The survey population and survey methods have been previously described in detail [18]. This study is reported according to the CHERRIES (Checklist for Reporting Results of Internet E-Surveys; Multimedia Appendix 1) [19]. Inclusion criteria included current US physicians within family practice, internal medicine, hospital medicine, critical care medicine, emergency medicine, and infectious disease. These specialties were selected to represent frontline physicians most likely to care for patients with acute COVID-19 infections. Early in the pandemic, it seemed COVID-19 infections were much less common and much less severe in children; hence, pediatricians were not included in the sample [20]. Specialties, such as obstetrics-gynecology and surgical specialties, that were less likely to be caring for patients with a chief complaint related to COVID-19 infection were considered outside the scope of this study.

A stratified national random sample of 10,000 US physicians was obtained from the American Medical Association (AMA) Physician Professional Data [21]. The AMA Physician Professional Data includes current and historical records for all physicians in the United States, including medical doctors (MDs) and osteopathic doctors (DOs). Formerly known as the AMA Physician Masterfile, it has been used in numerous studies of the physician workforce and physician surveys and found to be the best source of US physician sociodemographic and training information [22-26]. Obtaining a random sample from this comprehensive sampling frame ensured representativeness and provided data that enabled the analysis of both respondents and nonrespondents with weighting to address nonresponse bias. The sample was composed of 4000 primary care physicians (which included both family physicians and internists), 1000 hospitalists, 2000 critical care physicians (which included both critical care and pulmonary critical care physicians), 2000 emergency medicine physicians, and 1000 infectious disease physicians [21]. Hospitalists, intensivists, infectious disease physicians, and emergency medicine physicians were oversampled to help ensure adequate response rates. The survey was open only to this random sample of physicians. Because the response rate was unknown before fielding the survey, the sample size was estimated based on the assumption of a 10%-20% response rate [27].

The survey questions were developed through a series of focus groups and interviews held in May 2020 with physicians from around the United States representing the included specialties. Draft survey questions were developed by the authors (JM and MM) based on key questions identified by these physicians in the focus groups and interviews. Several physicians in the representative specialties, including some who participated in the focus groups or interviews, reviewed the draft survey and provided feedback. The Qualtrics survey was distributed by email through a unique link to each participant in June or July 2020. The survey was distributed through email to the sample described above 3 times over a 3-week period to maximize response rates. The introductory email explained the purpose of the survey, and physicians could respond by clicking the link to start the survey. Participation was voluntary. The survey consisted of 48 questions over 15 pages, with adaptive questioning used to reduce the number and complexity of questions. A back button was available, but respondents were not able to review a summary of responses. Embedded data including a unique study ID were associated with each participant before distribution. These embedded data were automatically tied to each survey response, and duplicates were removed before analysis. IP addresses were not tracked. Responses were automatically captured by Qualtrics. Partially completed surveys were still included in the analysis.

The survey aimed to describe changes to practice during the early period of the COVID-19 pandemic. Primary outcomes were the impact of the pandemic on patient care and practice structure, with a major focus on the adoption of telemedicine. We collected data on factors that might affect telemedicine implementation, including type of practice and confirming specialty. Gender, years in practice, and zip code were provided on the sample by the AMA Physician Professional Data. The survey asked about prepandemic telemedicine use rates and how these changed during the initial period of the pandemic and specifically inquired about telemedicine use rates within the first month of the pandemic and the month preceding survey completion (June 2020). The first month of the pandemic was defined as the first month of the pandemic in the physician’s local area, so varied based on individual physician location. In addition, free text responses were allowed in response to questions inquiring about how telemedicine had impacted physicians’ practice during the pandemic and the most important impacts the pandemic has had on the physicians’ practice.

Consistent with standard survey methodology, surveys returned due to an invalid email were removed from the denominator, as it could not be determined whether the email truly belonged to a person who should have been included in the sampling frame [28]. Responses were weighted to improve their representativeness and to reduce nonresponse bias as previously described [18]. Because some smaller specialties were oversampled to ensure representation, survey design weights were constructed as the inverse probability of selection into the sample. Entropy balancing, a nonparametric generalization of propensity score weighting, was applied to address nonresponse bias [29]. The nonresponse weights were constructed such that the mean, variance, and skewness of the DOs, females, years in practice, and type of practice among respondents matched the full stratified sample.

Data were analyzed in a weighted logistic regression to identify factors associated with the dependent variable: higher or lower odds of telemedicine use in the first month of the pandemic (for the respondent’s region) and in June 2020. Independent variables were included in the analysis based on the research team’s assessment (informed by physician focus groups and interviews) that they might be influential in the level of telemedicine adoption. For those with missing values for telemedicine use during the pandemic, if the respondent answered the first question of the telemedicine section of the survey and continued to complete the first question in the survey following the telemedicine section, we assumed that telemedicine use had remained unchanged and utilized prepandemic telemedicine use. To adjust for local pandemic-related confounding variation in the intensity of the COVID-19 pandemic, we linked each physician’s city and state of residence from the AMA Physician Professional Data to counties. County-level daily COVID-19 case count data were obtained from the COVID-19 Data Repository maintained by the Center for Systems Science and Engineering at Johns Hopkins University. SURVEYLOGISTIC procedure in SAS (SAS Institute) was used for weighted multivariate logistic regressions; SAS version 9.4 and Stata/MP 16 (StataCorp LLC) were used for data processing and analyses.

This study was reviewed by the University of California Davis Institutional Review Board and met the criteria for exemption (IRB ID 1593608). Therefore, informed consent was obtained through respondent review of an email cover letter containing the anticipated survey length, primary investigator contact information, study purpose, and the link to the physician survey. To protect participant privacy, data were collected anonymously, but physicians were offered the opportunity to provide contact information to be included in a lottery for a gift card or to participate in future studies. No other participant compensation was provided.

This cross-sectional national survey of frontline clinicians found that higher rates of telemedicine adoption by early June 2020 were associated with prepandemic telemedicine use and recent local COVID-19 case counts. This is the first study that we know of to compare pandemic telemedicine use across outpatient and inpatient frontline specialties. Increases in the use of telemedicine were noted in all frontline specialties but were most marked for infectious disease, critical care, and primary care physicians.

Across physician gender, specialties, census regions, and years in practice, this study found a substantial increase in telemedicine use in the early months of the pandemic. Previous studies have shown similar rapid telemedicine uptake, but these studies have focused primarily on clinic-based specialties [5,6,32,33]. This study shows telemedicine use rates increased with similar rapidity in the hospital-based specialty of critical care, with less substantial increases seen also in emergency and hospital medicine. Telemedicine use continued to rise through 2021, as pandemic fears persisted and telemedicine infrastructure continued to expand [16].

The increased use of telemedicine during the pandemic among those who had previously used telemedicine is not surprising, as this likely reflects preexisting local infrastructure for telemedicine. Previous use of telemedicine implies hospital and physician readiness to ramp up the use. Multiple previous telemedicine implementation models have emphasized the necessity of preexisting infrastructure, including coordinated hardware and software platforms, audiovisual integration, 24/7 information technology support, and clinician training in achieving telemedicine success [34-36]. Previous work has shown that successful completion of telemedicine relies, in part, on clinician comfort with technology [37].

If we consider telemedicine as a preventative strategy in the face of the pandemic threat to patient and clinician safety, our findings are consistent with the Protection Motivation Theory (PMT), a behavioral theory developed to understand human responses to fear [38]. PMT posits that response to fear-inducing situations is influenced by 2 main factors that are (1) threat appraisal (an individual’s perceived severity of and vulnerability to the threat), and (2) coping appraisal (an individual’s ability to respond to the threat with resources at hand). Applied to this study, telemedicine uptake may be influenced by an individual’s perceived threat from the pandemic, as well as their belief (or lack of belief) that telemedicine will help them respond to that threat, which is likely dependent on both environmental and individual factors. Furthermore, exploration through qualitative analyses may help more clearly explain how PMT factors of threat and coping appraisal impact telemedicine uptake and other adaptations in the face of pandemic threat. Such exploration may have important implications for the adoption of new technologies in responding to future public health emergencies.

It is notable that primary care, infectious disease, and critical care physicians reported higher pandemic telemedicine use than hospitalists and emergency medicine physicians. While several previous studies have evaluated telemedicine use across various outpatient specialties [16,33,39], in free text responses to our survey, several physicians reported that their hospital did not yet have the infrastructure to conduct telemedicine outside of the clinic setting, which could in part explain the higher rates of use among primary care and infectious disease physicians. However, an additional explanatory factor may lie within PMT, as previous studies have suggested primary care and critical care physicians were at the highest risk of contracting and dying from COVID-19 infection [40,41]. Notably, we found higher telemedicine adoption by physicians in regions where the 2-week moving average of COVID-19 cases was high, a situation that increased real and perceived threat, as well as an early rise in telemedicine adoption in the Northeast, the region that experienced pandemic surges before nationwide spread.

It is also notable that several studies have previously reported that telemedicine helps reduce physician burnout [12,14,15]. Therefore, our findings regarding factors influencing telemedicine uptake may have important implications for reducing physician burnout, which has known associations with physician turnover, mental health, and medical error [25,42].

Within the hospital setting, participants’ free text responses noted that telemedicine was often used for remote consultation and family communication. For example, one participant noted that telemedicine allowed for “improved communication with the family diaspora.” Consistent with our findings, another survey study of critical care physicians during the pandemic reported telemedicine was most frequently used in intensive care unit settings for clinician, nurse, and patient communication with patient families [43]. These communications varied from general updates on patient conditions to more in-depth goals of care discussions.

This study was limited by low survey response rates and the potential for selection bias. Our ability to weight responses based on known characteristics of the total sample minimized nonresponse bias, but there is the possibility of enduring bias in unmeasured characteristics. The number of respondents limited our ability to assess differences based on respondent characteristics. Another potential limitation is that of coverage bias, particularly with respect to the undeliverable surveys due to bounced emails. The characteristics of these individuals were unknown, including whether they were still in practice, and these units were therefore eliminated from the study sample, as is standard practice [28].

We did not explicitly collect information regarding prepandemic technology readiness but rather used prepandemic telemedicine as a proxy. Therefore, we cannot draw an explicit association, but rather can only infer that prepandemic technology readiness may have impacted pandemic telemedicine use. It is also possible that patient characteristics and preferences drove telemedicine uptake during the pandemic, but these factors were not evaluated in this study. Finally, since this study focused on the use of telemedicine in the early pandemic response for only a subset of specialties, we cannot provide information regarding telemedicine use in other specialties or regarding the longevity of telemedicine use throughout the pandemic, although other published works have addressed later time points [16].

The most important finding of this study was the capacity for rapid uptake of telemedicine under the right set of conditions—in particular, a preexisting telemedicine infrastructure combined with improved reimbursement and clear evidence of benefit given pandemic risks. Health care is notoriously slow to incorporate innovative, evidence-based technologies. However, our data show that telemedicine uptake in the early pandemic was rapid across genders, specialties, geographic regions, and experience levels. Rogers’ theory of diffusion of innovation posits a 5-step process involving knowledge, persuasion, decision, implementation, and confirmation [44]. This process puts the adopter (in this case, often the clinician) at the center of implementation. However, the pandemic-era implementation of telemedicine highlights the critical role of other factors, such as public policy, external supports, health system or practice and patient infrastructure, and customer demand, in the widespread adoption of a new technology.

Telemedicine use is at a critical juncture as the public health emergency has expired. Clinicians and patients alike have shared its benefits and have developed increasing comfort levels with telemedicine technology; studies during the pandemic showed that patients who received telemedicine visits had higher average patient satisfaction scores than those seen in-person [45]. However, recent studies suggest that both patients and physicians tend to prefer in-person care, seeking to move away from telemedicine in the postpandemic era [46]. Still, telemedicine may be particularly well suited to specific patient care scenarios, such as ongoing medical management of chronic conditions, behavioral health care, and communicating with families about hospitalized patients [9]. Future research must determine optimal applications of telemedicine within each specialty, and what factors will drive its continued use as pandemic fears recede and threat appraisal dissipates.

Continued use of telemedicine will also depend largely on enduring compensation policies. On November 1, 2022, the Centers for Medicare and Medicaid Services issued the 2023 Physician Fee Schedule Final Rule, which began peeling back some of the temporary telemedicine allowances passed in affiliation with the COVID-19 public health emergency. A summary of pandemic-era telemedicine policy changes with actual and impending end dates is provided in Table 3 [47-51].

The potential implications of these looming expirations are far-reaching given the widespread telemedicine uptake across specialties demonstrated in our study and others. The inability to bill for critical care telemedicine, which is largely leveraged to involve remote family members in decision-making and care coordination [43], could impact the ability to provide patient- and family-centered care. Elimination of direct online supervision will dramatically reduce the exposure of trainees to the practice of telemedicine. Loss of compensation for telephone visits will reduce access to care for low-income and rural patients who have less access to broadband internet. Meanwhile, the reinstatement of geographic and originating site restrictions, and the expiration of payment parity rules, could drastically limit telemedicine encounters even in the outpatient setting, resulting in a large-scale reduction in telemedicine use across specialties compared with pandemic levels. While many states have implemented policies requiring payment parity from private payors, other states have not yet implemented such requirements, and these requirements do not apply to Medicaid [52]. A final consideration in the roll-back of policies supporting telemedicine is the potential effect on readiness for the next pandemic. Maintaining support and infrastructure for telemedicine may enable rapid and life-saving transitions to remote care.

The COVID-19 pandemic brought about rapid incorporation of telemedicine across health care. This survey of frontline clinicians found higher rates of telemedicine adoption in response to the pandemic for physicians working in counties with higher COVID-19 case rates and for physicians with higher prepandemic telemedicine use, particularly in primary care, infectious disease, and critical care specialties. These findings have important implications for the ongoing adoption and maintenance of telemedicine to help reduce burnout, as well as key lessons for responding to public health emergencies. Future research must evaluate the use of telemedicine compared with in-person care on health outcomes and address the impact of policy changes on continued telemedicine use. To sustain the use of telemedicine across settings, the potential benefits of telemedicine in providing patient- and family-centered care and the importance of trainee experience in telemedicine must be communicated to policymakers and the public.