To our knowledge, this is the first study to assess real-world use of an industry-leading patient portal in a large, ethnically, culturally, socioeconomically, and educationally diverse sample of diabetes patients with a range of self-reported health literacy skills. We found clear racial/ethnic disparities in patient portal use, as have been previously documented (
Hsu et al., 2005), and differences by educational attainment as well. As expected, we found much lower rates of patient portal use among those reporting limited compared to adequate health literacy. The lower registration rates among those with limited health literacy underscores existing concerns about lack of computer/internet access among those with limited health literacy, i.e., literacy–digital divide (
Institute of Medicine, 2009;
Lenhart et al., 2000). Importantly, the disparities in use of the patient portal by health literacy, race/ethnicity, and education mirror the well documented disparities in diabetes outcomes (
Agency for Healthcare Research and Quality, 2004). As such, this represents a significant public health concern, especially as U.S. health policy increasingly emphasizes health information technology, potentially at the expense of alternative modes of service delivery. The inverse care law states that health care innovations disproportionately benefit healthier individuals (
Hart, 1971;
Schillinger, 2007) and thus exacerbate health disparities. Our findings serve as yet another manifestation of that phenomenon.
Even after we attempted to control broadly for computer/internet access, by including models involving only those who completed the internet-based registration process, we found that those with limited health literacy were less likely to activate their patient portal account, sign in using their personal login and password, and to use any of the functions. This suggests that those with limited health literacy may have difficulties with
navigation of an internet-based patient portal, in addition to having lower rates of access to computer/internet. Moreover, we found that the function most likely to be used, across all literacy levels, was the “Labs View” function, which is a relatively passive use function, a vehicle for delivering information, rather than more active and even interactive tasks like making an appointment or emailing one’s provider. This also suggests that navigation of more complex internet-based tasks (e.g., diabetes self-management instructions) constitutes a barrier, as others have suggested (
Norman & Skinner, 2006). Importantly, in our sample, among those with computer access, self-reported health literacy was more strongly and consistently associated with lack of patient portal use than was lower educational attainment.
There are a number of potential mechanisms to explain our findings. First, those with limited health literacy may not be aware that the online patient portal exists; tailored (literacy level appropriate) promotion of online patient portal use in this group may be needed. Second, those with limited health literacy may not be empowered or motivated to learn to use an online patient portal. For instance, inadequate social/family support for skill-building, culturally dependent health beliefs that do not promote patient activation, and lack of trust in the medical system may all influence use of an internet-based patient portal. Specific social messaging and other outreach efforts could combat these factors. Third, a major barrier to online patient portal use is internet access, which includes absent or limited computer access, lack of computer/internet training, and competing time demands. Increasing access to internet, computers, and training, particularly within health care settings, may reduce access barriers. Fourth, those with limited health literacy may find internet-based patient portals difficult to navigate. This can be addressed with specific, hands-on training, and design features that improve ease of navigation and are appropriate for limited-health-literacy populations, such as video, audio, and graphical information.
Study strengths include a large, diverse population with uniform access to care, and detailed assessments of use of an internet-based patient portal, among well characterized diabetes patients, within a real world setting. Nevertheless, this study has several limitations. First, although KPNC’s population is diverse, results may not generalize to the uninsured, those cared for in other systems, or different patient portals. Second, Kaiser Permanente (KP) has a much more robust and well established internet-based patient portal than most other U.S. health care delivery settings, and, as such, our finding may overestimate patient portal use compared with other settings. However, we anticipate that as internet-based patient portal become more the standard, these issues will emerge in other settings. Third, we did not directly measure participants’ computer access, although an internal survey of KPNC members suggests that 90% of members aged 25 to 64 years and 50% of members aged 65 to 79 years have internet/computer access (
Gordon, 2009). As another proxy for access, we inferred that those who registered on-line had some level of access to the internet. We acknowledge that a continuum of internet access exists; we could not capture extent of access, or access barriers, with the current data. Fourth, not all participants answered every question in this lengthy survey (184 questions, 52 pages). Nevertheless, this remains the largest observational study to date of internet-based patient portal use. Fifth, the nature of study design precluded our being able to directly measure health literacy in this sample; however, we used validated, widely used items to estimate health literacy (
Chew et al., 2004;
Chew et al., 2008;
U.Sarkar et al., 2008;
Wallace et al., 2006). Sixth, we did not assess the usability of the online patient portal through direct observation, user satisfaction evaluations, or other methods. Finally, it is possible that those with diabetes may be more likely to use an internet-based patient portal because of increased health care and self-management needs; these results may not apply to healthier adults. However, diabetes is common and costly; therefore, understanding how those with diabetes adapt to efficient new technologies is a public health and policy priority.
Future research should examine the possible links between patient portal use and processes of care, such as visit attendance, visit interval, attendance at screening such as ophthalmology, completion of ordered laboratory testing, and possibly clinical outcomes such as hemoglobin A1c and blood pressure. Moreover, patient portal usage data may be useful to clinicians and caregivers, perhaps as an indicator of relative internet navigational limitations.
Our study has important public health and policy implications. First, ensuring universal internet and computer access will become more important than ever. As services migrate to the internet, those most at risk for poor diabetes health outcomes are at risk of falling further behind as health systems increasingly rely on patient portal health service functions and limit the alternative modes of access and communication. Second, improving the usability of internet-based patient portals, such that those with limited health literacy can navigate them effectively, is critical to realizing health benefits. The internet has potential, via use of audio, graphic, video, and multiple languages, to greatly expand the capacity and reach of health care systems (
Institute of Medicine, 2009). However, in the real world, that potential has yet to be realized. Barriers to internet-based health care services require improved technology access as well as tailoring of design and services to reach those with limited health literacy.
The publisher's final edited version of this article is available at 
