Antimicrobial use before and during COVID-19: data from 108 Veterans Affairs medical centers.
Aug 7, 2024DOI: 10.1017/ash.2024.352Journal ArticleASHE
Written by: Matthew Bidwell GoetzTina WillsonMichael A RubinVanessa W StevensChristopher J Graber
Find these authors on Roon
MG
Dr. Matthew Goetz
Antibiotic Stewardship · Greater Los Angeles VA
CG
Dr. Christopher Graber
Antibiotic Stewardship · Greater Los Angeles VA
Summary
During the first two months of the COVID-19 pandemic at 108 VA hospitals, antibiotic use in infected persons was 30% higher than in uninfected persons; use quickly decreased and was 9% less than in uninfected persons in 2022.
Introduction
Prompted by concerns about potential bacterial coinfection and preliminary reports of the benefit of azithromycin in the treatment of COVID-19, during the early months of the COVID-19 pandemic 65–80% of patients with confirmed or suspected COVID-19 received antibiotic therapy.1–5 Following data demonstrating the lack of effectiveness of azithromycin and the low prevalence of documented community onset and secondary bacterial coinfection among persons with COVID-19,6 the rate of antibiotic administration among persons with COVID-19 diminished in the latter part of 2020, a trend that has continued.2,7,8 However, no large-scale studies have compared the long-term patterns of antibiotic use in persons with and without COVID-19. In this paper, we compare inpatient antibiotic use among hospitalized persons with and without COVID-19 as clinical care evolved from March 2020 to October 2022 within Veterans Affairs (VA) healthcare facilities.
Methods
Data on antibacterial use, patient days present, and COVID-19 care for acute care units defined as inpatient admission, observation, intensive care and acute care medical or surgical units in 108 VA facilities were extracted from the Corporate Data Warehouse through the VA Informatics and Computing Infrastructure; facilities which provide limited acute inpatient services (e.g., VA complexity level three facilities), were excluded.9 Days of therapy (DOT) per 1,000 days present (DP) were calculated and stratified by Centers for Disease Control and Prevention (CDC) defined standardized antimicrobial administration ratio (SAAR) antibiotic classes (Supplementary Table 1).10COVID-19 diagnoses were identified using the VA COVID-19 Shared Data Resource. This resource contains information on all veterans with a confirmed laboratory diagnosis of SARS-CoV-2 infection within the VA and those who were tested outside the VA with a VA clinical note confirming the diagnosis. The VA COVID-19 Shared Data Resource contains extensive demographic, clinical, pharmacologic, laboratory, vital sign, and clinical outcome information derived from multiple validated sources, including the VA Corporate Data Warehouse and the VA electronic medical record. The analysis plan for this study classified patients as having COVID-19 based on the earliest positive laboratory-based diagnosis or evidence of positivity in clinical notes extracted using Natural Language Processing. Patients with a positive COVID-19 test obtained on hospital days 1–5 were assumed to have been COVID-19 infected at admission. Individuals were considered to have acute COVID-19 for the first 30 days after their diagnostic test. Persons were considered to have a new episode of COVID-19 if a subsequent test was positive ≥90 days after the initial test.
Because the large sample size could result in the achievement of statistical significance despite non-meaningful differences, an a priori decision was made that differences <5% in the characteristics of patients with and without COVID-19 were not meaningful.
This study was approved by the VA Central Institutional Review Board.
Results
From January 2016 through February 2020, there were 2.3 million hospitalizations among 1.05 million unique patients seen at 108 inpatient VA facilities (Table 1). Subsequently, from March 2020 through October 2022 there were 1.1 million admissions of persons without COVID-19 and 84,000 admissions of persons with COVID-19 among 620,000 and 79,000 patients, respectively; 6,320 patients had ≥1 more admission associated with a new episode of COVID-19.
Table 1.Patient characteristics 1/1/2016–10/31/2022Jan 2016–Feb 2020Mar 2020–Oct 2022Non-COVID-19Mar 2020–Oct 2022COVID-19Invariant dataN = 1,046,290 distinct patientsN = 619,658 distinct patientsN = 78,717 distinct patientsSex (N,%)Missing20.0%10.0%10.0%Female71,5106.8%44,6647.2%4,8186.1%Male974,77893.2%574,99392.8%73,89893.9%Race (N,%)American Indian or Alaska Native7,5070.7%4,6400.7%6980.9%Asian5,6760.5%3,8670.6%5530.7%Black or African American223,32321.3%144,43323.3%20,56226.1%Two or more races8,9630.9%5,5000.9%6880.9%Native Hawaiian or Other Pacific Islander7,3030.7%4,6280.7%6700.9%White753,54672.0%427,65969.0%51,27365.1%Missing/Uknown39,9723.8%28,9314.7%4,2735.4%Ethnicity (N,%)Hispanic or Latino67,0816.4%41,8906.8%6,7108.5%Not Hispanic or Latino955,66291.3%553,35189.3%68,71387.3%Missing/unknown23,5472.3%24,4173.9%3,2944.2%Admission specific data2,264,797 admissions1,129,736 admissions84,452 admissionsAge (Mean SD)67.0312.9067.913.068.014.3Charlson comorbidity index score (Mean, SD)3.73.13.83.23.23.0Charlson comorbidities (N, %)Myocardial infarction218,4559.6%121,01910.7%6,5147.7%Congestive heart failure584,93325.8%308,40027.3%18,13721.5%Peripheral vascular disease452,10220.0%241,70821.4%13,35515.8%Cerebrovascular disease336,61914.9%179,01815.8%11,57713.7%Dementia288,11612.7%154,80213.7%12,24814.5%Chronic pulmonary disease765,16333.8%359,12131.8%22,69626.9%Rheumatologic diseases50,4572.2%27,1782.4%1,8742.2%Peptic ulcer disease65,0102.9%30,0082.7%1,4131.7%Mild liver disease136,2216.0%94,0988.3%5,1586.1%Diabetes without complications313,47513.8%149,01713.2%12,37014.6%Jan 2016–Feb 2020Mar 2020–Oct 2022Non-COVID-19Mar 2020–Oct 2022COVID-19Admission specific data2,264,797 admissions1,129,736 admissionsN84,452 admissionsDiabetes with complications655,53328.9%343,60130.4%24,11628.6%Hemiplegia or Paraplegia59,0202.6%31,9832.8%2,1712.6%Renal disease720,64131.8%394,63834.9%25,58130.3%Any malignancy438,44119.4%223,17619.8%12,95215.3%Moderate-severe liver disease186,6118.2%81,4627.2%4,2785.1%Metastatic solid tumor100,6774.4%58,8445.2%2,5923.1%HIV or AIDS24,2131.1%12,6531.1%8201.0%Drug resistance In past year (N,%)MRSA colonization or infection170,7587.5%57,4055.1%3,2233.8%VRE13,1240.6%5,2300.5%2810.3%ESBL12,4190.5%6,9520.6%4200.5%Admission type (N, %)Through ED1,550,03268.4%842,16674.5%74,08087.7%Severity of illness (N,%)ICU careEver397,05117.5%198,97617.6%24,36828.9%Within first 24 hours329,29014.5%164,73914.6%18,83922.3%PressorsEver23,9951.1%13,7211.2%2,6033.1%Within first 24 hours7,5110.3%4,2710.4%3170.4%Admitting Service (N,%)Medical1,746,05477.1%849,51375.2%68,74281.4%Surgery518,74322.9%240,92021.3%8,2239.7%Unknown39,3033.5%7,4878.9%Prior hospitalization within 90 days (N,%)652,30228.8%319,13928.2%15,21718.0%Long-term care exposure (N,%)Current (within prior 7 days)9,6290.4%4,0080.4%3450.4%Within prior 90 days46,0252.0%15,1081.3%1,0401.2%Infectious diagnosis* (N,%)1,144,89450.6%591,23452.3%45,27353.6%COVID-19 diagnosis (N,%)1080.0%14,0181.2%74,35088.0%28-day mortality109,9304.9%59,4785.3%10,31312.2%Jan 2016–Feb 2020Mar 2020–Oct 2022Non-COVID-19Mar 2020–Oct 2022COVID-19Admission specific data2,264,797 admissions1,129,736 admissionsN84,452 admissionsLength of stayTotal days (Mean, SD)5.47.36.211.29.213.3Total days(Median, IQR)42–642–663–11ICU days (Mean, SD)0.83.10.83.92.46.5ICU days(Median, IQR)00–000–000–2All data are collected at the time of admission.Drug resistance was defined by report in the VA laboratory information system in the prior year indicating the detection of a methicillin-resistant Staphylococcus aureus (MRSA), a glycopeptide-resistant Enterococcus spp. (VRE) or Enterobacteriaceae isolate that produced an extended-spectrum beta-lactamase.*Infection codes were determined as previously reported.(32)
Patient characteristics 1/1/2016–10/31/2022
All data are collected at the time of admission.
Drug resistance was defined by report in the VA laboratory information system in the prior year indicating the detection of a methicillin-resistant Staphylococcus aureus (MRSA), a glycopeptide-resistant Enterococcus spp. (VRE) or Enterobacteriaceae isolate that produced an extended-spectrum beta-lactamase.
Infection codes were determined as previously reported.(32)
In comparison to patients hospitalized during the pre-COVID-19 period, patients admitted without diagnoses of COVID-19 from March 2020 to October 2022 were more often admitted through the emergency department but did not differ in other demographic characteristics (sex, age, or race and ethnicity), comorbidities, severity of illness, mortality, or median length of stay. In contrast, compared with period-matched non-COVID-19 patients, patients with COVID-19 were more likely to be admitted through the emergency department or require ICU-level care, less often had prior diagnoses of congestive heart failure, peripheral vascular disease, or renal disease, less often received care on a surgical service, and were less often hospitalized in the prior 90 days. Patients with COVID-19 also had higher 28-day mortality.
Mean system-wide rates of antibiotic use declined from 575 DOT/1,000 DP in 2016 to 538 DOT/1,000 DP in 2019 (Figure 1). In contrast during the first two months of the COVID-19 pandemic, the overall rate of antibiotic use for all patients increased to 588 DOT/1,000 DP; rates for patients with and without COVID-19 were 745 and 572 DOT/1,000, respectively (Figure 2, panel F). Subsequently, rates of antibiotic use quickly declined for patients both with and without COVID-19 such that antibiotic use for 2020 was less than in prior years; this downward trend continued in 2021 and 2022 (Table 2). Later surges in COVID-19 admissions during periods of peak transmission of the Delta and Omicron variants were not accompanied by substantial increases in total antibiotic use in persons with or without COVID-19 infection (Figures 1 and 2).
Figure 1.The bars represent total monthly antibiotic use, shown as Days of Therapy per 1000 Days present, for patients hospitalized on acute care and intensive care units at all included VA facilities. The solid, dashed and dotted lines represent the number of monthly admissions for all patients and patients without and with acute COVID-19, respectively.
The bars represent total monthly antibiotic use, shown as Days of Therapy per 1000 Days present, for patients hospitalized on acute care and intensive care units at all included VA facilities. The solid, dashed and dotted lines represent the number of monthly admissions for all patients and patients without and with acute COVID-19, respectively.
Figure 2.In each of the panels, the dotted and solid lines represent monthly antibiotic use by CDC NHSN antibiotic class, for hospitalized patients with and without acute COVID-19.
In each of the panels, the dotted and solid lines represent monthly antibiotic use by CDC NHSN antibiotic class, for hospitalized patients with and without acute COVID-19.
Table 2.Mean antibiotic use (DOT/1,000 DP) versus patient-level COVID-19 statusPeriodNon-COVID-19COVID-19AllImmediately before, during and after onset of COVID-19 pandemicNovember 2019—February 2020534NA534March 2020—April 2020572745588May 2020—December 2020512534515Yearly ratesJanuary 2019—December 2019NANA539January 2020—December 2020NANA528January 2021—December 2021491497492January 2022—October 2022475432470
Mean antibiotic use (DOT/1,000 DP) versus patient-level COVID-19 status
Evaluation of CDC National Health Safety Network (NHSN) antibiotic classes showed that the use of broad-spectrum agents used for community onset infections (e.g., ceftriaxone) and antibiotics not falling within one of the CDC SAAR classes (e.g., azithromycin) was greater and the use of narrow spectrum ß-Lactam agents was less in persons with COVID-19 than in those without COVID-19 with the largest differences being in the first few months of the pandemic. There were no consistent differences in the use of Resistant Gram Positive agents or Broad-Spectrum agents for hospital-onset infections in persons with and without COVID-19 (Figure 2).
To assess whether increased antibiotic use in patients with COVID-19 was due to therapy of community onset as opposed to healthcare-associated infections, we compared rates of antibiotic use during the first ten days of hospitalization with antibiotic administration on hospital days 11–30. In the first four months of the COVID-19 pandemic, the rate of antibiotic therapy during the first ten days of hospitalization was 733 DOT/1,000 DP for patients with COVID-19 vs 548 DOT/1,000 DP for patients without COVID-19. In contrast, during hospital days 11–30, the rate of antibiotic therapy was similar in COVID-19 patients (526 vs 571 DOT/1,000 DP for patients with and without COVID-19, respectively. Patients with a second or later episode of COVID-19 had similar antibiotic use patterns to patients with their first COVID-19 episodes (data not shown).
Finally, we assessed facility-level variances in antibiotic use for persons with and without COVID-19 (Table 3 and Figure 3, Panel A). In the four months prior to the pandemic, the median facility antibiotic use was 537 DOT/1,000 DP. In the first two months of the pandemic median use increased to 591 DOT/1,000 DP for all patients; rates for patients with and without COVID-19 were 739 and 574 DOT/1,000 DP. During the remaining 8 months of 2021, the median rate was 504 DOT/1,000 DP for all patients, 494 for patients with COVID-19, and 508 DOT/1,000 DP for patients without COVID-19.
Median (IQR) facility-level antibiotic use versus patient-level COVID-19 status
Figure 3.Panel A shows the interquartile range (25%–75%) of facility-level antibiotic use at all included VA facilities. Panel B shows the interquartile range of the facility-specific difference between antibiotic use in persons with and without acute COVID-19. Values greater than zero represent instances where antibiotic use was greater in persons with COVID-19 than in persons without COVID-19.
Panel A shows the interquartile range (25%–75%) of facility-level antibiotic use at all included VA facilities. Panel B shows the interquartile range of the facility-specific difference between antibiotic use in persons with and without acute COVID-19. Values greater than zero represent instances where antibiotic use was greater in persons with COVID-19 than in persons without COVID-19.
After the first two months of the pandemic, the median within facility-level rate of antibiotic use in persons with COVID-19 fell sharply and was less in patients with COVID-19 than in uninfected individuals (Table 3 and Figure 3, Panel B). However, in 2022 antibiotic use remained higher in persons with COVID-19 in 25% of facilities.
Discussion
We previously reported that the rate of inpatient antibiotic use in VA facilities during early months of the COVID-19 pandemic was substantially higher than in previous years, thereby reversing year-to-year declines observed since 2015.11 We now demonstrate that rates of antibiotic use subsequently decreased substantially even as the proportion of all hospital days due to care of patients with COVID-19 increased during the Delta and Omicron surges in hospital admissions.
Increased antibiotic use during the first months of the COVID-19 pandemic was driven by high use of ceftriaxone and azithromycin among patients with COVID-19. Lesser increases were seen in the use of broad-spectrum agents for hospital-onset infections; no meaningful change occurred in the use of agents for resistant gram-positive pathogens (e.g., anti-methicillin-resistant Staphylococcus aureus (MRSA) therapy). Antibiotic use during hospital days 11–30 (i.e., for healthcare-associated infections) in patients with COVID-19 was similar to that in patients without COVID-19.
Our data expand upon the findings of others by critically evaluating the differences in antibiotic use among persons with and without COVID-19 during the pandemic. Kim et al. reported that among 1.1 million adults discharged with a COVID-19 diagnosis in 711 hospitals, the overall frequency of antibiotic use for non-critically ill hospitalized patients with COVID-19 decreased from 71% in 2020 to 61% in 2021.8 At those facilities, the median peak adjusted DOT/1,000 DP among persons with COVID-19 was over 1,300 DOT/1,000 DP and decreased by 215 DOT/1,000 DP from 2020 to 2022 with a median rate of over 600 DOT/1,000 DP in 2022.
Using data from the NHSN, O’Leary et al. demonstrated that the rate of total antibiotic use by all hospitalized patients regardless of COVID-19 status peaked in April 2020 and rapidly returned to baseline; during subsequent surges of COVID-19 smaller, transient increases in total antibiotic use were observed.12 However, neither of these studies directly compared rates of antibiotic therapy for persons with and without COVID-19.
Within the VA, the median facility rate of antibiotic use among persons with COVID-19 was 30% higher than in uninfected persons in March–April 2020. Subsequently, antibiotic use decreased such that total DOT/1,000 DP was 2% less in 2020 than in 2019, a trend that continued through 2022. Similar to O’Leary et al., who found a 30% decrease in inpatient days in April 2020,12 we found that admissions for persons without COVID-19 were 36% less in April 2020, when the peak in antibiotic use was observed, than in January 2020. Thus, it is likely that different patterns of seeking care and acuity of illness contributed to increased use of antibiotics in persons not known to have COVID-19.
This study has several limitations. First, the participating institutions were all VA facilities, which may limit generalizability. Although we assessed aggregate patient data at the facility level, we did not analyze patient-level data and or the appropriateness, indication, or duration of therapy; furthermore, we did not correct for differences in patient characteristics at the individual or facility level. We also may not have captured all initial SARS-CoV-2 diagnose for patients had initial positive SARS-CoV-2 tests at non-VA locations.
Many factors, including guidelines recommending against routine antibiotic use,13,14 increased awareness of low rates of coinfection,6 codification of risk factors for bacterial infection,7,15 increased willingness to obtain lower respiratory tract specimens for culture with better infection control practices, and changes in disease severity due to increased levels of immunity and changes in viral virulence,16 are likely to have contributed to decreased antibiotic utilization after the initial months of the COVID-19 pandemic. Nonetheless, our findings point to the resiliency of antimicrobial stewardship within VA, which has been a product of several factors: the establishment of an Antimicrobial Stewardship Task Force in 2011,17 organizational directives establishing and updating recommendations for antimicrobial stewardship program staffing, system-wide medication use evaluations that identify opportunities for improvements in antimicrobial use,18–20 development of regional antimicrobial stewardship collaboratives, and strong implementation science engagement and utilization of informatics-based tools.21
However, despite system-wide quality improvement initiatives, policies and staffing guidelines, variances in antibiotic use persist in the VA. Reported contributors to variation within the VA to include differing levels of staffing for stewardship programs,22,23 availability of infectious diseases specialists, variations in “antibiotic prescribing etiquette,”24,25 and implementation of nationally recommended programs.26–28 Further adding to the variations in stewardship practices during the period in question was the impact of COVID-19 on infection control and antimicrobial stewardship programs.29,30
In summary, the rapid normalization of antibiotic use in persons with COVID-19 over the course of the pandemic and the continued downward trend in antibiotic use throughout the VA is reassuring, especially since many antimicrobial stewards’ usual activities were co-opted by pandemic-related tasks.29,30 However, substantial variations in antimicrobial use persisted across institutions. While the most recent rates of antibiotic use were substantially lower in persons with COVID-19 than in non-COVID-19 patients, ongoing interfacility variance in antibiotic use for persons with COVID-19 suggests that despite guidelines to the contrary and the low prevalence of community-onset infections in persons with COVID-196,13,14 substantial opportunities remain for improved antibiotic stewardship in some facilities. Further research will allow for ongoing refinement of antimicrobial stewardship goals and priorities within VA over the next several years.31
Supporting information
Goetz et al. supplementary material
CG
Dr. Christopher Graber
Infectious Disease · Antibiotic StewardshipVA Greater Los Angeles Healthcare Systemchristophergraber