The Consensus-Based Checklist for Reporting of Survey Studies (CROSS) [15] was used for reporting this study. In June 2019 (t₁), an anonymous, quantitative cross-sectional survey both on the web [16] and offline through a paper-based version was conducted among all German dermatologists that provided care in medical practices and clinics. The questionnaire contained items on knowledge, interest, expectation, and use of digital applications. The following demographic data were collected: the dermatologist’s age, sex, and postal code of the office or clinic. The survey was also conducted in June 2021 (t₂), including a few additional items; the t₁ survey had 23 items, while the t₂ survey had 29 items (23 items were identical to t₁ and 6 items were related to developments due to the COVID-19 pandemic, medical device–classified smartphone apps, ie, DiGAs, AI, the electronic health care professional card [elektronischer Heilberufeausweis; eHBA], and the physician’s place of work (outpatient medical practice or hospital clinic).

The majority of items were on a 5-point Likert scale. Further items with 2-4 response categories were used (eg, great importance, moderate importance, and no importance). The full questionnaires are provided in Multimedia Appendix 1. Participants had the opportunity to provide written reasons for their responses.

At both time points, all members of the Federal Association of German Dermatologists (BVDD) and the German Dermatological Society (DDG) providing care in medical practices and clinics were invited to anonymously answer one of the questionnaires (on the web or offline). During the recruitment period of 4 weeks, the dermatologists received multiple reminders for participation. The total pool of addresses in both years was approximately 5900 dermatologists, but this also included members who did not practice dermatologic care anymore or had an outdated address. After considering 350 undeliverable mails or emails and an additional 30% of retirees or nonactive dermatologists in the pool, we assumed a sample of 3500 active dermatologists. No sample size calculation was conducted. To reduce errors in data entry regarding the paper-based questionnaires, 25% of all data entries were randomly checked. No errors were identified. During data cleaning, we also checked demographics for duplicates in the web-based survey–generated data.

Statistical analysis was performed with SPSS (version 27; IBM Corporation) for Windows, starting with descriptive statistics. For items with more than 2 answer categories, responses were binary-coded: for items with 5 answer categories, “very often” and “often” were coded as 1, while “rarely,” “never, but planned,” and “never, not planned” were coded as 2; for items with 4 answer categories, “yes and “no, but ordered” were coded as 1, while “no, not ordered” and “no, I reject a connection” were coded as 2; for items with 3 answer categories, “great importance” was coded as 1, while “moderate importance” and “no importance” were coded as 2.

Based on the postal code, participants were assigned to urban and rural regions using data from the Federal Institute for Research on Building, Urban Affairs, and Spatial Development (Bundesinstitut für Bau-, Stadt- und Raumforschung; BBSR) [17]. Age was categorized as 25-39 years, 40-49 years, 50-59 years, 60-66 years, and ≥67 years. For subgroup analysis, age was binary coded: ≥50 years and <50 years.

To test for differences between t₁ and t₂ (the primary aim of the study), t tests (2-tailed) for interval-scaled variables and chi-square tests for ordinal and nominal-scaled variables were performed (significance level of P=.05). Due to the exploratory nature of this research, no correction of the P value for multiple testing (eg, Bonferroni adjustment) was applied to prevent an increased likelihood of type II errors [18,19]. Additional adjustment for demographic and geographic differences between t₁ and t₂ was done through multivariate logistic regression models. Each model included age, sex, urban or rural area, and type of survey (paper or internet-based). Derived from the models, adjusted odds ratios (aORs) and the associated 95% CI are presented. Items that were introduced at t₂ were analyzed descriptively.

To reach the secondary aim of the study, unadjusted differences between subgroups (age (≥50 vs <50 years, sex, and urban or rural) were calculated at t₂. Additionally, aORs and 95% CI were obtained using logistic regressions.

Missing data are reported for items with a rate >5%. We identified high missing values for some variables (age, sex, and region; Table 1). The Little test of missingness [20] was performed and was significant (χ²₃₀₇₂=3,704,038; P<.001). Hence, it was assumed that data were not missing completely at random but were missing at random [20,21]. As recommended, 20 data sets were imputed with fully conditional specification through the multiple imputation algorithm within SPSS (version 22) [22]. Thereby, linear regression was used for age, and logistic regressions were used for binary-coded variables (eg, urban or rural and sex), using all variables. After imputation, all analyses were run on each data set, and the results were pooled in accordance with the Rubin rules [23]. As there were minimal differences in the distribution of values before and after imputation, the imputed values are reported.

For this study, no ethical review was necessary as the study was anonymous and noninterventional. The Ethics Committee of the Medical Association in Hamburg states that no ethical approval is necessary for studies in which data have already been collected anonymously (Ethik-Kommission; Sonstige Studien: Ärztekammer Hamburg [24]). The study was conducted in accordance with the principles of the Declaration of Helsinki and the Guidelines for Good Clinical Practice.

This trend analysis is the first to measure developments in knowledge, interest, expectation, and use of digital applications in dermatology during the events of the COVID-19 pandemic. First, we identified a moderate increase in the use rate of digital applications in dermatological care, while interest in the topic decreased slightly; one-third of practitioners expect great risks with the introduction of digital medicine. Second, we noticed regional and age-related disparities in self-assessed knowledge, confidence, expectations of digital medicine, and the use rate of digital applications. The survey also revealed differences between sexes in self-assessed knowledge and confidence in digital medicine among dermatologists in Germany.

The decline in interest in using digital applications might be explained by the more widespread implementation of digital applications, as verified by this survey. We hypothesize that digital applications may have moved from extraordinary to ordinary daily routines, resulting in a reduction of interest. Confidence and knowledge may have declined as practitioners were more often confronted with the actual use of applications in their daily routine, potentially disclosing the limits of their knowledge in digital medicine.

Interestingly, only one-third of physicians stated that digital applications support their daily activities, which is one of the main promises of health care digitalization. The outcome could be caused by the long, costly, and not yet fully implemented nationwide eHealth initiatives in Germany [25]. The overall perception will, however, differ depending on the type of digital application. The benefit of AI for diagnostic purposes, for example, might be high, as the technology could potentially improve the efficiency of screening and the diagnostic confidence of practitioners [26].

We observed an increase in real-time communication technologies to communicate with patients and other physicians, which could also be shown by other studies [14]. Compared to the United States, where nearly 42.6% of offices stated using video consultations, this service is rarely offered by dermatologists in Germany, as identified within our results (160/792, 7.6%) [27]. Reasons for the low use may involve insufficient reimbursement for providers and a 2019-loosened ban on remote treatments, which also explains the low use rate for this year [14]. In the future, it can be assumed that video consultation with patients may decline after the pandemic fades into the background again, as the main reason for its use was protection from infections [2]. Lower use of video consultations in times of low COVID-19 infection rates could already be shown in Germany [28]. Nevertheless, it is not assumed that the use rates of video consultation will go back to the use rates of the prepandemic era.

On the contrary, the use of asynchronous communication technologies was not affected by the pandemic or other changes during the observed period, which contradicts other results from Elsner [14] at first glance, wherein 75% of dermatologists in 2020 did introduce the technology with the onset of the pandemic. Within our survey, all purposes—administrative (eg, appointments and invoices), medical-related (eg, diagnosis), and technologies (eg, email, SMS text messages, or teledermatological platforms)—were covered by the item, whereas the item in the survey by Elsner [14] was specific to store-and-forward teledermatology, so a medical-related purpose. As the number of dermatologists adopting store-and-forward teledermatology substantially increased, we assume that the purpose of asynchronous communications has partly shifted from administrative purposes to medical-related purposes with the onset of the pandemic.

In terms of the TI, dermatologists are a bit behind other specialist groups: 14% of all practitioners that responded to a nationwide survey [2] and 21% (166/792) of dermatologists within our survey are not yet connected to the TI. Regarding DiGAs, they see a similar importance to other physicians: 40% nationwide of all physicians versus 38.2% (303/792) in our survey. We found a reduction in use of the electronic physician’s letter in dermatology, although most outpatient physicians stated that it eases the administrative workload in clinical practice [2].

Within the survey, age, sex, and regional differences were identified. The lower self-assessed knowledge and confidence in digital medicine by female dermatologists could be confounded by social expectations. In other areas of life, no or only marginal differences in digital literacy between sexes could be measured when accounting for education and employment status [29]. Nevertheless, female dermatologists could thus have similar knowledge of digital medicine, resulting in similar use rates, as shown.

The phenomenon of regional and age disparities, noted in our survey, has generally been noted with the advent of digitalization in all areas of life and is known as the digital divide. Thus, not only access to the internet, internet speed, and availability of appropriate hardware, but also digital competencies among practitioners and patients are related to inequalities in the use of digital tools [30]. In Germany, the average use rates of the internet, as well as digital competence and openness to use digital applications, are more pronounced in urban regions and younger age groups in comparison to rural regions and older groups [31], partially explaining the differences identified within our survey. Consequently, digital applications have the potential to increase unequal access to care, at least in the short term. To ensure equity in the delivery of care in dermatology, all disparities related to sex, age, or region should be addressed by legal, infrastructure, and reimbursement initiatives. In addition, implementation strategies for evidence-based digital applications should be customized to specific target groups [32].

The response rates of our surveys were slightly lower than in previous surveys among German dermatologists [33]. Hence, a certain selection bias among participants cannot be excluded. Considering the official physician statistics, dermatologists aged between 50 and 59 years were slightly overrepresented, whereas dermatologists aged 35 years or younger were slightly underrepresented. In addition, female dermatologists were underrepresented in the survey in comparison to the federal statistic (360/736, 48% in our survey vs 60% at the national level) [34]. Although our surveys were not representative, the sample sizes were large enough to detect trends and variations among demographic and geographic parameters. To improve comparability between survey waves, an adjustment of demographic and geographic differences was performed through multiple logistic regressions. Groups at t₁ and t₂ may differ for unmeasured characteristics (eg, professional expertise, number of consultations per week, socioeconomic patient groups in the area, and the number of private and statutory insured patients), thus resulting in underestimated or overestimated differences. Nevertheless, the same population was contacted at both time points, and important baseline characteristics did not present any larger differences.

The repeated survey revealed an increase in the adoption of the majority of digital applications in dermatology, amplified by the COVID-19 pandemic. However, regional and age-related disparities in the use of digital applications exist, reducing equal access to innovative health care solutions for patients. In the future, policy strategies must be developed to counteract these disparities and improve nationwide implementation. Developers of digital innovations should incorporate physicians’ perspectives more commonly to ensure feasibility and use. Future research should support policymakers and developers by identifying barriers to implementation, monitoring disparities, and ensuring an evidence base for digital applications. Widespread implementation and acceptance among patients and physicians can then improve the efficiency, equity, and effectiveness of care.