Purposive sampling was used to select hospitals of the 3 different types existing in the Netherlands. In total, 2 university medical centers (UMCs), 3 teaching and 2 general hospitals (including one collaborative oncology hospital comprising 3 general hospitals) were included. Hospitals were selected by means of convenience sampling using the authors’ network or by Web searching, and hospitals in various phases of implementation (contemplation, preparation, or implementation) were included. Contact persons in the hospitals were approached by phone or email. Snowball sampling was used for the selection of respondents, meaning that we informed the contact persons about the objective of the study and also asked them for contact information for 3 stakeholders, including (1) medical professionals (doctor or nurse practitioners [Advanced Practice Registered Nurses]) [30], (2) managers, and (3) IT employees.

If the contact person belonged to one of these groups, they were also asked to participate. Once the stakeholders had agreed to participate, an interview was scheduled with each person individually. In total, 8 hospitals were approached, of which 7 agreed to participate, and 21 subjects participated in the study. No ethical review is needed for this type of study. All participants were informed about the purpose of the study, and participation was voluntary. Verbal consent for audio recording the interviews was obtained for every participant. All data were analyzed and presented anonymously.

The interviews were conducted by the first author (LK). A few days before the interview, each participant received a confirmation email suggesting a scheduled date and time. A document was attached describing the objectives of the study and a topic list for the interview. We also added our own definition of a typical patient portal: “a personal digital environment, facilitated by a health care institution, for example a hospital. Patients need to login to the portal to get access to, for example, their medical file (with results), patient information and appointments. Patients can also fill in questionnaires and receive personalized advice regarding, for example, quality of life and physical activity.” We used a semistructured interview that was structured by applying the comprehensive model of Grol and Wensing [26] that summarizes the barriers to and facilitators of change in health care practice. This model describes 6 levels at which barriers and facilitators can occur: (1) innovation: patient portal; (2) individual professional; (3) patient; (4) social context; (5) organizational context; and (6) economic and political context. All these barriers and facilitators are described in Table 1.

All interviews were performed by telephone and lasted for, on average, 20 min. Participants were first asked for their consent to make audio recordings of the interviews. Then, the purpose of the interview was introduced, and subjects were asked if they received the introductory email. This email was then briefly discussed such that the subjects were aware of the topics to be discussed. After that, questions were asked about participants’ characteristics, such as their age and work experience. To make sure an unambiguous definition of a patient portal was used, participants were asked what their definition of a patient portal was, and if necessary, it was complemented with our definition. Then, we asked them about their perceived barriers to and facilitators of patient portal implementation at all 6 levels [26]. If necessary, for example, if the question was unclear, the interviewer provided examples (and these were also sent per email). At the end of the interview, the participants were asked to suggest additional topics or issues, if any, that had not yet been covered. The interviews were in Dutch, and the questions in Multimedia Appendix 1 are translations.

The first author transcribed all interviews verbatim. Two researchers (LK and WG) independently selected text fragments that reflected a barrier to or facilitator of portal implementation and coded the transcripts in Excel according to the model of Grol and Wensing [26]. A directed content approach was used, which is mainly a deductive approach as a pre-existing model is used for coding [31]. If quotes did not fit into the Grol and Wensing model [26], we looked for categories from the McGinn model [21], which was developed for implementation of EHRs. These models have considerable overlap, but the Grol and Wensing model [26] mainly covers socio-dynamic factors, whereas the McGinn model [21] also covers technical and portal characteristics. For the remaining quotes we created new categories, which is an inductive approach. To enhance clarity and unambiguity of the categories, we renamed them to better reflect the nature of being a barrier or a facilitator. A complete overview of the categories is presented in Multimedia Appendix 2. Coding was discussed between LK and WG until consensus was reached. Saturation of the data was checked by the first author by assessing (post hoc) the percentage of new categories appearing with the analysis of every subsequent hospital.

In total, we interviewed 21 stakeholders from 7 hospitals. We included 3 from each hospital including medical professionals (n=7), managers (n=7), and IT employees (n=7). The stakeholder group labeled medical professionals consisted of medical specialists (n=4) and nurse practitioners (n=3). The group of managers included a medical director (n=1), hospital division or department managers (n=5), and a project manager (n=1). IT employees were application specialists or managers (n=3), an IT manager (n=1), an IT architect and information manager (n=1), and a patient portal project manager (n=2). Mean age was 44.8 years (SD 6.7; range 25-61) and 57% (12/21) were female. We included 6 respondents (6/21, 29%) from UMCs, 9 respondents (9/21, 43%) from teaching hospitals, and 6 (6/21, 29%) from general hospitals. Participants’ work experience varied from 6 years or less (10/21, 48%) to more than 21 years (3/21, 14%). An overview of participants’ characteristics is listed in Table 2.

In total, we selected 376 quotes and identified 26 barriers and 28 facilitators. The results are presented according to the 6 levels of the Grol and Wensing model [26]. The full list of all barriers and facilitators—including the number of subjects for each stakeholder group—is presented in Multimedia Appendix 3. After the inclusion of 7 hospitals (using purposive sampling), we analyzed the data saturation. The data were found to be saturated, meaning that after analyzing the first 6 hospitals, no new categories emerged from the transcripts of the final hospital. We therefore did not include further hospitals.

Due to the high number of identified barriers and facilitators, only those common to all stakeholder groups (medical professionals, managers, and IT employees) are presented here. To demonstrate the similarities and differences between stakeholder groups and between hospitals types, their most mentioned barriers and facilitators are presented as well.

In total, 13 barriers and 12 facilitators (Table 3) were identified that were common to all stakeholder groups. The most relevant barriers and facilitators for each level are presented based on the number of subjects (and percentage of the total subjects) and are highlighted in italics. Quotes are used to illustrate the barriers and facilitators for each level that were mentioned by the majority of the subjects.

We found similarities between stakeholders, for example, regarding perceived usefulness, but also differences (Table 4). Overall, the findings regarding lack of resources were fairly similar among the groups, although the majority (5/7, 71%) of the IT employees also mentioned that there are sufficient resources available. Guaranteeing privacy and security was mentioned by both managers (5/7, 71%) and IT employees (5/7, 71%) as a barrier. The majority of medical professionals (4/7, 57%) and managers (5/7, 71%) mentioned lack of sufficient eHealth literacy of patients as a barrier.

However, we also found differences between stakeholder groups. The negative attitude or opinion of medical professionals was often seen as a barrier, especially by medical professionals. They were most often negative about providing patients with medical information via the patient portal because they were afraid it would lead to more work (such as more questions from patients), and they were worried about losing control. A remarkable finding is that all the managers (7/7, 100%) see their own positive attitude as a facilitator; however, this is true for only less than the half (3/7, 43%) of the other groups. All the medical professionals mentioned the perceived usefulness of the portal, but they (4/7, 57%) also indicated a lack of perceived usefulness because they think that the portal can lead to practical problems. However, the majority of this group is motivated to change (4/7, 57%) compared with only a minority in the other 2 stakeholder groups (both 2/7, 29%).

In Table 5, the top 3 barriers and facilitators for each hospital type are listed. A complete overview of all barriers and facilitators—including the number of subjects for each hospital type—is presented in Multimedia Appendix 4. Differences were found in the barriers mentioned by subjects from different hospital types. The majority (5/6, 80%) of subjects from UMCs mentioned lack of accessibility as a barrier, and the difficult login method was especially seen as a barrier in these hospitals. In general hospitals, most subjects think that the positive attitude or opinion of medical professionals will facilitate implementation because medical professionals are enthusiastic. Lack of time and increased workload is also an important barrier in general hospitals because everybody is already always busy. Along with the differences, we also found similarities between the 3 hospital types. Perceived usefulness was mentioned by all subjects (21/21, 100%), but also lack of resources was seen in every hospital type as an important barrier. The UMCs and general hospitals see that the lack of sufficient eHealth literacy can hinder patient portal use. The most similarities were found between the teaching and general hospitals. Positive attitude, guaranteeing privacy and security, and financial difficulties were mentioned by the majority of subjects in both teaching and general hospitals. This is an important difference from the UMCs, which can perhaps be explained by differences in the financing of these hospital types.

Although we did not explicitly ask the included hospitals in which phase of implementation they were, we could deduce this from the interviews. In total, we included 7 hospitals. Two of these hospitals had no patient portal but were planning implementation. Three hospitals had minimal experience with portals—small pilots with limited functionalities or a classic portal version—but were also in the implementation phase. Only 2 hospitals had an active patient portal; however, stakeholders of one hospital mentioned they were still implementing to extend their current functionalities. In Table 6, we list the barriers and facilitators that were mentioned by (at least one stakeholder) all the included hospitals both with a patient portal (n=2) and without a patient portal (n=5). A complete overview is presented in Multimedia Appendix 5. Although there were similarities (eg, financial difficulties, lack of sufficient eHealth literacy), we also found differences. All hospitals without a patient portal mentioned negative attitude or opinion of medical professionals and lack of specialist staff as barriers. These factors could negatively influence implementation. Although the hospitals with a patient portal see barriers for the implementation of their patient portals, they also mentioned multiple facilitators, for example, perceived ease of use, motivation to change, and sufficient resources. The barriers lack of a generic guideline (n=1) and participation of end users during implementation (n=1) were only mentioned by hospitals with a patient portal. Lack of a generic guideline was a barrier expressed by a manager (n=1), meaning that it could have been beneficial for implementation if there would have been coordination or a standard format. All stakeholders of one hospital that had implemented a portal noticed participation of end users during implementation. In that case, they referred back to the implementation and stated that it was useful to involve end users—both patients and health care professionals—during implementation and for each hospital division to be well represented in the project organization.

In this study, we have presented an overview of the barriers and facilitators related to patient portal implementation among various stakeholders within the hospital organization. In total, we identified 26 barriers and 28 facilitators. Positive factors related to perceived usefulness (eg, cost savings, accessibility for patients to their information) were mentioned by all subjects. The facilitators individuals’ positive attitude and management support (eg, strategy plan for eHealth and patient portals) were also mentioned by majority of the subjects. The main barriers reported were lack of resources (especially lack of staff), financial difficulties (high costs, lack of reimbursement), and guaranteeing privacy and security (eg, strict regulations). We want to emphasize that no inferences can be drawn about the prevalence of phenomena observed beyond the current sample.

We found several similarities between stakeholders (eg, regarding perceived usefulness) but also remarkable differences that highlight the importance of involving multiple stakeholders. One interesting finding is that approximately half the medical professionals see their own positive attitude and motivation to change as facilitators. Although medical professionals’ motivation to change is the highest of all stakeholder groups, lack of time and increased workload was perceived by them as a barrier. Apparently, they are willing to change, but at the same time, they assume that they do not have enough time to achieve implementation and portal use. The barriers guaranteeing privacy and security and lack of resources were mentioned by the majority of IT employees. This shows the challenges this group is dealing with when implementing a secure portal. Managers were the only group of which all (7/7, 100%) stated that they had a positive attitude. This is in clear contrast with the proportion of medical professionals and IT employees (both 3/7, 43%). Managers also stand out in their statements about the culture with more than the half of the managers (4/7, 57%) thinking the culture is hindering implementation, whereas only a minority of both the medical professionals (1/7, 14%) and IT employees (1/7, 14%) stated this. Managers mentioned that hospital culture is conservative and slow to change.

Koivunen et al [28] identified nurses’ barriers and facilitators regarding portal implementation. Their findings were comparable with ours; for example, concerning the barriers lack of resources and lack of time. However, in their study, nurses were included and were mainly negative because they had doubts about the benefits of the portal; moreover, they were unwilling to use a new technical tool because they believed that their primary tasks are to be more important. This differs from our findings as we found positive attitudes among all included stakeholders (medical professionals, managers, and IT employees), and all our subjects mentioned perceived usefulness as a facilitator for patient portal implementation. One reason for these differences may be the selection of stakeholders, as we focused on those directly involved and did not include nurses, only medical doctors and nurse practitioners (“Advanced Practice Registered Nurses”) [30]. Keplinger et al [27] also considered physicians’ attitudes regarding patient portal implementation. Some of their findings are in line with ours, for example, the expected increase in workload and positive attitudes regarding the patient portal. However, they also found differences in attitudes both before and after implementation. For example, before implementation, more than half of the physicians assumed that their workload would increase, whereas only one-third actually experienced such an increase in workload.

McGinn et al [21] showed the relevance of including the perspectives of various stakeholders regarding EHR implementation. Their results are both similar and different from our results. They found that the main factors common to all stakeholder groups were found at various levels and included “perceived ease of use,” “costs,” “motivation to use EHR,” and “privacy and security concerns.” These findings are similar to ours perhaps because financial difficulties, guaranteeing privacy and security, and positive attitude were mentioned by the majority of our subjects. The use of the internet and other electronic applications is becoming increasingly common in health care [33], and patients’ eHealth literacy needs to be taken into account. Participation of end users during implementation was mentioned as a facilitator and can be used to focus on the eHealth literacy of the users.

McGinn et al [21] argue that the consideration of various stakeholder opinions may contribute to successful EHR implementations. Similarities with and differences from our results were found. The main factors common to all stakeholder groups were found at various levels and included “design and technical concerns,” “costs,” “lack of time and workload,” and “privacy and security.” The findings are similar to ours, and this can be the case because both EHRs as well as patient portals are complex technologies that affect multiple levels of an organization. However, we also found differences because in our study, perceived usefulness and lack of sufficient eHealth literacy (patients) were mentioned by the majority of the subjects. Lack of accessibility (because of login methods perceived as difficult) was mentioned by almost half of the subjects. This difference can be due to an EHR being primarily aimed at professionals and a patient portal being primarily intended as a service for patients. The differences found among these implementation studies highlight the importance of identifying barriers and facilitators for each technology separately taking into account the perspectives of the several stakeholder groups that are involved.

There are many implementation models, and they have considerable overlap [34]. A combination of 2 models was used for categorization of the selected quotes, that is, the model of Grol and Wensing [26] for socio-dynamic factors and by McGinn [21] mainly for portal characteristic and technical factors. Although this combination of frameworks appeared to be a feasible approach, we also added categories and renamed existing ones, so they better match with our findings. An essential difference between our approach and, the CFIR framework is that in our study, patients are included as a separate factor, whereas in the CFIR framework, they are part of the “outer setting” [24]. In the FITT framework, separate categories such as “social context” and “organizational context” are missing, and the aspects related to social interaction, for example, are categorized under “individual” within the FITT model. We found these categories to be relevant as a separate level because many subjects reported on them [25]. In the McGinn model [21], a subcategory is “participation of end users during the design,” which does not cover all the input we received, particularly because it is not aimed at the complete implementation process. One of the added categories is participation of end users during implementation. Another new category is sufficient eHealth literacy, which encompasses the skills and knowledge necessary to use electronic applications [33]. The models we used only address patients’ skills and knowledge [26] and applicability—of EHR implementation—to patients’ characteristics [21]. Patients’ lack of eHealth literacy was identified as a barrier by the majority of the subjects.

Our findings suggest that implementation is affected by barriers and facilitators at various levels. McGinn et al [21] describe 3 key levels: the macro, meso, and micro levels. We present some suggestions and insights for organizations that intend to implement a patient portal.

This study has several limitations. First, we used semistructured interviews in which we provided participants with prompts/examples for each level. Providing subjects with examples may have restricted participants in their answers about new barriers and facilitators or to “think outside the box” on these topics, so we might have missed factors. However, we used the combined models of Grol and Wensing [26] and McGinn et al [21], and many stakeholders mentioned barriers and facilitators that fell outside our scope. Although we have confidence in the richness of the current data, we already reached data saturation after 6 hospitals, limiting the total number of hospitals and subjects. There were also differences in the included hospitals with regard to the phase of patient portal implementation. Some had already provided a portal, whereas others were in the middle of the implementation process or had no portal at all. Although we found only limited differences between the hospitals with and without an implemented patient portal, this could still have introduced bias into the responses because of the recall or the imagination of information. This means that the results might have been influenced by the current state of hospitals because participants sometimes had to recall information from the time of implementation or had to imagine an implementation process (if there is no portal or no implementation).

Although we presented many different types of barriers and facilitators, we acknowledge that quantity should not be taken as a proxy for importance. We therefore added quotes to the results so as to highlight the specific nature of specific barriers and facilitators. For data analysis, we used a directed content analysis (deductive) approach. This can be a possible limitation because we started with an already existing model with defined categories. However, as the methods allows, we did not completely hold on to the categories in the models as we added additional categories ourselves and renamed the existing (generic) categories to barriers and facilitators that better fit our findings. Despite these limitations this is, to the best of our knowledge, the first qualitative study to identify barriers and facilitators for patient portal implementation involving multiple stakeholder groups.

Instead of organizing health care around professionals and institutions, some contend that it should increasingly be arranged around patients [2]. In a recent review, we found little evidence for the efficacy of IT-supported shared care [38]; however, many initiatives exist that may facilitate patient-centered or shared care. We already see movement in this direction as information systems are evolving from purely organizational to regional and even international systems [39]. For instance, a PHR is an example of an application in which patients can access their health information that has been collected from various health care institutions but is controlled by the patients [40]. In several European countries, these national systems have already been introduced. For example, in France, there is a national initiative called “Dossier Médical Personnel,” which is accessible over the internet. The information is uploaded by the involved clinicians; however, patients are in charge about what is included in the portal and who is authorized to access it. In Estonia, health professionals transfer information into a system called the “Estonian Health Information System,” providing patients with information via a patient portal [41]. These initiatives show a shift from hospital-financed, -owned, and -managed health records for which access is granted through portals, toward PHRs in which providers upload the data and ownership by patients is facilitated. The present uptake/compliance rates of portals are however still rather low (seldom above 50%), so this is an aspect that should receive attention if widespread use is foreseen.

Future research is necessary to confirm the practical utility of our proposed model when used among various stakeholder groups and to test whether it is useful to tailor implementation strategies to these various stakeholders, and organizations, taking possible development routes into account. In addition, there is a lack of knowledge regarding the association between patient portal implementation and patient portal adoption (ie, actual uptake and use by patients). One important element we identified is eHealth literacy, and this should ideally be included in the implementation and evaluation strategies for health technology tools. Moreover, the expectations before implementations and the experiences afterward can vary among health care professionals [27] and patients [11]. Further research into “satisfiers” determining the attitude of professionals toward using these technologies is recommended because evidence of the effectiveness of technology-related aspects on patient empowerment and on health outcomes is a strong facilitator.

Patient portal implementation is a complex process that is not just a technical process, but it also affects an organization and its staff. We found barriers and facilitators at various levels that differed depending on hospital types (eg, lack of accessibility) and stakeholder groups (eg, sufficient resources) in terms of several factors. Our findings underscore the importance of involving multiple stakeholders in portal implementation projects. We identified a set of barriers and facilitators, which are likely to be useful in making strategic and efficient portal implementation plans.