In the United States, ∼60% of antimicrobial expenditures are associated with the outpatient setting, Reference Suda, Hicks, Roberts, Hunkler, Matusiak and Schumock1 and at least 30% of outpatient antibiotic prescriptions are potentially unnecessary. Reference Chua, Fischer and Linder2 Thus, targets of antimicrobial stewardship programs (ASPs) should also focus on outpatient antibiotic prescribing.
Telemedicine is the provision of health care remotely using various telecommunication tools such as phone visits or mobile devices with or without a video connection. Reference Dorsey and Topol3 Before the coronavirus disease 2019 (COVID-19) pandemic, telemedicine was relatively uncommon in the United States, although its use was gradually increasing. Reference Barnett, Ray, Souza and Mehrotra4 However, since the beginning of the COVID-19 pandemic, telemedicine use has rapidly increased. Reference Alexander, Tajanlangit, Heyward, Mansour, Qato and Stafford5,Reference Joshi and Lewiss6 Although telemedicine could provide an effective and safer alternative to face-to-face visits in many clinical contexts during the pandemic, there is not enough evidence of how provider antibiotic prescribing varies according to the mode of care delivery. Diagnostic uncertainty may increase for some infections because reliable physical examination and diagnostic tests are not always available via telemedicine. Providers may overprescribe antibiotics via telemedicine due to anxiety about missing bacterial infections. Furthermore, patients’ demand for antibiotics and providers’ perceptions of this demand may be different via telemedicine. Previous studies have suggested that telemedicine was associated with increased antibiotic prescribing. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8 One systematic review conducted in early 2020 suggested an association between telemedicine and outpatient antibiotic prescribing in primary care settings. Reference Han, Greenfield, Majeed and Hayhoe9 However, the study results were mixed and too heterogeneous to conduct a meta-analysis. Additionally, that study excluded ambulatory care settings other than primary care.
To address this knowledge gap, we conducted a systematic literature review and meta-analysis of the frequency of outpatient antibiotic prescribing via telemedicine versus face-to-face visits by including all visit settings not limiting to primary care settings to better describe variations in antibiotic prescribing according to the mode of care delivery (PROSPERO registration no. CRD42021228585).
Methods
Systematic literature review and search strategies
This systematic literature review and meta-analysis were conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement Reference Liberati, Altman and Tetzlaff10 and Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines. Reference Stroup, Berlin and Morton11 Search strategies were developed with the assistance of a health sciences librarian with expertise in searching for systematic reviews in December 2020 and January 2021. The literature search included publications from database inception to January 15, 2021. Comprehensive strategies, including both index and keyword methods, were devised for the following databases: PubMed, CINAHL, Embase (Elsevier platform) and Cochrane CENTRAL. To maximize sensitivity, no pre-established database filters were used. The full PubMed search strategy (Supplementary Table 1 online) was adapted for the other databases. In addition to the database searches, references of 14 relevant papers were located using the Scopus database. Publications were included if they evaluated the frequency of antibiotic prescribing in outpatient settings via telemedicine. Studies were excluded if they did not have a control group (ie, face-to-face visits). Titles and abstracts of the studies identified by the initial literature search were screened (by H.S.) to assess inclusion criteria. The authors of 6 studies were contacted to provide additional information needed for meta-analysis. Among them, 1 author provided additional information, and that study was included in the meta-analysis. Reference Tan, Mason and Gonzaga12
Table 1. Summary of Study Characteristics
Note. UTI, urinary tract infection; URI, upper respiratory infection; OM, otitis media; ARI, acute respiratory infection; D&B score, Downs and Black score; PCP, primary care physician.
Data abstraction and quality assessment
Of 3 independent reviewers (H.S., A.R.M., and S.H.), 2 abstracted data for each article using a standardized abstraction form. The reviewers abstracted data on publication year, study location, study setting, study design, study period, inclusion of adults and/or children, type of telemedicine, type of face-to-face visits, infectious diagnoses for which antibiotics were indicated, a definition of guideline-concordant antibiotic management, and an assessment of the potential risk of bias. Our primary outcome was the frequency of antibiotic prescribing via telemedicine and face-to-face visits, defined as the proportion of total visits in which an antibiotic was prescribed. As a secondary outcome, we evaluated guideline-concordant antibiotic management. We decided to conduct meta-analyses for individual diagnoses but not all diagnoses together.
The risk of bias was assessed by independent reviewers using the Downs and Black scale. Reference Downs and Black13 All questions of the original Downs and Black scale were answered as intended except a categorical question that we changed to a dichotomous answer for convenience. The maximum score was 28 points. Studies that scored 18 points or more were considered high quality. For data abstraction and quality assessment, inconsistent assessments were resolved by discussion.
Statistical analysis
To estimate the pooled odds ratio (OR) and 95% confidence interval (CI) for each infection, we used random-effects models with inverse variance weighting. We performed stratified analyses by the modes of telemedicine, location of face-to-face visits, adults or children, year of publication, and risk of bias according to the Downs and Black scale. Heterogeneity was evaluated with I Reference Chua, Fischer and Linder2 estimation and the Cochran Q statistic test. We used the Cochrane Review Manager (Revman) version 5.4 (The Nordic Cochrane Centre, The Cochrane Collaboration. Copenhagen, 2014). Publication bias was assessed using funnel plots.
Results
Systematic literature review of antibiotic prescribing in telemedicine versus face-to-face visits
Among 3,106 studies screened, 23 studies met the inclusion criteria and were included in the systematic literature review (Fig. 1). Of these 23 studies, 19 were retrospective cohort studies, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14–Reference Schmidt, Spencer and Davidson29 2 were cross-sectional studies, Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31 1 was a case–control study, Reference Miller and Bhattacharyya32 and 1 was a randomized controlled trial Reference McKinstry, Walker, Campbell, Heaney and Wyke33 (Table 1). Of the 22 observational studies, 7 studies used matching between the exposed group (telemedicine) and the nonexposed group (face-to-face visits). Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Yao, Clark, Gogia, Hafeez, Hsu and Greenwald17,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21–Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31 Of 23 studies, 21 were conducted in the United States, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14–Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Hersh, Stenehjem and Daines25–Reference Miller and Bhattacharyya32 1 was conducted in Denmark, Reference Huibers, Moth, Christensen and Vedsted24 and 1 was conducted in the United Kingdom. Reference McKinstry, Walker, Campbell, Heaney and Wyke33 Of the 21 studies conducted in the United States, 4 studies used a claim-based database Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 and the others were conducted either in a single healthcare system or in a primary care network.
Fig. 1. Flow diagram of literature search adapted from PRISMA flow chart.
Of the 23 studies, 11 included only adults, Reference Tan, Mason and Gonzaga12,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14–Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Murray, Penza, Myers, Furst and Pecina26,Reference Penza, Murray, Myers, Furst and Pecina27,Reference Miller and Bhattacharyya32 7 investigated both adults and children, Reference Bruxvoort, Bider-Canfield and Casey19,Reference Ewen, Willey, Kolm, McGhan and Drees20,Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Huibers, Moth, Christensen and Vedsted24,Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31,Reference McKinstry, Walker, Campbell, Heaney and Wyke33 and 3 involved only children. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Hersh, Stenehjem and Daines25,Reference Penza, Murray, Myers, Maxson, Furst and Pecina28 For telemedicine modalities, 17 studies evaluated synchronous video and/or phone visits, Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Davis, Marzec and Blea15–Reference Yao, Clark, Gogia, Hafeez, Hsu and Greenwald17,Reference Ewen, Willey, Kolm, McGhan and Drees20–Reference Schmidt, Spencer and Davidson29,Reference Lovell, Albritton, Dalto, Ledward and Daines31,Reference McKinstry, Walker, Campbell, Heaney and Wyke33 7 evaluated asynchronous text or internet visits, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Murray, Penza, Myers, Furst and Pecina26–Reference Schmidt, Spencer and Davidson29 and 3 neither specified nor separated those 2 modalities. Reference Bruxvoort, Bider-Canfield and Casey19,Reference Gordon, Adamson and DeVries30,Reference Miller and Bhattacharyya32 Also, 16 studies evaluated clinic visits, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18–Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Huibers, Moth, Christensen and Vedsted24,Reference Murray, Penza, Myers, Furst and Pecina26–Reference McKinstry, Walker, Campbell, Heaney and Wyke33 9 evaluated urgent care, Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Davis, Marzec and Blea15,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Hersh, Stenehjem and Daines25,Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31 5 evaluated emergency departments, Reference Yao, Clark, Gogia, Hafeez, Hsu and Greenwald17,Reference Bruxvoort, Bider-Canfield and Casey19,Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31 and 4 evaluated retail clinic visits. Reference Murray, Penza, Myers, Furst and Pecina26–Reference Penza, Murray, Myers, Maxson, Furst and Pecina28,Reference Gordon, Adamson and DeVries30
The most commonly evaluated indication was sinusitis, which was reported in 10 studies, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Davis, Marzec and Blea15,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21–Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Penza, Murray, Myers, Furst and Pecina27,Reference Gordon, Adamson and DeVries30,Reference Miller and Bhattacharyya32 followed by upper respiratory infection (URI), which was reported in 6 studies, Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 urinary tract infection (UTI), which was reported in 5 studies, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Bruxvoort, Bider-Canfield and Casey19,Reference Murray, Penza, Myers, Furst and Pecina26,Reference Gordon, Adamson and DeVries30 pharyngitis, which was reported in 5 studies, Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 and otitis media, which was reported in 4 studies. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22 Finally, 15 studies earned 18 points or more in the Downs and Black scale and therefore were considered high-quality studies (Supplementary Table 2 online). Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Yao, Clark, Gogia, Hafeez, Hsu and Greenwald17–Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Murray, Penza, Myers, Furst and Pecina26–Reference Penza, Murray, Myers, Maxson, Furst and Pecina28,Reference Gordon, Adamson and DeVries30,Reference Lovell, Albritton, Dalto, Ledward and Daines31,Reference McKinstry, Walker, Campbell, Heaney and Wyke33
Stratified analyses based on type of infection
Otitis media
Four retrospective cohort studies evaluated antibiotic prescribing for patients with otitis media. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22 In total, 1,033 patients (range, 8–603) with otitis media were treated via telemedicine and 71,919 patients (range, 28–41,966) were treated via face-to-face visits. Antibiotic were prescribed for 67.3% of telemedicine and 59.3% of face-to-face visits, respectively. Among these 4 studies, 3 studies Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22 used claims-based data. Also, 2 studies Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21 found more antibiotic prescribing via telemedicine and the other 2 studies Reference Norden, Wang, Desai and Cheung16,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22 did not find a significant difference between the 2 modalities. When those 4 studies were analyzed by meta-analysis, telemedicine use was associated with significantly more antibiotic prescribing compared to face-to-face visits (pooled OR, 1.26; 95% CI, 1.04–1.52) with mild-to-moderate heterogeneity (P = .23; I Reference Chua, Fischer and Linder2 = 31%) (Fig. 2).
Fig. 2. Forest plots for antibiotic prescribing among studies with mild to moderate heterogeneity.
Pharyngitis
Five observational studies compared antibiotic prescribing for patients with pharyngitis via telemedicine and face-to-face visits. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 In total, 1,378 patients (range, 40–669) with pharyngitis were treated via telemedicine and 66,841 patients (range, 82–28,433) were treated via face-to-face visits. Antibiotics were prescribed for 63.4% of telemedicine visits and 61.3% of face-to-face visits. Furthermore, 2 studies Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 found more antibiotic prescribing via telemedicine, and the other 3 studies Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21 did not find a significant difference between the 2 modalities. Those 5 studies were highly heterogenous (Supplementary Fig. 1 and Supplementary Table 3 online). When limited to more recent studies conducted in 2018 and after, studies were homogenous (P = 0.66; I Reference Chua, Fischer and Linder2 = 0%), and telemedicine was associated with more antibiotic prescribing compared to face-to-face visits (pooled OR, 1.16; 95% CI, 1.01–1.33) (Fig. 2). In addition to antibiotic prescribing, the utilization of streptococcal testing was evaluated in 2 studies. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21 Although streptococcal testing was ordered in ∼70% of face-to-face visits, it was ordered in only 1%–4% of telemedicine.
Urinary tract infection
Four observational studies were included in the meta-analysis for UTI. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Murray, Penza, Myers, Furst and Pecina26,Reference Gordon, Adamson and DeVries30 In total, 858 patients (range, 98–243) with UTI were treated via telemedicine, and 5,815 patients (range, 150–2,855) were treated via face-to-face visits. Antibiotics were prescribed for 85.4% of telemedicine and 62.6% of face-to-face visits, respectively. Those 4 studies were highly heterogenous (Supplementary Fig. 1 and Supplementary Table 3 online). When the analysis was limited to 3 high-quality studies, Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Murray, Penza, Myers, Furst and Pecina26,Reference Gordon, Adamson and DeVries30 studies were homogenous (P = 0.41; I Reference Chua, Fischer and Linder2 = 0%), and there was no significant difference in antibiotic prescribing between telemedicine and face-to-face visits (pooled OR, 1.12; 95% CI, 0.87–1.43) (Fig. 2). In addition to antibiotic prescribing, utilization of urinalysis and urine culture was evaluated in 3 studies. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Murray, Penza, Myers, Furst and Pecina26 Both urinalysis (0%–2.7% in telemedicine and 93%–97.1% in face-to-face visits) and urine culture (0%–7% in telemedicine and 21%–73.1% in face-to-face visits) were utilized less frequently in telemedicine.
Sinusitis
Ten observational studies compared antibiotic prescribing for patients with sinusitis via telemedicine and face-to-face visits. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Davis, Marzec and Blea15,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21–Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Penza, Murray, Myers, Furst and Pecina27,Reference Gordon, Adamson and DeVries30,Reference Miller and Bhattacharyya32 In total, 21,640 patients (range, 57–16,168) with sinusitis were treated via telemedicine visits and 588,749 patients (range, 100–469,828) were treated via face-to-face visits. Antibiotics were prescribed for 64.8% of telemedicine and 65.9% of face-to-face visits, respectively. The association between antibiotic prescribing and telemedicine compared to face-to-face visits varied among studies. Also, 5 studies Reference Davis, Marzec and Blea15,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Penza, Murray, Myers, Furst and Pecina27,Reference Miller and Bhattacharyya32 reported more antibiotic prescribing in face-to-face visits; 2 studies Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Gordon, Adamson and DeVries30 reported more antibiotic prescribing in telemedicine; and 3 studies Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Halpren-Ruder, Chang, Hollander and Shah23 did not find a statistically significant difference. Those 10 studies were highly heterogenous (Supplementary Fig. 1 and Supplementary Table 3 online). When studies were limited to 4 high-quality studies that used synchronous telemedicine, Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23,Reference Penza, Murray, Myers, Furst and Pecina27 studies were still moderately heterogeneous (P = 0.11; I Reference Chua, Fischer and Linder2 = 50%), and no statistically significant difference in antibiotic prescribing was observed between telemedicine and face-to-face visits (pooled OR, 0.89; 95% CI, 0.79–1.02) (Fig. 2).
Upper respiratory infection
Six observational studies evaluated antibiotic prescribing for patients with URIs between telemedicine and face-to-face visits. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Norden, Wang, Desai and Cheung16,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22,Reference Gordon, Adamson and DeVries30 In total, 20,668 patients (range, 132–15,852) with URI were treated via telemedicine and 838,116 patients (range, 85–460,646) were treated via face-to-face visits. Antibiotics were prescribed for 39.9% of telemedicine visits and 29.2% of face-to-face visits, respectively. The association between antibiotic prescribing and telemedicine compared to face-to-face visits varied among studies. In addition, 4 studies Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Tan, Mason and Gonzaga12,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Gordon, Adamson and DeVries30 reported more antibiotic prescribing in telemedicine, 1 study Reference Uscher-Pines, Mulcahy, Cowling, Hunter, Burns and Mehrotra22 reported more antibiotic prescribing in face-to-face visits, and 1 study Reference Norden, Wang, Desai and Cheung16 did not find a statistically significant difference. Studies were highly heterogenous, and stratified analyses did not identify any homogenous subgroups (Supplementary Fig. 1 and Supplementary Table 3 online).
Guideline-concordant antibiotic management
Guideline-concordant antibiotic management for patients with sinusitis was compared between telemedicine and face-to-face visits in 5 studies. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23 Guideline-concordant management was assessed by the choice of guideline-concordant antibiotics in 4 studies, Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21 and antibiotic prescribing only for complicated sinusitis (diagnosed based on history) in 1 study. Reference Halpren-Ruder, Chang, Hollander and Shah23 Also, 2 studies found more guideline-concordant management in telemedicine, Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23 1 study found more guideline-concordant management in face-to-face visits, Reference Mehrotra, Paone, Martich, Albert and Shevchik7 and another 2 studies did not find a significant difference between the 2 modes of delivery. Reference Ray, Shi, Gidengil, Poon, Uscher-Pines and Mehrotra8,Reference Johnson, Dumkow, Burns, Yee and Egwuatu18 These 5 studies were highly heterogenous (Supplementary Fig. 1 and Supplementary Table 3 online). When the analysis was limited to 3 high-quality studies, Reference Johnson, Dumkow, Burns, Yee and Egwuatu18,Reference Shi, Mehrotra, Gidengil, Poon, Uscher-Pines and Ray21,Reference Halpren-Ruder, Chang, Hollander and Shah23 there was moderate heterogeneity (P = .12; I Reference Chua, Fischer and Linder2 = 53%), and telemedicine use was associated with significantly more guideline-concordant management (pooled OR, 1.33; 95% CI, 1.01–1.76) (Fig. 3). Guideline-concordant management for other diagnoses was not investigated due to a small number of studies.
Fig. 3. Forest plot for guideline concordant antibiotic management for sinusitis, limited to high-quality studies.
Publication bias
We assessed publication bias by creating funnel plots for studies evaluating each diagnosis (sinusitis, URI, UTI, pharyngitis, and OM) (Supplementary Fig. 2). Aside from studies with extreme odds ratios (<0.2 or >5), studies were reasonably balanced around the pooled odds ratios, and studies with null results were included. Thus, there was little evidence of publication bias.
Discussion
In this systematic literature review and meta-analysis, antibiotics were more frequently prescribed via telemedicine compared to face-to-face visits for patients with otitis media and pharyngitis. Telemedicine was associated with more guideline-concordant management for patients with sinusitis. Nevertheless, the overall interpretation of those results requires caution because there was substantial heterogeneity among studies.
The decision to prescribe antibiotics is a complex process involving provider factors, patient factors, and external factors. Reference Rose, Crosbie and Stewart34 Outpatient providers’ antibiotic prescribing can be driven by the provider’s anxiety or fear regarding diagnostic uncertainty, complications from an infection, and lack of continuity of care. Additionally, outpatient providers may try to maintain good relationships with patients and increase patient satisfaction by prescribing antibiotics. Reference Björkman, Erntell, Röing and Lundborg35 It is also suggested that outpatient providers may feel that antimicrobial resistance is related to transmission in hospital settings and is not driven by outpatient antibiotic prescribing. Reference Kotwani, Joshi, Jhamb and Holloway36 Additionally, patients’ demand for antibiotics could pressure a provider to prescribe an antibiotic. Reference Kumar, Little and Britten37 Outpatient antibiotic overprescribing is probably a result of tightly interacting provider and patient factors, as well as external factors such as organizational pressures for time and financial incentives. Reference Rose, Crosbie and Stewart34 Through telemedicine, some of these factors may be stronger and others may be weaker. For example, a thorough physical examination is lacking with telemedicine. Also, there would be higher thresholds for ordering lab tests or imaging during telemedicine visits compared to face-to-face visits. The lack of physical examination and diagnostic modalities may increase providers’ anxiety about diagnostic uncertainty; therefore, antibiotics may be prescribed more often. In contrast, patients’ demand for antibiotics and providers’ perception for that might be weaker via telemedicine, especially in the case of asynchronous telemedicine.
In our study, telemedicine visits for pharyngitis and otitis media were associated with higher rates of antibiotic prescribing. One possible explanation for this is the lack of physical examination in telemedicine, which is necessary to make a correct diagnosis for streptococcal pharyngitis. Reference McIsaac, White, Tannenbaum and Low38 The availability of streptococcal rapid testing was lower via telemedicine. Providers may feel that it is easier to prescribe antibiotics to treat pharyngitis, rather than pursuing a time-consuming process to obtain reliable physical examination or rapid testing to make a correct diagnosis. Similar to the diagnosis of streptococcal pharyngitis, the diagnosis of otitis media requires an otoscopic examination. Reference Lieberthal, Carroll and Chonmaitree39 Although new technologies, such as digital videoscopes and smart phones, may enable remote ear and oropharyngeal examinations, they are not yet routinely available in primary care settings. Reference Cai, Zheng and Gulati40 Thus, it is possible that otitis media is overdiagnosed in telemedicine settings given the lack of otoscopic examination, and this could be driving increased antibiotic prescribing. Improvements in remote otoscopic examination may eliminate this barrier, but that would require further study.
Contrary to pharyngitis and otitis media, we did not observe a significant difference in antibiotic prescribing for patients with sinusitis or URI when care was delivered during telemedicine or face-to-face visits. For the management of sinusitis, antibiotics are only indicated in cases of severe disease, worsening course, or persistent illness, which can be differentiated with clinical history alone without the need for physical examination or diagnostic tests. Reference Wald, Applegate and Bordley41 Therefore, diagnostic uncertainty may not be greatly different for sinusitis between the 2 modes of delivery. Interestingly, telemedicine was associated with more guideline-concordant management for patients with sinusitis. Although it is possible that there was less patient demand for antibiotic prescribing in telemedicine, the true reason for that observation remains unclear. Diagnosis of URI is ultimately made after excluding other diagnoses that mimic URI. Patients with URI are probably a more heterogeneous group than those with other diagnoses, and it is difficult to make conclusions about antibiotic prescribing for URI with the heterogeneity of the included studies. A significant proportion (30%–40%) of patients with URI received antibiotics even though antibiotics are almost never indicated for URI, indicating room for improvement in future ASP activities.
We did not detect a significant difference in antibiotic prescribing for patients with UTI in our meta-analysis. UTI is a diagnosis for which treatment with antibiotics is almost always indicated. Reference Gupta, Hooton and Naber42 Therefore, it is not surprising that there was not a significant difference in antibiotic prescribing between telemedicine and face-to-face visits. On the other hand, the appropriateness of treatment may be affected by the mode of care delivery because significantly fewer urinalyses and urine cultures were ordered during telemedicine visits. Interestingly, 2 studies that evaluated either first-line antibiotics or guideline-recommended antibiotics showed that telemedicine provided more appropriate treatment. Reference Mehrotra, Paone, Martich, Albert and Shevchik7,Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14 Moreover, studies that investigated revisit as a marker for treatment failure did not report an increase in revisiting after a telemedicine encounter. Reference Johnson, Dumkow, Salvati, Johnson, Yee and Egwuatu14,Reference Murray, Penza, Myers, Furst and Pecina26 Although it is possible that telemedicine can provide similarly effective but lower-cost care for UTI, this hypothesis will need to be validated by future studies.
Our systematic literature review and meta-analysis have several limitations. First, due to the heterogeneity among studies, our findings should be interpreted with caution. The studies varied in the study settings, population, and type of telemedicine and face-to-face visits. Due to the heterogeneity among studies, we elected not to perform a meta-analysis including all diagnoses but rather to conduct meta-analyses for each diagnosis. We also tried to determine the sources of heterogeneity by conducting several stratified analyses, but we could not conduct some of the stratified analyses due to the limited number of studies. Therefore, it is possible that there remained substantial residual heterogeneity. Although we acknowledge this limitation, we believe our findings will provide very important preliminary information for future studies. Second, most of the included studies used administrative codes to identify infections without confirmation by chart review. Administrative codes are not always accurate, but the reported positive predictive values for common infection such as pharyngitis or bronchitis were fairly good, ∼80%. Reference Maselli and Gonzales43 Third, there may be some bias due to lack of information for inclusion in meta-analysis. We asked the corresponding authors of 6 studies to provide additional information, but we could include only 1 study with additional information. Fourth, many of the included studies had significant imbalance in sample size between telemedicine and face-to-face visits. Unmeasured biases may have affected the selection of patients seen via telemedicine. Finally, we did not fully investigate the appropriateness of treatment and follow-up of care, which are also very important components in assessing care variation between telemedicine and face-to-face visits. It is challenging to assess appropriateness of antibiotic prescribing using a retrospective study design without extensive chart review.
The use of telemedicine may change after the COVID-19 pandemic has been controlled, but the adaptation trajectory of these technologies has been forever changed. Reference Weiner, Bandeian, Hatef, Lans, Liu and Lemke44 As we expect continued high-volume use of telemedicine in outpatient settings, it would be very important to correctly understand how telemedicine affects antibiotic prescribing. To validate or further investigate our preliminary findings, large-scale, well-designed studies with comprehensive assessments of antibiotic prescribing for common outpatient infections will be warranted.
In conclusion, our systematic review and meta-analysis found that telemedicine visits were associated with higher rates of antibiotic prescribing in some diagnoses, such as otitis media and pharyngitis. It seemed that providers overprescribed antibiotics for patients with diagnoses where antibiotics were not always indicated via for both modes of delivery. The interpretation of these findings requires caution due to substantial heterogeneity among available studies. Large-scale, well-designed studies with comprehensive assessment of antibiotic prescribing for common outpatient infections comparing telemedicine and face-to-face visits are needed to validate our findings.
Acknowledgement
We thank Dr. Lo Fu Tan for providing additional information about his study. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans’ Affairs or the United States government.
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ash.2021.179
Open access
