Abstract
BACKGROUND: Many COVID-19 studies are constructed to report hospitalization outcomes, with few large multi-center population-based reports on the time course of intra-hospitalization characteristics, including daily oxygenation support requirements. Comprehensive epidemiologic profiles of oxygenation methods used by day and by week during hospitalization across all severities are important to illustrate the clinical and economic burden of COVID-19 hospitalizations.
METHODS: This was a retrospective, multi-center observational cohort study of 15,361 consecutive hospitalizations of patients with COVID-19 at 25 adult acute care hospitals in Texas participating in the Society of Critical Care Medicine Discovery Viral Respiratory Illness Universal Study COVID-19 registry.
RESULTS: At initial hospitalization, the majority required nasal cannula (44.0%), with an increasing proportion of invasive mechanical ventilation in the first week and particularly the weeks to follow. After 4 weeks of acute illness, 69.9% of adults hospitalized with COVID-19 required intermediate (eg, high-flow nasal cannula, noninvasive ventilation) or advanced respiratory support (ie, invasive mechanical ventilation), with similar proportions that extended to hospitalizations that lasted ≥ 6 weeks.
CONCLUSIONS: Data representation of intra-hospital processes of care drawn from hospitals with varied size, teaching and trauma designations is important to presenting a balanced perspective of care delivery mechanisms employed, such as daily oxygen method utilization.
Introduction
Throughout the pandemic, observational studies that evaluated data across integrated health systems1,-,3 in patients with COVID-19 reported hospital mortality and hospitalization characteristics, including length of stay, ICU admission rates, and the varying levels of oxygen support required for treating respiratory failure.2,4 These epidemiologic insights into broad processes of care during hospitalization have generated consensus around risk factors associated with ICU admission, in-hospital mortality, and post-hospitalization debility attributed to COVID-19.5,6 Many studies, however, are constructed to report hospitalization outcomes, with few large multi-center population-based reports on the time course of intra-hospitalization characteristics, including daily oxygenation support requirements. Further, the incidence and prevalence and characteristics of patients with chronic critical illness, broadly defined as organ failure that required prolonged ICU stays of >6 d,7,8 are difficult to discern from existing aggregate reports of COVID-19 hospitalizations.
Hence, high-quality integrated health-system data offer insights into community-level contributions that may otherwise be hidden in prominent datasets, which largely report on academic medical centers.9 Thus, we describe catchment-area factors (ie, facility volumes, location, trauma center status) and patient characteristics, including daily oxygenation support, and outcomes associated with critical illness and mortality for adults hospitalized with COVID-19.
Quick Look
Current Knowledge
Many observational studies are constructed to report hospitalization outcomes, with few large multi-center population-based reports on the time course of intra-hospitalization characteristics. During the COVID-19 pandemic oxygen therapy and mechanical ventilation reached peak usage prompting conservation measures in individual facilities.
What This Paper Contributes to Our Knowledge
We present epidemiologic profiles of daily oxygenation methods by day and by week to describe these intra-hospital processes of care for a consecutive case series of 15,361 hospitalized adults diagnosed with SARS-CoV-2. Measures of daily oxygen support methods can serve as a comprehensive system-level lens into the epidemiology of demand for oxygen specific to adults hospitalized with COVID-19.
Methods
We conducted a retrospective, multi-center cohort study of adult subjects admitted to 25 hospitals within an integrated health system over 12 months (March 13, 2020 – February 28, 2021). This research was approved by the Baylor Scott & White Research Institute’s institutional review board (020–119) with a waiver for informed consent. The ClinicalTrials.gov identifier is NCT04323787, and the reporting of this study conforms to the STROBE statement.10
Study Population, Setting, and Data Collection
Baylor Scott & White Health hospitals participating in the Society of Critical Care Medicine Discovery Viral Respiratory Illness Universal Study (VIRUS) COVID-19 registry for patients hospitalized with SARS-CoV-2 were included in this study.11 Inclusion criteria were hospitalized, adult (age ≥18 y) patients with either a positive polymerase chain reaction test result for SARS-CoV-2 during their admission or within the preceding 14 d or diagnosed with COVID-19. Our patient selection criteria were consistent with approaches applied by large United States networks to extract COVID-19 illness diagnoses (eg, COVID-19, respiratory failure, or pneumonia) or related signs or symptoms (eg, cough, fever, dyspnea, vomiting, diarrhea) by using ICD-10 diagnosis codes in combination with COVID-19 test results and identifiers.12 The exclusion criteria were hospitalizations that were missing discharge status or missing patient age.
Twenty-five hospitals that serve rural, suburban, and urban settings within an integrated health system contributed data to this analysis. As an integrated health system, participating hospitals were directly linked to resource- and knowledge-sharing via an institutionally embedded logistic distribution center and system acute care council networks, which were matured before the pandemic. For example, regional forecasting of hospitalization demand paired with just-in-time institutional oxygen delivery equipment (eg, high-flow nasal cannula [HFNC], invasive mechanical ventilation) among hospitals occurred at least daily to enable anticipatory oxygen device inventory readiness for all sites throughout the pandemic. This approach enabled clinicians to select from the full array of oxygen methods to match patient acuity at their discretion, so oxygen method selections were neither affected by equipment scarcity nor tethered to medication delivery methods.
Variables were defined by the international multi-center VIRUS COVID-19 registry,11 with core elements drawn from the World Health Organization templates.13 Comorbidities were classified by using validated ICD-10 algorithms14 for the Charlson comorbidity index.15 We accessed the electronic health record (Epic Systems, Verona, Wisconsin) by using structured query language to obtain all variables. Query development was an iterative process between the principal investigator (VD) and the programming team, which focused on continuous quality improvement.16 To validate the integrity of data, consecutive subsets of cases were verified by using manual data abstraction procedures. Study data were collected and managed by using REDCap17 electronic data capture tools hosted at Baylor Scott & White Health. Consecutive cases are reported at the hospitalization level. Data elements in the VIRUS COVID-19 registry have been previously reported.11 For this report, data elements include demographics, comorbidities, and hospitalization characteristics with daily oxygenation support methods. Oxygenation support methods were classified as low (ie, conventional nasal cannula), intermediate (ie, HFNC, noninvasive ventilation) and advanced respiratory support (ie, invasive ventilation).
Statistical Analysis
Descriptive results are reported by using medians (quartile 1 to quartile 3) for continuous variables, frequencies and percentages for categorical variables, and data visualizations via box plots and histograms. Proportions are computed on known values; thus, missing data were removed from denominators. We tested for differences between groups by using the Cochran-Armitage test for trend, chi-square, Fisher exact, Wilcoxon rank sum, and Kruskal-Wallis tests, as appropriate. Statistical analyses were conducted by using SAS Enterprise Guide Version 9.4 (SAS Institute, Cary, North Carolina) software. All statistical tests were 2-sided, with a statistical significance level set at P < .05.
Results
There are 26 adult acute care hospitals within the Baylor Scott & White Health system in Texas with a shared electronic health record. One of the facilities was excluded due to incomplete data reporting. The 25 participating hospitals represent diversity in trauma-level designation (16% levels 1 and 2, 16% level 3, 68% level 4), teaching hospitals (4), and hospital size (11 hospitals with <100 beds; 8 hospitals with 101–200 beds; 3 hospitals with 201–399 beds; and 3 hospitals with >400 beds), which spanned a 36,000 square mile service area (Fig. 1). Bed counts ranged from 15 to 914 beds per hospital (mean ± SD, 150 ± 188). Of a total of 4,457 licensed beds, almost half (46%) represent non-teaching hospitals.
Geographic distribution of hospitals (no. = 25).
Among 18,267 hospitalization records of patients with COVID-19 at 25 hospitals during the study period, 2,759 hospitalizations with unknown discharge status and 147 patients < 18 y old were excluded, thus 15,361 hospitalizations met eligibility criteria and were included in analyses. Subject characteristics stratified by the highest level of oxygenation support required during hospitalization are presented in Table 1. The overall sample was predominantly male (53.0% [n = 8,145]) and white (74.1%, [n = 11,383]). Significant differences were observed across the age groups in every characteristic examined, including lower body mass index in older adults and fewer comorbidities in younger adults. The proportion of hospitalizations for respiratory failure increased from younger (18–49 y) to older age groups (50 y and older) (5.7% vs. 11.0% vs 12.7%, P <.001). (Supplementary Table 1 [see the supplementary materials at http://www.rcjournal.com]). Similarly, the proportion of ICU admissions was lowest for younger adults, and their ICU and hospital survival rates were higher than for adults ages ≥ 65 y (Supplementary Tables 2 and 3 [see the supplementary materials at http://www.rcjournal.com]).
Subject Characteristics by Highest Level of Oxygen Support
Methods of oxygenation support required by adults hospitalized with COVID-19 were stratified by day (Fig. 2) and by waves of hospital admissions (Supplementary Fig. 1 [see the supplementary materials at http://www.rcjournal.com]) to present distributions over time. On the first day of hospitalization, 44.0% required basic oxygen support (ie, nasal cannula), 12.9% required intermediate respiratory support (ie, HFNC, noninvasive ventilation), and 2.8% required advanced oxygen delivery methods (ie, invasive ventilation). By day 4 of hospitalization, oxygen support methods reflected higher proportions of intermediate and advanced respiratory support, from a combined 15.7% at hospital admission to 34% of subjects still admitted. Overall, the proportion of high-acuity oxygenation support demand increases over time, which reflects a decreasing denominator of subjects with longer hospitalization durations (Fig. 3). After 4 weeks of acute illness, 69.9% of adults hospitalized with COVID-19 required intermediate or advanced respiratory support, with similar proportions that extended to hospitalizations that lasted ≥ 6 weeks.
Oxygen support by day of hospitalization. Overall < 3% missing daily data: 0% on admission day, 3% (n = 445) on day 1, 3.1% (n = 450) on day 2, 2.4% (n = 304) on day 3, 2.1% (n = 223) on day 4, 2.4% (n = 213) on day 5, 2.4% (n = 178) on day 6, and 3.1% (n = 186) on day 7.
Oxygen support by week of hospitalization. Overall < 3% missing weekly data: 0% on week 1, 1.2% (n = 74) on week 2, 1.4% (n = 30) on week 3, 2.9% (n = 29) on week 4, 2.2% (n = 11) on week 5, 1.9% (n = 5) on week 6, and 2.6% (n = 4) on week 7.
Discussion
In this diverse 25-hospital consecutive case series that describes all COVID-19 hospitalizations irrespective of oxygenation requirements and acuity, we observed that oxygen support was required in 60% of adults on the day of hospital admission, which reflected varied initial presentations of hypoxemia. At 1 week, >70% of subjects hospitalized with COVID-19 required oxygen support (Fig. 2). Intermediate respiratory support (ie, HFNC or noninvasive ventilation) represents an increasing proportion of oxygenation method utilization across the time course of COVID-19 hospitalizations, increasing from 13% of hospitalized adults at admission to 32% at 1 week, and to 70% of adults hospitalized at 1 month.
Similarly, in prolonged hospitalizations that lasted 5, 6, or 7 weeks, invasive ventilation utilization represents the oxygenation method in > 50% of subjects (Fig. 3). We illustrate that, whereas the number of subjects decreased over time due to hospital discharge or death, the prevalence of daily oxygenation support remained high (78–88%) when hospitalizations were > 6 d. The utilization of oxygenation methods influenced patient outcomes, clinician work load, and health system resourcing. To the best of our knowledge, this was the first multi-site study conveying basic, intermediate, and advanced oxygenation support requirements as a time course and inclusive of hospitalization stays > 2 weeks.
Despite the intuitive nature of day-by-day oxygenation support methods required throughout hospitalizations across all severities of COVID-19 illness, few cohort analyses demonstrated the persistence of high oxygen demands that extend into chronic critical illness (>6 d) with oxygen methods needed during very long hospitalizations (eg, >2 weeks). Thus far, oxygen support requirements of hospitalized adults with COVID-19 represent a patchwork of populations, geographies, and exposures, with most reports aggregating oxygenation methods used at the hospitalization level, which mask the system-level epidemiology of demand. For example, in the few very large cohort studies that described >10,000 hospitalizations, the methods of oxygen support during hospitalization included reports of 17.0–55.0% of subjects who required invasive ventilation, as few as 2.8% who required HFNC, 56.0% required nasal cannula, and reports of 13.1% of the hospitalized patient population not requiring any oxygen support.5,18
In contrast to these reports, we present comprehensive epidemiologic profiles of oxygenation methods used by day and by week during hospitalization across all severities of COVID-19. We demonstrate that a clinically important proportion of 44% of hospitalized adults across 25 hospitals relied on low levels of oxygenation support (conventional nasal cannula) at admission. In general, the diversity of oxygenation support methods needed by subjects hospitalized with COVID-19 were relatively similar during the first 3 d of hospitalization, with fewer than 1 in 4 who required intermediate or advanced oxygen support (HFNC, noninvasive ventilation, invasive ventilation) each day. In contrast, half the adults hospitalized with COVID-19 required intermediate or advanced oxygen support on day 14. Generally, patients with chronic critical illness require longer hospitalizations and thus need more resources allocated for in-patient care.7,19,20 For the subset of subjects with hospitalizations that persist to 6 weeks, >70% were relying on intermediate and advanced oxygen support methods.
Trends in oxygenation support method utilization over the daily and weekly time course of COVID-19 hospitalizations can contribute to high-level resource allocation and health-care delivery planning to help inform long-range equipment and staffing needs. In addition, all patients with a COVID-19 diagnosis require equipment and clinician time-related resources, including personal protective equipment21 and visitation restrictions, which alter the usual frequency and communication patterns with family members.22,23 Overall, care processes are affected across the continuum of daily oxygenation method utilization, with time costs and resources tied to COVID-19 hospitalizations. The risk factors associated with ICU admission and in-hospital mortality specific to COVID-19 are aligned with previous reports,5,18 but the inclusion of oxygen support methods by day and inclusive of prolonged hospitalizations contributes to conveying the prevalence pattern over time.
Daily oxygenation method utilization illustrates the clinical and economic burden of COVID-19 hospitalizations. The high proportion of subjects who required HFNC, noninvasive ventilation, and invasive ventilation oxygen support throughout the time course of COVID-19 hospitalizations corresponds with high-intensity clinician-delivered care by nurses, respiratory therapists, and physicians for near-continuous monitoring and frequent assessments of patient condition and equipment (eg, settings, tubing, documentation).24 Unlike previous pandemics, COVID-19 has been associated with resource allocation burdens specific to oxygen scarcity.25 The oxygen flows for HFNC (10–60 L/min) are high compared with conventional nasal cannula (0–6 L/min) and compared with advanced oxygen support via invasive ventilation (20–30 L/min).26,27 At the integrated health system level, the overall consumption of oxygen used for patient care increased substantively, with monthly oxygen volume utilization up to 220% higher during the pandemic, representing more than three times the amount consumed during the same month one year prior to the pandemic (volume purchased and used, January 2020 vs January 2021) (personal communication, May 2022, T. Williams-Dennis, Supply Chain Director). Our analysis of daily oxygenation methods needed by adults hospitalized with COVID-19 contributes to conveying the scope of oxygen consumption specific to COVID-19.
Focused evaluations of hospital-level variation in the context of COVID-19 describe mortality differences despite similar interventions (eg, renal replacement therapy, invasive ventilation).5,28,29 The data representation of intra-hospital processes of care drawn from small (<100 hospital beds) and medium (100–399 beds) hospitals, with varied teaching and trauma designations paired with large and very large academic medical centers, are important to presenting a balanced perspective of care delivery mechanisms used. In very large cohort studies (>10,000 hospitalizations), hospital characteristics specific to size, case volume, and teaching status are more often reported, but the inclusion of the day-by-day intra-hospital process of care variables are rare. For example, in 11,721 hospitalizations derived from commercial insurance claims from 245 hospitals, the proportion of subjects who received invasive ventilation reported by academic status (16% of subjects in academic centers vs 19% in non-academic centers)18 but is not inspected for variability.
Changes in clinical practice over time may have influenced outcomes of subjects hospitalized with COVID-19 across waves. Reports from Spain describe differences in medication management between waves with increased use of HFNC as a first-line therapy in waves 2 and 3 (July 1, 2020 – February 28, 2021), while describing no significant association found between COVID-19 waves and mortality.30 Beyond pharmacologic management variations between waves, the consideration of controversial practice pattern differences of earlier intubation during the first wave of COVID-19 hospitalizations reported in critical care settings merits attention in the absence of large-scale empirical evidence.31,32 In our sample of 15,361 consecutive hospitalizations within an integrated health system, fewer subjects required oxygen support at hospital admission in wave 1 (March 1, 2020 – June 30, 2020) compared with waves 2 and 3 (July 1, 2020 – February 28, 2021). In the later waves 2 and 3, overall oxygen method utilization reflects higher proportions of subjects managed by using HFNC during the first week of hospitalization. These results add a broader perspective of hospitalizations inclusive of all care units.33 Further analyses that describe practice patterns and assessing outcomes associated with delayed intubation specific to COVID-19 are warranted.
Future research can contribute beyond prediction modeling to include subtype discovery and trajectory analyses34,35, about oxygenation methods for respiratory failure (eg, HFNC) and to advance knowledge about clinically relevant catchment-area factors (ie, facility volumes, trauma center status) that inform non–COVID-1936 and COVID-19 care delivery.29 Furthermore, examining prolonged hospital and ICU length of stay contributes to controversial topics of futility, palliative care services, hospital-level capacity, and triaging decisions in the setting of resource constraints inherent to a pandemic of the COVID-19 scale.
Limitations
Our first limitation is due to the study design, because we were unable to make any causal inferences, and there is potential residual confounding by unmeasured variables. Second, although we report that 40% of adults did not require oxygen support on the day of hospital admission, half were hospitalized at 4:00 pm or later, which reflects ≤ 8-h periods and not reflective of full calendar days. Oxygen support required within the first 24 h of hospitalization was thus distributed for most subjects across the first and second day of hospitalization due to reporting by calendar day increments. Ultimately, 80% of adults received oxygen support during hospitalization (Table 1). Third, the utilization of concomitant practices, such as prone positioning or inhaled vasodilators, likely contributed to the selection of oxygen delivery devices, which affected the interpretation of our results. Fourth, our population-based approach is reflective of a single large integrated health system, with representation specific to the southwest region of the United States. The reporting of oxygenation methods as a static time course during hospitalization could be viewed as a weakness in that the dynamic surge capacity demands are not directly reflected, while also serving as a strength, by offering a broad epidemiologic view of these important intra-hospital processes of care. Overall, cohorts from other health systems and regions are needed to confirm findings.
Conclusions
This consecutive case series of >15,000 hospitalizations describes daily oxygenation measures as intra-hospital processes of care that conveys clinical characteristics and COVID-19 implications, which are traditionally masked in cohort studies. We present epidemiologic profiles of oxygenation methods by day and by week to demonstrate that 44% of hospitalized adults across 25 hospitals relied on basic oxygenation support at admission, with intermediate and advanced oxygen support over the course of hospitalization required by 34% of subjects by day 4 of hospitalization. In prolonged hospitalizations, 70% of adults required intermediate or advanced respiratory support at 4 and 6 weeks. This study serves as a comprehensive system-level lens into the epidemiology of demand for oxygen specific to adults hospitalized with COVID-19. Knowing about the prevalence of a specific disease can help with understanding the demands on health services, thus these findings underscore the importance of the intra-hospital organization of care in contributing to epidemiologic insights, health system science, and health-care policy and planning.
Footnotes
- Correspondence: Valerie Danesh PhD, Center for Applied Health Research, Critical Care, Baylor Scott & White Health, 3434 Live Oak St, Dallas, TX 75204. E-mail: Valerie.Danesh{at}BSWHealth.org
The location of the institution was Baylor Scott & White Health, Texas.
This work was partially funded by the Cardiovascular Research Review Committee of the Baylor Healthcare System Foundation, Society of Critical Care Medicine, and the Gordon and Betty Moore Foundation.
The authors have disclosed no conflicts of interest.
Supplementary material related to this paper is available at http://www.rcjournal.com.
ClinicalTrials.gov registration NCT04323787.
- Copyright © 2023 by Daedalus Enterprises
