INTRODUCTION

Substantial resources and attention have been invested recently in health information technology (IT).1 – 4 The meaningful use incentive program administered by the Centers for Medicare and Medicaid Services (CMS) outlines requirements for electronic medical records (EMRs), among which is the requirement that eligible providers will “provide patients the ability to view online, download, and transmit information” from their medical records.5

It is unclear, however, whether elderly patients will be able to take advantage of the opportunity afforded by EMRs. While Internet use among older adults is increasing rapidly, as of 2013, only 59 % of adults aged 65 and older reported that they were online.6 Moreover, the proliferation of health IT may exacerbate the digital divide.7 – 9 Vulnerable groups such as African-Americans or those with low socioeconomic status are less likely to use EMRs or patient portals,9 – 16 in part because they are less likely to use the Internet at all.17 – 20

Health literacy may also play an important role in determining whether elderly patients are willing and able to use EMRs and other Internet-based health tools. Surprisingly, given the substantial body of research that documents the importance of health literacy in health behaviors and outcomes,21 – 23 it has been largely overlooked as a determinant of health IT use.24 , 25 One study16 used a large sample of patients with diabetes enrolled in a single health plan in northern California to document the importance of health literacy as a predictor of Internet-based patient portal use, even after controlling for basic demographic characteristics. This finding supports the idea that individuals with low health literacy are on the wrong side of the digital divide; however, no population-based study to date has estimated the relationship between health literacy and use of the Internet to obtain health information in the general population of older Americans. The current study fills that gap by exploring both how low health literacy is related to use of the Internet for any purpose and whether, conditional on being Internet users, those with low health literacy are less likely to use the Internet for the purpose of obtaining health or medical information.

METHODS

Setting & Participants

We used data from the Health and Retirement Study (HRS), an ongoing nationally representative prospective cohort study of approximately 22,000 individuals aged 51 and older. The HRS is conducted by the University of Michigan under a cooperative agreement with the National Institute on Aging. The design of the HRS is described elsewhere in greater detail.26 , 27 Core surveys are administered in both English and Spanish, with half of the sample receiving an in-person interview and half receiving a telephone interview in each wave. Recent waves of the HRS have achieved response rates above 88 %,26 and bias from nonrandom sample attrition is low.28 , 29 The use of survey weights to adjust for the complex sample design and non-response yields estimates that are nationally representative.30 , 31

The current study makes use of data from a special module on health literacy that was administered to a random subset of HRS respondents in 2010 and a supplemental Internet-based survey administered in 2009 to respondents who reported in 2008 that they regularly used the Internet.

In 2010, health literacy questions were slated to be administered to a subsample of 1,168 respondents aged 65 and older randomly drawn from 2010 HRS core respondents who (1) responded to the 2010 core survey themselves, as opposed to having a proxy respondent answer for them, and (2) were included in the random half-sample of respondents receiving face-to-face rather than telephone interviews in 2010. For this analysis, we considered 1,121 community-dwelling respondents 65 years of age or older; of these, 898 completed the health literacy module, for an overall module response rate of 80 %. However, 74 of these individuals ultimately completed the module via a telephone interview rather than a face-to-face interview, so they lack results from one of the health literacy measures. Our final sample for analysis, therefore, comprised 824 respondents, for an effective module response rate of 74 %. We did find some evidence that in both of the subsamples we used, individuals in worse health or with lower cognitive ability were less likely to provide complete data, an issue we discuss in more detail below.

The 2009 Internet survey was administered to a random subsample of respondents to the 2008 core survey that had reported regular Internet use in that wave. Among those aged 65 and older, 3,406 individuals were contacted for the Internet survey, of whom 2,367 responded, for an overall response rate of 69 %. Among those who responded, 1,617 individuals were assigned the health literacy measures from the survey, of whom 1,584 gave valid responses.

Measures

Internet Use

We characterized Internet use in two ways. First, all participants in the core HRS were asked, “Do you regularly use the World Wide Web—or the Internet—for sending and receiving e-mail or for any other purpose, such as making purchases, searching for information, or making travel reservations?” We considered respondents who answered “yes” to this question as Internet users. Second, participants in the 2009 Internet survey were asked, “How often do you do each of the following activities on the Internet?” Options included “get medical or health information online,” among others. Possible responses included: never, rarely, sometimes, or often. We defined those who responded “sometimes” or “often” as individuals who used the Internet for medical or health information.

Health Literacy

Participants in the 2010 health literacy module and the 2009 Internet survey were asked a single health literacy question: “How confident are you filling out medical forms by yourself?” with response options: extremely, quite, somewhat, a little, or not at all. Studies validating this question with more commonly used tests of health literacy, including the Rapid Estimate of Adult Literacy Measure (REALM) and the Short Test of Functional Health Literacy in Adults (S-TOFHLA), have suggested a response of “somewhat confident” or less as a proxy for low health literacy.32 – 35 This measure has been validated in Spanish.36 The 2010 health literacy module also included the revised, shortened version of the Rapid Estimate of Adult Literacy in Medicine (REALM-R); we followed established convention defining low health literacy as seven or fewer words pronounced correctly.37 , 38

Cognitive Function

We used three measures of cognitive function included in the HRS: serial sevens subtraction (7 from 100 successively), immediate and delayed recall (from a list of common words), and counting backwards. We defined a composite measure from 0 to 27, and defined cognitive impairment as a score of 11 or lower, an approach that has been validated using these data.39

Health and Function

Self-reported health was assessed with the question, “Would you say your health is excellent, very good, good, fair, or poor?” Impaired function was defined by a self-report of difficulty with activities of daily living (ADLs: dressing, walking, bathing, eating, getting into or out of bed, or using the toilet) and instrumental activities of daily living (IADLs: preparing meals, grocery shopping, using the phone, taking medication, or handling money). Previous research using the HRS has established the significance of functional impairment with regard to Internet use.18

Demographic and Socioeconomic Characteristics

We also included demographic characteristics (age, race, ethnicity, gender, marital status, and educational attainment) and family income relative to poverty status.

Results were weighted to be nationally representative of community-dwelling Americans aged 65 and older. All standard errors were adjusted to account for the complex sampling design of the HRS. Statistical analyses were performed using Stata version 13 (StataCorp LP, College Station, TX).

RESULTS

Table 1 summarizes key demographic characteristics in the 2010 health literacy module (general population) sample and the 2009 Internet sample. Internet users were younger, more highly educated, and in better physical and cognitive health than the general population sample. In our general population sample, about 20 % of respondents had low health literacy as measured by the REALM-R, while 40 % had low self-assessed health literacy. In the sample of Internet users (for whom we have only the self-assessed measure of health literacy), the fraction with low health literacy was about half of what it was in the general population (Table 1).

Table 1 Characteristics of Study Participants, N (%)*
Full size table

Individuals with adequate health literacy were three times more likely than those with low health literacy to use the Internet to obtain health information. Only 9.7 % of older individuals with low health literacy regularly used the Internet for health information, compared with 31.9 % of those with adequate health literacy (Table 2). These differences in health-related Internet use were driven both by higher rates of any Internet use among the more literate—only 22.0 % of those with low health literacy regularly used the Internet for any purpose, while 54.8 % of those with adequate literacy did—and a higher likelihood of using the Internet for health information among Internet users with adequate health literacy. These differences suggest the presence of a significant digital divide between individuals with low and high health literacy.

Table 2 Health Literacy and Internet Use
Full size table

Table 3 presents results of a multivariate analysis relating Internet use to low health literacy, after controlling for other factors likely to influence Internet use. Low health literacy, whether self-assessed or measured using the REALM-R, significantly reduced the odds of regular Internet use (for self-assessed health literacy, OR = 0.36 [95 % CI 0.24 to 0.54]; for health literacy measured using REALM-R, OR = 0.25 [95 % CI 0.15 to 0.41]). Other covariates were as expected: age greater than 75 years, education less than high school, and low cognitive function were also associated with a lack of regular Internet use. Interestingly, although existing evidence suggests that the relationship between health literacy and other health-related measures in many cases is rendered insignificant by the inclusion of measures of general cognitive function,40 , 41 our results remained robust even after adjusting for these. We also repeated the regression analyses entering scores on each of the three cognitive tests separately (rather than in the combined 27-point scale), and our results were unchanged, so we are reporting the results using the combined 27-point scale.

Table 3 Multivariate Determinants of Internet Use
Full size table

Among regular Internet users, 55.4 % (95 % CI 52.1 to 58.9 %) reported using the Internet sometimes or often to get medical or health information online. Low health literacy significantly reduced the odds of Internet use for medical or health information (OR = 0.60 [95 % CI 0.47 to 0.77]). Low cognition had relatively less impact on the likelihood of using the Internet for medical or health information, conditional on use of the Internet. That is, our results suggest that once older adults do use the Internet for any task, poor cognition is a smaller barrier than low health literacy with respect to Internet for health-related tasks.

DISCUSSION

In a nationally representative sample of older U.S. adults, we found that health literacy was a significant predictor of Internet use. For those who did use the Internet, low health literacy was also predictive of what they did once they were online; individuals with low health literacy were significantly less likely to use the Internet for medical or health information. These patterns persisted even after adjusting for demographic covariates including age, sex, and race, as well as clinical covariates such as chronic medical conditions, functional impairment, and cognition. Interestingly, though past evidence has suggested that health literacy may be a proxy for general cognition,40 , 41 our results suggest that each has an independent relationship with likelihood of regular Internet use. We found that health literacy was a more important predictor of Internet use for medical or health information than was level of cognitive function, suggesting that interventions specifically targeting health literacy among Internet-using older adults may be effective for narrowing the digital divide by facilitating their ability to obtain medical information online.

This study was the first to examine the importance of health literacy as a predictor of Internet use for obtaining health information, a topic that, surprisingly, has been overlooked in nearly all prior studies of Internet use. In contrast to an earlier study documenting the importance of self-assessed health literacy as a predictor of patient portal use in a sample of diabetes patients enrolled in a single health plan in northern California, we used data that were nationally representative of older Americans. We also measured health literacy using the REALM-R in addition to self-assessed health literacy.

Our study did have a number of potential limitations. Our analysis was limited to individuals 65 and older. Older cohorts of adults have been slower than younger cohorts to adopt Internet technology related to health42; the effect of health literacy on the use of these technologies could be either smaller or larger among younger cohorts. As mentioned above, individuals in worse physical or cognitive health appeared less likely to provide complete data for the survey components that we used, raising concerns about selection bias. We believe, however, that any bias is unlikely to affect the differential in Internet use between individuals with low and high health literacy.

Another limitation was that other important patient attributes such as patient activation43 , 44 or self-efficacy45 were not available in our data; omitting these variables from the analysis may have overstated the true relationship between health literacy and Internet use—if they are, indeed, positively correlated with both health literacy and use of the Internet. In this case, our estimates overstated the true independent effect of low health literacy on Internet use, and future work with a more extensive set of covariates will be needed to address this problem.

A final concern is that the outcome measures that we used—use of the Internet for any reason and, conditional on using the Internet at all, using it to get health or medical information—are more general than, for example, whether individuals in a particular health plan use that plan’s patient portal. This generality is both a strength and weakness. It allows us to understand the extent to which older individuals with low health literacy are not even at the starting gate in utilizing specialized applications such as a patient portal, in that they do not use the Internet at all. We found that this fundamental obstacle was highly prevalent: more than three-quarters of older individuals with low health literacy did not use the Internet for any purpose. At the same time, our measure of health-related Internet use—obtaining any health or medical information on the Internet—likely overstated the extent to which some of these individuals would, in fact, be able to navigate a patient portal or other more specialized Web-based health interventions. The small fraction of older individuals with low health literacy who used the Internet to obtain health or medical information—9.7 %—is almost certainly higher than the fraction that would be able to adopt sophisticated health information technologies to effectively interact with their EMRs. Further research on such interventions using population-based samples should be a high priority, with a focus on identifying features of interventions that make them accessible to older individuals with low health literacy.

CONCLUSION

Individuals with low health literacy represent a vulnerable population who are at high risk of being left behind by the advance of technology. Our results suggest that a simple measure of health literacy—confidence filling out medical forms—is effective at identifying individuals who are less likely to use the Internet to obtain information about health. Interventions to improve health literacy may have a spillover benefit of lowering barriers to effective use of information technology. Even in the absence of such interventions, screening for low health literacy in the clinical setting can help clinicians identify patients who are likely to have difficulty adopting electronic health technology.

Health information technology, like any innovation in health care,46 offers both the promise of significant benefits and the risk that these benefits will not be shared equally. Low health literacy may attenuate the effectiveness of Web-based interventions to improve the health of vulnerable populations. As Internet use becomes increasingly relevant to the provision of health care,42 programs must address barriers to substantive use among vulnerable populations, or otherwise risk deepening the existing disparities in access and outcomes.