The study was conducted at San Francisco General Hospital (SFGH), a safety net hospital in the San Francisco Health Network system. From December 2013 to September 2014, we recruited participants to gain a pool of individuals with a wide range of health knowledge and engagement. Recruitment sites included (1) the General Medicine Clinic (GMC), a primary care clinic serving more than 6500 patients, most of whom are uninsured (32%) or on Medicaid (39%); (2) a diabetes support group led by diabetes nurses; and (3) a diabetes education class that GMC patients were referred to at the hospital. GMC began exclusively using an EHR in June of 2013. At the time of the interviews, the SFGH-wide patient portal was not yet launched; rollout was scheduled for early 2015.

We recruited patients through an electronic query of patients with upcoming clinic or diabetes group appointments. To recruit participants identified as having upcoming clinic appointments, study staff approached potential participants before or after their appointments, explained the study, and recruited interested participants. To recruit participants identified as being enrolled in group sessions, study staff attended the sessions, described the study to the group, and recruited interested participants. Caregivers, defined as someone playing a role in the management of a patient’s health other than the patient or the medical provider, were recruited by provider referral of someone who attended medical visits with or communicated with a provider on behalf of a patient. Participants were eligible for the study if they were (1) English-speaking, (2) not cognitively impaired, and (3) diagnosed with a chronic disease or the caregiver of such a patient. We focused on patients with chronic illnesses because portal use may be particularly useful in supporting ongoing self-management. We included only participants who expressed some interest in using the Internet overall to manage their health care, unless accompanied by a caregiver who expressed such interest, because we felt that this represented a realistic sample of individuals who would be potentially interested in and able to use the portal when it launched.

During recruitment, we administered a short questionnaire to gather information on demographics (age, race/ethnicity, gender); diagnosis of a chronic disease (heart disease, diabetes, high blood pressure, heart failure, asthma/chronic obstructive pulmonary disease [COPD], and/or chronic kidney disease); interest in using the Internet to manage health care at SFGH (high, some, none, or don’t know/need more information); and frequency of current Internet use (daily, weekly, monthly or less, or none). Finally, we administered a previously validated one-item health literacy scale regarding how confident participants were filling out medical forms on their own (not at all, a little bit, somewhat, quite a bit, extremely) [28] because this has been shown to be predictive of portal use in our previous quantitative work [14]. We classified participants noting any lack of confidence in filling out forms as having limited health literacy.

We conducted semi-structured in-depth interviews with 11 patients and 5 caregivers, two of which were dyad interviews with both a patient and his or her respective caregiver (Multimedia Appendix 1, Interview Guide). Although the interviews included discussion of current health status, health behaviors, and health care utilization, emphasis was placed on prior use of the Internet and specific interest in the use of a patient portal website for health management, informed by the Theory of Acceptance and Use of Technology [29,30]. To provide a visual example of a patient portal interface, participants were shown screenshots of a sample patient portal interface on paper, including the log-in, test results, and visit summary features. Participants were asked to state whether they thought they would use the portal website in their own care and what features of the portal website were of most interest to them. Caregivers were asked to discuss the potential impact proxy access to a patient’s personal health records would have on their role. The interviews were transcribed and deidentified before analysis.

Using data from the questionnaires, we summarized the participant demographics and Internet and portal use responses.

Authors CL and LT read the interview transcripts in their entirety before independently analyzing them. We used an interpretive description approach [31,32] to analyze the transcripts, using inductive and deductive coding techniques. Coding was done using Atlas.ti 7 software [33]. First, we used inductive open coding to identify all emerging themes and subthemes that participants mentioned during the discussions [34,35]. To assure quality of the analysis and to uphold the constant comparison open coding approach, CL and LT met regularly to discuss the thematic findings. When there was disagreement, US established agreement on codes. The entire team reviewed and provided comments on the final codebook.

A main goal in this study was to determine if the barriers and facilitators in our safety net setting were similar or different from previously published literature on portal use in other settings—the majority of which have been conducted in integrated health care settings. For this reason, we recoded all the transcripts in a deductive manner to be able to determine if the categories and severity of barriers and facilitators that emerged from our analysis were comparable to previously published work [5,13]. Study staff halted further enrollment in the study after a consensus that thematic saturation had been reached.

Because there was a clear indication of clustering of themes by health literacy status, the transcripts were re-examined by self-reported health literacy status in a secondary exploratory analysis. Although we did not purposively sample for limited health literacy status, we were able to generate some hypotheses for additional types of barriers for patients in this group. We also examined the caregiver transcripts independently from the patient interviews in a similar exploratory fashion.

Finally, we summarized basic usability and accessibility comments as participants looked at screenshots of a hypothetical portal website to understand how they might use the website in the future. This included identifying patients reporting interest in using specific portal features.

A total of 45 individuals were approached about the study. Of those approached, 25 (56%) expressed interest in the study, 7 (16%) declined citing lack of interest in computer use, and 13 (29%) declined due to unknown or other reasons (too busy, uninterested in research). Of the 25 who expressed initial interest, 3 subsequently could not be reached, 5 stated they were too busy to schedule an interview, and 1 declined an interview due to emerging health issues. We enrolled 16 participants in the study before reaching thematic saturation.

Participants in the study were predominately male (10/16, 63%) and ethnically diverse (50%, 8/16 African American; 19%, 3/16 Latino; 19%, 3/16 Asian or Pacific Islander; 13%, 2/16 white). All patients in the sample were diagnosed with diabetes, 60% (3/5) of caregivers cared for individuals with diabetes and 20% (1/5) of caregivers cared for individuals with multiple chronic conditions, including hypertension, heart disease, chronic kidney disease, and COPD. The mean age of the sample was 56 years (SD 11). More than half (10/16, 63%) of participants had limited health literacy. Overall, participants reported high experience and interest in Internet use: 56% (9/16) expressed high interest in using the Internet to manage their health care and 69% (11/16) were daily users of the Internet. All but 2 participants reported at least occasional Internet use (Table 1).

Overall, the 5 major categories characterizing the barriers and facilitators for portal use were similar to the previously published research on this topic: (1) computer or Internet access, (2) technological skills and interest, (3) security and privacy of information, (4) patient-provider relationship, and (5) chronic illness self-management.

The secondary analysis by health literacy status showed a much higher prevalence of barriers for participants with limited health literacy within several of these categories. These findings are highlighted subsequently.

Among a diverse group of patients and caregivers in a safety net clinic, we identified significant barriers to portal use, including concerns about security and privacy, limited technological proficiency, and a desire to preserve in-person aspects of existing patient-provider relationships—most often among patients with limited health literacy. Recruiting only those who expressed at least some interest in using the Internet for health management, our findings are likely conservative in that they represent some of the more engaged patients within our safety net health care system. The majority of participants in our study were African American and male, characteristics which have both been associated with lower portal use [13,14,17,26,36]. Nevertheless, it is important to note there was overall enthusiasm among these participants about the potential of a patient portal to improve aspects of health monitoring, patient-provider relationships, and caregiver burden. This is consistent with interest [21,26,27] and benefits described among low-income patients in past studies [5,21,27].

The overall categories of barriers to portal use in our study were consistent with previous studies: concerns about security [5,37], difficulty understanding medical information [5,21], the desire to preserve verbal communication and in-person contact [5,20,38], and the burden of portal use on clinician workloads [5]. However, because our sample included predominantly patients with limited health literacy, our findings uncovered more pronounced aspects of these barriers in safety net settings, such as access to computers in public settings; negative past experiences with technology, including security breaches and viruses; and a lack of more rudimentary computer skills. In particular, a distrust of potential security measures to prevent access of personal health information by hackers, researchers, and others unauthorized by the patient may hinder patient portal use among safety net patients. This is consistent with past studies indicating that individuals with limited health literacy are less likely to sign onto a portal website [14,39], use portal messaging functions [39], identify blood sugar test results as out of range [19], and contact a provider to discuss abnormal test results from an online portal [19]. Public computer use coupled with relatively widespread security concerns may be particularly relevant to address among patients with lower socioeconomic status, especially because previous studies have suggested that only a minority of a general patient population express hesitancy to use portals because of security [37,40]. Past research has shown that older, low-income patients desire assistance in interpreting their personal health information [41]. Our findings show that health literacy is a major barrier among younger populations within the safety net as well.

Although our sample of caregivers was small (limiting our ability to make strong inference), our findings also imply that there is a potential supportive role for caregivers to facilitate portal use in a safety net setting. Particularly for patients lacking adequate health literacy, English proficiency, and/or the technological know-how or interest to access and interpret their personal health information, there is potential for caregivers to use patient portals to improve their ability to interpret health information, coordinate care, and assist with medical decision making [22]. Past studies have found high interest among patients in sharing their personal health information with caregivers [42,43] and among caregivers in accessing patient health information through health technology such as portals [44,45]. Caregivers in our study expanded on the utility of having access to patient health information to describe a deeper role in caregiving, noting their role as interpreters of this information, guides in decision making and behavior change, and advocates in ensuring quality of care. Our findings illustrate the need for safety net health systems implementing patient portals to address caregiver needs—including strategies for formal proxy processes for patients to officially grant others access to their portal account as well as awareness of informal sharing of username/passwords between family members that is also likely to occur. Because the caregivers in our sample were already highly involved in health care management tasks, further research is needed to understand perspectives on the levels of caregiver access to personal health information [40,42].

Because our small study examined the in-depth perspectives of patients receiving care from one urban safety net hospital, our findings are likely not generalizable to patients receiving care from larger health systems or networks. Furthermore, because our eligibility criteria required participants to speak English and express at least some interest in using the Internet to manage their health, our findings may not be generalizable to those with limited English proficiency or needing even more basic computer or technology training. Finally, our study did not incorporate perspectives on the actual usability of a live portal website. Instead, we focused on gaining in-depth information about the barriers and facilitators to portal use in advance of its rollout to support a patient-centered approach to implementing our portal system-wide within the San Francisco Health Network.

Our findings indicate that interest in using patient portals may not always match the technological proficiency of more vulnerable patients. This indicates a need for safety net health systems or other social service providers (eg, library, adult literacy classes) to provide training not only in portal use, but also in equipping patients with the basic computer and health literacy to effectively use a portal. To address patient concerns, it is important for health care systems implementing portals to assess the potential effects of the replacement of in-person or verbal communication resulting from portal use and establish high levels of online security.

From a national perspective, our findings suggest that widespread EHR and portal implementation may be hindered by patient engagement challenges in the coming years, especially with respect to health literacy and language proficiency status. Although incentives to promote meaningful use have been successful at driving health systems to implement health information technology, these standards do not guarantee that safety net health systems will adopt the newest or most accessible technologies on the market [46,47] or that patient portals will be accessible or useful to all patients, especially those who may face additional limitations in literacy and technology experience. Addressing health literacy and other barriers may best be achieved through patient-centered approaches to the adoption of health information technology at the planning, implementation, and evaluation stages [48]. If implemented with patient perspectives in mind, patient portals have the potential to be a convenient and effective way to improve self-management and quality of care for patients and caregivers receiving care from safety net settings.