In December 2019, a new type of coronavirus causing acute respiratory syndrome (COVID-19) was discovered in Wuhan. COVID-19 spread rapidly around the world and was designated as a Public Health Emergency of International Concern by the World Health Organization (WHO) on January 30, 2020 [1]. By 2 AM CEST (Central European Summer Time) on June 29, 2020, there had been 9,962,193 confirmed cases and 498,723 confirmed deaths [2]. Although lockdown measures have been eased in many countries, a WHO COVID-19 situation report (June 28) still shows a rising trend [3]. This worldwide pandemic has had a far greater impact than expected, and the upward trend is likely to continue in the near future until effective vaccines and antivirals are introduced.

Looking back at the early months of the outbreak, the rapid spread of COVID-19 and the growing number of patients placed a burden on unprepared medical systems worldwide [4,5]. Some countries such as China and Spain set up mobile cabin hospitals to relieve the pressure on their hospitals [6-8]. However, to ensure that the limited health care resources were spent on urgent cases involving severe symptoms, most countries decided that persons with mild symptoms could be isolated at home for a self-recovery trajectory [9]. According to the early WHO studies, about 80% of COVID-19 cases present mild symptoms, and most of these patients should typically be able to recover at home [10-13]. Therefore, disease control centers in various countries (eg, the United Kingdom, the United States, the Netherlands, Italy, and Canada) directed persons with mild COVID-19 symptoms to stay at home and contact their general practitioner by phone instead of directly visiting the hospital. New guidelines for home care were developed by the WHO and countries’ public health departments to present the proper measures for the home care of patients [13-17].

As an alternative solution to conventional health care services, the uptake of eHealth services rapidly expanded during the COVID-19 pandemic [18]; the main fields of application have been telemedicine, remote patient monitoring, and triage and risk assessment [19]. Enabling better response to the pandemic, such digital health care solutions not only reduce the risk of disease transmission thanks to the provision of remote medical care services but also hold the promise to improve the mental health of isolated patients with distance guidance [18-20]. Initial studies have shown that, since the outbreak began, the frequency of internet searches related to “online medical care” has increased significantly, and the public’s interest in eHealth is rising as the number of infected cases climbs [21,22]. Although the limited capacity of the existing eHealth service systems cannot immediately meet this growing demand [22], there is no doubt that the COVID-19 pandemic will impact the current service provision of medical institutions and lead to an accelerated transition to digital health care.

The large number of digital health applications that have been released in response to the COVID-19 outbreak includes a growing number of artificial intelligence (AI) tools; these include tools that make use of natural language data processing and machine learning with big data lakes, such as in decision support agents, advanced self-diagnoses tooling, and AI-enabled mental health interventions [23-27]. These AI tools have the potential to add new service options to remote health care modes such as remote assessment, remote diagnosis, remote interaction, and remote monitoring [24]. The main touchpoints of AI technology appeared to be mobile phone apps, wearable devices, and chat robots [25]. Prior research in the context of public eHealth and disease prevention has found that AI technology improves patients’ health conditions more efficiently [24]. Moreover, application of AI appeared to enable more personalized care pathways based on personal health profiles [18]. The COVID-19 pandemic requires existing health care models to have better integration, delivery, and distribution capabilities, and thus, new requirements for AI in eHealth have been put forward [18,28,29]. Thus far, the main fields that AI technology has been applied to during the pandemic is early detection and diagnosis of infections, personal contact tracking, case and mortality prediction, drug and vaccine development, reducing the workload of medical staff, and other aspects of controlling and managing the spread of the virus [23]. Overall, the emerging AI tools have the potential to perform a useful role in combating COVID-19, and in particular, AI has the potential to improve the quality of home care by providing more personalized, sophisticated, and continuous medical services [18,24].

Despite the potential of AI in digital health services, many attempts to integrate AI technology into health have failed [24,30]. The primary reason for this is that the development of AI tools and applications is predominantly focused on technical and functional aspects, and largely ignores the demands of users and contextual aspects [31]. Data scientists and health professionals usually start developing AI technology without exploring the patient perspective in advance, that is, the health experience and needs of the users. Often users are only involved in providing feedback after a system solution test is carried out or after the final digital service has been released [30]. However, without sufficient understanding of the user experience, AI service applications will not be capable of solving the real and serious problem of a lack of useful value [32,33]. In other words, the real needs of patients with mild symptoms of COVID-19 have not yet been identified. Illustrative of this lack of understanding the real needs, most of the current machine learning algorithms behind decision support tools are too opaque and difficult for users to reconstruct [34]. When AI machines provide treatment suggestions, the mysteriousness of this process basically causes users, including health professionals and patients, to question the result [35]. If AI technology is expected to be used to improve existing eHealth service capabilities, the key actors should focus more on the users rather than technology to reduce the gap between technology and user, and to improve the usefulness of AI [24]. Corresponding findings related to cancer conditions indicated that AI technologies provide a way to transition from a traditional aperiodic “snapshot” monitoring approach to a continuous and longitudinal monitoring paradigm, increase patients’ engagement in their care, and facilitate doctor-patient interaction pathways [36]. In particular, a study that applied machine learning and natural language processing techniques on social media data from online cancer support groups provided new insights toward informed decision making on personalized health care delivery [37]. Likewise, an increasing amount of AI demonstrators evidence a new service potential of AI applications that are hardly used yet in enhanced service providing. To be able to meet personal patients’ needs, in-depth research is required.

To overcome the barrier of the lack of knowledge on what is useful and what is not, AI research and development should involve user-centered design methods to gain insights into the real experiences and needs of users [38], as well as apply a person-centered perspective to construct an explainable AI and make the AI process more transparent and comprehensible to multiple users, including patients and health professionals [39]. As Xu [32] stated, “a useful AI is defined as an AI solution that can provide the functions required to satisfy target users' needs in the valid usage scenarios of their work and life,” which means the user experience should be deeply understood before the AI development starts. However, most of the current research is mainly focused on persons with severe COVID-19 or patients treated at hospitals. We found no literature that explicitly describes the home isolation experience of persons with mild COVID-19 symptoms. Thus, knowledge of this area is still required. This paper intends to address this gap by elaborating on the entire process of the home isolation experience of people with mild COVID-19 symptoms—from infection to recovery—and then extracting in-depth insights for AI concepts in eHealth. Specifically, our research question is how can we improve the home isolation experience of persons with mild COVID-19 symptoms through eHealth services with AI technology?

We employed a design research method in which a qualitative approach was used to explore the home isolation experience of persons with mild COVID-19 symptoms because this is a relatively new field [40]. Our design research had a phenomenology perspective that rests on the philosophical assumptions of studying people’s experiences in their daily living, viewing these experiences as conscious [41]. This phenomenon study provides a real and comprehensive description of the home isolation experience of persons with mild COVID-19 symptoms, which is needed to obtain insights into their user needs and tasks [40], and to find touchpoint interaction needs for the useful application of AI in eHealth.

Data on the home isolation experiences of persons with mild COVID-19 symptoms was collected from the top viewed personal video stories on YouTube and their comment threads. As researchers, the global pandemic meant that we were bound by the necessity to engage in social distancing and limit interpersonal contacts. Therefore, we chose to use personal video stories instead of interview techniques. YouTube, one of the major online video sharing platforms, has become recognized as a valuable social media source for personal stories about health and disease [47]. We chose YouTube as the data source because, compared to other social media platforms such as Twitter, Instagram, and Tik Tok, its content richness and the completeness of the stories presented on it enables more detailed data collection about an entire journey experience. Videos also allowed us as researchers to better understand the feelings of the patients from their nonverbal movements and expressions [48]. The personal video stories and comment threads provided us with a new research opportunity to investigate actively shared experiences instead of relying on actively obtained experiences from interviews. From a researcher’s perspective, YouTube videos eliminated research bias and brought to light unexpected information that those posting them consider important from their personal perspective. The difficulties involved in the use of YouTube concerned the analysis of different narrative structures that posed challenges in extracting, coding, and synthesizing commonly shared experiences.

In the data analysis, triangulation was used by clustering the data from the observation and the transcripts of videos and the comment threads [54].

The patient journey in Multimedia Appendix 2 maps the commonly shared home isolation experiences of persons with mild COVID-19 symptoms. The first key findings concerned an extensive awareness period before testing positive, experiences of less typical and more personal symptoms, a severe negative mood experience curve, and inadequate home health care service support for patients with mild COVID-19 through all stages. Second, the key finding from the analysis of the video’s purpose and the comment threads concerned the benefits and drawbacks of social media support for patients with mild COVID-19. With the third and final key findings, main touchpoint needs during home isolation were synthesized to provide opportunities for AI eHealth concepts.

The stories on personal experiences revealed a considerable period during which the YouTubers became aware of the outbreak of the virus and its public health impact before they related their symptoms to COVID-19. Although most of these persons (P2, P3, and P4) became highly aware of the public health threat (after the prestage of unconsciousness) in less than a week, some of them had low awareness (P1 and P5) and took much longer to do so—from 2 weeks to as long as 2 and a half months (stage 1). All of them went through a period of up to 4 days during which they related the public health situation to their personal condition and symptoms (stage 2), followed by 1-2 days for the testing stage (stage 3). The home isolation period (stage 4) lasted at least 1 and a half weeks but was around 1 month for most (P1 and P2). As none of the YouTubers had yet fully recovered at the conclusion of their video stories, the self-recovery period is expected to last even longer.

Based on the overall analysis of similarities and differences, the symptoms reported in the personal stories appeared to be different from one another. Each of the patients appeared to experience their own specific symptoms. All in all, almost 50 different symptoms were reported, ranging from a mild headache, loss of smell, a stomachache, high temperature, and dizziness to the more critical symptoms of fainting, shortness of breath, and high heart rate (Multimedia Appendix 1, bottom layer). In addition, the occurrence of similar symptoms also appeared to be different over time. For instance, P1 and P2 only had a fever in stage 2, while P4 had a continuous fever until the end. In contrast to these individually experienced physical symptoms, a general consensus was found on negative feelings and deteriorating mood experiences.

The consensus on the negative feelings that all the YouTubers experienced concerned severe anxiety about dying and related feelings of depression and despair.

From the moment that they experienced their first symptom, they experienced severe negative moods that became dramatically worse when the symptoms continued to deteriorate, reaching the lowest level just before testing positive. (The mood experience curve is diagrammed on the top layer of the patient journey in Multimedia Appendix 2.) Surprisingly, testing positive was commonly experienced as an emotional relief. After this, their overall negative mood improved slowly but gradually during the home isolation period of self-recovery. That said, some of them experienced another period in which they felt emotionally broken again and then improved afterward. It is worth noting that, although not all of the patients went through the same ups and downs, overall, they all faced severe feelings of depression and mood fluctuations throughout the journey and especially in stages 2, 3, and 4.

The main differences were that P3 and P5 did not repeatedly look for medical help with no improvement in the second stage while P1 and P2 did. P3 was the only one who had not been tested and did not worry about having limited access to resources because her sister is a doctor and can get timely 24-hour professional help.

As shown in Multimedia Appendix 2, the patient journey followed several distinct stages: prestage with unconscious and low awareness of the public health risk posed by the virus outbreak, experiencing suspected symptoms, relating symptoms to COVID-19, testing and confirmed positive, and quarantine and self-recovery.

The findings from the analysis of the YouTubers’ purposes for sharing their videos and the comments made by viewers confirmed the benefits and drawbacks of the social sharing of public health experiences. The video purpose analysis revealed the commonly shared purpose of going through a difficult time together and receiving support from the audience. Reasons for sharing their personal story were to encourage viewers to pay more attention to protective measures and take social distancing seriously in public; to help viewers relieve their excessive anxiety and fear, and gain a better understanding of mild COVID-19; and to help others with similar experiences by sharing their real condition and self-recovery advice.

The findings from the comment analysis showed that the majority of the 250 commenters (n=166, 66.4%) expressed their likes and thanks to the YouTubers for sharing real COVID-19 experiences and encouraged and blessed them. Of the commenters, 25.2% (n=63) also shared their experiences and feelings, indicating that they can relate to the YouTubers. After watching the video, 13 of them said that they actually realized that they might have mild COVID-19 too. Some (n=26, 10.4%) indicated that they became scared and depressed. A minority (n=48, 19.2%) talked about the public health response to COVID-19 from governments, media, and the public, and asked people to take it more seriously. Home remedies such as vitamin C, elderberry syrup, lemons, and honey were suggested by 7.2% (n=18) of the commenters. Only 5 health care professionals commented. Inappropriate behavior such as going for a walk before having fully recovered were pointed out by 9.2% (n=23) of the commenters, and some (n=10, 4%) of the commenters made jokes.

In summary, the positive influence of personal video stories is that they reach people who are not familiar with the disease yet, they encourage viewers to take mild COVID-19 more seriously, and they provide some emotional relief; their negative influence is that they can spread disinformation and panic.

Based on these key insights of the patient journey map and interaction between YouTubers and viewers, three needs were identified.

This study clarifies the stages, symptoms, mood curve, and touchpoint needs of persons with mild COVID-19 symptoms through mapping the patient journey. This design research of the systematic and in-depth analysis of how patients with mild COVID-19 told their personal stories in their self-shared videos and the comment threads found that persons with mild COVID-19 usually took an extensive period to realize that they were personally experiencing the public health threat of the virus outbreak. They all faced the same problems of severe negative and fluctuating moods while dealing with different symptoms. They lacked adequate and effective home health care service to overcome adversity. The home-isolated persons with mild COVID-19 symptoms turned to their family and friends not only for social support but also for medical assistance and obtained additional emotional support by sharing their stories on social media. Three principal touchpoint needs were identified. First, there is a need to relieve the anxiety caused by the virus by providing reliable public health information. Second, more personal health monitoring and guidance is needed to address personal symptoms. Third, more mental health guidance and social support is required to positively influence the severe moods and emotional problems of those with mild COVID-19.

The theoretical implications concern a new contribution to better understand underserved persons with mild COVID-19 symptoms during their home isolation. As a contribution to the field of AI in eHealth, we propose taking the user-centered findings and embedding them in AI eHealth service touchpoints to improve the home isolation experience.

As the number of patients with COVID-19 is still increasing and some countries are still conducting limited testing, the shortage of global medical resources will persist in the near future, and the demand for more and better eHealth services for patients will continue to rise. To meet the urgent need for public health, it is time to put AI technologies into practice. Although persons who are unfamiliar with new technologies may be less willing to use them, research also shows that, as long as they feel that a specific eHealth service has the ability to improve the quality of treatment, they will intend to try it. Therefore, the target users for eHealth innovation with AI are those who have urgent needs for better medical service—in this study, the target was home-isolated patients with mild COVID-19 symptoms in the context of limited medical resources. In addition, as most of those with mild COVID-19 symptoms are young people, the application of AI can be easier to promote because they are generally more receptive to new technologies [50].

We translated the patient journey insights into value creation for AI innovations in eHealth [43] and designed 3 initial service concepts for an AI application. We used the insights to improve a patient’s experience with eHealth services. The three concepts are based on the premise of an eHealth app used on smart mobile devices.

Although the patient journey mapping is grounded on rigorous and systematic analysis of the qualitative data on the experiences of multiple persons with mild COVID-19 symptoms, this study has several limitations. In this design research, we used videos shared by people on YouTube as the main data source. The advantage of this self-shared data is that these patients have not been influenced by the researcher in advance, and the data is guaranteed to represent the patient perspective, which to a certain extent led the amount of information to exceed the researchers’ expectations. The disadvantage is that unilateral dialogue without questions from researchers also meant that much of the data was irrelevant to the research question, which required the researchers to spend more time on sampling, extracting, and managing the risk that the personal video stories would not provide in-depth answers to some of the subquestions on the patients’ experiences. In this regard, future research that includes additional face-to-face verification procedures is recommended to further enhance the robustness and reliability of the results. Concerning the sample strategy, the limitation of the current sample is that the participants are all YouTubers from the Americas and Europe, who tend to actively share and seek social support more easily than others. Further limitations relate to the racial and socioeconomic status disparities in online narratives that have been documented; in particular, stories by minorities are underrepresented on the internet, including on YouTube [47]. For future research and eHealth service design, more types of personal experience need to be considered. In addition, compared with other videos that are not shared on the social media platform, YouTube videos have comments from viewers under each video. From those comments, we could easily collect different viewers’ opinions on the video content, and we gained insights by analyzing those data. As most comments are composed of short sentences, future research could include AI technologies such as natural language processing and machine learning to efficiently analyze a larger number of comments.

The design of the patient journey map and the underlying insights into the home isolation experience serve to uncover new knowledge and enhance the professional understanding of persons with mild COVID-19 symptoms. The journey mapping synthesized urgent needs for eHealth service touchpoints, for instance, that patients require reliable public health information, personalized health monitoring guidance, and social support. To overcome the inadequate service provision challenges that became apparent in mapping the journey, initial service concepts were proposed for new AI eHealth services to improve the experience of patients with COVID-19 by providing effective health care guidance.