Beyond access: barriers to internet health information seeking among the urban poor

Abstract

Objective Communication inequalities deepen health disparities even when internet access is achieved. The goal of this study is to understand how a range of barriers may inhibit individuals from low socioeconomic position (SEP) from engaging with online health information even when it is freely available.

Materials and methods Detailed data were collected from 118 low-SEP individuals from a randomized controlled trial providing internet access. Measures triangulated the health-seeking experience through internet use tracked in real-time, call log data, and self-reported barriers. Negative binomial regression models were fitted with technology and perceived predictors, and our outcome, health information seeking, and then stratified by medical status.

Results Participants experienced a median of two computer issues (median 6 days) and two internet issues (median 6.5 days). Duration of internet problems was associated with a decrease in the rate of internet health information seeking by a factor of 0.990 (P = .03) for each additional day. Participants with a medical problem who were frustrated in their search for health information had half the rate of health information seeking of those who were not frustrated (incidence rate ratio = 0.395, P = .030).

Discussion Despite IT support, participants still experienced internet connectivity issues that negatively impacted their health information seeking. Frustration in their search to find information may serve as an additional barrier to those who have medical issues.

Conclusion After initial internet access, a second-level digital divide emerged due to connectivity issues, highlighting the need to understand the complex network of barriers experienced by low-SEP internet users.

BACKGROUND AND SIGNIFICANCE

Information and communication technologies hold the potential to engage populations from low socioeconomic position (SEP)¹ in their own care and link them to health resources.² However, as health information-gathering increasingly takes place through information and communication technologies, the impacts of communication inequalities may be profoundly felt among low-SEP individuals. Communication inequalities reflect unequal distribution of communication resources,³ and fall along the same sociodemographic lines as health inequalities, leading to pronounced differences in communication behaviors and outcomes along education and income gradients.⁴ For example, low-SEP individuals may not be able to use the web in the same way as their high-SEP counterparts, potentially encountering difficulties with computer technologies due to a lack of relevant skills and resources.^5–7 Such barriers may prevent internet-related benefits from being realized equally among population subgroups and may serve to deepen health disparities by exacerbating gaps in health-related knowledge and outcomes.^3,8 This study represents some of the first research to triangulate the internet health-seeking behavior of low-SEP individuals from 1) health information seeking on the web data recorded in real time, 2) actual observed technological barrier data in a home-based setting, and 3) subjective, self-reported measures to fill a key gap in the internet engagement literature.

Online health information seeking has been linked with improved health outcomes⁹ and greater patient engagement.¹⁰ Although 72% of internet users have looked for online health information,¹¹ there is widespread evidence that high-SEP individuals use the internet more frequently for health information seeking purposes than their low-SEP counterparts.^12–16 Less educated individuals also gain fewer positive outcomes from their online engagement in terms of health care needs and behaviors,¹⁷ which may lead to gaps in health-related knowledge and outcomes for low-SEP groups.⁸ Although the digital divide is narrowing,¹⁸ access alone is not sufficient to guarantee equal internet engagement,¹⁹ particularly given that the evolution of online infrastructure further exacerbates existing health disparities, including the presence of difficulties in maintaining, updating, and navigating technology and online resources.²⁰ Given the challenges that low-SEP individuals may face when using the internet for health information, future digital interventions must be informed by the barriers to be overcome for low-SEP individuals to fully benefit from online resources.

“Digital inequalities” refer to differences in the ability to engage with information content among persons with formal access to the internet^12,21–25 and fall under the broader communication inequalities that consider individuals’ ability to access, process, and use the resources various communication platforms including the web.^14,20 Low-SEP individuals may face perceived and technological obstacles to continued internet use. Low confidence in internet abilities may influence how beneficial individuals rate their online experience.²¹ Low-SEP users in particular express concerns about their ability to find high-quality health information on the internet.^26,27 Individuals who use the internet for health information more frequently, such as those have a medical problem,^16,28,29 may be particularly susceptible to such barriers.

Computer-based barriers may include technical problems, including issues with computer hardware, software, and connectivity quality, which may frustrate users and inhibit internet use.^2,7,21 Nearly two-thirds of users in a North American study reported an internet virus, with half reporting the virus led to loss of information or damaged software.³⁰ Issues with internet providers and their service availability are particularly salient for low-SEP computer users, particularly those in high-poverty urban neighborhoods,³¹ due to interruptions in internet connections caused by frequently moving residences.^2,32

Although these perceived and technology-based barriers may contribute to communication inequalities, data are often collected through self reports and may lead to a lack of specificity on the information engagement behaviors of the urban poor.³³ Moreover, challenges with collecting data from low-SEP populations have led to low numbers of low-SEP individuals on national surveys.^33–35 Once recruited into a study, directed, intensive effort is needed to retain and track the population.³⁶ Furthermore, studies about internet use are often not refined enough to explore all the ways in which it may contribute to social inequality.²³ Study settings that observe low-SEP individuals’ internet barriers in controlled environments³⁷ or through online surveys³⁸ do not mimic how barriers are experienced at home naturally over time. Surveys frequently rely on self-reported accounts of past internet health seeking and may also be subject to bias and inaccuracy, particularly over longer periods of recall.¹⁶ These gaps in information fail to provide the full frequency of internet use and the actual consumption of online health materials among low-SEP groups.

OBJECTIVE

The purpose of this study was to examine the influence of internet and computer barriers on the internet health information seeking behaviors of low-SEP individuals. The data come from the Click to Connect (C2C): Improving Health Literacy through Computer Literacy study, a randomized controlled trial providing internet access to the urban poor. C2C was designed to elucidate how obviating access barriers to internet, conceptualized as the classic digital divide, would lead to greater use of internet for health among other subjects. The study was designed to meet the critical gap in the literature on how people from low-SEP groups use the internet and help influence design of platforms and services to enhance usage. The intervention group received free computers and internet access whereas the control group did not. Given the focus of the current paper on understanding barriers to online health information seeking, the data for the current paper comes from only the intervention group.

We explored perceived and staff-documented barriers to determine whether they impact observed, unbiased accounts of internet health information seeking recorded through tracking software and whether the impact of these barriers differs by medical status. This study used real-time web tracking data from low-SEP individuals to more fully contextualize these barriers.

MATERIALS AND METHODS

Data for this study were drawn from the intervention group of C2C, a randomized control trial funded through the National Cancer Institute to understand computer use patterns among a low-SEP population.

Intervention Details

The original parent C2C study was a randomized controlled trial focused on providing computers and internet access to a group of urban poor in Boston, MA. The study was comprised of an intervention and control group that were randomized through a computer program to either receive the intervention (described below) or to serve as a control, completing the pre- and post-test surveys only. Intervention participants for C2C received a computer, broadband internet access (18 months for wave 1, and 9 months for waves 2 and 3), access to a web portal designed specifically for low-literacy populations, ongoing technical support, and 9 monthly computer training classes at a local college that taught basic computer and internet skills. For the purposes of this study, we have broken the randomization to focus solely on the data gathered from the intervention group since they received several additional tracking measures of online activity.

Recruitment

We recruited participants for the parent study through presentations at adult education centers in General Educational Development, pre-General Educational Development, or English for speakers of another language classes in order to target low-SEP individuals who had not received a US high school degree. Recruitment efforts used a proactive strategy including outreach and continuous phone contact designed to recruit low-income, urban poor who are seldom represented in national surveys.³⁶ Three waves of the study were conducted due to the resource-intensive nature of the intervention, with groups of participants enrolled in the study from 2007 to 2010 and randomized to the intervention and control conditions in each wave. Inclusion criteria were enrollment in a basic adult education class, ability to understand and speak English, age 25–60, no internet access in the home, and not being computer savvy, as determined by having limited prior experience with computers. In the parent study, 175 participants were randomized to the intervention group, with 20 lost post-randomization.³⁶ There was a final intervention population of 133 with complete pre- and post-data. Tracking data were overwritten for 15 additional participants due to a malfunction in the tracking software, leading to a final sample of 118 for the current study.

Data Sources

The data came from four sources: 1) self-report surveys; 2) IT support call log data; 3) documented moving data; and 4) website tracking data. First, a telephone survey containing measures of information seeking and demographics was conducted with all participants at baseline and at 12 months follow up. A detailed call log documented every IT call placed by participants. A separate form tracked participant address changes during the study period. Internet use throughout the intervention period was tracked using Spector 360, software that logs each uniform resource locator visited into a secure server through a virtual private network. This method allowed us to capture real-time data of websites visited. For reasons of confidentiality per Institutional Review Board (IRB) mandate, we tracked household browsing activity rather than individual browsing information. Data sources were merged and were analyzed using STATA (Release 13, College Station, TX, USA).

Measures

Outcome-internet health information seeking

Internet health information seeking was conceptualized as the purposeful seeking of health information through visiting health websites. The creation of the outcome variable involved a multi-step process of coding uniform resource locator s of sites visited to create a final list of health websites. The detailed coding process for this measure may be found in Appendix 1. Each “hit,” or visit to a particular health-related website, was considered an instance of information seeking.

Predictors-barriers to internet health information seeking

Technology-based barriers captured the issues that were recorded by the study staff over the course of the intervention. These include:

• Computer hardware barriers included problems with the CPU/monitor/computer, keyboard/mouse, and virus or spyware issues. Three facets of computer issues were documented: number of separate incidents reported on computer malfunctions, the amount of time in days each incident took to resolve, and the nature of the incident. Logs kept by the study staff described each call placed by a participant and assigned a problem code (computer or internet problem), which was used to categorize each issue. Dates of the calls were tracked to identify the start and resolution of each issue. Qualitative call notes written by the study staff were used to provide more information on the nature of the call.

• Internet connectivity barriers were also measured through the call log process data, and included internet provider issues and modem problems. Data on this variable were gathered in the same fashion as the computer barriers.

• Relocation barriers were documented through a log indicating each date a change of address occurred.

• Perceived barriers were measured through self-reported data through the follow-up 12-month telephone survey.

• Perceptions of computer barriers asked participants about their difficulties using computers.³² Questions inquired how often the following prevented them from using a computer: not having enough free time; family and friends using computer; family and friends not wanting the participant to use the computer; internet connectivity problems; computer skills; and literacy skills. Response options were Likert-type responses on a 5-point scale, and were dichotomized into categories of always/often/sometimes and rarely/never.

• Perceptions of information barriers asked participants to indicate how much they agreed or disagreed with the following when they looked for health information: it took a lot of effort to get information, they felt frustrated during their search for information, they were concerned with the quality of the information, and the information was hard to understand, with categories dichotomized by strongly/somewhat agree and strongly/somewhat disagree.³⁹

Additional variables

Medical Problems

We created a variable indicating whether the participant currently had one of the following medical problems (yes/no): cancer, diabetes, heart disease, stroke, inflammatory bowel disease, or another medical condition.

Confounders

In the baseline telephone survey, we measured sex, race/ethnicity, age (under 35, 35–49, 50 or older), and native language. The participant’s study wave was also included in the analysis. Income and education were not included as confounders due to our recruitment of low-SEP subjects with a restricted income and education range. To account for differences that may have occurred due to differences in study timing, length, or service provision, we also controlled for variables that account for a) the wave of the study the participant was a part of and b) and the time that they remained in the study (9 or 18 months).

Analysis

We analyzed the predictor variables and the outcome, including frequencies and descriptive statistics, to gain a full picture of the participants’ perceived and experienced barriers. We assessed the distribution of our outcome variable, which was a count variable with non-negative, integer values. After observing over-dispersion of the data,⁴⁰ a goodness-of-fit test determined that a negative binomial regression model was the most appropriate analysis tool. We used the exposure command to account for time differences by wave.

To address the influence of each barrier on information seeking, a negative binomial regression was run between each individual predictor and the outcome. We identified confounders for each predictor by running bivariate models with the potential confounder and the outcome, and then with the confounder and the predictor of interest. Confounders that were significant at the P ≤ .25 level with both predictor and outcome were included in the multivariate model. Each individual model was then stratified by medical illness (yes vs no) to determine if information seeking varied by health status.

RESULTS

Table 1 shows key demographic characteristics. Demographic data from other nationally representative datasets have been provided in the table to illustrate how the C2C data provide a richer dataset of low-SEP individuals in comparison with large surveys. The majority of the sample had less than a high school education (77%) and had an income of less than $20 000 a year (65%). Over half (55%) of the sample was African American, and 68% were age 35 or older. Seventy-five (64%) reported a medical problem such as cancer, diabetes, heart disease, stroke, or other condition, and 62 (52%) spoke English as their primary language.

Table 1:

Demographic Comparisons between Click to Connect (C2C) and Selected National Surveys

	C2C % (n)	US Census 2010	Hints 2014 Cycle 1	PEW Internet Tracking Survey 2013
Total n	118	308 745 538 (population estimate)	3959	4178
Sex
Male	38 (45)	49	39	49
Female	62 (73)	51	59	51
Age
18-34	32 (38)	31	15 (18-34)	22 (<30)
35-49	46 (54)	34 (35 to 59)	24	32 (30-49)
50-64	22 (26)	13 (55 to 64)	46 (50-64)	45
Race/Ethnicity
African American	55 (65)	13	15	12
White	7 (8)	78	61	75
Hispanic	19 (23)	16	12	13
Income
Less than 10,000	33 (39)	8	16 (<$20K)	9
$10,000-19,999	31 (37)	6 ($10K-<15K)		11
$20,000-29,999	17 (19)	11 ($15K-<$25K)	15 ($20K-<$35K)	10
$30,000-39,999	8 (9)	10 ($25K-<$35K		11
$40,000-49,999	3 (3)	14 ($35K-<50K)	13 ($35K- <$50K)	7
$50,000-74,999	2 (2)	18	17	15
$75,000+	0 (0)	32	29	28
Education
Grade School or Less	14 (16)	6	10 (High school or less)	7 (Less than high school)
Some High School	63 (74)	8
High school graduate/ GED	3 (3)	50	20	34
Some College	0 (0)	21	31	31
Bachelor’s degree or highera	8 (9)	28	39	27

^aFor C2C – college completed in another country.

Internet Health Information Seeking

In total, there were 25 322 health information hits out of 5 084 901 hits across all websites, accounting for 0.5% of all website hits across the study period. Per participant household, there was a median of 85.5 health website hits (mean 214.59 hits, SD 411.65), with a range of 0 to 3537 hits. Participants who reported a medical problem had 2.1 times the rate of internet health information seeking compared to those without a medical problem after accounting for other factors in the adjusted model (P = .012, 95% CI, 1.176, 3.626).

Technology Barriers

Computer Hardware Barriers

There were a total of 298 computer issues reported over the course of the study period. Each household had a median of two calls (mean 2.37, SD 2.27), with a range of no calls to nine calls (Table 2). Computer issues lasted a median of 6 days (mean 14.78 days, SD 20.84), with a range of 0 to 100 days per issue. Within the call logs, 41% of calls were described as general computer issues that were addressed via remote connection or home visit by IT staff. The most common specific problem reported was infection by virus or spyware (32%). An additional 11% of calls referred specifically to rebuilding or restoring the machine to address problems. Six percent referred to machine slowness, while 5% dealt specifically with replacing parts, and 5% dealt with electrical power issues.

Table 2:

Call Log and Self-Reported Information Barriers (n = 118)

Variable	% (n)
Computer Problem Calls (median, range)	2 (0–9)
Computer Problem Duration, in Days (median, range)	6 (0–100)
Internet Problem Calls (median, range)	2 (0–9)
Internet Problem Duration, in Days (median, range)	6.5 (0–120)
Moves
0 moves	76 (90)
1 move	14 (17)
2+ moves	10 (11)
Concerned about Information Quality
Yes	54 (64)
No	46 (54)
Information was Hard to Understand
Yes	28 (33)
No	72 (85)
Took a Lot of Effort to Find Health Information
Yes	43 (51)
No	57 (67)
Frustrated in Search for Health Information
Yes	30 (35)
No	70 (83)
Self-reported Computer Use Barriers
No Free Time to Use Computer
Always/often/sometimes	70 (83)
Rarely/never	30 (35)
Lack of Internet Connection
Always/often/sometimes	39 (46)
Rarely/never	61 (72)
Other Family Uses Computer
Always/often/sometimes	35 (41)
Rarely/never	65 (77)
Family Does not Want Them to Use Computer
Always/often/sometimes	23 (27)
Rarely/never	77 (91)
Computer Skills
Always/often/sometimes	37 (44)
Rarely/never	63 (74)
Literacy
Always/often/sometimes	31 (36)
Rarely/never	69 (82)

Internet connectivity barriers

There were a total of 293 internet issues over the study period. There was a median of two calls (mean 2.29, SD 2.01) with a range of no calls to nine calls per household (Table 2) lasting a median of 6.5 days (mean 17.85 days, SD 28.47, range 0 to 120 days) per issue. Sixty-five percent of these calls were described as general internet issues, such as maintaining connectivity. Fourteen percent of calls were attributed to issues in scheduling installation appointments with the internet provider. Ten percent of problems were directly linked with internet termination due to nonpayment of a bundled account (e.g., participants were responsible for other services on the account such as cable and did not pay, therefore suspending the entire account), while 5% cited issues with a certain site, and 4% cited modem issues.

Relocation barriers

During the study period, 24% of study participants experienced at least one move, with 10% moving twice or more.

Perceived Barriers

Notably, over half (54%) of participants indicated that they felt concerned by the quality of information they found about health, and 43% felt it took a lot of effort to find acceptable information (Table 2). Thirty percent felt frustrated in their search for information. In regards to computer use, the majority of the sample (70%) indicated lack of free time as a barrier to computer use at least some of the time, with other barriers including problems with internet connectivity (39%), lack of computer skills (37%), other family members using the computer (35%), and literacy (31%).

Barriers and Internet Health Information Seeking

Internet health information seeking was negatively associated with the number of days an internet problem was experienced (Table 3); the rate of internet health information seeking decreased 1% for each additional day a problem persisted when other variables were held constant (incidence rate ratio (IRR) = 0.990, P = .030, 95% CI, 0.980-0.999). This trend was also observed in the stratified models. Among participants with a medical problem, moving to a different residence was significantly associated with a decrease in the rate of internet health information seeking by 38.7% for each additional move, accounting for other variables in the model (IRR = 0.613, P = .034 95% CI, 0.389-0.965). Participants with a medical problem reporting frustration (Table 3) searched for health information at less than half the rate of those who were not frustrated (IRR) = 0.395, P = .030 95% CI, 0.171-0.916). Conversely, participants without a medical problem who reported frustration when looking for information searched for internet health information at 2.856 times the rate of those who were not frustrated (P = .036, 95% CI, 1.072-7.605).

Table 3.

Technology and Perceived Predictors of Internet Health Information Seeking in Unadjusted, Adjusted, and Stratified Models (by Medical Problem) (n = 118)

	Model 1 (unadjusted)	Model 2 (adjusted)	Model 3	Model 4
	(n = 118)	(n = 118)	Reported Medical Problem	Did Not Report Medical Problem
			(n = 75)	(n = 43)
	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)	IRR (95% CI)
Medical Problem (yes)	2.263 (1.313-3.901)**	2.065 (1.176-3.626)*
Days with Internet Issues	0.992 (0.983-1.001)	0.990 (0.981-0.999)*	0.990 (0.978-1.002)	0.981 (0.966-0.996)*
Number of Moves	0.700 (0.481-1.019)	0.700 (0.481-1.019)	0.613 (0.389-0.965)*	1.113 (0.554-2.238)
Other Family Uses Computer	1.066 (0.592-1.921)	0.990 (0.551-1.777)	0.874 (0.428-1.779)	2.317 (0.766-7.013)
Literacy Barriers	1.403 (0.779-2.529)	1.403 (0.779-2.529)	1.825 (0.837-3.797)	1.381 (0.494-3.861)
Greater Effort to Find Health Information	0.677 (0.388-1.182)	0.671 (0.364-1.238)	0.563 (0.272-1.161)	1.975 (0.677-5.766)
Frustrated in Information Search	0.643 (0.348-1.189)	0.686 (0.362-1.301)	0.395 (0.171-0.916)*	2.856 (1.072-7.605)*
Hard to Understand Information	0.602 (0.332-1.092)	0.627 (0.343-1.145)	0.488 (0.235-1.015)	1.361 (0.503-3.685)

*P < .05, **P < .01.

Each model adjusted for: wave, race, employment status, native language when variable met criteria of

P ≤ .25 in bivariate models.

Only predictors that had a P-value of 0.15 or below in at least one model are shown. The bold text indicates that the value was significant at p<0.05.

DISCUSSION

We used detailed, multifaceted data from a randomized controlled trial to provide a picture of the internet health information seeking behaviors of the urban poor.³³ Numerous internet connectivity barriers were found, with each household averaging two issue calls and each issue lasting a median of 6.5 days. Over half of participants reported concern about information quality and 43% reporting that it took a lot of effort to find health information. A significant, negative relationship was found between number of days experiencing internet issues and internet health information seeking. Frequent moves were also significantly associated with lower rates of seeking. While frustration was associated with lower rates of health seeking among those with a medical condition, it was associated with higher rates of seeking among those without a medical condition.

Health Information Seeking

This study reported actual, observed levels of internet activity in a natural environment. Although there was a wide range of hits per household to health websites, the low median number of hits reflects other studies that have found health information seeking is often not the top online activity for low-SEP individuals.^16,27 This finding is also supported by other analyses of the intervention’s online tracking data, which indicate participants spent the majority of their online time on social networking sites and internet portal sites such as MSN.com.³³ However, we are not able to determine whether they were exposed to health information through social network posts or general purpose sites; participants may be exposed to a range of health topics through these sources that are not specifically health-designated websites. Further research is warranted to determine how to link health promotion efforts with other types of websites to better reach low-SEP individuals with health information outside of traditional health websites.³³

Technological Barriers

This study found that internet barriers, and therefore communication inequalities, persisted after initial access was granted. Despite the provision of free internet access and having a team of IT and study staff on call throughout the study,³⁶ thus “ensuring” internet access and overcoming the digital divide, a second level divide occurred as the struggle to maintain connectivity began, as has been suggested elsewhere.²¹ Although the magnitude of the association was small, possibly due to the ongoing IT support to mitigate these issues, the unit of observation was per day; some participants experienced up to 100 days of service disruption. Furthermore, given that this effect was still observed in a population that had access to consistent IT support, it may be assumed that these effects would be amplified in the absence of a support system.

The frequency with which participants encountered viruses and spyware and its potential to damage computer software³⁰ may indicate the need for more training on safe web browsing as low-SEP participants become more active in the online world. Frequent moves may have the potential to exacerbate barriers and contribute to lower rates of heath information seeking as moves require more interaction with service providers to schedule install and set up billing.^31,32 These additional demands may be particularly difficult for those with a medical problem, since they may have competing concerns due to scheduling medical visits or other interactions with the healthcare system. Furthermore, despite the study staff serving as advocates for the participants, often acting as liaison between participant and the internet provider, problems with billing and scheduling persisted and often resurfaced as recurring issues. It is unknown what the impact of technical problems may have been without the presence of the staff and IT support. Thus, these data point to the need for ongoing IT support in eHealth interventions with low-SEP individuals to monitor potential problems with connectivity.

Perceived Barriers

In light of the range of skills required to engage with the internet,^37,38 and reports of perceived barriers to computer use corroborate previous evidence that once online, low-SEP users face a number of challenges with information gathering.²⁰ Findings of participant frustration with finding information reflects literature that low literacy users may be overwhelmed by the staggering amount of health information available on the internet, with many sources of questionable quality or trustworthiness.^2,3 Although our initial results indicated that like other studies, individuals with medical problems are more likely to search for health information,²⁹ these findings suggest that frustration in finding this information may serve to discourage them from further seeking. Conversely, in the face of these barriers, those without a health problem increased their seeking behaviors when frustrated; dissatisfaction with the information found may serve to push individuals to search for information that better meets their needs. This discrepancy may be due to frustration being more emotionally taxing for those searching for their own health condition information. It is also of note that 70% of the sample reported that no free time as a barrier to using computers. This may be indicative of other competing concerns, life concerns (such as moves and work). concerns that limit scheduling internet provider service calls, that take precedence over computer use. Other studies have shown that people from low-SEP groups are unable to take advantage of interventions that directly benefit them because of other daily life challenges.⁴¹

Limitations

Not all instances of computer issues may have been reported, leading us to underestimate barriers. Our sample may not be representative of other low-SEP groups in Boston or the United States. Though the restricted range of our sample precluded gauging differences by income or education, the nature of the sample made it ideal for studying the communication behaviors of a low-SEP group. We recruited participants through presentations at adult education centers, which may have led to selection bias because these individuals may have been more interested in or able to use computers and the internet compared to those who were not enrolled in such classes, limiting the generalizability of our findings. Per our IRB mandate, we tracked internet data at the household level, which obscured individuals’ online behavior and our ability to know how many individuals in the household were using the internet. To address this discrepancy, self-reported internet use data from each participant was cross-checked against website tracking data.

CONCLUSION

This study examined the myriad barriers that face the urban poor as they embark upon internet health information seeking. Although internet access was granted through intervention procedures, a second level digital divide emerged as households with difficulty maintaining internet connectivity experienced significantly lower rates of internet health information seeking, reflecting the persistence of communication inequalities. Additionally, those with a medical problem may be particularly vulnerable to frustration in searching for information, becoming discouraged in information seeking when encountering difficulties in finding needed information. This study highlights the need for ongoing IT support for eHealth studies with low-SEP individuals and using continuous, real-time web tracking data to understand the complex network of barriers that may impact internet health information seeking for the urban poor.

Funding

This work was supported by NIH grant number R25 CA057711 and R01 CA 122894. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

Competing Interests

The authors have no competing interests to declare.

Contributors

All authors contributed to this manuscript. R.McC. served as the lead author, conducting data analysis and leading manuscript preparation and writing. C.O. and G.S. provided analytical guidance, feedback, and edits during the process. K.V. was the PI for the C2C project, and provided substantive guidance through the development, analysis, and writing of the manuscript. All authors have approved this work.