The coronavirus disease 2019 (COVID-19) pandemic and resultant mitigation strategies, social distancing, and mask mandates, have had a substantial impact on the epidemiology of other pediatric infections. Reference Hatoun, Correa, Donahue and Vernacchio1–Reference Haddadin, Schuster and Spieker3 At the same time, primary-care delivery transformed with the rapid increase in use of telemedicine. Reference Fiks, Jenssen and Ray4 More recently, children returned to in-person school for the fall semester of 2021 during a surge in COVID-19 cases due to the severe acute respiratory coronavirus virus 2 (SARS-CoV-2) δ (delta) variant and an atypical summer–fall peak of respiratory syncytial virus (RSV). Reference Brown, Vostok and Johnson5,Reference Agha and Avner6 Antibiotic use is an important contributor to the development of antimicrobial resistance, and ∼30% of outpatient antibiotic prescribing is unnecessary. 7,Reference Fleming-Dutra, Hersh and Shapiro8 Understanding changes to antibiotic prescribing due to nontraditional viral respiratory patterns and novel care-delivery strategies is valuable for understanding and implementing outpatient antibiotic stewardship efforts.
In many outpatient settings serving children, the COVID-19 pandemic resulted in rapid implementation of telehealth visits. Reference Fiks, Jenssen and Ray4 Virtual encounters, while offering numerous benefits, also have the potential to impact the quality of antibiotic prescribing, especially for conditions that benefit from physical exam and laboratory testing. Reference Suzuki, Marra and Hasegawa9 Early studies showed significant reductions in antibiotic prescribing associated with the COVID-19 pandemic in the United States, Reference Katz, Spencer, Zhang and Banerjee10–Reference Buehrle, Wagener, Nguyen and Clancy12 and a recent study found sustained reductions in antibiotic prescribing in pediatric primary-care offices through June 2021, primarily driven by reductions in respiratory encounters. Reference Dutcher, Li, Lee, Grundmeier, Hamilton and Gerber13 However, less is known about alternative outpatient settings, the influence of more recent changes in viral respiratory patterns (RSV and COVID-19 variants), return to in-person school, variations by patient or visit characteristics, and appropriateness of antibiotic prescribing. The primary objective of this study was to describe changes in antibiotic prescribing during the COVID-19 pandemic across a range of outpatient site and provider types. Secondary objectives were to evaluate changes in visit types, including (1) visits where antibiotics are sometimes, always, or never appropriate; (2) respiratory visits, the most common indications for outpatient antibiotic prescribing in children; and (3) telehealth visits and associated in-person follow-up visits.
Methods
Study design and setting
This retrospective control study utilized electronic health record (EHR) data from pediatric outpatient visits at Norton Healthcare, a large healthcare system serving the greater area of Louisville, Kentucky, and southern Indiana. Outpatient visit data were included for children aged 0–19 years from 27 family medicine or primary-care clinics, 27 pediatric clinics, and 26 urgent-care or retail clinics. In March 2020, 3 academic pediatric clinics were added to Norton Healthcare; in addition to expanding pediatric outpatient care, these are the only practice sites that include medical trainees. No formal outpatient antibiotic stewardship programs or interventions were conducted at Norton Healthcare during the study period. Visits were excluded if the provider type of record did not have authority to prescribe antibiotics (eg, nurse, medical assistant, registered dietician) or for encounter types not typically associated with medication prescribing (eg, imaging, injection visit, laboratory visit, social work). Outpatient visits were defined as virtual based on encounter and visit types, as coded in EHR variables, and we did not include telephone encounters. Virtual visits were counted as having a follow-up visit if any in-person visit type (eg, hospital, office visit, imaging, laboratory visit) for the same patient occurred within 48 hours.
A state of emergency was declared in Kentucky on March 6, 2020, in response to the COVID-19 pandemic. 14 Beginning March 16–20, 2020, schools and daycare facilities were closed, and social distancing recommendations were implemented. 14 In January 2021, schools were permitted to reopen for in-person learning, and legislation required in-person classes to resume by March 29, 2021. 14 We defined a “baseline” (ie, prepandemic) period of January 1, 2019, to March 31, 2020. The “COVID-19 era” included April 1, 2020, to October 31, 2021. To calculate an absolute reduction in visits and antibiotic prescribing, we also compared corresponding periods of April–October in 2019, 2020, and 2021.
Antibiotic prescribing and diagnoses
All included outpatient visits were queried for associated antibiotic prescribing through the EHR. The primary outcome of this study was percentage of antibiotic visits, defined as any visit with an associated oral antibiotic or select intramuscular antibiotic agents commonly used for pediatric infections (Supplementary Table A). Additional secondary outcomes of interest included changes in proportion of visit types, assessed by antibiotic appropriateness, all respiratory visits, and virtual visits. Visits were further sorted by diagnostic categories using International Classification of Diseases, Tenth Edition (ICD-10) codes. Visit appropriateness for antibiotic prescribing was defined using a previously described, mutually exclusive, ICD-10 classification scheme Reference Chua, Fischer and Linder16 in which a diagnosis is classified as “always” appropriate for an antibiotic (eg, bacterial pneumonia, streptococcal pharyngitis), “sometimes” appropriate for an antibiotic (eg, acute sinusitis, acute otitis media), or “never” appropriate for an antibiotic (eg, acute upper respiratory infection, acute bronchitis). To identify a single diagnosis for each visit, modified clinical classification software categories 15 were prioritized using a similar tiered classification scheme. Classification was hierarchical such that if any of the listed medical diagnoses should always be treated with an antibiotic, the “always” diagnosis was prioritized over diagnoses for which antibiotics are sometimes or never indicated, respectively. For indications for which antibiotics are never indicated, we prioritized reporting of the following infectious categories of interest: acute upper respiratory infection, acute bronchitis (including bronchiolitis), influenza, and COVID-19. If >1 diagnosis was identified per tiered category, the first reported diagnosis field was selected. COVID-19 visits were defined using ICD-10 codes for diagnosis, exposure, screening, or personal history (ie, U071, Z20822, Z1152, Z8616, respectively). Respiratory visits of interest were identified using modified clinical classification software categories 15 and included bronchitis, COVID-19, influenza, otitis media, pharyngitis, pneumonia, sinusitis, tonsillitis, and other upper respiratory infections (eg, laryngitis). A full list of diagnosis classifications is available in Supplementary Table B.
Data collection
This study was deemed exempt from approval by the Institutional Review Board of the University of Louisville and the Norton Healthcare Research Office. Patient and visit characteristics, antibiotic prescriptions, and ICD-10 codes were extracted from the EHR database. Race and ethnicity were assigned using the primary variable collected in the EHR and were recategorized as White, Black, Hispanic, other, and unknown.
Data analysis
Descriptive statistics were used to summarize outpatient visits and antibiotic prescribing stratified by period (ie, baseline vs COVID-19 era), month, visit diagnoses of interest, and virtual visits with in-person follow-up within 48 hours. Reductions in the percentages of antibiotic visits in the COVID-19 era for respiratory visits were calculated as relative reductions compared to the baseline era.
We applied multivariable logistic regression models to identify which characteristics were associated with antibiotic prescribing in respiratory visits during the study period. The following factors were included in the analysis: period (baseline vs COVID-19 era); visit type (in-person vs virtual); sex, age group (0–2 years, 3–9 years, or 10–19 years); race or ethnicity (White, Black, Hispanic, or other); insurance (Medicaid vs private); department (family medicine, pediatrics, retail clinics, or urgent care); and provider type (physician, nurse practitioner, or trainee). Odds ratios (ORs) and their 95% confidence intervals (95% CIs) were calculated. All statistical analyses were carried out using SAS version 9.4 software (SAS Institute, Cary, NC).
Results
During the study period, 984,244 outpatient visits were included among 198,315 unique patients. From April to October 2019, there were 197,253 outpatient visits, compared to 158,497 visits in the same period of 2020 and 244,189 visits in 2021. In the same monthly periods (April–October), there were 40,140 visits during which antibiotics were prescribed in 2019, compared to 16,396 antibiotic visits in the same period of 2020. These data represent a 59.2% reduction in antibiotic prescribing from 2019 to 2020. During the follow-up period of April–October 2021, there were 32,820 antibiotic visits, an 18.2% reduction in antibiotic prescribing compared to the baseline. The percentage reduction in antibiotic prescriptions during respiratory visits during only the corresponding April–October periods were larger: 71.2% in 2020 and 23.2% in 2021.
The proportion of visits by children who were younger, non-White, and insured by Medicaid increased during the COVID-19 era, as did visits to pediatric clinics (Table 1). The percentages of antibiotic visits decreased for all patient characteristics in the COVID-19 era. Frequency of visits over time by appropriateness for antibiotics (always, sometimes, or never), and rates of antibiotic prescriptions per 1,000 visits are summarized in Figure 1a. This figure shows sharp declines in response to initial stay-at-home orders (beginning March 2020) and subsequent increases in visits beginning around March 2021.
Note. The baseline period was January 1, 2019, to March 31, 2020, and the COVID-19 era was defined as April 1, 2020, to October 31, 2021.
a Antibiotic visit percentages calculated as percentage of visits where an antibiotic was prescribed (row percentages).
Relative changes in the proportion of visits that were “always,” “sometimes,” or “never” appropriate for antibiotics are presented in Figure 1b. “Sometimes” visits gradually increased from 10% to 20% of all visits in the COVID-19 era, but this did not reach the peak seasonal highs of 30% observed prior to the pandemic. “Always” visits remained reduced throughout the COVID-19 era. Antibiotic prescribing for “sometimes” visits decreased from 51.7% to 48.4% from baseline to the COVID-19 era, whereas prescribing for “never” visits decreased from 5.4% to 2.86%. Reductions over time were greater for acute upper respiratory infections, but nonrespiratory infections (eg, skin and soft-tissue infection and urinary tract infection) remained stable over the study period (Fig. 2).
Subanalyses were performed for pediatric respiratory visits. We detected a 29.4% relative reduction in percentage of antibiotic visits from baseline to the COVID-19 era (Table 2). Reductions in percentage of antibiotic visits were highest in the following groups: children aged 3–9 years (36.4%) and 10–19 years (42.7%), Black children (32.4%), privately insured children (31.7%), and children seen at urgent-care or retail clinics (39.6%). A multivariable regression model revealed that the odds of receiving an antibiotic prescription for a respiratory visit decreased ∼50% during the COVID-19 era versus baseline (OR, 0.548; CI, 0.539–0.557) and 15% in virtual visits versus in-person visits (OR, 0.855; CI, 0.755–0.969), adjusted for patient and visit characteristics (Table 3).
Note. The baseline period was January 1, 2019, to March 31, 2020, and the COVID-19 era was defined as April 1, 2020, to October 31, 2021.
a Antibiotic visit percentages calculated as percentage of respiratory visits where an antibiotic was prescribed (row percentages).
Virtual visits
There were 8,652 virtual visits during the study period: 437 (5.1%) were in the baseline period and 8,215 (95.0%) were in the COVID-19 era. Also, 715 (8.3%) of all virtual visits had in-person follow-up visits within 48 hours (Table 4). The rates of follow-up were higher for female patients (9.2%), for children aged 0–2 years (10.8%), for visits at urgent-care clinics (15.8%), and for visits with nurse practitioners (14.0%). Visits for viral infections (including acute bronchitis and acute upper-respiratory infections) had a higher follow-up rate (17%), as did visits for pharyngitis (including Streptococcus and non-Streptococcus diagnoses, 16.9%) and COVID-19 (20.6%). However, AOM had a follow-up rate of 6.1%, sinusitis had a follow-up rate of 7.1%, and other diagnoses had a follow-up rate of 7.1%.
Note. The baseline period was January 1, 2019, to March 31, 2020, and the COVID-19 era was defined as April 1, 2020, to October 31, 2021.
a Follow-up included all visit types (laboratory, imaging, office visits, hospitalizations, etc).
b Viral diagnoses include acute bronchitis and acute upper respiratory infection.
c Pharyngitis diagnoses include streptococcal pharyngitis and pharyngitis not specified.
Discussion
Our study describes a 59.2% reduction in antibiotic prescribing to children during the height of social isolation due to COVID-19 in 2020 and an 18.2% reduction in antibiotic prescribing in the same seasonal period in 2021, despite an overall increase in the number of outpatient visits in 2021. We detected greater reductions in antibiotic prescribing for respiratory illnesses: 71.2% in 2020 and 23.2% in 2021. This partially sustained reduction in antibiotic prescriptions occurred in the setting of return to in-person learning and likely substantial increases in social gatherings outside school. A portion of this reduction in antibiotic prescribing could be explained by continued masking in schools and/or public settings. However, this period also included an atypical RSV season; local cases reached a higher peak in August 2021 than typically seen in winter months, suggesting effective transmission of respiratory pathogens in the community. Typical respiratory viruses (eg, RSV and influenza) have been shown to directly correlate with seasonal increases in antibiotic prescribing both before and during the COVID-19 pandemic. Reference Lepak, Taylor and Stone17
Our findings show dramatic changes in respiratory diagnoses over the study period but minimal changes to nonrespiratory diagnoses (Fig. 2). Similarly, these findings are consistent in our analyses of appropriateness of antibiotic prescribing based on visit diagnoses (“always,” “sometimes,” and “never” visits). Sharp declines in diagnoses and antibiotic visits in March–May 2020 have been previously explained. Reference Katz, Spencer, Zhang and Banerjee10–Reference Dutcher, Li, Lee, Grundmeier, Hamilton and Gerber13 However, the changes detected in May–September 2021 are novel. For instance, initial declines in AOM diagnoses were likely because AOM is most frequently a consequence of viral upper-respiratory infection, Reference Lieberthal, Carroll and Chonmaitree18 and viral infections decreased due to social distancing and masking. AOM cases increased again in summer 2021when acute upper-respiratory infections peaked.
Previous studies showed similar declines in respiratory diagnoses and antibiotic prescribing during earlier stages of the COVID-19 pandemic. Reference Katz, Spencer, Zhang and Banerjee10–Reference Dutcher, Li, Lee, Grundmeier, Hamilton and Gerber13 Evaluating national prescribing data, Chua et al Reference Chua, Volerman and Conti19 reported a 55.6% reduction in antibiotic prescribing to children in April–December 2020 compared to the corresponding period in 2019. More recently, investigators described a 72.7% reduction in antibiotic prescriptions within a study period from January 2019 through June 2021 in pediatric primary-care practices associated with the Children’s Hospital of Philadelphia. Reference Dutcher, Li, Lee, Grundmeier, Hamilton and Gerber13 Our study expands on the existing literature by including fall 2021 data, which included return to school and increases in circulating respiratory viruses as previously described. We also included nonpediatric primary-care settings such as family medicine and urgent care and retail practices.
Additionally, we identified disproportionate changes in antibiotic prescribing by patient characteristics, prescriber type, and practice settings. Earlier publications evaluating antibiotic prescribing during the pandemic have reported on monthly trends and relative changes in antibiotic prescribing beyond expected, using monthly matched time periods. Reference King, Lovegrove and Shehab11,Reference Buehrle, Wagener, Nguyen and Clancy12 However, in 2021, the RSV peak and SARS-CoV-2 δ (delta) variant produced unseasonably high rates of viral illness in summer–fall months. Therefore, we chose to report percent reductions in antibiotic prescribing in respiratory visits, to account for fluctuations in respiratory visits throughout the study period.
The COVID-19 pandemic has disproportionately affected adults and children by race and social status with regard to illness, use of telemedicine, and ability to practice social distancing. Reference Mackey, Ayers and Kondo20–Reference Brown, Montez and Amati23 Additionally, prior to the COVID-19 pandemic, White children were prescribed antibiotics at higher rates than Black or non-White children. Reference Gerber, Prasad and Localio24–Reference Wattles, Vidwan, Feygin, Jawad, Creel and Smith26 Although the reduction in antibiotic visits could represent reduced inappropriate prescribing, it could also potentially represent undertreatment and/or limitations in access to care. The racial discrepancy identified in our study may represent another example of how racism has caused widened health disparities through the COVID-19 pandemic Reference Milano27 and warrants further study because this factor could have greater implications for other areas of pediatric care.
Our study also showed greater decreases in antibiotic visits for older children aged ≥3 years, possibly due to virtual learning and masking or shifts in parental expectations for antibiotics in the setting of viral illness. For instance, caregivers could be more focused on obtaining testing for COVID-19 for return to school and less likely to expect antibiotics for a child’s respiratory illness. These findings suggest that there could be a favorable shift in culture and expectations surrounding outpatient visits for acute upper respiratory infections, representing a reduction in unnecessary antibiotic prescribing. Continued monitoring and reporting of antibiotic prescribing trends are warranted. We also saw disproportionate changes in practice settings, likely reflective of the addition of academic pediatric practices that joined the healthcare system in March 2020. Similarly, trainees are only involved in outpatient care in the newly added pediatric clinics and therefore had an increase in respiratory visits during the COVID-19 era.
Differences in relative reduction of antibiotic visits by practice setting and provider type could be attributable to several changes throughout the pandemic. For instance, reductions were greatest in urgent care and retail clinics which could be due to such settings being more frequently visited for COVID-19 tests as the primary focus of the visit. Physician assistants also experienced greater reductions in antibiotic visits compared to nurse practitioners and physicians; however, physician assistants are uncommon in pediatric primary care at our institution and results likely reflect the prescribing practices of a few providers.
Prior to the COVID-19 pandemic, studies evaluating antibiotic prescribing in virtual visits compared to in-person visits had conflicting and controversial results. Reference Hersh, Stenehjem and Daines28–Reference Ray, Martin and Wolfson30 Although virtual visits in our study sample were low, our findings suggest that antibiotic prescribing for respiratory diagnoses was less common in virtual visits (Table 3). Additionally, a subanalysis of virtual visits showed that rates of in-person follow-up were not high for common bacterial diagnoses of AOM and sinusitis (6.1% and 7.1% follow-up, respectively). Alternatively, we did see higher rates of in-person follow-up visits (16%–20%) for diagnoses that could benefit from laboratory testing, such as pharyngitis, COVID-19, and other respiratory viral infections. Given the low rates of virtual and follow-up visits, additional analyses of antibiotic prescribing in these visits were not performed.
This study had several limitations. Our study sample included pediatric visits to a single healthcare organization, and findings may not be generalizable to other organizations or geographic settings. We were unable to quantify any changes in pediatric visits due to patient relocation or care-seeking behaviors. We utilized ICD-10 codes to identify visit diagnoses, a common practice for database research in antimicrobial stewardship; however, accuracy between diagnostic billing and clinical diagnosis cannot be guaranteed. We utilized respiratory visits to compare changes in antibiotic prescribing and in the regression model, to standardize time periods for seasonal variations. We chose to include COVID-19 visits because COVID-19 may present with respiratory symptoms similar to other viral indications for which antibiotics are commonly prescribed inappropriately. However, inclusion produces lower rates of antibiotic visits in the COVID-19 era. We included medical practices that were added to our organization during the study period; however, sensitivity analyses revealed that the addition of these practices did not impact odds of an antibiotic visit. Finally, we included all visit types for our virtual follow-up visit analysis, which could have included routinely scheduled visits and, therefore, we may have overestimated the rate of in-person follow-up visits.
In conclusion, we report decreases in antibiotic prescribing during the COVID-19 pandemic, with disproportionate changes among patient groups by age, race or ethnicity, insurance status, and practice setting. Additional qualitative studies are needed to better understand changes in culture and expectations surrounding antibiotic prescribing during this unprecedented time. Tracking and reporting of antibiotic prescribing is a key component of outpatient stewardship initiatives. Efforts should be made for organizations to share such data with prescribing clinicians to continue to promote favorable trends in judicious antibiotic use.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ash.2022.235
Acknowledgments
The authors acknowledge and thank Joe Head for assisting with data retrieval and analysis and Dr Deborah Davis for research and leadership support.
Financial support
This project was supported by the Kentucky Cabinet for Health and Family Services Department for Medicaid Services under the State University Partnership contract titled “Improving Care Quality for Children Receiving Kentucky Medicaid”; the Norton Children’s Hospital Research Institute; the University of Louisville School of Medicine, Department of Pediatrics, and School of Public Health and Information Sciences; and the Duke University School of Medicine. B.W., K.J., Y.F., J.S., N.V., M.D.S, and M.J.S. also received additional grant support from the Merck Investigator Studies Program (study no. 60356).
Conflicts of interest
B.W., K.J., Y.F., J.S., N.V., M.D.S, and M.J.S report having received grant support from Merck & Co.