Investigating Learning Effects Through the Implementation of Teledermatology Consultations Among General Practitioners in Germany: Mixed Methods Process Evaluation

Abstract

Background: The increasing prevalence of dermatological diseases will pose a growing challenge to the health care system and, in particular, to general practitioners (GPs) as the first point of contact for these patients. In many countries, primary care physicians are supported by teledermatology services. Objective: The aim of this study was to detect learning effects and gains among GPs through teledermatology consultations (TCs) in daily practice. Methods: As part of a mixed methods study embedded in a cluster-randomized controlled trial (TeleDerm), a full survey and semiguided face-to-face interviews were conducted among GPs of participating intervention practices using the telemedicine approach. A TC assessment tool (TC-AT) was developed to evaluate the quality of clinical data and images of TCs conducted during the run-in and intervention phases, with a score ranging from 0 (lowest quality) to 10 (highest quality). Mixed methods analysis triangulated qualitative content analysis, survey data with a growth curve model calculated from TC-AT data, comparing subjective experiences of GPs with objective process data. Results: A total of 487 TCs of 33 practices were analyzed. Questionnaires from n=46 GPs (practice-level response rate: 69.9%) were included in the quantitative analysis. Two-thirds of the GPs (n=31; 67.4%) in the written survey rated the TCs as helpful for differential diagnosis and treatment management. Improved self-reported confidence in diagnosing skin diseases due to the timely clinical feedback from dermatologists was reported by more than half of the responding GPs (n=25; 54.3%). In the interviews (n=13), teleconsultations were mainly seen as a learning opportunity by the GPs. Regarding the quality of TCs, a mean TC-AT score of 7.4 (SD 1.7, range 0‐10) was observed. In the growth curve model, a simple linear time trend provided the best fit to the TC-AT score trajectory across the observed study period. A significant time * TC-AT start score interaction was found (F452=30.66, P