We conducted a qualitative inductive interview study using thematic analysis combined with sentiment analysis to explore patients’ experiences with using the digital platform. Inclusion criteria were having used the digital platform at least once during 2024 and being available for an online interview in Swedish. Sentiment analysis was used to identify differences in user experience based on the frequency of using health care services [36,37].

Approximately 120 PHCCs serving 1.1 million residents in 3 regions in the southeast part of Sweden constituted the population from which participants were recruited. These 3 regions collaborated closely before and after the launch of the digital platform regarding technical optimization, training, and support. The digital platform was accessible via computer, tablet, or smartphone during the opening hours of the affiliated PHCCs, putting patients in touch with their own PHCC (Figure 1). In one of the regions, the digital platform was also staffed during evenings and weekends, but this was by a central regional unit. The platform was neutral in terms of registration, but patients could be referred back to their own PHCC, if necessary. Typically, a contact was initiated by the patient, either by choosing “Health care” or “Administrative matters” from the web portal. Following a patient-initiated contact, the patient was asked to complete an automated symptom checker, which comprised a set of predefined health concerns that generated predefined outcomes based on patient responses. In the subsequent automated triage, the information from the symptom checker was summarized and recorded in a table with a proposed order of priority. An assigned nurse staffing the platform assessed and managed the enquiry. Images could be uploaded by the patient, and additional information about the patient such as patient records could be obtained by the nurse. The nurse could make a second assessment and, if deemed necessary, readjust the order of priority before proceeding. Administrative matters were dealt with separately. In case a staff member was not available to deal with the enquiry immediately, a message would be sent to the patient informing them that the chat conversation would start later. The chat conversation could result in, for example, an appointment with a physician, if possible, the patient’s named accountable care contact (defined as the general practitioner officially assigned to a patient and responsible for that patient’s overall care within a PHCC) or other staff members at the PHCC relevant to the case. These appointments could either be online (chat conversation or video meeting) or at the PHCC. Staff members handled several enquiries on the digital platform simultaneously.

A total of 152,960 contacts have been made via the digital platform in the 3 regions since the launch in 2023 until the end of the study period in December 2024. This was an estimated share of 2% of the total number of contacts with PHCCs in these regions based on information in Vården i siffror, a national website providing current statistics from health care services all over Sweden [38]. Contacts include physical and digital visits. Administrative inquiries, laboratory tests results, or advice are not included. Of these contacts, 61% (92,999/152,690) were made by women, and 39% (59,961/152,960) were made by men (Table 1). The estimated mean age was 37 years. It was a phased introduction of the digital platform; all regions were fully operational by June 2023 [39].

Patients who used the digital platform at least once, starting from January 2024, were asked if they would like to participate in the study. In 2 of the regions, triage nurses at 20 selected PHCCs were contacted and asked to invite patients to participate by having a nurse forward the invitation in the chat conversation at the end of the session. The selection of PHCCs to be approached was supposed to be purposefully sampled based on size, location, and socioeconomic index among the registered population. However, in the end, the selection of PHCCs was limited to those PHCCs that agreed to participate in the study. In the region where the digital platform had a central unit during out-of-office hours, the invitation was offered in the same way, regardless of the PHCC at which the patient was registered. During the recruitment of participants, strategic purposeful sampling was attempted. In the invitation from the nurse, the potential participants were asked to share their name and mobile telephone number, which were then forwarded to the research team who looked for characteristics such as sex and age in order to intentionally include heterogeneity in the group. The participants were then contacted by the research team via SMS text messaging and asked to provide an email address where further information about the research project could be sent. Two reminders were sent. All participants provided their oral informed consent, which was registered.

A semistructured interview guide was developed for the interviews. The guide was compiled by the research team based on evidence from evaluations, reports, and scientific articles regarding similar digital platforms [21,22,25] and was structured around 3 sets of questions: experiences and expectations before using the digital platform, experiences when using the digital platform, and experiences and effects after using the digital platform (Multimedia Appendix 1). Participants were also asked to state age, sex, time of contact with the PHCC, and how often they had visited health care in the past 12 months; rate their digital literacy by choosing one of 6 response options regarding self-rated digital ability (excellent, good, moderate, fair, low, none); and to state whether they would recommend the digital platform to others or not. The numerical cutoff to be considered a frequent user was set at 5 or more contacts per year following the methodology developed by Smith et al [40]. Three of the authors (PE, HTN, FGJ) conducted the interviews. Even though the interviews were structured in line with the interview guide, the interviewers were actively seeking an open conversation in which the participants were free to raise issues not covered in the interview guide [41]. The interviews were conducted using standard video conferencing platforms (eg, FaceTime) via smartphone or computer, but only the audio files were saved and stored. All interviews were transcribed verbatim by a professional transcriber. After the analysis, the interview data were deidentified and given an interview person number based on the time and date of the interview.

Thematic analysis was used, as described by Braun and Clarke [41,42]. An inductive approach was used in which the themes and subthemes were derived directly from the data without any preconceived theories or expectations. Inductive thematic analysis is also referred to as bottom-up analysis as it starts with the data and builds up to themes. A thematic description of the entire data set was developed, maintaining a rich overall description of the predominant and important themes. This is a useful method for investigating under-researched areas or working with participants whose views on the topic are not known [41]. Three of the authors (PE, HTN, FGJ) read all the transcripts. After that, each identified sentences or phrases from their own interviews to form meaning units corresponding to the study aim. The first iteration of coding meaning units was carried out by PE. Every meaning unit was given a code that reflected the original content as much as possible. Codes were translated into English, while subthemes and themes were all formulated directly in English. Quotes were translated by PE. The translation of quotes came after the analysis was completed. Only those quotes that were selected for inclusion in the manuscript were translated, resulting in a total of 19 quotes. Translation was supported by Google Translate. The entire manuscript was revised by a professional language reviewer before submission. The results of the first iteration were jointly reviewed by 3 of the authors (PE, HTN, FGJ). The second iteration included a second reading of the interviews by PE. The codes were revisited, and subthemes and early main themes were suggested. A mind map was also generated by PE to support the process of grouping and refining the subthemes and themes.

The results of the analysis were discussed and commented on by PE, HTN, and FGJ during a first joint workshop. PE continued refining the analysis and presented revised results in a second joint workshop where preliminary major themes were formed and a more consolidated mind map was developed together with a revised report. Based on the outcome of the second workshop, PE continued consolidation and added sentiment analysis as a new feature. Each identified sentence or phrase was identified as being positive, negative, or neutral [43]. Following a lexicon-based method for sentiment analysis [44], positive experiences described satisfaction, benefits or, for example, recommendations to others to use the digital platform. Negative experiences described dissatisfaction, challenges, or unfavorable outcomes. Neutral experiences were neither explicitly positive nor explicitly negative. A ratio of positive to negative sentiments was created for each participant. A ratio was also created for the 2 groups of frequent and infrequent users of health care services to explore the differences between the 2 groups. A chi-square test was used to examine the differences in ratios. A box plot was used to visualize the distribution of differences in sentiment scores between frequent and infrequent users of health care services.

A further refined and developed report was discussed at a third and final workshop attended by the entire research team (PE, FGJ, HTN, EN). During the final workshop, the UTAUT was discussed and identified as a useful lens to better understand and contextualize the findings. The UTAUT constructs include performance expectancy, effort expectancy, social influence, and facilitating conditions. All constructs were used in the analysis. They were applied to support the interpretation of the themes, rather than guiding the initial coding process. Comments and suggestions were included in the final report. MS Excel for Mac 16.93.1 was used for analysis of qualitative data, and STATA 19.5 was used for quantitative data.

Braun and Clarke [41] emphasized the importance of trustworthiness during analysis, which refers to the credibility and reliability of the findings. Trustworthiness was strengthened by a continuous dialogue between the authors and their peers. The first author (PE) performed the analysis, assisted by HTN and supervised by FGJ and EN. An example of the analysis is provided in Table 2.

Ethics approval was obtained from the Swedish Ethical Review Authority (Dnr 2023-05555-01). Supplementary ethics approval was obtained to also include young people who are not of legal age but older than when their parents or guardians have full access to their health records (Dnr 2024-03480-02). Written information about the study was given to all the participants, and all of them gave their informed consent. No compensation was provided to participants for their participation. All data were kept on a secure server in accordance with Linnaeus University guidelines. The study was carried out in compliance with relevant data protection regulations, including the General Data Protection Regulation (GDPR).

A total of 23 patients agreed to participate. Their ages ranged from 16 years to 70 years with a median age of 39 years, and 65% (15/23) were women (Table 3). Some were using the platform on behalf of their children. The 23 interviews lasted from 24 minutes to 48 minutes, and 648 quotes were identified and labelled as being positive, negative, or neutral. Three main themes, “Digital technology's impact on access to services,” “Perceived digital platform functionality and usability,” and “Shifts in the patient-provider dynamic through digital communication,” emerged, with 9 subthemes (Table 4).

Early in the process of forming the themes and subthemes, it became clear that the data were more scattered in content and meaning than expected. The respondents differed in opinions about similar experiences, and the same respondent might report both positive and negative sentiments regarding the same experience. The respondents commented on their own experiences but also on potential consequences for themselves as well as for others using the platform. The wide spectrum of opinions was not a problem but rather reflects the complexity of digital care. Opinions could differ depending on, for example, prior experiences, expectations, and health status. However, the spread created methodological challenges. Including the sentiment analysis resolved some of these challenges. The UTAUT model also provided a structure to better understand and appreciate the complexity of the data. During the analysis, 3 main themes and 9 subthemes emerged. Finally, each subtheme was linked to one or more of the UTAUT constructs.

All of the 648 quotes were coded for sentiment: positive, negative, or neutral. Of the total number of identified meaning units, 307 were categorized as positive, 219 were categorized as negative, and 122 were categorized as neutral. A sentiment score was then calculated for each participant (Figure 2). To explore differences between the 2 groups of frequent and infrequent users, a ratio of positive to negative sentiments was calculated for the 2 groups (Figure 2). The ratio of positive to negative sentiments among frequent users was 1.94, while the ratio of positive to negative sentiments among infrequent users was 0.96. This means that frequent users were more positive in their comments related to their experiences using the digital platform than infrequent users. This difference was found to be statistically significant (n=526; χ²₁=19.08, P<.001). Furthermore, a box plot was used to compare the distribution of the data (Figure 3). It indicates a wider spread among frequent users than among infrequent users, meaning that participants in this group expressed both positive and negative sentiments. Among the group of infrequent users, the spread is narrower, and participants expressed mainly negative sentiments.

This study is one of the few studies to investigate patients’ experiences using a digital platform for chat-based consultations between patients and health care professionals in primary health care services in Sweden.

Overall, the participants appreciated the quick access and swift handling of enquiries through the platform. Most of the participants found the interface to be user-friendly, especially when using a laptop or computer as opposed to a smartphone. Contacting the PHCC using the platform offered discretion and convenience, which was perceived as added value for the participants. The digital contact was also perceived as changing the balance of interaction between patients and health care professionals in favor of the patient, making it easier for certain patient groups, such as patients who felt embarrassed or experienced prejudice about their condition, to get in touch with their PHCC. In this study, IT security was not found to be a major issue. Easy access to EHR notes from the conversations was appreciated.

On the other hand, the participants found it challenging to know when the platform would be their best option, compared with other means of contact. The automated symptom checker was perceived by many participants to be cumbersome and not suitable for its purpose. For some participants, the range of symptoms was too broad, while for other participants, it was too specific and therefore did not offer the support they expected. Participants perceived a conflict between contact through the digital platform and relational continuity. It was perceived as negative and counterproductive to continuity of care. The sentiment analysis of the responses showed that infrequent users of health services were, overall, more negative than frequent users.

Due to the small number of PHCCs participating in the study, the risk of selection bias, which affects transferability and representativeness of the findings, is acknowledged. Ultimately, however, the participating PHCCs exhibited a satisfactory variation in size, location, and socioeconomic status in the catchment area, which contributes to greater reliability in the results. Another limitation is that the study primarily included patients who were able and willing to use a digital platform when contacting primary health care. Patients who are less knowledgeable about digital technology or who have functional impairments were not represented. Their experiences and perspectives are important and may differ from those captured in this study. However, including under-resourced groups would require a different type of recruitment process. Studies investigating under-represented and under-resourced groups (eg, low socioeconomic status, people with impairments, older adults) and patients’ reasons for not using a digital platform for contact if such a platform is available should therefore be encouraged. Such studies can complement our findings and could further deepen the understanding of acceptance and adoption of digital technology such as a digital platform for contact in primary health care.

Much of what we know about digital health care today emphasizes the importance of a person-centered approach that tailors services to the patient’s specific need, rather than relying on standardized solutions to address simple problems [20,22,24,45,46]. In a systematic review, Ladd et al [47] concluded that system issues, particularly relating to triage and access, affected the patient’s ability to sustain relational continuity when using digital platforms for contact with their health care provider. This could cause frustration among patients and ultimately lead to inefficiencies in the provision of care. These results are consistent with the findings in this study. In a study from Canada looking at the impact of virtual visits on primary care use, McGrail et al [28] found that virtual visits are used for less complex issues and therefore appear to complement rather than replace in-person visits. McGrail et al [28] also concluded that, although the digital format improves accessibility, particularly for patients who prioritize convenience, it may be less suited to those requiring patient-centered care. A careful integration into existing systems and structures with a focus on continuity of care is therefore required.

To better understand the factors influencing patients’ use of digital health care services in this study, the discussion was structured using the UTAUT model [33]. This model is useful for examining how performance expectancy, effort expectancy, social influences, and facilitating conditions shape user behavior and the adoption of digital technology in health care [48].

Performance expectancy is the degree to which using technology will provide benefits to users when they perform activities. Regarding performance expectancy, this study indicates that participants value quick and easy access [22,23,49]. The participants expect that contact via the platform will provide a quicker and more convenient path to their PHCC [20,21,23]. For some participants, it did, but the study also indicates that the perceived quick access was challenged when the actual waiting time was not necessarily shorter, rather just transferred downstream. Therefore, participants perceived the actual time it took from initiating contact and completing the consultation to be longer than what they had expected. This is an important finding, as it may negatively influence patients’ willingness to use similar digital platforms. For some, it introduces a sense of uncertainty, leading to skepticism about the effectiveness of the services.

Furthermore, the results highlight patient expectations for more personalized contact, particularly for returning users. The standardized protocol for accessing services through the digital platform did not facilitate such personalization. Although it remains unclear whether this discouraged patients from using the platform, there is an obvious risk that potential efficiency gains could be lost if this is not addressed. Given the platform’s low uptake, this warrants careful consideration.

As aforementioned, most participants expressed both positive and negative sentiments regarding the same experience. Therefore, a sentiment analysis was conducted, to look for patterns in data that could shed more light on the experience. The results indicate that frequent users of health care services were more positive in their responses than infrequent users. These findings are interesting and have not been reported in other similar studies. Patient satisfaction has been found to be closely linked to initial expectations and subsequent perception of performance [50]. It is reasonable to assume that returning patients have expectations that are adjusted based on their previous experiences. Therefore, preferences of returning patients should be carefully considered to achieve efficiency gains from adopting a new technology. On the other hand, it has been reported that patients who value continuity of care are more likely to seek physical care rather than use a digital platform [24]. Similarly, older patients who wanted their requests to be handled by a known health care professional have been found to be less prone to use a digital platform [23]. Failure to recognize the needs of the group of returning patients could reduce the level of uptake [48].

Effort expectancy describes the degree of ease associated with the use of the digital platform. Most participants found the digital platform to be user-friendly and easy to navigate. The participants also expected the contact via the digital platform to be convenient in that it would enable a consultation from home or at work and that it would give them more autonomy compared with other forms of contact. Time to reflect before replying in a chat conversation was also mentioned as something positive.

Effort expectancy was also affected by the fact that the participants found it difficult to understand the kind of situations and enquiries for which the digital platform should be used. Combined with some participants finding it difficult to access the digital platform, this could negatively impact their willingness to use it.

Another concern was the automated symptom checker and the extent to which it was perceived as relevant to the enquiry placed by the patient [22,25,47]. The fact that the automated symptom checker had to be completed before a chat conversation could start caused frustration and sometimes discontinuation among participants. It must therefore be assumed that effort expectancy was negatively affected. The automated medical history-taking was not perceived as supportive when the patient made contact via the digital platform but was rather perceived as an obstacle. It was also perceived as blunt in the sense that it did not interpret and contextualize patient responses but expected the patient to make their own interpretation. Perhaps this is what comes across in the results, when patients stated that they felt uncomfortable not knowing what their responses would lead to next and that they felt that the responsibility was with themselves as patients to interpret their problems. On the other hand, an automated symptom checker can also act as a form of quality control, ensuring that standard questions are asked, which has been found in previous studies [25,27]. This was also acknowledged by some of the participants in this study, and for them, it might improve effort expectancy.

The UTAUT model also includes social influence as an important factor for behavioral intention. This is about users adopting the digital platform due to personal choice, recommendations, or perhaps social pressure. Some participants felt obliged to use the digital platform based on recommendations from their health care provider, even though they themselves did not see any obvious benefits. This is an interesting finding that has not been as prominent in earlier studies. It reveals a complex dynamic between health care providers’ recommendations and patient perceptions, suggesting that, although endorsement from providers can drive adoption, addressing patient concerns and ensuring clear communication about the benefits are crucial for ensuring sustained engagement.

Facilitating conditions are about perceptions of the resources and support available to perform a behavior. In this study, using a computer or tablet was found to be easier for the participants than using a smartphone, and access to and familiarity with how to use these devices were found to be facilitating factors. Most participants scored high on self-reported digital literacy. This is not surprising given that Sweden scores high on digital literacy in general. In 2023, the Swedish Internet Foundation reported that 94% of the population was using the internet on a daily basis [51]. On the other hand, some participants experienced a lack of technical support, which, in some cases, seems to have negatively affected their behavior. For example, some participants discontinued contact and switched to other forms of contact (telephone or physical visit) or even turned to other digital care providers. Other studies have found that digital communication between patients and health care professionals is subject to certain limitations, such as the patients’ level of knowledge [24,31] as well as health care professionals’ competence and ability to master the “digital encounter” [20,27]. Different types of user support integrated into the digital platform, clarifying when and how to best use it, could potentially solve some of the problems identified.

As the uptake of digital health services among users of primary health care is low, innovations such as digital platforms will only have limited effects on efficiency. During the study period, the proportion of contacts via the digital platform in the 3 regions was only about 2% of the total number of contacts. A key issue for health policymakers should be to identify where potential efficiency gains can be found and how digital health technology can capture these gains. This study suggests that frequent users have a more positive attitude toward the digital platform than infrequent users. It is reasonable to assume that these frequent users often include returning patients and those who require continuity of care. A digital platform design that does not prioritize the needs of this group risks missing important opportunities for improving care as well as for achieving much-needed efficiency gains in primary health care.

This study reveals a wide variation in user experiences, with most participants encountering both opportunities and challenges. A sentiment analysis provides more insights into user behavior not reported in similar research earlier. For digital platforms to improve efficiency in primary care, the following key areas should be considered: ease of access, time-saving aspects, and a more individualized contact process. Failure to recognize the needs of returning patients may hinder higher uptake of digital primary care.