key: cord-0903001-rcmmzmrn
authors: Shrestha, Sundar S; Kompaniyets, Lyudmyla; Grosse, Scott D; Harris, Aaron M; Baggs, James; Sircar, Kanta; Gundlapalli, Adi V
title: Estimation of Coronavirus Disease 2019 Hospitalization Costs From a Large Electronic Administrative Discharge Database, March 2020–July 2021
date: 2021-12-20
journal: Open Forum Infect Dis
DOI: 10.1093/ofid/ofab561
sha: 71bc9d0db9c55f7846396dc139350734bffd12b7
doc_id: 903001
cord_uid: rcmmzmrn

BACKGROUND: Information on the costs of inpatient care for patients with coronavirus disease 2019 (COVID-19) is very limited. This study estimates the per-patient cost of inpatient care for adult COVID-19 patients seen at >800 US hospitals. METHODS: Patients aged ≥18 years with ≥1 hospitalization during March 2020–July 2021 with a COVID-19 diagnosis code in a large electronic administrative discharge database were included. We used validated costs when reported; otherwise, costs were calculated using charges multiplied by cost-to-charge ratios. We estimated costs of inpatient care per patient overall and by severity indicator, age, sex, underlying medical conditions, and acute complications of COVID-19 using a generalized linear model with log link function and gamma distribution. RESULTS: The overall cost among 654673 patients hospitalized with COVID-19 was $16.2 billion. Estimated per-patient hospitalization cost was $24 826. Among surviving patients, estimated per-patient cost was $13 090 without intensive care unit (ICU) admission or invasive mechanical ventilation (IMV), $21 222 with ICU admission alone, and $59 742 with IMV. Estimated per-patient cost among patients who died was $27 017. Adjusted cost differential was higher among patients with certain underlying conditions (eg, chronic kidney disease [$12 391], liver disease [$8878], cerebrovascular disease [$7267], and obesity [$5933]) and acute complications (eg, acute respiratory distress syndrome [$43 912], pneumothorax [$25 240], and intracranial hemorrhage [$22 280]). CONCLUSIONS: The cost of inpatient care for COVID-19 patients was substantial through the first 17 months of the pandemic. These estimates can be used to inform policy makers and planners and cost-effectiveness analysis of public health interventions to alleviate the burden of COVID-19.

can be substantial [9, 10] , information on those costs, a significant component of healthcare burden of COVID-19, is limited. Previous studies on costs have been limited to the early months of the pandemic or to patients with specific insurance types [11] [12] [13] [14] , based on flu or pneumonia diagnoses as proxies for COVID-19 [9, 12] , or limited to small numbers of hospitals that reported costs [15] . Detailed information on hospitalization costs of COVID-19 is important not only to understand the economic burden of disease but also to assess economic benefits of COVID-19 prevention efforts such as vaccination efforts in the United States. In this study, using the most recent data from >800 hospitals covering all payers in the United States, we assess the average per-patient cost of inpatient care for adult COVID-19 patients, overall, and by severity, age, sex, underlying medical conditions, and acute complications.

The data for this study was obtained from the Premier Healthcare Database Special COVID-19 Release ([PHD-SR] release date: September 14, 2021) . The PHD-SR is a large, geographically diverse US all-payer administrative database that contains information on inpatient discharges from more than 800 hospitals located in 48 states [16] .

This activity was reviewed by the Centers for Disease Control and Prevention (CDC) and was conducted consistent with applicable federal law and CDC policy.

Patients aged ≥18 years with ≥1 hospitalization with a COVID-19 diagnosis code in PHD-SR during March 2020-July 2021 were included. Coronavirus disease 2019 diagnosis was based on an International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis of U07.1 ("COVID-19, virus identified") during April 2020-July 2021 or B97.29 ("Other coronavirus as the cause of diseases classified elsewhere", recommended before the April 2020 release of U07.1) during March-April 2020 [17, 18] . Patients whose sex was not recorded (n = 309 [0.05%]) and those with zero costs (n = 1174 [0.18%]) were excluded from the study. The final sample included 654 673 unique patients (Supplementary Figure 1 ).

The outcome variable was the estimated cumulative per-patient hospitalization cost of all hospitalizations with a COVID-19 discharge diagnosis code for each patient in PHD-SR. We used validated costs that passed all validity checks, including financial reconciliation in PHS-SR; otherwise, costs were calculated using reported charges multiplied by estimated hospital-specific cost-to-charge ratios (CCRs) when available or aggregate CCR otherwise (Supplementary Figure 2) . Costs and charges reported by hospitals are facility fees, inclusive of salaries but excluding professional fees for services billed to patients by independent physicians.

Covariates of interest were patient and hospital characteristics, severity indicators, underlying medical conditions, acute complications of COVID-19, and patient demographic characteristics. All patient and hospital information (sex, age, race/ethnicity, payer type, hospital US Census region, hospital urbanicity) was obtained from patient medical records during a patient's first COVID-19 hospitalization in PHD-SR. Severity indicators included the following: ICU admission, IMV, and death. Intensive care unit admission and IMV were defined using patient billing records, and death was determined at patient discharge. Underlying medical conditions were captured by ICD-10-CM diagnosis codes from inpatient or outpatient medical records in PHD-SR from January 2019 up to and including a patient's first inpatient encounter with COVID-19. Conditions included in the analysis were those defined by the CDC as risk factors for severe COVID-19 illness [19] . Acute complications were obtained from all hospitalizations with COVID-19 for each patient [20] .

We used 3 multivariable generalized linear models with log link function and gamma distribution, with the outcome variable being the cost of COVID-19 inpatient care per patient. All models (1) used robust standard errors clustered on hospital identifier and (2) controlled for patient age group, sex, race/ ethnicity, payer type, hospital urbanicity, hospital US Census region, and admission month as a linear and quadratic term (to account for potential nonlinear unobservable changes in the severity of illness over the course of the pandemic). Model 1 included severity indicators (ICU admission, invasive mechanical ventilation, death) and their interactions as the covariates of interest; the model estimated the per-patient cost of COVID-19 inpatient care overall and by each severity indicator. Model 2 estimated differences in the cost of COVID-19 inpatient care by age, sex, race/ethnicity, and underlying medical conditions as the covariates of interest. Model 3 estimated the differences in the hospitalization costs for patients with COVID-19 by acute complications of COVID-19 as the covariates of interest. Finally, mean and median costs per COVID-19 patient were compared with hospitalization costs of influenza (ICD-10-CM codes J09-11) or pneumonia (ICD-10-CM codes J12-J18) in 2019, adjusted to 2020 US dollars using the Personal Consumption Expenditures: Health care index (https: www.// apps.bea.gov) [21] .

We performed 3 supplemental analyses, modified from Model 1. The first analysis was performed among surviving patients. The second analysis was performed among patients with COVID-19 as a primary diagnosis. The third analysis used the cost of the first hospitalization with COVID-19 as the outcome variable.

All analyses were conducted using R 4.0. Figure 5) . The average estimated per-patient cost was lowest for patients with no ICU admission, IMV, or death and highest for surviving patients with IMV.

The estimated cost was $13 543 higher for patients with ICU admission (95% CI, $12 474-$14 612) compared to those without ICU admission, $39 840 higher for patients who needed IMV (95% CI, $36 159-$43 520) compared to those without IMV, and not significantly different for patients who died (absolute difference = $955; 95% CI, −$192 to $2102) compared to patients who survived ( Figure 1 ).

Based on Model 2, compared to female patients, male patients had 14% higher estimated cost ($3280; 95% CI, $2983-$3577) ( Figure 2 ). There was a nonlinear association between age and cost of inpatient care, with the incremental difference in hospitalization costs rising with older age until the ages of 60-69 years, and then decreasing slightly with age groups over 70 years. The incremental difference in hospitalization costs ranged from $2044 (95% CI, $1496-$2592) for patients aged 30-39 years to $9750 (95% CI, $8986-$10 515) for patients aged 60-69 years, compared to those aged 18-29 years.

The estimated costs also varied by race/ethnicity. The highest adjusted cost differentials compared to non-Hispanic white patients were observed among non-Hispanic Asian and Hispanic patients, with absolute differences of $7014 (95% CI, $4534-$9493) and $5720 (95% CI, $4183-$7256), respectively ( Figure  2 ). In contrast, non-Hispanic black patients had a differential per-patient cost of −$1169 (95% CI, −$1920 to −$418) compared to non-Hispanic white patients.

Based on Model 2, per-patient hospitalization cost was the highest among patients with chronic kidney disease (by 64% or $12 391; 95% CI, $11 522-$13 260), followed by liver disease (by 37% or $8878; 95% CI, $7879-$9878), cerebrovascular disease (by 30% or $7267; 95% CI, $6316-$8219), and obesity (by 26% or $5933; 95% CI, $5445-$6422), compared to patients without the corresponding condition ( Figure 2 ).

Based on Model 3, acute complications of COVID-19 were associated with higher costs of inpatient care (Figure 3 ). The cost of per-patient inpatient care was 3.09 times as high for patients with acute respiratory distress syndrome (95% CI, 

This study provides estimates of the cost of inpatient care for adult patients hospitalized with COVID-19 during March 2020-July 2021, using data from >800 geographically dispersed US hospitals covering all payers. This study builds on a prior study using data from the PHD-SR [15] by creating the costto-charge ratios to obtain costs when they were not verified in the data, by including an additional 10 months of data, adding death as a severity indicator, estimating costs by underlying medical condition and acute complication of COVID-19, and by providing adjusted estimates of costs. The unadjusted mean cost of $24 717 per patient hospitalized with COVID-19 in this study is comparable to the mean cost of $24 571 from a prior study using the Premier Healthcare COVID-19 database during April 1-October 31, 2020 [15] and mean Medicare fee-for-service payment of $24 454 per COVID-19 hospitalization using the Centers for Medicare & Medicaid Services (CMS) data during January 1, 2020-July 24, 2021 [13] , but slightly higher than mean hospitalization cost of $21 752 among Medicare free-for-service beneficiaries during April 1-December 31, 2020 [14] . Our estimated and unadjusted mean per-patient cost of $24 826 and $24 713 are much higher than the unadjusted mean cost of influenza-related inpatient care in 2019 ($13 483 in 2020 dollars) and slightly higher than that of pneumonia-related inpatient care in 2019 ($21 447, in 2020 dollars) in Premier Healthcare Database Special COVID-19 Release database (Supplementary Table 1 ). The finding that patients needing ICU admission and IMV for COVID-19 have higher costs of care is comparable to that of a previous study [15] . Our finding of a nonlinear positive adjusted association between age and cost of inpatient care shows the need for more intensive use of resources with older age, and it is comparable to previously reported patterns of unadjusted mean or median hospitalization costs by age group [15] and median estimated allowed amount by age group [11] . Our finding of higher COVID-19 hospitalization costs among males than females could be due to a combination of factors, including relatively higher odds of readmission and having higher proportion of clinical severity such as ICU admission and IMV among males than females [8, 15, 22] .

Our supplementary analyses showed that the estimated cost of the first COVID-19 hospitalization was comparable to that from the main analysis because most patients (94.9%) had only 1 hospitalization with COVID-19. The estimated per-patient cost among 410 489 patients with a primary diagnosis of COVID-19 was slightly lower, but it was comparable to estimates from the main analysis. This indicates that the per-patient hospitalization cost from the main analysis is likely a close estimate of the hospitalization cost attributable to COVID-19. On the other hand, as expected, the estimated per-patient cost of surviving patients with COVID-19 was slightly lower than per-patient cost estimated including patients who died of COVID-19. This could be due to differences in the average unadjusted number of hospital days per patient (8 days vs 13.5 days) and prevalence of IMV (6.9% vs 60.6%) among surviving patients and those who died, respectively (results not shown). Our finding shows modestly lower COVID-19 hospitalization costs among non-Hispanic black patients compared to non-Hispanic white patients. Previous studies suggest that non-Hispanic black patients have higher rates of underlying medical conditions, acute complications, and greater severity than non-Hispanic white patients [23] [24] [25] and greater odds of hospital admission at the time of presenting with COVID-19. Those patterns possibly reflect delays in seeking care due to barriers to accessing care [26] . Some studies suggest that once hospitalized with COVID-19, there were no differences in risk of poor outcomes [27] , or of mortality between non-Hispanic black and non-Hispanic white patients [28] , whereas others show a higher risk of poor outcomes and death among non-Hispanic black patients [29] . The modestly lower average costs among non-Hispanic black patients compared to non-Hispanic white patients, controlling for underlying medical conditions, could be a result of a combination of factors, including barriers in access to care, lower odds of readmission [8] , and higher odds of COVID-19 mortality [29, 30] . Further research on the relationship between COVID-19 hospitalization cost and race/ethnicity may be warranted.

Our findings of variations in the cost of COVID-19 hospitalization by patients' underlying medical conditions and acute complications could be due to several factors. Those include differences in relative risk severity, readmission, and length of stay by conditions or complications. For example, the higher cost for COVID-19 patients with chronic kidney disease could be because of (1) relatively higher odds of readmission [8] and (2) COVID-19 severity and ICU admission [31, 32] .

The findings in this study have limitations. First, although the PHD-SR data cover more than 20% of inpatient admissions in the United States [17] , the results may not be generalizable to all hospital patients. For example, the PHD-SR database underrepresents hospitals in the Western states. Second, because the cost in PHD-SR data only represents the facility or institutional fees and excludes noninstitutional (professional) fees, our estimated hospitalization costs would be underestimated. A study using the MarketScan databases calculated that including professional fees increased total per-admission payments in 2012 by 26.4% for commercially insured admissions and by 17.7% for Medicaid admissions relative to facility-only payments [33] . Third, in our cost estimation, we did not adjust 2020 costs to 2021 equivalents because suitable index data were not available for all months. Fourth, there is likely incomplete and potentially biased reporting of race/ethnicity, which complicates the interpretation of the obtained associations of cost with race/ethnicity. For example, just one third of patients hospitalized with COVID-19 in the PHD-SR were classified as either non-Hispanic black or Hispanic. Hospitalization data from states that have reported complete data on race/ ethnicity have consistently reported that non-Hispanic black individuals accounted for much larger shares of hospitalized patients than in the general population [34] . Furthermore, the exclusion of public hospitals in the sample and lack of representativeness of hospitals in the sample would have reduced the share of disadvantaged minority populations in the sample, complicating the interpretation of our results. Fifth, the reliance on ICD-10-CM diagnosis codes for identification The figure represents the results from Model 3, a single multivariable generalized linear model, with log link and gamma distribution function, and acute complications as covariates of interest, adjusted for other control variables (age group, sex, race/ethnicity, payer type, hospital US Census region, hospital urbanicity, and admission month as a linear and quadratic term). Relative and absolute adjusted differences in costs are presented. Error bars represent 95% confidence intervals. ARDS, acute respiratory distress syndrome; COPD, chronic obstructive pulmonary disease; DIC, disseminated intravascular coagulation.

of COVID-19 patients is subject to risk of misclassification. However, although COVID-19 diagnoses were not laboratory confirmed, the risk of misclassification of COVID-19 cases is low. A study that linked PHD-SR data to laboratory test results for a subset of 150 hospitals found specificity of 99% and sensitivity of 98% for the U07.1 ICD-10-CM code for COVID-19 [17] . Similarly, the identification of underlying medical conditions and acute complications was based on ICD-10-CM codes, and the co-occurring conditions may therefore be misclassified. The accuracy of hospital diagnoses for other conditions is heterogeneous [35] . Sixth, the per-patient hospitalization cost may be underestimated because of likely exclusion of patients who received care in other hospitals, and only some professional services, those provided by hospitals, are included in the hospital charges. Seventh, we did not estimate the excess costs attributable to COVID-19 because we did not estimate costs for patients with non-COVID-19 hospitalizations. In a supplementary analysis, we estimated costs among patients with COVID-19 as a primary diagnosis, which can reflect potential costs attributable to COVID-19. However, further research would be needed to estimate costs attributable to COVID-19. Finally, the costs in this study represent the resource costs incurred by hospitals in providing inpatient services. These costs may be either higher or lower than the reimbursements received by hospitals for the care provided, which can vary greatly by payer type. Because the PHD-SR database does not report payments received by hospitals, it is impossible to infer from these estimates the financial impact of COVID-19 care on participating hospitals. Similarly, these estimates of costs should not be confused with expenditures incurred by health plans or patients, information that was not available in the PHD-SR data.

Costs of hospitalized patients diagnosed with COVID-19 were estimated using a large observational cohort of all-payer electronic health administrative data from >800 US hospitals during the first 17 months of the pandemic in 2020 and 2021. The results indicated that the average per-patient cost incurred by hospitals for inpatient care of COVID-19 patients in the United States was substantial. The cost was higher with an increasing level of clinical severity among surviving patients, with the highest cost associated with IMV. The cost estimates by patients' profile provide additional insight into patient groups needing special attention. These cost estimates could be used in combination with trial-based estimates of averted COVID-19 adult hospitalizations to assess the largest component of the economic benefits of COVID-19 vaccination efforts to the US healthcare system and the cumulative economic impact of COVID-19 on US hospitals. The results could also be used to assess the resources required to treat COVID-19 patients in a hospital setting with specific underlying or acute medical conditions or demographic characteristics.

Demographic trends of COVID-19 cases and deaths in the U.S. reported to CDC

New admissions of patients with confirmed COVID-19 per 100,000 population by age group

COVID-19 data in motion

COVID-NET: COVID-19-Associated Hospitalization Surveillance Network

Death certificate-based ICD-10 diagnosis codes for COVID-19 mortality surveillance -United States

Weekly hospitalization trends

Body mass index and risk for COVID-19-related hospitalization, intensive care unit admission, invasive mechanical ventilation, and death -United States

Characteristics of hospitalized COVID-19 patients discharged and experiencing same-hospital readmission -United States

The potential health care costs and resource use associated with COVID-19 in the United States

The COVID-19 pandemic and the $16 trillion virus

Available at: https:// s3.amazonaws.com/media2.fairhealth.org/brief/asset/Key%20Characteristics%20 of%20COVID-19%20Patients%20-%20Profiles%20Based%20on%20Analysis%20 of%20Private%20Healthcare%20Claims%20-%20A%20FAIR%20Health%20Brief

Incremental costs associated with length of hospitalization due to viral pneumonia: impact of intensive care and economic implications of reducing the length of stay in the era of COVID-19

Preliminary Medicare COVID-19 Data Snapshot

Patient characteristics and costs associated with COVID-19-related medical care among Medicare fee-for-service beneficiaries

Health outcomes and economic burden of hospitalized COVID-19 patients in the United States

Premier Healthcare Database (COVID-19): Data that informs and performs

Uptake and accuracy of the diagnosis code for COVID-19 among US Hospitalizations

Centers for Disease Control and Prevention. New ICD-10-CM code for the 2019 novel coronavirus (COVID-19). Available at

Scientific evidence for conditions that increase risk of severe illness

Respiratory and nonrespiratory diagnoses associated with influenza in hospitalized adults

Adjusting health expenditures for inflation: a review of measures for health services research in the United States

Risk of clinical severity by age and race/ethnicity among adults hospitalized for COVID-19-United States

Risk of severe illness or death from COVID-19: Racial and ethnic health disparities

Risk for in-hospital complications associated with COVID-19 and influenza -Veterans Health Administration

Racial and ethnic disparities in hospital admissions from COVID-19: determining the impact of neighborhood deprivation and primary language

Disparities in outcomes among COVID-19 patients in a large health care system in California

Risk factors for intensive care unit admission and in-hospital mortality among hospitalized adults identified through the US Coronavirus Disease 2019 (COVID-19)-Associated Hospitalization Surveillance Network (COVID-NET)

Hospitalization and mortality among black patients and white patients with Covid-19

National disparities in COVID-19 outcomes between black and white Americans

Comorbidities associated with mortality in 31,461 adults with COVID-19 in the United States: a federated electronic medical record analysis

Effect of underlying comorbidities on the infection and severity of COVID-19 in Korea: a nationwide case-control study

Clinical characteristics and morbidity associated with coronavirus disease 2019 in a series of patients in metropolitan Detroit

Professional fee ratios for US hospital discharge data

Assessment of COVID-19 hospitalizations by race/ethnicity in 12 states

Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in a unique dually coded database

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.