications, chronic kidney disease, congestive heart failure, cancer, hypertension, neurological conditions, substance use disorders, and human immunodeficiency virus infection) identified by the CDC as high-risk comorbidities for COVID-19 were associated with an increased risk of readmission in our study. These results are consistent with those of previous studies. Reference Lavery, Preston and Ko10,Reference Atalla, Kalligeros, Giampaolo, Mylona, Shehadeh and Mylonakis14,Reference Ramzi16 A postulated explanation for the association between comorbidities and readmission is the worsening of complications of underlying comorbidities due to COVID-19 or collateral damage caused by the COVID-19 pandemic. For instance, patients may avoid seeking medical attention, and thus, their comorbid conditions may be under poor control, or they may receive less specific postdischarge care than they need because of the burden on the healthcare system. Reference Chopra, Flanders, O’Malley, Malani and Prescott13,Reference Masroor38
Although the CDC has identified obesity as a strong risk factor for severe COVID-19 in our study, obesity was associated with a 9% decrease in adjusted odds of all-cause readmission. Our results are consistent with prior studies in which obesity was associated with a decreased risk for readmission. Reference Lavery, Preston and Ko10,Reference Verna, Landis and Brown17 This phenomenon, sometimes referred to as the “obesity paradox” is explained by a potentially protective role of excessive fat accumulation, which may contribute to a more favorable environment to withstand breakdown in patients caloric intake frequently occurring in ICU settings. Reference Biscarini, Colaneri and Ludovisi39 Another possible explanation for the reduced readmission risk is that physicians tend to consider obese patients to be at a higher risk of worse outcomes, resulting in earlier admission and more aggressive, timely management and postdischarge instructions.
In our study, ∼1 in 9 COVID-19 hospitalized patients were readmitted within 30 days after discharge. These findings were comparable to those reported in the literature. Reference Horwitz, Partovian and Lin6,Reference Taupin, Anderson and Merchant11,Reference Mozaffari, Liang, Stewart, Thrun, Hodgkins and Haubrich18,Reference Yeo, Baek and Kim21 Also, 30.9% of all-cause 30 day readmissions were due to COVID-19 in our study, which was less than that in a prior study conducted to evaluate all-cause 60-day readmission using the Premier database, in which 45% of the readmitted patients had COVID-19 as their primary readmission discharge diagnosis. Reference Lavery, Preston and Ko10 A possible explanation for this discrepancy could be the varying duration of follow-up and timing of the prior study (March 2020 to August 2020), when COVID-19 discharges may have been rushed, resulting in a higher COVID-19 readmission rate. Readmissions of patients with COVID-19 as the primary discharge diagnosis within 30 days of discharge were much lower than all-cause 30-day readmission. Only 1 in 30 patients discharged with COVID-19 was readmitted with COVID-19 as the primary discharge diagnosis. Some postulated explanations for readmissions with COVID-19 include the continuation of the initial disease process from the index COVID-19 hospitalization and clinical worsening due to underlying comorbidities.
As the clinical and epidemiological features of COVID-19 may have many parallels with influenza, it is important to monitor both diseases to ensure the optimal management of resources, as we anticipate their continued cocirculation. The 30-day all-cause readmission rate for influenza in a 2018 US study using administrative data was 11.4%, similar to the 11.2% 30-day all-cause readmission for COVID-19 in our study. Reference Yandrapalli, Aronow and Frishman40 However, in our study, ∼17% of patients died, which was higher than the 6.5% of patients who died during readmission for influenza. Reference Yandrapalli, Aronow and Frishman40 Also, all-cause mortality during readmission was higher than the mortality during the index COVID-19 hospitalization (11%), suggesting that poor outcomes are associated with COVID-19 readmission within 30 days. Given the seasonality of influenza, we observed temporal variations in 30-day readmissions following COVID-19 hospitalization, which could be attributed to a surge in COVID-19 cases or the emergence of newer variants overwhelming the healthcare system, resulting in increased readmissions.
Our study had several limitations. The readmissions recorded in our study could have been underestimated because the Premier database captures readmissions only if patients were readmitted to the same hospital as their index hospitalization. Although we adjusted for several potential confounders in our analysis, residual confounding may persist owing to variables not being available in the database or not being analyzed. Our cases of COVID-19 were not laboratory confirmed and body mass index was not included in the database. Another limitation of our study was the inability to test the association between vaccination history, various COVID-19 variants, and the odds of readmission. Finally, we used ICD-10 CM codes to identify comorbidities; therefore, if a patient’s comorbidity is incorrectly assigned, this could lead to misclassification bias.
Unlike most previous COVID-19 readmission studies that assessed selective comorbidities, our large sample size allowed us to study the association between readmissions for all 29 Elixhauser comorbidity categories for all-cause and COVID-19 specific readmission within 30 days. Another strength of our study is the utilization of a large, representative administrative database collected using validated methods to minimize selection bias. To our knowledge, this is the first study performed in the United States to date using an administrative database to study 30-day readmission for COVID-19 as a primary readmission diagnosis and to identify comorbidities associated with readmission.
Although recent studies have shown that only ∼25% of 30-day hospital readmissions following admission for COVID-19 are potentially preventable, these results highlight opportunities for improvement in reducing COVID-19 readmissions. Reference Taupin, Anderson and Merchant11 Understanding the comorbidities associated with readmission may help quality leaders to define optimal timing for discharge of COVID-19 patients, ensure safe care transition, and postdischarge care for COVID-19 patients.
With the continued emergence of COVID-19 variants and the increased risk of readmission among COVID-19 patients, hospital epidemiologists and infection prevention and control teams will continue to play an integral role in pandemic preparedness and response. Identification of COVID-19 patients who may be high-risk for readmission prior to discharge may aid in more strategic allocation of resources and could ultimately facilitate the delivery of high-quality patient care.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2022.232
Acknowledgments
We thank Gwen Robinson for her assistance in creating the Clinical Classification Software Refined (CCSR) categories.
Financial support
Access to the Premier Healthcare Database was purchased through departmental funds provided by the Department of Epidemiology and Public Health of the University of Maryland School of Medicine. The authors received no financial support for the research, authorship, or publication of this article. Premier, Inc, did not participate in the design, analysis, or drafting of the manuscript.
Conflicts of interest
A.D.H. serves as an infection control section editor for UpToDate outside the submitted work. All other authors declare no relevant conflicts of interest.
