Telehealth visits increased significantly during the coronavirus disease 2019 pandemic without consensus on the appropriate scope of telehealth antibiotic prescribing within pediatric primary care. We describe telehealth antibiotic prescribing patterns within our statewide pediatric primary care network during the coronavirus disease 2019 pandemic.
In a retrospective observational study of a large statewide pediatric primary care network, we identified and analyzed telehealth and in-person encounters with oral antibiotics prescribed from March 2020 to July 2021. We focused on the top 5 general diagnosis groupings using International Classification of Disease 10 codes.
Of the 55 926 encounters with an oral antibiotic prescribed, 12.5% were conducted via telehealth and 87.5% in person. The proportion of telehealth antibiotic encounters varied significantly according to diagnosis category (P <.001): ear (30.8%), skin and subcutaneous (21.8%), respiratory (18.8%), genitourinary (6.3%), and Lyme disease infections (3.8%). The proportion of telehealth antibiotic encounters for all diagnosis categories peaked in spring of 2020. The greatest proportion of telehealth antibiotic prescribing during the most recent 4weeks of the analysis were Lyme disease infections (11.7%) and for skin and subcutaneous tissue infections (3.1%).
Telehealth continues to be used to prescribe antibiotics even after the initial stage of the pandemic. Clinicians and patients would benefit from clearer guidelines about the appropriate use of antibiotics prescribed during telehealth encounters.
Much effort has gone into reining in the overuse of antibiotics for children, but research on antibiotic prescribing practices in telehealth compared with in-person care reaches varying conclusions and is still maturing.
There continues to be telehealth antibiotic prescribing during the coronavirus disease 2019 pandemic, particularly for diagnoses that can be made visually such as skin, soft tissue, and Lyme disease infections.
Telehealth use within pediatric primary care has been described in the literature as early as the 1970s.1 Initial use cases included psychiatric consultation, nurse practitioner oversight, remote monitoring of chronic conditions, and school-based care.1–3 Its subsequent slow uptake had been largely attributed to lack of or low reimbursement.4 Telehealth uptake expanded rapidly during the coronavirus disease 2019 (COVID-19) pandemic,5 and it is not expected to return to prepandemic rates. Although telehealth encounters offer many benefits, 1 area of potential concern with expanded use of telehealth is its effect on antibiotic prescribing practices. Much effort has gone into reining in the overuse of antibiotics for children over many years, but research on antibiotic prescribing practices in telehealth compared with in-person care reaches varying conclusions and is still maturing.6–9 Concerns about antibiotic prescribing via telehealth include the lack of access to basic laboratory testing, such as rapid streptococcal antigen tests and urinalysis, as well as limitations to the physical examination, such as visualization of tympanic membranes.3,6,10 These concerns are heightened within pediatric care because of children’s communication limitations and greater clinician reliance on the physical exam.11
Much of the literature on telehealth antibiotic prescribing describes prescribing practices from before the pandemic and is restricted to direct-to-consumer (DTC) and adult telehealth.11–18 Although DTC telehealth may offer convenient on-demand patient scheduling, the DTC model is generally outside of the patient’s medical home, and thus, clinicians lack access to medical records and continuity of care.11,12 A prepandemic systematic review of studies of adult and pediatric patients found an overall increase in antibiotic prescribing via various types of synchronous telehealth (phone or video-based) compared with in-person care, with variation in findings by diagnosis.12 In 1 study of children with acute respiratory infections, children presenting via DTC telehealth were more likely to be prescribed antibiotics (52%) compared with children presenting via in-person urgent care (42%) and primary care (31%) visits, P <.001.11 In a study of adults with acne, patients presenting via telehealth (43%) were more likely to be treated with oral antibiotics compared with those seen in-person (28%), P <.001.13 Other studies found no significant difference in antibiotic prescribing for pediatric conjunctivitis, common adult dermatologic conditions, or adult urinary tract infections between telehealth and in-person encounters.14–18
Understanding the role of telehealth in antibiotic prescribing has become even more critical as telehealth use for pediatric illness encounters changed dramatically with the COVID-19 pandemic. Firstly, there was a shift toward expansion of virtual encounters within primary care practices. Secondly, our network and others have observed a significant decrease in diagnosis of common childhood infectious illnesses and antibiotic prescribing when coronavirus rates are highest.19,20 Finally, 1 study of pediatric antibiotic prescribing during the pandemic described a generally high level of appropriate antibiotic prescribing for acute respiratory tract infections, albeit with an initial surge of telehealth antibiotic prescribing for acute otitis media and streptococcal pharyngitis without appropriate physical exam or laboratory testing in the early phase of the pandemic.21
To date, national associations and agencies have not published guidelines for appropriate telehealth antibiotic prescribing for common pediatric illnesses.22,23 In their latest telehealth position statement from 2019, the Infectious Diseases Society of America outlines a limited list of established use cases including: human papillomavirus management, tuberculosis management, remote infectious disease physician-led antimicrobial stewardship programs, and outpatient parenteral antibiotic therapy.23 The American Telemedicine Association guidelines from 2017, endorsed by the American Academy of Pediatrics, recommends that commercial DTC telehealth should not be used for children less than 2 years old.24
Our aim in the current study is to describe the experience of a large pediatric primary care network’s rapid implementation of telehealth video visits during the COVID-19 pandemic and how it affected antibiotic prescribing patterns. We are particularly interested in describing use beyond the initial 3 months of the pandemic in which there were many restrictions on businesses, schools, and healthcare facilities. By describing what may be viewed as a “steady state” after the initial rollout, we hope to inform the development of clinical guidance for telehealth applications of pediatric infectious disease management and protect antimicrobial stewardship.
Methods
Context
We conducted a retrospective observational study of outpatient encounters in a network of 77 independent pediatric primary care practices across Massachusetts. The network consists of over 500 primary care clinicians affiliated with Boston Children's Hospital and provides primary care for approximately 400 000 children. Our practice sizes consist of 1 to 2 PCPs (24%), 3 to 7 PCPs (38%), and ≥8 PCPs (36%). Practices are in suburban (60%) and urban (38%) settings.
Patient Characteristics
We identified patient age, sex, insurance status, and preferred language using insurance claims data. We measured patient complexity using the Pediatric Medical Complexity Algorithm, which defines 3 levels of complexity: no chronic disease, noncomplex chronic disease, and complex chronic disease.25 We compared these patient characteristics between populations that had at least 1 in-person encounter versus at least 1 telehealth video visit encounter with an antibiotic prescribed using χ2 test for categorical variables and Kruskal-Wallis test for the continuous variable.
Data Collection and Analysis
We extracted encounter data from the network’s electronic health record using Structured Query Language and SAS version 9.4 (SAS Institute Inc., Cary, NC) for dates of service from March 2, 2020 to July 31, 2021, reflecting the start of the COVID-19 pandemic through the most recent data available at the time of the analysis. First, encounter type of in-person versus telehealth was defined by billing codes (Virtual encounter Current Procedural Terminology codes G2010, G2012, 98970, 98971, 98972, 99421, 99422, 99423, or E and M encounter with modifier codes 95, GT, GQ). Second, for each encounter, we identified whether an oral antibiotic was prescribed and, if so, to what diagnosis it was linked. Encounters were classified by condition by using the first diagnosis on the antibiotic order when available or the diagnoses on the encounter when no antibiotic was ordered. Third, we used the first 3 digits of the International Classification of Disease 10 (ICD-10) code to create diagnosis groups of related infections, as determined by the authors with clinical experience. Fourth, we focused on the top 5 general groupings by encounter frequency. Encounters that were not mapped to an ICD-10 code or “not elsewhere classified” were excluded. Finally, weekly numbers of in-person and telehealth encounters, as well as count and percent of each with antibiotic ordered were calculated and analyzed graphically to identify patterns and trends over the measurement period.
χ2 testing was conducted to compare visit proportions between diagnosis groupings using Stata version 7.0 (Stata Corp, College Station, TX).
Ethics
This project is limited to secondary use of deidentified healthcare data and was approved by the Boston Children’s Hospital Institutional Review Board review.
Results
In the 74-week (17 month) timeframe of this study, there were a total of 413 565 active patients within the network. Of this total, 42 518 (10.3%) unique patients met the inclusion criteria, 37 772 (88.8%) with at least 1 in-person encounter with an antibiotic prescribed and 6298 (14.8%) with at least 1 telehealth encounter with an antibiotic prescribed. Patient characteristics differed across groups (Table 1). Patients in the telehealth group were older (median age 11) than the in-person group (median age 9) (P <.0001). Those in the telehealth group were more likely to be commercially insured, White non-Hispanic, prefer English language, and have complex chronic disease (P <.0001).
. | All Patients . | Patients With ≥1 In-Person Encounter With Antibiotic Prescribed . | Patients With ≥1 Telehealth Encounter With Antibiotic Prescribed . | P . |
---|---|---|---|---|
Number of patients | 413 565 | 37 772 | 6298 | |
Age, y, median (IQR) | 11 (5–16) | 9 (4–16) | 11 (6–17) | <.0001b |
Sex, n (%) | <.0001c | |||
Female | 204 911 (50) | 20 493 (54) | 3388 (54) | |
Male | 208 654 (50) | 17 279 (46) | 2910 (46) | |
Insurance status, n (%) | <.0001c | |||
Commercially insured | 284 061 (69) | 23 641 (63) | 4368 (69) | |
Publicly insured | 129 504 (31) | 14 131 (37) | 1930 (31) | |
Race and ethnicity, n (%) | <.0001c | |||
Asian, non-Hispanic | 19 037 (5) | 1078 (3) | 175 (3) | |
Black, non-Hispanic | 14 012 (3) | 920 (2) | 100 (2) | |
Hispanic | 41 038 (10) | 4402 (12) | 496 (8) | |
White, non-Hispanic | 242 605 (59) | 23 922 (63) | 4201 (67) | |
Another race or multiracial, non-Hispanic | 63 621 (15) | 4762 (13) | 833 (13) | |
Unknown | 33 252 (8) | 2688 (7) | 493 (8) | |
Preferred language, n (%) | <.0001c | |||
English language preferred | 394 211 (95) | 36 217 (96) | 6084 (97) | |
Non-English language preferred | 19 354 (5) | 1555 (4) | 214 (3) | |
Pediatric medical complexity algorithm category,an (%) | <.0001c | |||
No chronic disease | 263 411 (64) | 21 618 (57) | 3441 (55) | |
Noncomplex chronic disease | 102 817 (25) | 10 514 (28) | 1787 (28) | |
Complex chronic disease | 47 337 (11) | 5640 (15) | 1070 (17) |
. | All Patients . | Patients With ≥1 In-Person Encounter With Antibiotic Prescribed . | Patients With ≥1 Telehealth Encounter With Antibiotic Prescribed . | P . |
---|---|---|---|---|
Number of patients | 413 565 | 37 772 | 6298 | |
Age, y, median (IQR) | 11 (5–16) | 9 (4–16) | 11 (6–17) | <.0001b |
Sex, n (%) | <.0001c | |||
Female | 204 911 (50) | 20 493 (54) | 3388 (54) | |
Male | 208 654 (50) | 17 279 (46) | 2910 (46) | |
Insurance status, n (%) | <.0001c | |||
Commercially insured | 284 061 (69) | 23 641 (63) | 4368 (69) | |
Publicly insured | 129 504 (31) | 14 131 (37) | 1930 (31) | |
Race and ethnicity, n (%) | <.0001c | |||
Asian, non-Hispanic | 19 037 (5) | 1078 (3) | 175 (3) | |
Black, non-Hispanic | 14 012 (3) | 920 (2) | 100 (2) | |
Hispanic | 41 038 (10) | 4402 (12) | 496 (8) | |
White, non-Hispanic | 242 605 (59) | 23 922 (63) | 4201 (67) | |
Another race or multiracial, non-Hispanic | 63 621 (15) | 4762 (13) | 833 (13) | |
Unknown | 33 252 (8) | 2688 (7) | 493 (8) | |
Preferred language, n (%) | <.0001c | |||
English language preferred | 394 211 (95) | 36 217 (96) | 6084 (97) | |
Non-English language preferred | 19 354 (5) | 1555 (4) | 214 (3) | |
Pediatric medical complexity algorithm category,an (%) | <.0001c | |||
No chronic disease | 263 411 (64) | 21 618 (57) | 3441 (55) | |
Noncomplex chronic disease | 102 817 (25) | 10 514 (28) | 1787 (28) | |
Complex chronic disease | 47 337 (11) | 5640 (15) | 1070 (17) |
Per Simon et al, Pediatric Medical Complexity Algorithm.25
P value determined by Kruskal-Wallis test.
P value determined by χ2 test.
These patients had 1 332 685 total encounters, of which 55 926 (4.2%) had an oral antibiotic prescribed. Of these encounters with an antibiotic prescribed, 48 939 were in person (87.5%) and 6987 were via telehealth (12.5%) (Table 2). The proportion of encounters with an antibiotic prescription that were performed via telehealth peaked at 52.1% the week of April 27, 2020, then declined to an average of 2.4% over the last 4 weeks of the study period (Fig 1). The mean number of telehealth visits with antibiotics prescribed per provider is 8.9 (standard deviation 13.2).
. | Encounters With Antibiotic Prescribed, N (%) . | In-person Encounters With Antibiotic Prescribed, N (%) . | Telehealth Encounters With Antibiotic Prescribed, N (%) . |
---|---|---|---|
Total | 55 925 (100) | 48 938 (87.5) | 6987 (12.5) |
Ear infections | 17220 (30.8) | 16325 (94.8) | 895 (5.2) |
Skin and subcutaneous tissue infections | 12203 (21.8) | 9700 (79.5) | 2503 (20.5) |
Respiratory infections | 10506 (18.8) | 9380 (89.3) | 1126 (10.7) |
Genitourinary infections | 3550 (6.3) | 3289 (92.6) | 261 (7.4) |
Lyme disease | 2120 (3.8) | 1397 (65.9) | 723 (34.1) |
All other diagnoses | 10326 (18.5) | 8847 (85.7) | 1479 (14.3) |
. | Encounters With Antibiotic Prescribed, N (%) . | In-person Encounters With Antibiotic Prescribed, N (%) . | Telehealth Encounters With Antibiotic Prescribed, N (%) . |
---|---|---|---|
Total | 55 925 (100) | 48 938 (87.5) | 6987 (12.5) |
Ear infections | 17220 (30.8) | 16325 (94.8) | 895 (5.2) |
Skin and subcutaneous tissue infections | 12203 (21.8) | 9700 (79.5) | 2503 (20.5) |
Respiratory infections | 10506 (18.8) | 9380 (89.3) | 1126 (10.7) |
Genitourinary infections | 3550 (6.3) | 3289 (92.6) | 261 (7.4) |
Lyme disease | 2120 (3.8) | 1397 (65.9) | 723 (34.1) |
All other diagnoses | 10326 (18.5) | 8847 (85.7) | 1479 (14.3) |
There were 45 600 (81.5%) of total encounters with an antibiotic prescribed that were accounted for by the top 5 diagnosis groupings: (1) ear infections (17 220; 30.8% of all encounters with an antibiotic prescribed), (2) skin and subcutaneous tissue infections (12 203; 21.8%), (3) respiratory infections (10 506; 18.8%), 4) genitourinary infections (3550; 6.3%), and 5) Lyme disease (2120; 3.8%). We excluded encounters that were not mapped to an ICD-10 code or “not elsewhere classified” (18.5% of total encounters with an antibiotic prescribed). The number and proportion of in-person versus telehealth encounters with an antibiotic prescribed for each diagnosis grouping is shown in Table 2.
For all the diagnosis groupings, there was an initial surge of antibiotic prescribing via telehealth encounters early in the COVID-19 pandemic (March–May 2020). Antibiotic prescribing for all diagnosis groupings declined after that time. The proportion of antibiotic encounters that were done via telehealth varied significantly according to diagnosis group (P <.001).
We analyzed in-person versus telehealth antibiotic prescribing by each diagnosis group separately (Fig 2). For ear infections, the peak proportion of visits with an antibiotic prescribed conducted over telehealth was 40.0% in April 2020 and declined to an average of 0.9% over the most recent 4 weeks of the analysis. The analogous proportions for the other diagnosis groups were: for skin and subcutaneous tissue infections, a peak of 73.6% in April 2020 and an average of 3.1% over the most recent 4 weeks; for respiratory infections, a peak of 57.9% in April 2020 and an average of 0.9% over the most recent 4 weeks; for genitourinary infections, a peak of 36.7% in April 2020 and an average of 1.6% over the most recent 4 weeks; and for Lyme disease infections, a peak of 92.9% in March 2020 and an average of 11.7% over the most recent 4 weeks.
We analyzed the proportion of telehealth visits for each diagnosis group to determine how frequently patients who are seen via telehealth had a subsequent in-person visit with an antibiotic prescribed for the same condition. Overall, in 1.2% of cases, a telehealth visit was followed within 7 days by an in-person visit for the same condition with an antibiotic prescribed. This rate varied from a high of 2.1% for ear infections and 0.7% for skin and subcutaneous infections.
Discussion
Similar to other studies of telehealth completed in the pediatric population during the pandemic, within our large statewide pediatric network we observed a rapid uptake of telehealth early in the COVID-19 pandemic, including for acute illness visits associated with antibiotic treatment. In fact, at the peak of telehealth use in the first 3 months of the pandemic, more than half of antibiotic prescribing was done by telehealth. However, quite rapidly, the proportion of antibiotics prescribed via telehealth declined. It is highly likely that this initial surge in telehealth antibiotic prescribing in spring of 2020 was caused by emergency measures being taken by practices rather than a long-term practice change, because rates normalized as practices became more comfortable with managing in-person visits through the pandemic. It appears that over the spring and summer of 2021, the practices reached a fairly steady state, wherein only about 2% of antibiotics were prescribed via telehealth. This suggests that pediatricians are fairly self-regulating in their antibiotic prescribing practices and have developed their own norms and practices in which antibiotic prescribing is reserved for in-person encounters in the vast majority of cases.
There was, however, some variability in telehealth prescribing of antibiotics according to condition. For ear, respiratory, and genitourinary infections, antibiotic prescribing via telehealth became a fairly rare phenomenon later in the pandemic. This is presumably because these diagnoses rely on physical exam or in-office testing, such as otoscopy, pharyngeal exam, lung auscultation, rapid streptococcal antigen test, or urinalysis, to make a reliable diagnosis. But for skin and subcutaneous infections and Lyme disease visits, there was a greater persistence of antibiotic prescribing via telehealth of 3.1% and 11.1% respectively, presumably because clinicians feel comfortable making a diagnosis of a skin or soft tissue infection or of Lyme disease rash or exposure via telehealth in at least some cases.
There were also differences in antibiotic prescribing settings by patient demographics. Patients with a telehealth antibiotic prescription were older than those seen in-person. This is presumably because of younger children’s communication limitations and thus, greater clinician reliance on physical exam. We also found a small but statistically significant difference in antibiotic prescription setting in which Hispanic, non-English preferred language, and publicly insured children were less likely to have antibiotics prescribed via telehealth compared with in-person encounters. These differences may be attributed to disparities in broadband coverage and language interpreter integration in telehealth workflows, which may affect clinician confidence in diagnosis and antibiotic prescription.26–28 Finally, 17% of children with a complex chronic disease received an antibiotic remotely. This is not surprising given telemedicine has been found to provide more convenient accessible care and reduce infectious exposures for this population.29
This analysis of prescription practices during the natural experiment of a shift to a large proportion of telehealth visits in the setting of a pandemic in a large pediatric primary care network may help guide policy making around appropriate use of telehealth to treat infections in children. Telehealth is likely to remain important postpandemic to meet consumer demand for access and comfort. If antibiotic stewardship is to be maintained in such a setting, the question arises of how to approach telehealth antibiotic prescribing. Studies of adult teledermatology prescribing for common dermatologic conditions, for instance, found high agreement between diagnosis and antibiotic management plans for both telehealth and in-person care.17,18 Common pediatric dermatologic infections should be further studied to determine whether such diagnosis and treatment concordance translates to children as well. Several previous studies were of DTC telehealth, which is outside of the primary care medical home. It is unclear whether there should be more antibiotic prescribing latitude for providers practicing within a medical home with greater knowledge of the patient and family and access to previous history and laboratories. Continued telehealth prescribing highlights the need for guidelines regarding appropriate diagnostic and prescribing criteria. Although the pediatricians in this study were fairly self-regulating in their antibiotic prescribing, guidance is needed for the residual telehealth prescribing seen, as well as to determine whether more liberal prescribing for certain diagnoses and patient demographics may be appropriate. Our study is not without limitations. We are studying a changing landscape. Vaccination against COVID-19 became available for individuals 16 years of age and older in December 2020 and was expanded to include ages 12 through 15 years in May 2021. Children were also not yet fully back in school, so many of the common respiratory infections that are precursors to antibiotic-requiring infections had not returned to prepandemic levels.19 We also used ICD-10 diagnosis codes and not more detailed chart review and thus, we may not be capturing the full clinical presentation of encounters.
Our findings imply that telehealth antibiotic prescribing is here to stay in at least a limited form after a surge in its use during the initial phases of the COVID-19 pandemic. Skin and soft tissue and Lyme disease diagnoses remained the most common diagnoses associated with telehealth antibiotic prescribing after the acute pandemic phase and therefore, may be targets for further research and guideline development on appropriate telehealth antibiotic prescribing. More robust guidelines are needed to guide pediatric clinicians on the appropriate use of antibiotic prescribing via telehealth.
Conclusions
In this statewide, large pediatric network study, we found an initial surge in telehealth antibiotic prescribing during the early phase of the COVID-19 pandemic, and a quick decline thereafter, with only a small amount of ongoing antibiotic prescribing from telehealth encounters, particularly for diagnoses that rely on visual examination such as skin and soft tissue infections and Lyme disease. Efforts should be focused on these remaining areas of antibiotic prescribing for targeted clinical guidance.
Dr Payvandi performed the analysis and drafted the initial manuscript; Ms Correa extracted the data and performed the analysis; Drs Vernacchio, O’Donnell, and Hatoun conceptualized and designed the study, reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.
FUNDING: No external funding.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no conflicts of interest to disclose.
Comments