Learnings from the design and acceptance of the German COVID-19 tracing app for IS-driven crisis management: a design science research

The world has been facing the COVID-19 pandemic since 2019 (amounting to 140,322,903 confirmed COVID-19 infection cases¹ worldwide). While governments are striving to achieve the most promising strategies to lower infection rates, such as via lockdowns, the full potential of digital tracing apps has not yet been realized in many countries. One existing solution for contact tracing lies in the Exposure Notification Framework (ENF) published by Apple and Google.² In Germany, by its federal state structure, all corona-related non-pharmaceutical interventions, such as local curfews or rules for schools and kindergartens, are decided by each federal state [2]. Thus, the Corona-Warn-App (henceforth the CWA), which is built on the ENF, is the only nationwide response to the current situation in Germany. Therefore, the CWA faced the highest expectations from the beginning and raised several ethical or privacy-related concerns prior to its launch [40]. With a target group of more than 80 million users, the technical aspects, such as the requirements for accessibility, connectivity, and user experience, were quite ambitious. Thus far, the recommended mark of 60% of users (to achieve a significant benefit regarding the infection tracing) among the population has not yet been reached, with the current 23.8 million users [19, 46].³ Furthermore, epidemiologists recommend lowering the incidence value (i.e., new coronavirus disease 2019 infections in a given timeframe for a population) [45, p. 77], to an incidence value of around 25 to disburden local health departments in contact tracing, which could serve as the primary benefit of a digital tracing solution [26]. Thus, there remain open research fields in terms of mass acceptance and the design criteria of the CWA. From a practical viewpoint, the Robert Koch Institut (RKI) published a job vacancy for the CWA advancement during the research timeframe of this paper, indicating needed support in this area.⁴ Complementary research from different disciplines and perspectives, such as epidemiology, mathematical modeling, or information systems, is required to help fight the pandemic [53] and is highly welcome, as proclaimed at the top of the official CWA website (see Fig. 7, “Appendix”).⁵

As COVID-19 tracing is a highly sensitive topic, a large bandwidth of publications concerns possible failures in terms of privacy, security, data accuracy, or epidemiological effects (e.g., [20, 21]). Nevertheless, we identified a research gap in a design-oriented, forward-driven approach in developing, designing, and testing concrete enhancements of existing COVID-19 tracing apps (here, the German Corona-Warn-App). Various authors (e.g., [4, 54]) examine prototypes and vignettes regarding their acceptance, but field studies based on the released application have not yet been published. Therefore, to enhance the design and features of this app and its acceptance, this paper addresses the research gap after the launch and diffusion of the German COVID-19 tracing app.

The potential for improvement of the CWA is evident by comparing the shared tests (56,681) and not-shared positive tests (43,631) for the CWA with the corona-positive tested persons (583,622) in Germany [8, 47]. After all, the number of shared test results outweighs the number of non-shared tests, but the difference in COVID-19-positive non-users of approximately 500,000 people is significant. Thus, this paper aims to address this problem by increasing the acceptance and knowledge regarding the functions and meaningfulness of the app based on a prototypical agile enhancement. In addition to our online survey consisting of 2616 participants from Germany in July 2020, generating plenty of feedback and ideas for further development, we have used renowned practical literature and podcasts from well-known epidemiologists (such as Prof. Dr. Drosten and Prof. Dr. Lauterbach) and the advice of primary experts from the healthcare and IT sectors. Therefore, we investigate the following research questions (RQs):

RQ1: Which design and functional elements of the German COVID-19 tracing app (Corona-Warn-App) should be enhanced to increase its acceptance?

RQ2: What lessons can be learned for applications in crisis management from the challenges of the Corona-Warn-App?

The remainder of this paper is structured as follows: First, we describe the research field and discuss related work. Following this, “Multi-methodological approach” section contains the method of this paper: a design science research approach. In “Results: artefact description” section, we present our prototype. Furthermore, the evaluation of the current CWA and our extended version within three cycles are described in “Evaluation as Part of the Development” section. Based on our results, “Discussion and implications” section provides the practical and scientific implications and a new user journey regarding an enhanced CWA. In addition, we derive general learnings for applications in crisis management. Finally, in “Conclusion and outlook” section, we summarize our results and implications, drawing a conclusion.

Research concerning COVID-19 tracing apps is rapidly expanding. In the field of information systems (IS), multiple research perspectives have evolved—for example, technical specifications such as infrastructure [1, 24, 55], acceptance (e.g., [4, 10, 38, 39, 54], and data security (e.g., [6, 11, 22, 50, 52]). Due to the topicality and the publication pipelines, the research data of these papers is primarily built upon vignettes or prototypes of hypothetical COVID-19 tracing apps (e.g., [4]). Jahnel et al. [23] discuss the data privacy dilemma of corona-tracing apps, particularly for the German COVID-19 tracing app, and conclude that the impact and effectiveness of the CWA for the pandemic situation cannot be measured. High privacy standards are needed to achieve a certain level of acceptance in the public, however, in retrospect, this high standard limits the evaluability of this technology. For future app development in this context, the authors suggest stronger user involvement and voluntary donation of data [23]. A clear focus on the deployed CWA acceptance, front-end, and use itself has been published by Lasarov [32]. In his research, Lasarov conducted a sentiment analysis of approximately 1000 comments regarding the CWA and derived three leveraging categories for the acceptance of the app: (1) usage benefit, (2) data protection (and the related factors privacy, transparency, and trust) and (3) social implications.

In global crisis management, the potential enhancement of the CWA with design science is a valuable investigation that should be shared with the community [53]. Many previous research articles primarily focus on privacy aspects, for example, Stroscher et al. [52] conduct a privacy assessment on the CWA, which could be treated as one aspect of CWA design. In contrast with other research methods, both the scientific outcome and the communication to the public are components of the entire research journey. As the need for broader scientific communication increased also in the IS discipline, design science research (DSR) offers, on one hand, a rigorous and well-elaborated process for IS research. On the other hand, scientific implications do not stand alone as the generated knowledge to be appended to the body of knowledge, but—according to Hevner et al. [18]—creating a communication concept is part of the scientific responsibility.

To strengthen the argument that a DSR approach is recommendable and reasonable in this context, Fig. 1 presents the original user journey⁶ of the CWA, published by the SAP⁷ developer team on Github. It is noteworthy that the CWA was the first worldwide open-source published implementation of the Google/Apple framework, and therefore, it can be understood as an important reference (and blueprint) for other countries. Thus, the limited scope of the user journey must be reflected critically within this paper, as it was not updated by the developers since the first design draft. We question whether this approach is still recommendable in the third wave of the pandemic. Regarding the rush within the development process, we build our research design on potential enhancements: First, as outlined in the user journey, the CWA has been designed in a linear, not a circular, flow. This can be described along a phenomenon in software development, Conway’s law [7]: Organizations design systems with the tendency to represent existing structures rather than follow the agile paradigm to design a user-centered application and adapt internal structures and data flows to these needs. Second, the current user journey simply neglects several use case scenarios, which are part of the medical path that users of the app experience: People overcome a coronavirus disease 2019 infection, but this scenario is not illustrated. Even the official CWA website admits that in the case of recovery, the user must reconfigure the app manually.⁸

As presented in “Background” section, several investigations concerning tracing apps already exist. Nevertheless, there is a research gap for DSR applied to crisis management. This motivates us to investigate feedback regarding the CWA from users, epidemiologists, other health experts, and IT experts. This DSR project is part of a 7-month investigation examining the technology acceptance of the CWA and furthering the development of a prototype with additional functionalities, for which it was valuable to build upon the results of the acceptance study in German society. We followed the DSR paradigm of Hevner et al. [18] and derived meta requirements and design principles. Furthermore, we enriched this approach by preparing and presenting the results in agile development terms for further procedures. Thus, in addition to deriving meta requirements, we included more detailed, functional requirements so that these results could be directly applied in practice. In doing so, we added the DSR view of Peffers et al. [41] to our research: Communicating research artifacts—to the public and professionals—has great importance in the Peffers et al. [41] DSR approach. In the given context, the contribution of this research should be communicated toward app- and policymakers and the open-source community on Github. Figure 2 presents three design cycles and respective evaluation cycles to ensure the identification of any potential for improvement, especially for the enhancement of the functionalities and design of the CWA. The applied procedure contains (1) a survey featuring 1992 participants from Germany for evaluating the current CWA, (2) a gathering of recommendations from epidemiologists in practical literature and from a focus group discussion with IT and health experts, prioritizing and identifying sophisticated functional requirements, and (3) a survey for evaluating the presented enhanced CWA. The survey structures and the interview guideline can be seen externaly (see Additional file 1).

Regarding the first iteration cycle, our initial starting point was the current version 1.0 of the CWA (until 12 August). We surveyed 2616 participants from Germany, in which we examined user acceptance, focusing on reasons for usage and non-usage. The data collection ended in the middle of July 2020; thus, the feedback pertains to the first month of using the CWA, since the app was published in June for the first time. We cleaned the data results by deleting incomplete questionnaire answers and those in which the time needed for the participant to complete the questionnaire was less than four minutes, yielding 1,992 relevant respondents of our survey. On one hand, we analyzed the data of our technology acceptance model (TAM) according to Davis [9] using partial least squares structural equation modeling (PLS-SEM) according to Hair et al. [17] using the analysis tool WarpPLS [27]. On the other hand, we investigated the descriptive results (see Table 4, “Appendix”). As a result, we received feedback and recommendations for improvement from a user’s perspective (see Table 3).

In the second iteration cycle, a set of features was derived to enhance the effectiveness of the CWA from an epidemiological perspective: We extended the feedback of the first round with expert opinions, such as epidemiologists from podcasts (NDR Corona-Virus-Update, UKW, Kekulés Corona Kompass) and online news (e.g., [33]). Therefore, to obtain a comprehensive understanding of possible features, an open Google search was performed using the following search string: ‘“Corona-Warn-App” AND update OR feature* OR “Design Science Research” OR DSR’. Subsequently, prior to performing the in-depth analysis, we coded each feedback into a matrix and assigned each feedback to a newly defined abstracted area (a functional requirement) and a category (an epic). This coding was conducted by two authors independently of each other with an inter-coder agreement of 0.81, which is greater than the defined threshold of 0.8 [30]. Different results were discussed and resolved by the authors. In total, summarizing our feedback into functional requirements for an enhanced CWA served as our basis for development (see “Prototype development: an enhanced COVID-19 tracing app” section  and Table 2). For further evaluation, we discussed and ranked CWA functions and potential future use with a focus group workshop according to Gläser and Laudel [14] (see “Evaluation cycle 2: usefulness and usability” section). Table 1 displays the seven participants involved in this focus group. Based on the ranked functional requirements of the second evaluation cycle, we developed our COVID-19 tracing app prototype using Adobe XD (see “Prototype development: an enhanced COVID-19 tracing app” section). This prototype contains new content management, crowd notification, mask obligation, vaccination, and corona test results that are not yet included in the current CWA. In the final evaluation, we conducted an online experiment with 53 end users. The questions were focused on the navigation of the app regarding the user experience according to Schrepp et al. [49]. Here, the participants were asked to solve five different tasks in the prototype (see “Evaluation cycle III: usability” section).

For both surveys and all items, we used a seven-point Likert-type with multiple-item scales [36] and conducted pretests before the public survey began. As a result, we corrected linguistic mistakes and added more open questions in the event that the participants wanted to share their opinions. Regarding the data collection for all three evaluation cycles, we acquired people through social networks, personal email, and the newspaper.

Our study provides implications on the application level for COVID-19 tracing apps, such as the CWA. From a broader perspective, we can derive learnings for an iterative design approach—as presented in this research—in the context of crisis management and pandemic countermeasures. We believe that this perspective matters substantially under the predictable continuation of global pandemic diseases but even further for the design of digital crisis management artifacts of any kind, whether refugee, climate, or any other global crisis.

In this study, we present concrete enhancements for the German Corona-Warn-App transferable to other pandemic apps, based on an iterative DSR approach, and derived further learnings for crisis management applications. Throughout three rounds of iterations and evaluation, we presented, prototyped, and tested user-centered ideas of improvement for increasing the perceived usefulness of the CWA.

Our findings suggest a strict agile development pipeline. The CWA began from an application engineering perspective, which is today completely solved by the operating system and the ENF. Indeed, the entire user journey appears in a completely different picture during the current second pandemic wave. While the current development team appears to already work in agile settings, we further recommend an agile and therefore ongoing understanding of the German COVID-19 tracing application manager role (i.e., Scrum Product Owner instead of a reporting- and documentation-oriented manager). Thus, it might be recommendable for the involved decision-makers and stakeholders to switch from classic application management to an agile setup, which allows for more flexible reaction to upcoming changes. This argument is strengthened by our data from medical experts acquired in the second evaluation cycle, in which our research shall create the sensitivity in science, public, and government institutions that this pandemic will neither be over soon nor be the last one, as history has demonstrated.

As with every study, this paper is limited by multiple factors. The current situation makes it almost ethically and practically impossible to include physicians in the IS research. Using highly flexible scheduling and digital scheduling tools, we attempted to include experts from the broader health sector regarding shifting hours. Although we appreciate the size and in-depth quality and experience of the achieved focus group, one could argue that such a setting might be biased. Furthermore, as this research is part of a longer DSR project, the initial study was focused worldwide and localized only for Germany. Thus, the primary sample is independent of the third sample of participants for the online survey combined with testing our prototype.

While this paper offers clear recommendations for the feature set of the CWA and some slight UI/UX enhancements, future research might focus on interaction or even gamification aspects to increase the user interaction inside the app.

In all, we must be mindful that the current pandemic situation will not be the last crisis. Thus, it is even more important for governments to have a well-established, flexible, design-oriented process for creating and adapting technology to handle a crisis.