COVID-19 and Antibiotic Prescribing in Pediatric Primary Care

We conducted a retrospective, observational study to assess the impact of the COVID-19 pandemic on encounter volume and antibiotic prescribing in the pediatric primary care setting. This study was conducted in 27 practices in the Children’s Hospital of Philadelphia (CHOP) primary care network, a not-for-profit health care organization located in the greater Philadelphia region. All practices primarily see patients with health insurance, and all were open during the entire study period. We included encounters from January 1, 2018, through June 30, 2021. COVID-19 cases were first diagnosed in Philadelphia in March 2020, and CHOP rapidly expanded ambulatory telemedicine capabilities at that time. In the Philadelphia Public School District, in-person learning ceased in mid-March 2020; hybrid in-person and remote learning was resumed for a subset of schools in younger students in March 2021, but otherwise remained remote through the end of the study period. Other regional school districts had variable policies regarding remote learning, although many did not resume exclusively in-person learning until 2021. Patients <19 years of age were included.

We defined billed encounters as in-person and telemedicine visits with an associated Current Procedural Technology (CPT) evaluation and management billing code, and all other encounters as nonbilled encounters (eg, telephone calls). To distinguish office visits from telemedicine visits, modifier codes were used (GT, 95). The primary outcome was antibiotic prescriptions per 1000 patients, assessed by month. For the outcome numerator (antibiotic prescriptions), we included oral antibacterial agents commonly used for acute infectious indications as well as several common outpatient intramuscular antibiotics, and we excluded antimycobacterial agents as well as all intravenous antibiotics (Supplemental Table 2). All antibiotics prescribed at billed and nonbilled encounters were included in the primary outcome. The population cohort (primary outcome denominator) was defined as any patient who had 1 billed encounter during the 2018 and 2019 calendar years. We used a prepandemic population to define the cohort that would be eligible to receive an antibiotic prescription during the entire study period to avoid an artifactual underestimation of the overall cohort during the pandemic (ie, patients eligible to be seen in primary care in the included practices) because of COVID-19 lockdowns leading to avoidance of typical well-care and acute encounters. Thus, our assumption was that the number of children who would have been patients of these practices (the population cohort denominator used for defining the rate of antibiotic prescribing per population) would have been stable during calendar years 2018, 2019, 2020, and 2021.

We also assessed the primary outcome in a subset of encounters with common specific infectious disease diagnoses, including RTI, urinary tract infection (UTI), and skin and soft tissue infection (SSTI) defined by International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes (Supplemental Tables 3 and 4). Encounters could be classified into more than 1 disease group. Additionally, the primary outcome was assessed in 4 age group strata (selected as groupings that generally reflect the differing epidemiology of infections in children) and by antibiotic spectrum of activity; penicillin and amoxicillin were classified as narrow-spectrum, and all other antibiotics were classified as broad-spectrum.

Several secondary outcomes were evaluated. To assess whether patients who sought care during the pandemic may have been sicker than those before the pandemic and more likely to receive an antibiotic, we examined the likelihood of each encounter resulting in an antibiotic prescription by assessing antibiotic prescriptions per 1000 billed encounters. We also examined billed encounter volume by month, stratified by telemedicine and in-person encounters.

Demographic characteristics, antibiotic prescriptions, and ICD-10-CM code data were obtained from the electronic health record database. This study was approved by the institutional review board at CHOP.

We report descriptive statistics for the study population. We plotted antibiotic prescriptions per 1000 patients by month from January 2018 through June 2021 (primary outcome). Similar plots were generated for the secondary outcomes. To account for seasonal variation, we compared the monthly antibiotic prescriptions for the same months in 2019, 2020, and 2021. We then used interrupted time series analysis to assess the both the immediate and longer-term impact of the COVID-19 pandemic on monthly antibiotic prescription per 1000 patients. April 2020 was designated as the first month of the pandemic period because it was the first full month during which social distancing measures were enacted and telemedicine was available, although these were both initiated in the latter half of March 2020. This analysis allowed assessment of the immediate impact on the antibiotic prescription trend from the pandemic onset as well as the enduring time effect. The immediate impact was estimated by measuring the immediate change in the antibiotic prescription per 1000 patients in April 2020, and the enduring time effect was estimated by calculating the rate of change in monthly prescription after the onset of the pandemic. Interrupted time series analysis was also performed for the monthly encounter volume secondary outcome. For all calculations, a two-tailed P value of .05 was considered statistically significant. All calculations were performed by using Stata v15.1 (Stata Corp, College Station TX).

We identified 276 678 unique patients in our population cohort (primary outcome denominator) who had at least 1 billed encounter from January 1, 2018 through December 31, 2019. Of these, 49% were girls, 53% were White, and 27% were Black, and the mean age was 7.9 years (SD 5.6) (Table 1). There were 69 327 total antibiotic prescriptions during the period from April through December in 2019 and 18 935 antibiotic prescriptions during the same months in 2020, a 72.7% reduction. For the RTI encounter subset, there were 50 876 total antibiotic prescriptions from April through December in 2019, compared to 6890 antibiotic prescriptions during the same period in 2020, an 86.5% reduction; as a result, the reduction in RTI encounter prescriptions accounted for 87.3% of the overall reduction in prescriptions from 2019 to 2020.

Prepandemic, seasonal variation in antibiotic prescriptions was seen, with peaks in the winter months for the overall group and RTI subset. A decrease in prescriptions was observed beginning in March 2020, lasting through February 2021, after which prescriptions started to increase slowly (Fig 1). Antibiotic prescribing associated with UTI and SSTI encounters was substantially less frequent than RTI encounter prescriptions (Fig 1). When comparing across years, there were 35.7 overall antibiotic prescriptions per 1000 patients for April in 2019, vs 8.1 and 8.8 for the same months in 2020 (77% reduction) and 2021 (75% reduction), respectively. The 2019 to 2020 difference was at its narrowest in August (53% reduction), and then widened in December (84% reduction) (Supplemental Fig 7A). Similar differences were seen for the RTI subgroup, although smaller differences were noted across years for the UTI and SSTI groups, with less seasonal variation (Supplemental Fig 7 B–D).

Interrupted time series analysis demonstrated an immediate significant decrease in April 2020 from a predicted mean of 31.6 prescriptions per 1000 patients to 6.4 prescriptions per 1000 patients (difference of −25.2 prescriptions per 1000 patients; 95% CI: −32.9 to −17.5). This was followed by a nonsignificant monthly rise in antibiotic prescriptions in the pandemic period (slope = 0.25 prescriptions per 1000 patients; 95% CI: −0.01 to 0.51) (Fig 2A, Supplemental Table 5). For the RTI subset, a similar pattern was seen (Fig 2B, Supplemental Table 5). For the UTI subset, a much smaller relative immediate decrease in antibiotic prescriptions per 1000 patients was observed, followed by an essentially stable prescribing rate (Fig 2C, Supplemental Table 5). For SSTI encounters, there was no significant immediate or sustained change in prescriptions (Fig 2D, Supplemental Table 5).

Similar patterns were seen in antibiotic prescribing across age groups, although prescriptions were generally more frequent in the groups 1 to 4 and 5 to 11 years of age compared to other age groups in both the prepandemic and pandemic periods (Supplemental Figs 8–11). The rise in prescriptions (overall and RTI) in 2021 was most notable in the group 1 to 4 years of age (Supplemental Figs 8 and 9). During the pandemic period, broad-spectrum antibiotics comprised a larger proportion of antibiotics prescribed compared to the prepandemic period, primarily owing to the reduction in RTI encounter prescriptions, which were more frequently narrow-spectrum compared to other diagnosis groups (Fig 3).

In examining all antibiotic prescriptions per 1000 billed encounters, a sustained decrease was observed beginning in April 2020 (Fig 4A). In comparison, a smaller decrease in antibiotic prescriptions per 1000 encounters was noted for RTI encounters (Fig 4B). For UTI and SSTI encounters, smaller decreases were noted (Fig 4C–D). Overall, a majority of antibiotics were prescribed at in-person office encounters (Fig 4, Supplemental Fig 12).

Interrupted time series analysis demonstrated that in April 2020, there was an immediate decrease in encounter volume, followed by a significant monthly increase during the pandemic period (Supplemental Table 6, Supplemental Fig 13A). For RTI encounters, a similarly large immediate decrease was seen, although the subsequent monthly trend upward during the pandemic was smaller and more delayed, in contrast to all encounters (Supplemental Table 6, Supplemental Fig 13B). At the maximum, telemedicine encounters comprised 25% of all billed encounters in April 2020, with a gradual subsequent proportional decline (Fig 5). For RTI encounters, small shifts in specific RTI diagnoses were seen in the pandemic months compared to the prepandemic period, including a decrease in the proportions of “other upper respiratory infection” (URI) and pneumonia diagnoses, and an increase in the proportion of nonspecific respiratory symptoms (eg, cough) (Fig 6).

Among 27 pediatric primary care practices, we found a significant reduction in antibiotic prescribing beginning in April 2020, coinciding with the outbreak of COVID-19 in our region and the social distancing measures enacted in response to the pandemic. The overall decrease in antibiotic prescribing appeared to be driven largely by sustained decreases in RTI encounter volume. This conclusion is further supported by the observation that both RTI encounters and antibiotic prescribing remained relatively low through the end of 2020 and only began to substantially increase in 2021, unlike overall encounter volume, which rebounded more quickly. Although other studies demonstrated early reductions in RTIs and antibiotic prescribing during the COVID-19 pandemic, to our knowledge, this is the first study to demonstrate a sustained decrease in antibiotic prescribing in pediatric primary care throughout 2020 and early 2021, driven by a durable reduction in RTI encounters, followed by increasing prescribing in March through June 2021.^4,5,10–18 

Importantly, these findings suggest that decreases in viral (non-COVID-19) RTI transmission in the community as a result of public health measures played a large role in the decrease in RTI encounters and RTI-related antibiotic prescribing. Additionally, the observed rise in RTI encounters and antibiotic prescribing in 2021 coincided with loosening of various local pandemic restrictions and increasing vaccination in the adult population.²²  Furthermore, the small rise in RTI encounters in March through June 2021, when many students had not yet returned to in-person learning, suggests that community viral RTI transmission outside of schools contributed to this rise. We also observed a relatively small but significant immediate drop in UTI encounters and antibiotic prescriptions at these encounters. This decrease suggests that avoidance of care during the pandemic also may have occurred, as we would not expect UTI occurrence to be significantly impacted by public health pandemic measures. Although this would not explain why we did not observe a similar decrease for SSTIs, it suggests that changes in health care–seeking behaviors may account for at least a small proportion of the overall reduction in visits and antibiotic prescriptions.

Our findings supplement the results of Zhu et al and Gillies et al, who found a reduction in antibiotic prescribing in the United Kingdom and Australia, respectively, through the end of 2020, although changes in specific infectious diagnoses were not explored.^13,19  King et al also identified substantial reductions in antibiotic prescriptions at retail pharmacies in the United States early in the pandemic, particularly in the pediatric population.¹¹  However, antibiotic prescriptions were not tied to clinical diagnoses, limiting understanding of the causes of changes in antibiotic prescribing. Lepak et al identified reduced antibiotic prescribing associated with RTI reductions in the United States without increases in prescribing by the end of February 2021. In contrast to our study, however, they did not observe a significant reduction in ambulatory visits, although their population was not limited to primary care or pediatrics.⁹ 

Per-encounter antibiotic prescribing also decreased during the pandemic. This was a somewhat unexpected finding, because we had considered that for patients who sought (and were granted access to) in-person medical care, especially early in the pandemic, there might be a higher probability of a serious acute illness, and thus the likelihood of antibiotic prescribing for these encounters might actually increase. Whereas we did note a small relative increase in nonspecific respiratory symptom ICD-10-CM codes, it is unlikely that shifts in specific RTI diagnoses would entirely account for a decreased likelihood of antibiotic prescribing per encounter. Possible explanations include increased provider capacity for patient communication regarding appropriate antibiotic prescribing due to reduced encounter volume or an increase in patients and families primarily seeking reassurance regarding COVID-19 infection without seeking antibiotic prescriptions.

We also found that antibiotic prescribing primarily occurred at in-person visits rather than telemedicine visits during the pandemic period. Our findings are consistent with Ray et al, who found that guideline-concordant antibiotic prescribing was higher in telemedicine visits during the pandemic, and that overall prescribing was lower in telemedicine visits compared to in-person office visits.²³  Previous literature from the pre-COVID era has reported mixed results regarding antibiotic prescribing in telemedicine, although several key studies have reported increased prescribing at telemedicine encounters, particularly in direct-to-consumer telemedicine.^24–26  However, the COVID-19 pandemic represents the first period during which telemedicine was made widely available within primary care practices. Telemedicine within a primary care medical home, where patients have established relationships with their providers, likely differs from direct-to-consumer telemedicine.²⁶ 

Our findings have implications for antimicrobial stewardship in pediatric primary care. By looking at the impact of the public health response to the COVID-19 pandemic on antibiotic prescribing, our data suggest that reducing community viral RTI transmission through social distancing and masking corresponds with a reduction in antibiotic prescribing. Because many antibiotic prescriptions related to viral RTI are inappropriate, this further emphasizes the importance of judicious prescribing for RTIs and the need for antimicrobial stewardship in this setting to mitigate both short- and long-term adverse effects of antibiotic overuse in children.^27–29 

There are several limitations present in this study. To define our population cohort denominator for our primary outcome, we included patients seen at least once between 2018 and 2019 and assumed stability of this eligible cohort into 2020. Although it is impossible to precisely define the cohort who would present to primary care when needed, we did confirm that this cohort number remained stable over a longer 4-year period as well (2016 to 2019), supporting the face validity of this approach. In addition, ICD-10-CM codes were used to classify encounters into RTI, UTI, and SSTI disease groups, and use of these codes may lead to misclassification if providers entered inaccurate codes. We also assumed that an antibiotic prescribed at the same encounter in which a particular ICD-10-CM code was entered was for that infection, which may not always be the case. However, most encounters had a single infectious diagnosis code, and these potential misclassification events are not expected to differ before and during the pandemic. We also did not examine appropriateness of antibiotic prescribing, although we expect that much of the reduction in antibiotic prescribing represents a decrease in inappropriate antibiotic use, particularly given previous studies suggesting that at least some of the winter spike in prescribing for RTI is likely inappropriate.^20,30,31  However, we acknowledge that there was also likely a reduction in postviral bacterial infections, where antibiotic use is more often appropriate. Additionally, it is possible that some patients did not seek primary care during the pandemic for conditions that warranted an antibiotic prescription or urgent treatment. These patients may have sought care or received antibiotic prescriptions through other means, such as visits to urgent care or the emergency department, although it was previously reported that emergency department visits at CHOP decreased during the pandemic.³² 

Antibiotic prescribing in pediatric primary care practices decreased substantially during the COVID-19 pandemic, beginning almost immediately after widespread social distancing measures were enacted. This reduction was sustained through early 2021 and primarily driven by a reduction in prescribing associated with RTI encounters, likely in part related to reduced viral RTI transmission in the community. Future studies should examine changes in antibiotic prescribing with relaxation of social distancing and authorization of COVID-19 vaccination for children, as well as the impact on antibiotic resistance and other antibiotic-related adverse effects.

CHOP

Children’s Hospital of Philadelphia

CI

confidence interval

COPD

chronic obstructive pulmonary disease

COVID-19

coronavirus disease 2019

CPT

Current Procedural Technology

ICD-10-CM

International Classification of Diseases, Tenth Revision, Clinical Modification

RTI

respiratory tract infection

SSTI

skin and soft tissue infection

URI

upper respiratory infection

UTI

urinary tract infection