One health priority in the United States is to decrease rates of antibiotic prescribing for viral acute respiratory tract infections (ARTIs) and improve guideline-concordant management of ARTIs in outpatient settings (Centers for Disease Control and Prevention [CDC], 2019a). Inappropriate prescribing of antibiotics for ARTIs has been associated with increased rates of antibiotic resistance, increased health care costs, and adverse patient outcomes (Baillargeon et al., 2011; Chitnis et al., 2013; Dantes et al., 2015; Shehab et al., 2016; Thorpe et al., 2018). Inappropriate prescribing of antibiotics and failure to adhere to guidelines for treating common ARTIs has been shown to be more common in urgent care settings when compared with other outpatient settings (Ebell & Radke, 2015; Jones et al., 2015; Palms et al., 2018). Virtual urgent care, a form of telemedicine, is becoming a popular way for providers to care for patients with ARTIs remotely (Elliott & Shih, 2019; Elliott & Yopes, 2019; Ray et al., 2019). Although virtual urgent care is viewed as a convenient and inexpensive way to care for patients with low-acuity illnesses, concern has been raised regarding the potential for increased antibiotic prescribing practices in this clinical setting (Ashwood et al., 2017; Khairat et al., 2019; Martinez et al., 2018a). The purpose of this article is to summarize published research regarding antibiotic prescribing patterns for ARTIs in virtual urgent care settings and determine whether differences exist in antibiotic prescribing and guideline-discordant management of ARTIs in virtual urgent care compared with in-person visits. If virtual urgent care is to be a viable solution for providing cost-saving convenient care, then it must be demonstrated that the quality of care provided is comparable to that which patients receive in traditional in-person office visits. One way of measuring quality in this context is antibiotic prescribing and adherence to guidelines for the treatment of ARTIs.
Background
Clinical practice guidelines specify when it is appropriate to prescribe antibiotics for many common ARTIs seen in outpatient settings (Harris et al., 2016; Rosenfeld et al., 2015; Shulman et al., 2012). These guidelines are accessible on the CDC website and offer management recommendations for common ARTIs, such as acute rhinosinusitis, acute uncomplicated bronchitis, common cold, and pharyngitis (CDC, 2017). Guidelines indicate that although acute rhinosinusitis may be bacterial, the majority of cases are viral and do not require antibiotics (Rosenfeld et al., 2015). Similarly, guidelines indicate that patients with acute bronchitis, bronchiolitis, and nonspecific upper respiratory infections should not receive antibiotics at all (Harris et al., 2016). Guidelines for pharyngitis indicate that patients with two or more of the following should receive a rapid antigen detection test (RADT): fever, tonsillar exudates, tender cervical lymphadenopathy, and absence of cough (Shulman et al., 2012). Patients with a negative RADT should not receive antibiotics (Shulman et al., 2012).
Providers frequently prescribe antibiotics in ambulatory settings for patients presenting with respiratory conditions, even though many ARTIs are viral illnesses and will not respond to antibiotics. An estimated 30%-50% of antibiotic prescriptions in outpatient settings are not appropriate (Chua et al., 2019; Fleming-Dutra et al., 2016). In one US nationwide study, antibiotics were prescribed in 51% of adult outpatient visits for ARTIs (Shapiro et al., 2014). Of the 27 million adult outpatient visits with an antibiotic prescribed in this study, antibiotics were rarely appropriate (Shapiro et al., 2014). A second national study estimated that approximately half of all antibiotic prescriptions for ARTIs were unnecessary, accounting for 34 million antibiotic prescriptions annually (Fleming-Dutra et al., 2016). A third study used national private health insurance claims data for persons up to 64 years (Chua et al., 2019). In this study, inappropriate antibiotic prescribing occurred most frequently for acute bronchitis, acute upper respiratory tract infection, and respiratory symptoms such as cough (Chua et al., 2019).
Failure to adhere to clinical practice guidelines continues to contribute to antibiotic overprescribing practices in outpatient settings (Ebell & Radke, 2015; Fleming-Dutra et al., 2016; Schroeck et al., 2015). Despite the low incidence of Group A Streptococcus in adult patients, one national study examining antibiotic prescribing for 1,107 adults aged 20-64 years presenting with pharyngitis found that 72.4% of visits resulted in an antibiotic prescription (Fleming-Dutra et al., 2016). Although antibiotics are not recommended for patients diagnosed with acute bronchitis or bronchiolitis, 72.4% of 821 patients aged 20 through 64 years diagnosed with acute bronchitis or bronchiolitis received antibiotics for these conditions (Fleming-Dutra et al., 2016). Another study, using descriptive methods and review of administrative data from a regional health system in the Southeastern United States, found that antibiotics were prescribed in 67.8% (n = 11,039) of visits for persons diagnosed with acute bronchitis (Ebell & Radke, 2015). A third study examined antibiotic prescribing for respiratory infections in an outpatient setting among an adult veteran population (n = 1,662), who were otherwise healthy, using retrospective chart reviews (Schroeck et al., 2015). Authors reported that 1,067 (64.2%) of patients received inappropriate antibiotic treatment based on the Centers for Disease Control and Prevention Get Smart guidelines; 77.5% of patients overall received prescriptions for antibiotics (Schroeck et al., 2015). Each of these studies used standardized methods for assessing the appropriateness of treatment.
Virtual urgent care, also known as direct-to-consumer (DTC) telemedicine, is a subset of telehealth and telemedicine (Elliott & Shih, 2019). It allows patients seeking care for a minor illness to contact a provider from any location using their smartphone, tablet, or computer (Ashwood et al., 2017; Elliott & Shih, 2019). Virtual urgent care visits may be carried out between a patient and provider that already have an established relationship, a patient and provider from a practice where the patient has already been seen physically in-person, or a patient and provider that have no previously established relationship (Elliott & Shih, 2019). Virtual urgent care visits may be conducted through a hospital or health system's electronic medical record, or they may be conducted by a for-profit commercial DTC telemedicine company, such as Teladoc or American Well (Elliott & Shih, 2019). Virtual urgent care may be provided synchronously or asynchronously. Synchronous visits involve live video conferencing, chat rooms, or audio-only visits (Elliott & Yopes, 2019). Asynchronous encounters involve text messaging or email and may include audio, images, or other multimedia files (Elliott & Yopes, 2019).
As the telemedicine industry grows and expands, an increasing number of companies are starting to provide their employees with access to virtual urgent care services. The number of large employers offering their employees virtual visits tripled from 2010 to 2012 (Schoenfeld et al., 2016). An estimated 96% of large business insurance plans currently offer virtual urgent care service coverage to employees and their families, and by 2020, it is estimated that all large employers will provide virtual urgent care either directly or through their employee health plan (Ray et al., 2019). Additionally, more than 90% of chief executive officers in health services responding to a US survey reported that they have future plans to offer virtual care services (Elliott & Yopes, 2019). With the rapid growth of virtual urgent care services, it is important to evaluate the quality of health services provided in these settings.
Methods
Objective
Our objective is to summarize published primary scientific literature on antibiotic prescribing patterns for ARTIs and adherence to guideline-concordant management of ARTIs in virtual urgent care settings. Our review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (Moher et al., 2009).
Inclusion and exclusion criteria
We included randomized control trials, cluster randomized trials, controlled before-and-after, cohort, longitudinal, and case-control studies. Participants included adults, 18 years and older, who were treated for ARTI, excluding pneumonia, in a virtual urgent care setting, synchronously or asynchronously. Pneumonia was excluded because this diagnosis often requires an antibiotic (Ebell & Radke, 2015; Fleming-Dutra et al., 2016). Acute respiratory tract infections included sinusitis, suppurative and nonsuppurative otitis media, pharyngitis, URI, bronchitis, bronchiolitis, asthma, allergies, and influenza (CDC, 2019b). We included studies that evaluated synchronous or asynchronous virtual encounters using either DTC telehealth services, such as Teladoc and American Well, or a health system's virtual urgent care service. Comparisons included in-person care settings, including primary care practices, outpatient offices, urgent care centers, retail clinics, and emergency departments (EDs). Outcome measures included frequency of antibiotic prescribing for ARTIs, adherence to antibiotic prescribing guidelines for ARTIs, frequency of broad-spectrum antibiotic prescribing, and frequency of diagnostic testing for conditions that require antibiotics. Other outcomes included occurrence of follow-up visits. Outcome measures were reported as percentages or percentage point differences between study groups.
We excluded studies that only included pediatric patients or if data for pediatric and adult patients were not reported separately. We also excluded studies with patients treated for nonrespiratory infections, such as urinary tract infections or skin infections, studies conducted outside of the United States, studies not written in the English language, and articles not published in the past 10 years. We focused exclusively on the United States to examine prescribing practices in a developed country that was already using telehealth platforms prior to the coronavirus disease (COVID-19) pandemic and that had high rates of antibiotic overprescribing. In addition, United States is of particular interest because it has ranked consistently lowest overall in a series of health services delivery scorecards (Schneider et al., 2021).
Search methods and bibliographic databases
We conducted searches in PubMed, CINAHL, Embase, and Scopus during February 2020. These databases were selected because they are often used for conducting research in nursing and health care disciplines. Our search terms included the following: antibiotic, prescribing, stewardship, utilization, adherence, guidelines, acute respiratory tract infection, upper respiratory infection, sinusitis, bronchitis, outpatient setting, virtual urgent care, and telehealth. We entered included studies into Scopus to identify and screen other studies that cited these studies and reviewed reference lists of included studies to find additional studies that may meet inclusion criteria.
Study selection
We screened, reviewed, and selected studies using a series of steps. One author (K.T.) listed identified titles and abstracts in a table and screened them for potential eligibility, examining study design, participant characteristics, clinical setting, intervention, comparison, and study outcomes. Screening was sensitive rather than specific to avoid missing key articles. Studies were classified as follows: met screening criteria (yes); insufficient information (maybe); and did not meet criteria (no). We obtained full articles for those screened as yes or maybe. One author (K.T.) independently reviewed full-text articles to further assess eligibility.
Data abstraction and management
Two authors (K.T. and J.J.V.) independently read and abstracted included studies using a data abstraction tool based on the Cochrane Effective Practice and Organisation of Care Review Group (EPOC, 2017) data collection checklist. Data included the reference, study purpose, study design, country, study period, intervention and comparison descriptions, sample inclusion and exclusion criteria, characteristics of participants, sample sizes, measures and measurement, data collection and analysis methods, results, and limitations. We resolved abstraction disagreements between investigators by formal reconciliation to achieve consensus.
Two authors (J.J.V. and S.O.) assessed risk of bias of nonrandomized studies using the Risk of Bias in Non-randomized Studies-of Interventions (ROBINS-I) assessment tool (Sterne et al., 2016). We assessed risk of bias for confounding, participant selection, intervention classification, deviations from intended intervention-effect of assignment and effect of starting and adhering to intervention, missing data, outcomes measurement, and selection of the reported result. We scored each criterion, for each study, as low, moderate, serious, or critical risk of bias, or no information. For ROBINS-I, a low risk of bias score was interpreted as "the study is comparable to a well-performed randomized trial"; moderate as "sound for a non-randomized study [horizontal ellipsis] but [not] comparable to a well-performed randomized trial"; serious as "important problems"; critical as "too problematic [horizontal ellipsis] to provide any useful evidence on the effects of the intervention"; and no information means insufficient information to make a judgement (Sterne et al., 2016, p. 22).
Data analysis, synthesis, and presentation of findings
We present characteristics of included studies, including: study designs, participant characteristics, patient conditions, interventions and comparisons, modalities, settings, and outcomes. We report outcomes for individual studies; meta-analysis was not conducted because it was not expected that pooling of outcomes from included studies would be meaningful based on the number of studies and heterogeneity in some outcomes (Ryan & Cochrane Consumers and Communication Review Group, 2016). Outcomes are reported as frequencies, percentages or proportions, percentage point changes, and/or relative risks. We summarized findings by clinical setting type for antibiotic prescribing for ARTIs, adherence to ARTI antibiotic prescribing guidelines, broad-spectrum antibiotic prescribing, diagnostic testing for ARTIs that may require antibiotics, and follow-up visits. Findings are displayed in Supplemental Digital Content 1 (Table 1, http://links.lww.com/JAANP/A157). Excluded studies and reasons for exclusion are displayed in Supplemental Digital Content 2 (Table 2, http://links.lww.com/JAANP/A157).
Results
Search results
We identified 745 articles for potential inclusion, omitted 251 duplicates, and screened the 494 unique articles (Figure 1). We obtained and reviewed 31 full-text articles. Of these, 7 were included (Davis et al., 2019; Johnson et al., 2019; Shi et al., 2018; Tan et al., 2017; Uscher-Pines et al., 2015; Uscher-Pines et al., 2016; Yao et al., 2020, see Table 1, Supplemental Digital Content 1, http://links.lww.com/JAANP/A157) and 24 were excluded. Reasons for exclusion include ineligible study designs, outcomes, comparisons, settings, outcomes, treatments, or populations, and not a study (see Table 2, Supplemental Digital Content 1, http://links.lww.com/JAANP/A157).
Included studies
Study designs
All included studies used a retrospective cohort study design. Three studies used matched comparisons.
Patient visits and health conditions
The seven studies included 1,347,986 patient visits for ARTIs. Of these, 42,151 visits were delivered virtually and 1,305,835 in-person. All included studies assessed effects of telehealth patient visits for ARTIs. Two studies focused exclusively on visits for acute sinusitis visits (Davis et al., 2019; Johnson et al., 2019), and one for viral upper respiratory tract infections (Tan et al., 2017). Four studies included patients with broad ARTIs, including sinusitis, streptococcal pharyngitis, otitis media, bronchitis, bronchiolitis, and other viral ARTIs (Shi et al., 2018; Uscher-Pines et al., 2015, 2016; Yao et al., 2020).
Intervention settings
Three studies assessed virtual urgent care visits conducted by a hospital or health care system (Davis et al., 2019; Johnson et al., 2019; Tan et al., 2017). Three studies evaluated virtual urgent care visits conducted by commercial DTC companies (Shi et al., 2018; Uscher-Pines et al., 2015, 2016). One study evaluated telemedicine visits conducted within the ED (Yao et al., 2020). The studies were conducted in the United States, in Colorado, Michigan, Nevada, New York, California, and nationwide. Settings included university health systems, primary care provider networks, medical subspecialty group, large commercial insurer, and health maintenance organization.
Intervention modality
One study conducted virtual visits with an audio and video-enabled monitor (Yao et al., 2020). In three studies, virtual visits were preferentially performed using live interactive video; however, visits were conducted by phone if patients were unable to connect by video (Davis et al., 2019; Shi et al., 2018; Tan et al., 2017). One study conducted virtual visits using a text-based format available from a smartphone or computer (Johnson et al., 2019). Two studies assessed Teladoc, which primarily conducts telephone visits (Uscher-Pines et al., 2015, 2016).
Comparisons
Two studies compared virtual visits with in-person urgent care (Davis et al., 2019; Tan et al., 2017). Three studies included in-person office visits as comparisons (Johnson et al., 2019; Uscher-Pines et al., 2015, 2016). One study included two comparison groups, in-person primary care visits, and traditional urgent care visits (Shi et al., 2018). One study compared virtual visits conducted in an ED with in-person ED visits (Yao et al., 2020).
Risk of bias
Using ROBINS-I, we scored seven studies as low risk of bias for classification of interventions and missing data. We scored six of seven studies as low and one as moderate risk of bias for selection of participants and deviations from interventions. However, all studies involved self-selection in to the virtual care intervention group. Six of seven studies were scored as low risk of bias for the measurement of outcomes, with one study not providing information. We scored seven studies as moderate risk of bias for the selection of reported outcomes and seven studies as serious risk of bias for confounding (Table 1).
Outcomes
Six studies assessed oral antibiotic prescribing (Davis et al., 2019; Johnson et al., 2019; Shi et al., 2018; Tan et al., 2017; Uscher-Pines et al., 2015; Yao et al., 2020). Two studies assessed broad-spectrum oral antibiotic prescribing (Shi et al., 2018; Uscher-Pines et al., 2015). Three studies evaluated guideline-concordant antibiotic prescribing (Johnson et al., 2019; Shi et al., 2018; Uscher-Pines et al., 2016). Other outcomes included guideline-concordant diagnoses (Johnson et al., 2019), streptococcal testing with a streptococcal pharyngitis diagnosis (Shi et al., 2018; Uscher-Pines et al., 2016), and follow-up visits (Johnson et al., 2019; Shi et al., 2018; Tan et al., 2017).
Effects of virtual urgent care visits
Antibiotic prescribing
Six included studies reported prescribing of oral antibiotics, with mixed results (Davis et al., 2019; Johnson et al., 2019; Shi et al., 2018; Tan et al., 2017; Uscher-Pines et al., 2015; Yao et al., 2020) (Table 2). Differences in oral antibiotic prescribing between virtual and in-person care ranged between 25 percentage points lower to 4 percentage points higher with virtual care (Table 2). Two studies compared antibiotic prescribing for acute sinusitis between virtual urgent care and in-person visits within large health systems; both reported less frequent antibiotic prescribing during virtual visits (Davis et al., 2019; Johnson et al., 2019). In the first study, antibiotic prescribing for acute sinusitis was 25 percentage points lower for virtual care (67%, n = 39) than in-person urgent care visits (92%, n = 92) (Davis et al., 2019). Similarly, in the second study, antibiotic prescribing during virtual visits (68.6%) was 24.7 percentage points lower than for in-person visits (94.3%) (Johnson et al., 2019). Four other studies observed smaller differences in antibiotic prescribing between virtual and in-person care, ranging from 1 percentage point lower to 4 percentage points higher antibiotic prescribing rates in the virtual visits groups (Shi et al., 2018; Tan et al., 2017; Uscher-Pines et al., 2015; Yao et al., 2020). One study, conducted in a multispecialty group, reported that antibiotics were prescribed in 25% of virtual and 21% of urgent care visits for upper respiratory tract infection (Tan et al., 2017). Two studies of DTC virtual visits observed 1 to 3 percentage point differences in antibiotic prescribing for ARTIs between study groups. In one, antibiotic prescribing for ARTIs was 1-3 percentage points lower for DTC virtual care (52%) than primary care (53%; p < .01) and urgent care (56%; p < .001) (Shi et al., 2018). In the second study, antibiotic prescribing for ARTIs was 3 percentage points higher in DTC Teladoc visits (58%) than in-person visits (55%; p = .07) (Uscher-Pines et al., 2015). The last study compared antibiotic prescribing for ARTIs during virtual visits conducted within an ED with in-person ED visits; antibiotics were prescribed in 29% of virtual visits and 28% of in-person visits (p = .846) (Yao et al., 2020).
Broad-spectrum antibiotic prescribing was compared in two studies, with differences in prescribing ranging from 2 percentage points lower to 30 percentage points higher in virtual care visits compared with in-person visits. In one study, broad-spectrum antibiotic prescribing was marginally lower in DTC virtual visits (27%) when compared with primary care visits (29%, p < .001) and urgent care visits (28%, p < .001) (Shi et al., 2018). In a second study, broad-spectrum antibiotic prescribing was 30 percentage points higher in virtual Teladoc visits for ARTI (86%) than in-person visits (56%; p < .01) (Uscher-Pines et al., 2015).
Adherence to guidelines
Three studies assessed adherence to guidelines in diagnosing and treating ARTIs. Differences in guideline-concordant prescribing between study groups ranged from 2.7 percentage points higher concordance with virtual care to 11.2 percentage points lower (Table 2). One study reported a higher proportion of patients being diagnosed using appropriate guidelines among virtual visits for acute sinusitis (69.1%) when compared with in-person care (45.7%; p < .001) (Johnson et al., 2019). However, guideline-concordant antibiotic prescribing for sinusitis was similar for virtual (67.5%) and in-person visits (64.8%) (Johnson et al., 2019). A second study reported similar guideline-concordant antibiotic prescribing between DTC virtual (62%), primary care (60%, p < .001), and urgent care visits (59%, p < .001) (Shi et al., 2018). A third study examined guideline-recommended avoidance of antibiotics for acute bronchitis; this was observed in 16.7% of Teladoc encounters compared with 27.9% (p < .01) among in-person visits (Uscher-Pines et al., 2016) (Table 2).
Diagnostic testing
In both studies that examined streptococcal testing for patients with a streptococcal pharyngitis diagnosis, testing occurred less frequently for virtual care compared with in-person care (Shi et al., 2018; Uscher-Pines et al., 2016). In one claims-based study, rapid streptococcal tests were performed in only 1% of DTC virtual visits for pharyngitis, with appropriate testing completed during 67% (p < .001) of in-person primary care and 78% (p < .001) of urgent care visits for pharyngitis (Shi et al., 2018). The second study similarly observed streptococcal testing in only 3.4% of Teladoc virtual visits for pharyngitis compared with 49.5% of in-person visits (p < .01) (Uscher-Pines et al., 2016). In this study, patients were instructed to contact their primary care clinician or visit an ED if testing was needed (Uscher-Pines et al., 2016).
Follow-up visits
Unplanned follow-up visits occurred with greater frequency among virtual visits compared with in-person visits for two of three studies (Johnson et al., 2019; Shi et al., 2018; Tan et al., 2017). One study reported higher occurrence of unplanned revisits within both 24 hours and 30 days of the initial virtual visit (Johnson et al., 2019), with 6.3 (p = .006) and 7.5 (p = .027) percentage point differences between study groups, respectively (Johnson et al., 2019). A second study reported a 22-percentage point difference in follow-up visits made within 14 days of the initial visit between virtual visits (4%) and in-person urgent care visits (26%, p < .0002) (Tan et al., 2017). The third study did not report effect sizes for this outcome (Shi et al., 2018).
Discussion
In our systematic review, antibiotic prescribing patterns for ARTIs were similar between virtual urgent care and in-person visits for most included studies. However, in two studies, antibiotic prescribing frequency was approximately 25 percentage points lower with virtual care compared with in person care (Davis et al., 2019; Johnson et al., 2019). Differences in oral antibiotic prescribing between care settings ranged between 25 percentage points lower to 4 percentage points higher among the virtual care groups. Use of diagnostic testing for ARTIs was lower in virtual care visits. Additionally, differences in adherence to guidelines and occurrence of follow-up visits varied between studies.
Variations in findings between studies could be explained by several factors. The two studies with the largest gap in antibiotic prescribing between study groups exclusively evaluated visits for sinusitis (Davis et al., 2019; Johnson et al., 2019). Diagnosing sinusitis virtually may differ from other ARTIs in that a sinusitis diagnosis does not require testing or other diagnostics, such as x-rays (Chow et al., 2012; Rosenfeld et al., 2015). In addition, physical examination for sinusitis diagnosis is limited and does not require lung auscultation, but it is often based on history and review of systems, which can be communicated through a virtual visit (Chow et al., 2012; Rosenfeld et al., 2015). In two studies that stratified antibiotic use by diagnosis, one found that antibiotics were prescribed more often for sinusitis during in-person visits when compared with virtual visits (Shi et al., 2018); however, the second study reported no meaningful difference (Uscher-Pines et al., 2015).
The possible relationship between variations in study settings and outcomes is not clear. The two studies with large effect sizes focused exclusively on sinusitis and were conducted within virtual urgent care services embedded within larger health systems. These services had access to the health system's electronic medical record, which offered providers access to patients' medical information (Davis et al., 2019; Johnson et al., 2019). In contrast, providers may not have had access to patients' complete medical records in two studies that examined virtual care services delivered by commercial DTC companies, such as Teladoc (Shi et al., 2018; Uscher-Pines et al., 2015). Two other studies, conducted using platforms that allowed providers to obtain information from the patient's medical chart, reported minimal differences in antibiotic prescribing between study groups (Tan et al., 2017; Yao et al., 2020).
The higher frequency of guideline-concordant sinusitis diagnosis observed in virtual visits in one study may have occurred because providers conducting virtual visits were provided guideline-supported criteria for bacterial sinusitis diagnosis directly within the visit, whereas providers at in-person office visits did not routinely have readily accessible diagnostic guidelines (Johnson et al., 2019). Lower testing frequency in virtual care compared with in-person care may be attributed to the difficulty in conducting testing in most virtual care visits (Shi et al., 2018; Uscher-Pines et al., 2016). The mixed results observed in follow-up visit frequency between studies may partially be explained by variability in reasons for follow-up visits between studies. Two studies were not able to identify whether follow-up visits were planned or unplanned using claims data (Johnson et al., 2019; Shi et al., 2018). In these studies, higher frequency of follow-up visits in virtual care groups could indicate a misdiagnosis, worsening symptoms, or that patients' needs were not properly addressed at the first visit (Johnson et al., 2019; Shi et al., 2018). Alternatively, follow-up visits could have been recommended by providers in virtual care settings more frequently than in-person settings to confirm suspected diagnoses with in-person evaluations or to monitor patients for improvement or worsening symptoms after an initial visit (Johnson et al., 2019; Shi et al., 2018). One study that observed higher occurrence of follow-up visits following in-person visits identified the reasons as worsening condition, development of a complication, or misdiagnosis (Tan et al., 2017).
Comparison to other studies
We did not identify other systematic reviews that compared antibiotic prescribing in DTC virtual care with in-person visits for ARTIs among adults. We identified 1 systematic review of 18 studies that examined outcomes of telemedicine infectious disease consultations, including mortality, hospital readmission, antimicrobial use, cost, length of stay, adherence, and patient satisfaction (Burnham et al., 2019). Clinical outcomes included 30-day mortality after an infection, 30-day readmissions, patient adherence, patient satisfaction, cost-effectiveness, length of stay, and antimicrobial use (Burnham et al., 2019). Of the five studies that examined antimicrobial use, only one study examined ARTIs among adults in DTC telemedicine visits; this study was included in our review (Shi et al., 2018). We excluded several studies of antibiotic prescribing in virtual care settings because they included pediatric patients in their analyses (Gordon et al., 2017; Halpren-Ruder et al., 2019; Lovell et al., 2019; Mehrotra et al., 2013; Ray et al., 2019). Similar to our systematic review, results from these studies reported mixed results.
Limitations
Our review has several potential limitations. First, we limited study inclusion to those reporting outcomes for adults. However, results were also mixed for several studies we excluded because they included pediatric patients. Second, detailed information about experience or qualifications of care providers in each care setting was not described in several studies. It may be important to assess whether provider qualifications and experience are associated with outcomes. Third, virtual visits were conducted through a range of modalities, including telephone only, audiovisual teleconferencing, and text messages. This limits generalizability because variation in outcomes may be partially attributed to the virtual modality. Fourth, with a small number of included studies, we were unable to stratify findings by intervention characteristics to determine which may be more effective.
Included studies in our review also have limitations. Small sample sizes occurred in several studies. Self-selection of patients to virtual care occurred in all included studies; participants in the virtual care groups were typically younger, on average, than for in-person visits. One virtual care intervention was enhanced by guideline-supported diagnostic criteria presented to the provider during the visit, which may have confounded study results (Johnson et al., 2019). One study focused on patients presented to the ED; these patients may have been sicker and more likely to require antibiotics than persons requesting virtual care visits (Yao et al., 2020). Three studies only used data from patients with commercial health insurance (Shi et al., 2018; Tan et al., 2017; Uscher-Pines et al., 2015, 2016). Claims data included filled prescriptions but may miss prescriptions offered but not accepted or filled. Commercially insured patients may also differ from other populations because of the ability to pay and other factors (Barlam et al., 2016).
Implications for future research
This study has several implications for future research. Only seven studies met our inclusion criteria, and variability existed in the interventions, settings, and outcome measures. Additional studies, that compare virtual urgent care with in-person urgent care, are needed to better understand the conditions in which virtual care may be most effective. New studies are needed in diverse health systems, geographic areas, patient populations, clinical conditions, and telemedicine modalities. Future studies should consider comparing the effects of virtual care modality, such as audiovisual, telephone-based, and text-based telemedicine platforms, on outcomes, such as antibiotic prescribing and unplanned visits. Future research may focus on whether care is affected by provider qualifications or specialty when delivering by virtual platforms.
Implications for clinical practice
The identified published scientific evidence is not sufficient to conclude that virtual care visits for ARTIs differ from in-person visits with respect to antibiotic prescribing and adherence to guidelines. However, limited evidence suggests that giving providers easy access to evidence-based guidelines and recommendations for commonly encountered conditions directly within the virtual visit may allow for improved diagnosis and management in this setting (Johnson et al., 2019). Furthermore, practices that use telemedicine platforms for virtual urgent care visits should consider how to potentially improve diagnosis and management of conditions through use of home-based point-of-care testing or accessory "e-tools," such as electronic stethoscopes or otoscopes. Practices that offer virtual urgent care visits should also consider which patients are appropriate for these visits, whether providers will have access to the EMR, and how practices can improve health services access to those who may have broadband or connectivity issues (American Telemedicine Association [ATA], 2015). It is also important for providers to develop processes that facilitate timely access to follow-up care, if necessary, when caring for patients through virtual care platforms (ATA, 2015).
Conclusions
Antibiotic overprescribing and lack of guideline-concordant management is commonly seen when patients are treated for ARTIs in outpatient settings, especially in urgent care settings (Ebell & Radke, 2015; Jones et al., 2015; Palms et al., 2018). Antibiotic overprescribing practices results in increased rates of antibiotic resistance, increased health care costs, and adverse patient outcomes (Baillargeon et al., 2011; Bergman et al., 2004; Chitnis et al., 2013; Dantes et al., 2015; Shehab et al., 2016; Thorpe et al., 2018). With the rapid growth of virtual urgent care settings and the potential for virtual urgent care to improve access to care and patient satisfaction, it is important to identify whether quality is similar between virtual care and in-person care. Antibiotic prescribing is one aspect of ARTI-related quality of care. The results of our review indicate that antibiotic prescribing frequency may be similar between virtual urgent care and in-person care for adult patients treated for ARTIs. However, additional research is needed to further assess the quality of care provided in virtual care settings and interventions that support antibiotic stewardship.
References