714
Weaver et al., Internet Medical Information Users
Research Paper 䡲
Profiling Characteristics of Internet Medical Information Users
JAMES B. WEAVER, III, PHD, MPH, DARREN MAYS, MPH, GREGG LINDNER, MA, DOĞAN EROĞLU, PHD,
FREDERICK FRIDINGER, DRPH, JAY M. BERNHARDT, PHD, MPH
Abstract
Objective: The Internet’s potential to bolster health promotion and disease prevention efforts has
attracted considerable attention. Existing research leaves two things unclear, however: the prevalence of online health
and medical information seeking and the distinguishing characteristics of individuals who seek that information.
Design: This study seeks to clarify and extend the knowledge base concerning health and medical information use
online by profiling adults using Internet medical information (IMI). Secondary analysis of survey data from a
large sample (n ⫽ 6,119) representative of the Atlanta, GA, area informed this investigation.
Measurements: Five survey questions were used to assess IMI use and general computer and Internet use during
the 30 days before the survey was administered. Five questions were also used to assess respondents’ health care
system use. Several demographic characteristics were measured.
Results: Contrary to most prior research, this study found relatively low prevalence of IMI-seeking behavior.
Specifically, IMI use was reported by 13.2% of all respondents (n ⫽ 6,119) and by 21.1% of respondents with
Internet access (n ⫽ 3,829). Logistic regression models conducted among respondents accessing the Internet in the
previous 30 days revealed that, when controlling for several sociodemographic characteristics, home computer
ownership, online time per week, and health care system use are all positively linked with IMI-seeking behavior.
Conclusions: The data suggest it may be premature to embrace unilaterally the Internet as an effective asset for
health promotion and disease prevention efforts that target the public.
䡲 J Am Med Inform Assoc. 2009;16:714 –722. DOI 10.1197/jamia.M3150.
Introduction
Internet access is a widely diffused technology in the United
States. Almost 80% of adults aged 18 years and older access
the Internet,1,2 and the majority (58%) use high-speed broadband connections.3,4 Searching for and using health information online appears to be a prevalent activity among
United States adults, with some estimates at 40 –70%, or
90 –160 million people.3,5–7 Indeed, data suggest that from
Affiliations of the authors: National Center for Health Marketing,
Centers for Disease Control and Prevention (JBW, DM, DE, FF,
JMB), Atlanta, GA; Department of Behavioral Sciences and Health
Education, Rollins School of Public Health, Emory University (DM),
Atlanta, GA; Scarborough Research (GL), New York, NY.
The authors are indebted to Jennifer Cadoret, Richard E. Dixon, and
Marinella Macri for their significant contributions to this project.
This research was supported in part by Scarborough Research and
by the appointment of the first and second authors to the Research
Participation Program at the Centers for Disease Control and
Prevention administered by the Oak Ridge Institute for Science and
Education through an interagency agreement between the United
States Department of Energy and CDC.
The findings and conclusions in this article are those of the authors
and do not necessarily represent the views of the Centers for
Disease Control and Prevention or the United States Department of
Health and Human Services.
Correspondence: James B. Weaver, III, PhD, MPH, National Center for
Health Marketing, Centers for Disease Control and Prevention, 1600
Clifton Road, MS-E21, Atlanta, GA 30333; e-mail: ⬍Jim.Weaver@
CDC.GOV⬎.
Received for review: 01/21/09; accepted for publication: 06/09/09.
2001–2007 the number of United States adults who used the
Internet to seek health information nearly doubled8 and that
adults are more likely to seek health information online than
from interpersonal sources, such as doctors, friends, family,
and coworkers.9 As a result, the Internet is broadly recognized as a potentially important tool for transforming medical care and public health.10 –12 Researchers conclude that
the Internet offers tremendous promise as a health communication and education tool,13,14 is an economical and effective part of chronic disease management,15,16 and is a key
resource in health behavior change interventions and programs.17,18 In addition, information technologies such as the
Internet are receiving even greater attention for their potential impact on public health and health care in health care
reform policy discussions.19
Inconsistencies exist, however, in the research that profiles
individuals who engage in health information seeking behaviors (HISB) online, and these inconsistencies obstruct researchers from making accurate projections about the Internet’s
potential for health promotion and disease prevention. Both in
peer-reviewed10,20 –24 and self-published1,2,8,9 reports, for example, strong variability in HISB prevalence exists: estimates
range from 13% to more than 80%. One explanation may be the
different time frame referents used in questions about HISB.
In some studies,1,2 respondents were asked if they had
“ever” used Internet health information. The prevalence
estimates in these studies ranged from 70% to more than
80%. In studies8 –10,23–25 in which respondents had a
narrower time frame (e.g., “over the last 12 months”),
prevalence diminished substantially, from 40% to 60%.
Journal of the American Medical Informatics Association
Volume 16
Number 5
September / October 2009
715
In studies that asked respondents to consider both broad
and narrow time frames, wide discrepancies also emerged.
In one such study,1 prevalence for the “ever” referent was
71%, whereas for the “in the last month” referent, prevalence
was 53%, an 18% decrease. Another study20 reported a 20%
decrease in prevalence between “ever” and a time frame that
was “every 2–3 months or more frequently within the last
year.” While longer time frames are useful for formulating
broad prevalence estimates, requiring respondents to recall
exact behaviors over an extended period may introduce
considerable response bias. Rather, limiting responses to a
shorter time frame could help clarify online HISB and better
inform time-sensitive, health promotion and disease prevention efforts.
Despite these limitations, some characteristics of individuals
who seek health information online have emerged.30,32 For
instance, females are more likely to seek health information
online,9,33 as are those with higher educational achievement.8 –10,23,24 Specific health reasons (e.g., being diagnosed
with a new health problem, ongoing medical conditions) are
also linked with online information seeking.29,32,34 For other
sociodemographic characteristics, however, results are less
consistent. Age, income, and race/ethnicity, for instance,
have yielded mixed results in several studies.8 –10,21,23–25
Surprisingly, the basic measurements of computer and Internet use—factors thought to be strong determinants underlying online HISB2—are essentially missing in most
peer-reviewed investigations.10,25
Another inconsistency in past studies is in the way studies
define “health information.” Examples abound, including
“information or advice about health care”10,14,20, “information
about a personal health concern”8, “health- and wellnessrelated information”9, and “health or medical information”26.
Operationalizations as broad as these allow respondents to
perceive “health information” in vague and ambiguous terms.
Indeed, research9 shows that the “health- and wellnessrelated topics” that Internet users search for include many
areas related to health status, ranging from “symptoms”
(⬃60%), “treatment” and “diseases/conditions” (⬃55–58%),
“wellness” (i.e., nutrition, exercise, weight loss) and “drugs/
medications” (⬃40 – 46%), and “vitamins/supplements” and
“alternative medicines” (⬃20 –30%). The topics also included “health insurance/healthcare providers”, “doctor/
medical practices”, “care services” (i.e., hospitals, clinics,
managed care), and “pharmaceutical companies” (all ⬍20%;
p. 9). Equally revealing, emerging evidence shows that the
public distinguishes between general health information
(e.g., wellness, vitamins, etc) and medical information (e.g.,
treatment, symptoms, diseases/conditions, etc) in their online health information seeking behavior.27 Other evidence
shows that the topics sought in online HISB vary as a
function of a respondent’s age,28 health status,21 or chronic
conditions.8,10,22,29 In light of these observations, it can be
argued that many operationalizations of “health information” are problematic because they allow respondents tremendous latitude in ascribing meaning to a focal concept
and, consequently, invite tremendous sample-specific response variability.30
Taken together, these limitations impede our understanding
of the Internet as an asset for health promotion and disease
prevention. To shed light on the Internet’s potential as a
channel, this study examined online use of “medical” information—a more specific operationalization than “health”
information use—as the focal behavior. Several possible
determinants of Internet medical information (IMI) use that
have not been previously explored, including sociodemographics, computer and Internet use, and health care use,
were examined via secondary analysis of survey data from a
large sample representative of the Atlanta, GA, area.
At the same time, the body of research that informs our
understanding of Internet HISB reflects several major limitations, both in the targeted samples and in the sampling
procedures. Baker et al.10 for example, drew their conclusions from respondents to whom Internet access was provided as a participation incentive. Although the group took
care to account for any consequences, the potential for bias
cannot be overlooked. Sampling limitations also must shape
interpretations of data drawn from the Health Information
National Trends Survey 200523–26 and the Pew Internet and
American Life Project.31 For both, the solicitation of a
nationally representative sample of adults for an extensive
telephone interview yielded low response rates, about 21%
and 27%, respectively. Lastly, for other reports,1,8,9 sufficient
methodological details are not available to assess accurately
the sampling frame and its potential impact on findings.
Methods
Respondents and Setting
This study uses data collected in 2006 and 2007 from adult
men (n ⫽ 2,394) and women (n ⫽ 3,725) who reside in the
Atlanta, GA, designated market area (DMA). The data were
drawn from Scarborough Research,35,36 which conducts
local market, Media Rating Council-accredited, consumer
marketing surveys in 81 DMAs in the United States to
examine new and traditional media use, health care system
use, and lifestyle and consumer behaviors. Comparisons of
individual and household demographic variables from 2006
to 2007 reveal only one significant trend, a 3.1% decline in
home ownership, ␹2(1, n ⫽ 6,119) ⫽ 4.8, p ⬍ 0.05. Consequently, the samples were combined for subsequent analyses (2006, n ⫽ 3,077; 2007, n ⫽ 3,042).
Sampling Procedure
The sampling frame included all households in the Atlanta
DMA with a landline telephone and in which at least one
adult, aged 18 years or older, resided. A DMA, as defined by
Nielsen Media Research, is a measurement area to which a
county is exclusively assigned on the basis of marketspecific television viewing. The Atlanta DMA includes 52
counties in north Georgia, 2 counties in western Alabama,
and 1 county in North Carolina.
Within this DMA, two sampling strata were defined: the
Metro Survey Area (MSA) and the non-Metro area. The
MSA corresponds closely to the Office of Management and
Budget’s Metropolitan Area.37 Population estimates, annually updated by Claritas,38 were used in sample balancing.
Sample balancing variables included geography, age within
gender, household size, education, race and ethnicity. Data
were weighted and projected to adults aged 18 years or
older in the Atlanta DMA.
716
Weaver et al., Internet Medical Information Users
Scarborough Research used a two-stage, random sampling
procedure. In the first stage, households were selected with
a systematic random sampling technique. In the second
stage, an adult aged 18 years or older with the most recent
(last) birthday was chosen as the designated respondent. Up
to 16 attempts were then made to reach that individual.
Survey Methodology
Complete details about the survey methodology are available elsewhere.39 Briefly, an eligible adult from each selected
household completed a Computer Assisted Telephone Interview (CATI) that assessed information about demographics,
health care usage, and media use including Internet behaviors. For the 2006 and 2007 surveys, the response rates for
the CATI interview were 45.2 and 40.5%, respectively.
Measures
Demographics
Demographic variables were assessed by individual and by
household. Individual variables included gender, age, race/
ethnicity, level of education, employment status, and marital
status. Table 1 details the levels assigned within each measure. Household variables included annual household income, number of members, number of adults, number of
children (aged 17 yrs or younger), and home ownership.
These measures are detailed in Table 2.
Internet Medical Information Use
The use of IMI was assessed via a question set asking
respondents to “Check ways you used online/Internet services in the past 30 days.” “Medical services/information”
was one of almost 3-dozen online/Internet services including “download/listen to music”, “e-mail”, “financial information/services”, and “pay bills.” A single dichotomous
IMI use item was retained to indicate IMI use.
Computer and Internet Use
Four aspects of computer and Internet use were measured
(see Table 3). One item was used to assess Internet access:
“Do you have access to the Internet? Include computers at
work and home, portable electronic devices, etc.” Another
assessed where respondents accessed the Internet in the past
30 days. The choices included home, work, school, public
library, or other place; respondents checked all that applied.
A “did not access Internet” option was also provided.
Individuals who lacked Internet access or who had not
accessed the Internet in the past 30 days were excluded from
the study.
Table 1 y Individual Demographic Determinants Underlying Internet Medical Information Use
Individual Characteristics
Age (␹2[5] ⫽ 23.9**)
18–24 yrs old
25–34 yrs old
35–44 yrs old
45–54 yrs old
55–64 yrs old
65⫹ yrs old
Education (␹2[2] ⫽ 10.11*)
High school or less
Some post high school
College plus
Employment (␹2[6] ⫽ 6.05)
White collar
Blue-collar
Part-time, white-collar
Part-time, blue-collar
Homemaker
Retired or disabled
Unemployed
Race/ethnicity (␹2[4] ⫽ 22.96**)
White
African-American
Asian
Hispanic
Other
Gender (␹2[1] ⫽ 10.71*)
Female
Male
Marital status (␹2[2] ⫽ 10.97*)
Never married
Married
Unmarried
IMI Users (N ⫽ 807)
IMI Nonusers (N ⫽ 3,022)
Unadjusted Odds Ratio
95% Confidence Interval
8.1%
22.4%
24.2%
21.2%
17.4%
6.8%
14.6%
21.9%
25.7%
20.0%
11.6%
6.1%
Ref.
1.85†
1.69†
1.91†
2.69†
2.00†
1.15, 2.97
1.08, 2.67
1.22, 3.00
1.23, 4.24
1.23, 3.26
4.0%
55.0%
41.40%
4.7%
61.1%
34.2%
Ref.
1.06
1.41
0.59, 1.89
0.80, 2.50
45.1%
13.4%
10.0%
6.7%
8.6%
10.4%
5.7%
45.9%
15.0%
9.3%
5.5%
7.7%
8.5%
7.9%
Ref.
0.91
1.09
1.25
1.14
1.19
0.73
0.65, 1.28
0.78, 1.51
0.72, 2.00
0.81, 1.59
0.92, 1.55
0.47, 1.16
76.3%
14.6%
2.0%
3.3%
3.9%
67.1%
23.3%
2.2%
4.3%
3.0%
Ref.
0.55†
0.80
0.67
1.14
0.43, 0.71
0.38, 1.69
0.37, 1.23
0.66, 1.96
58.1%
41.9%
49.9%
50.1%
1.39†
Ref.
19.0%
68.3%
12.7%
26.3%
61.3%
12.4%
Ref.
1.54†
1.41†
IMI ⫽ Internet Medical Information.
Note: Chi-squares are from Wald test for independence based on the unadjusted log odds ratios.
†Confidence intervals (95% CI) not containing the null value (1.00) are statistically significant at p ⬍ 0.05.
*p ⬍ 0.01.
**p ⬍ 0.001.
1.14, 1.68
1.19, 2.00
1.02, 1.95
Journal of the American Medical Informatics Association
Volume 16
Number 5
September / October 2009
717
Table 2 y Household Demographic Determinants Underlying Internet Medical Information Use
Household Characteristics
Income (␹2[6] ⫽ 6.58)
ⱕ$24,999/yr
$25,000–$34,999/yr
$35,000–$49,999/yr
$50,000–$74,999/yr
$75,000–$99,999/yr
$100,000–$149,999/yr
ⱖ$150,000/yr
Home ownership (␹2[1] ⫽ 1.86)
Own
Rent/other
Number of HH Members (t [3827] ⫽ ⫺2.34*)
Number of HH adults (t [3827] ⫽ ⫺3.72**)
Number of HH kids (t [3827] ⫽ ⫺0.45)
IMI Users
(N ⫽ 807)
IMI Nonusers
(N ⫽ 3,022)
Unadjusted
Odds Ratio
95% Confidence
Interval
4.1%
7.0%
17.2%
20.7%
16.2%
21.4%
13.65%
4.3%
6.1%
18.9%
19.6%
19.9%
18.7%
12.5%
Ref.
1.21
0.95
1.11
0.85
1.20
0.13
0.67, 2.19
0.57, 1.59
0.67.1.84
0.51, 1.42
0.72, 2.00
0.68, 1.89
82.2%
17.8%
3.08
2.20
1.88
79.3%
20.7%
3.25
2.35
1.91
1.20
Ref.
0.92, 1.56
HH ⫽ household; IMI ⫽ internet medical information.
Note: Chi-squares are from Wald test for independence based on the log odds ratios.
*p ⬍ 0.05.
**p ⬍ 0.001.
Respondents also estimated time spent online in an average
week. The choices included none, less than 1 hour 1– 4 hours
5–9 hours 10 –19 hours, and 20 hours or more. Respondents
indicated if they had home broadband connection; yes if
their household used a cable modem or DSL or no if dial-up,
other connection, or no Internet connection. Computer ownership was confirmed if a respondent or other household
member owned a personal or home computer.
Health Care Use
In all, five measures were used to assess health care system use
(see Table 4). Three determined services use. Respondents
were asked if they or anyone in their household had used
services in the past 3 years at a hospital, at a nonhospital
medical facility, and if during the past 12 months they or
anyone in their household had visited a health care specialist
(respondents selected from a list of specialists). Two measures determined medication use in the past 12 months.
Respondents indicated if they had purchased any over-thecounter or prescription drugs (respondents selected from a list of
reasons), and if they had purchased medicine or prescriptions on
the Internet. For each of the five measures, a dichotomous variable
reflecting any or no use was computed.
Data Analyses
The SAS Version 9.2 (SAS Institute, Inc, Cary, NC) procedures
incorporating stratification and weighting to accommodate
complex sampling designs (e.g., PROC SURVEYFREQ, PROC
SURVEYLOGISTIC) were used for data analyses.40 Because
Scarborough Research stratified sampling based on Metro
Survey Area (MSA) and non-MSA residency status, residency within the MSA served as the sample stratification
measure to account for this aspect of the survey design.40
The data were also adjusted based on weights provided by
Scarborough Research to compensate for both differential
sample selection probabilities and population subgroup
response rates. This analytic approach yielded estimates
reflecting the adult population within the Atlanta DMA
based on the survey sample.
Missing data management was undertaken by Scarborough
Research which reported using a two-step approach. First,
Table 3 y Internet and Computer Use Determinants Underlying Internet Medical Information Use
Internet and Computer Use
Home computer ownership (␹2[1] ⫽ 16.82**)
Home broadband connection (␹2[1] ⫽ 6.92*)
Places accessed (past 30 days) (␹2[1] ⫽ 22.62**)
Away from home
Home
Home and work
Online Time per week (␹2[3] ⫽ 44.0**)
4 hours or less
5–9 hours
10–19 hours
20⫹ hours
IMI Users
(N ⫽ 807)
IMI Nonusers
(N ⫽ 3,022)
Unadjusted
Odds Ratio
95% Confidence
Interval
96.2%
77.1%
90.6%
71.5%
2.63†
1.35†
1.66, 4.17
1.08, 1.68
5.4%
46.5%
48.0%
7.8%
52.1%
40.1%
Ref.
1.28
1.72†
0.84, 1.95
1.13, 2.61
31.9%
26.0%
19.8%
22.2%
46.8%
24.0%
14.6%
14.7%
Ref.
1.59†
2.00†
2.22†
1.23, 2.04
1.53, 2.62
1.70, 2.90
IMI ⫽ internet medical information.
†Confidence intervals (95% CI) not containing the null value (1.00) are statistically significant at p ⬍ 0.05.
*p ⬍ 0.01.
**p ⬍ 0.001.
718
Weaver et al., Internet Medical Information Users
Table 4 y Health Care Utilization Determinants Underlying Internet Medical Information Use
Health Care Usage
Any
Any
Any
Any
Any
IMI Users
(N ⫽ 807)
IMI Nonusers
(N ⫽ 3,022)
Unadjusted
Odds Ratio
95% Confidence
Interval
80.3%
48.9%
89.3%
96.7%
14.1%
71.3%
33.2%
80.21%
89.8%
5.5%
1.70†
1.93†
2.06†
3.27†
2.82†
1.36, 2.13
1.59, 2.33
1.49, 2.84
2.06, 5.18
2.08, 3.81
hospital service (␹2[1] ⫽ 21.10**)
medical facility service (␹2[1] ⫽ 45.48**)
specialist visit (␹2[1] ⫽ 19.34**)
medication purchase (␹2[1] ⫽ 25.23**)
Internet medication purchase (␹2[1] ⫽ 42.59**)
IMI ⫽ internet medical information.
†Confidence intervals (95% CI) not containing the null value (1.00) are statistically significant at p ⬍ 0.05.
**p ⬍ 0.001.
respondents were required to provide answers to several
items (e.g., age, sex) for sample inclusion. Next, missing data
due to item nonresponse were imputed using a line-item
ascription procedure that apportioned estimates randomly
to maintain initial response distributions.
Included in analyses were respondents in the combined 2006
and 2007 dataset who indicated they had accessed the
Internet in the previous 30 days (62.6%; n ⫽ 3,829). Data
analysis was conducted in two steps. First, bivariate analyses were used to compare IMI users with nonusers on the
basis of demographics, Internet access, and health care use.
Wald ␹2 statistics were computed to examine the statistical
significance of each comparison, and unadjusted odds ratios
were calculated using logistic regression models.41 Next,
four logistic regression models were created to explain the
relationships between the dependent variable, IMI use, and the
independent variables including individual demographic characteristics (Model 1), household demographics (added in
Model 2), Internet and computer use (added in Model 3), and
health care use (added in Model 4). Log likelihood ratio tests
were used to assess whether including each group of independent variables added significantly to the subsequent model.42
The Akaike Information Criterion (AIC), a measure of goodness-of-fit of statistical models, was used to compare the
amount of information explained across the models, with
lower AIC models indicating improved model fit.41
Results
Internet medical information (IMI) use during the previous
30 days (n ⫽ 807) was reported by 13.2% of the entire sample
(n ⫽ 6,119). When results were limited to those respondents
who accessed the Internet in the past 30 days (n ⫽ 3,829),
21.1% of respondents reported IMI use.
Comparing Internet Medical Information Users
and Non-users
Demographic Determinants
Significant differences in demographic characteristics emerged,
including in age, gender, race/ethnicity, and marital status (see
Table 1). On the basis of unadjusted odds, respondents in older
age groups were 1.5–2.7 times more likely to report IMI use
than those in the referent group aged 18 –24 years. Others more
likely to report IMI use included females more than males and
white respondents more than African Americans. Additionally, respondents who were married or currently unmarried
were more likely to report IMI use than those who were
never married. Education, income, and employment, by
contrast, yielded no significant relationships. Differences in
household demographics can be found in Table 2. The
results of t tests revealed that IMI users lived in households
with fewer adults and fewer people overall than nonusers
(Table 2).
Internet Determinants
As Table 3 illustrates, all four measures of computer and
Internet use yielded significant differences between IMI
users and nonusers. For example, IMI users were more likely
to own a home computer and have a household broadband
connection. They also tended to spend 5 hours or more
online per week. In general, the relationship between IMI
use and time spent online per week grew stronger the more
time respondents spent online (Table 3). Finally, based on
unadjusted odds, respondents who accessed the Internet in
places other than work or home were less likely to report IMI
use than those who accessed the Internet at work (Table 3).
Health Care Use
As Table 4 illustrates, respondents who reported IMI use also
reported greater health care system use across all five measures. Specifically, IMI users were more likely to have used
a hospital or other medical facility services during the past 3
years, were more likely to have visited a specialist during
the previous 12 months, and were more likely to have
purchased medications both overall and via the Internet.
Explaining Internet Medical Information Use
In Model 1, IMI use was regressed on the individual
demographic characteristics that bivariate analyses showed
were significantly associated with IMI use: gender, age,
race/ethnicity, and marital status. A respondent’s level of
education, which past research consistently associates with
online health information seeking, was also included. When
controlling for other variables in the model, results from
Model 1 showed that those more likely to report IMI use
included females, respondents aged 55– 64 years and white
respondents (Table 5).
In Model 2, household demographics were added to the
logistic regression model. Three included the characteristics
that bivariate analyses showed were significantly associated
with IMI use: home ownership, number of adults, and
number of children. The fourth was household income,
which past research consistently associates with online
health information seeking. When controlling for other
variables in the model, results from Model 2 showed no
significant associations between these variables and IMI use
(Table 5). Adding household demographics resulted in a
minimal change in the AIC (AIC ⌬ ⫽ 0.004%). The significant relationships across individual demographics remained
between Models 1 and 2.
Journal of the American Medical Informatics Association
Volume 16
Number 5
September / October 2009
719
Table 5 y Multiple Logistic Regression Models Examining Internet Medical Information Use
Female
Age
18–24 yrs old
25–34 yrs old
35–44 yrs old
45–54 yrs old
55–64 yrs old
65⫹ yrs old
Education
High school or less
Some posthigh school
College plus
Race/ethnicity
White
African American
Asian
Hispanic
Other
Marital status
Never married
Married
Unmarried
Income
ⱕ$24,999/yr
$25,000–$34,999/yr
$35,000–$49,999/yr
$50,000–$74,999/yr
$75,000–$99,999/yr
$100,000–$149,999/yr
ⱖ$150,000/yr
Home ownership
Number of HH adults
Number of HH kids
Home broadband connection
Places accessed (past 30 days)
Away from Home
Home
Home and work
Home computer ownership
Online time per week
4 hours or less
5–9 hours
10–19 hours
20⫹ hours
Any hospital
Any medical facility
Any specialist
Any medications
Internet Rx purchases
Model-specific
Wald ␹2 statistic
AIC§
⫺2Log likelihood test
Model 1 OR (95% CI)
Model 2 OR (95% CI)
Model 3 OR (95% CI)
Model 4 OR (95% CI)
1.46* (1.19, 1.78)
1.43* (1.17, 1.75)
1.60* (1.30, 1.98)
1.46* (1.18, 1.81)
Ref.
1.59 (0.96, 2.62)
1.42 (0.86, 2.34)
1.57 (0.95, 2.59)
2.18* (1.31, 3.62)
1.56 (0.91, 2.69)
Ref.
1.39 (0.84, 2.31)
1.25 (0.74, 2.09)
1.44 (0.86, 2.42)
1.96* (1.15, 3.35)
1.38 (0.78, 2.45)
Ref.
1.33 (0.79, 2.22)
1.20 (0.72, 2.02)
1.40 (0.83, 2.36)
1.99* (1.16, 3.41)
1.56 (0.88, 2.77)
Ref.
1.23 (0.73, 2.08)
1.10 (0.64, 1.87)
1.25 (0.73, 2.13)
1.75* (1.01, 3.04)
1.31 (0.72, 2.37)
Ref.
0.86 (0.48, 1.56)
1.09 (0.61, 1.97)
Ref.
0.86 (0.47, 1.57)
1.07 (0.58, 1.99)
Ref.
0.78 (0.42, 1.45)
0.87 (0.46, 1.64)
Ref.
0.71 (0.38, 1.33)
0.77 (0.41, 1.47)
Ref.
0.58* (0.45, 0.76)
0.86 (0.40, 1.81)
0.81 (0.44, 1.48)
1.18 (0.68, 2.07)
Ref.
0.58* (0.44, 0.77)
0.89 (0.41, 1.91)
0.85 (0.46, 1.56)
1.23 (0.69, 2.18)
Ref.
0.59* (0.45, 0.78)
0.77 (0.36, 1.62)
0.92 (0.50, 1.71)
1.30 (0.71, 2.35)
Ref.
0.62* (0.46, 0.82)
0.85 (0.40, 1.80)
1.00 (0.53, 1.89)
1.31 (0.72, 2.38)
Ref.
1.13 (0.84, 1.52)
1.01 (0.71, 1.44)
Ref.
1.20 (0.89, 1.63)
0.96 (0.66, 1.38)
Ref.
1.25 (0.92, 1.70)
0.97 (0.67, 1.41)
Ref.
1.09 (0.79, 1.50)
0.94 (0.64, 1.36)
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
Ref.
1.15 (0.63, 2.12)
0.97 (0.57, 1.67)
1.08 (0.63, 1.85)
0.81 (0.46, 1.40)
1.07 (0.61, 1.89)
0.96 (0.54, 1.71)
0.93 (0.69, 1.26)
0.86 (0.75, 1.00)
1.02 (0.92, 1.12)
⫺
Ref.
1.18 (0.64, 2.18)
0.92 (0.54, 1.59)
0.96 (0.56, 1.65)
0.75 (0.43, 1.31)
0.93 (0.52, 1.66)
0.83 (0.46, 1.49)
0.91 (0.67, 1.23)
0.86 (0.75, 1.00)
1.03 (0.93, 1.14)
1.18 (0.92, 1.51)
Ref.
1.28 (0.70, 2.34)
0.97 (0.57, 1.65)
0.99 (0.58, 1.70)
0.79 (0.46, 1.37)
0.96 (0.56, 1.70)
0.86 (0.48, 1.56)
0.89 (0.65, 1.21)
0.87 (0.75, 1.01)
1.00 (0.90, 1.10)
1.14 (0.89, 1.46)
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
Ref.
0.79 (0.50, 1.25)
0.97 (0.61, 1.54)
2.18* (1.36, 3.52)
Ref.
0.73 (0.46, 1.29)
0.80 (0.50, 1.11)
2.21* (1.36, 3.62)
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
Ref.
1.57* (1.22, 2.04)
1.87* (1.42, 2.47)
2.18* (1.66, 2.85)
⫺
⫺
⫺
⫺
⫺
Ref.
1.67* (1.23, 2.16)
1.85* (1.40, 2.45)
2.11* (1.60, 2.78)
1.58* (1.24, 2.02)
1.60* (1.30, 1.97)
1.50* (1.07, 2.10)
1.92* (1.16, 3.19)
2.23* (1.61, 3.09)
␹2[14] ⫽ 63.31**
5,428,017.0
n/A
␹2[23] ⫽ 73.01**
5,406,111.1
␹2[9] ⫽ 21,923.9**
␹2[30] ⫽ 146.71**
5,270,740.7
␹2[7] ⫽ 135,384.4**
␹2[35] ⫽ 244.28**
5,090,585.4
␹2[5] ⫽ 180,165.3**
IMI ⫽ internet medical information; HH ⫽ household; OR ⫽ odds ratio; CI ⫽ confidence interval.
*Confidence intervals (95% CI) not containing the null value (1.00) are statistically significant at p ⬍ 0.05.
**p ⬍ 0.0001 for ␹2 tests.
§The Akaike information criterion (AIC), a measure of statistical model fit, was used to compare the amount of information explained across
the logistic regression models. A lower AIC value indicates a model is a better fit for the observed data.
Model 3 incorporated Internet and computer use measures
into the logistic regression model. The results showed that,
when controlling for other variables in the model (individual and household demographics), respondents who owned
a computer were more than twice as likely to report IMI use
as those who did not (OR ⫽ 2.18, 95% CI ⫽ 1.36 –3.52, p ⬍
0.05). Similarly, respondents who spent more than 4 hours
online per week were more likely to report IMI use (Table 5).
720
Model 3 showed no significant associations between IMI use
and either home broadband connection or place of Internet
access. The significant relationships evident in Model 1
proved consistent in Model 3. The addition of computer and
Internet use variables in Model 3 significantly improved the
explanatory power of the model, resulting in an AIC ⌬ of 2.5%.
Including health care use variables into Model 4 revealed
that respondents who used hospital or other medical facility
services, or who visited a specialist, were more likely to
report IMI use than those who did not (Table 5). Furthermore, in the fully controlled model, respondents who purchased medications were more likely to report IMI use:
those who had purchased any medications were nearly
twice as likely (OR ⫽ 1.92, 95% CI ⫽ 1.16 –3.19, p ⬍ 0.05) and
those who had purchased medication on the Internet were
more than twice as likely (OR ⫽ 2.23, 95% CI ⫽ 1.61–3.09,
p ⬍ 0.05). Though health care use measures did not change
substantially the significant relationships found in Model 1
or Model 3, the addition did significantly improve the
explanatory power of the model, resulting in an AIC ⌬ of
3.4%. Because a small number of respondents indicated
purchasing prescription medications online in the past 12
months and did not use IMI in the past 30 days, we also
created the regression models with these respondents removed from the analysis. The point estimates and statistical
decisions were unchanged, indicating that these respondents did not bias the results of the models reported.
Discussion
The Internet’s potential to bolster health promotion and
disease prevention efforts has attracted considerable attention, but existing research leaves two things unclear: the
prevalence of online medical information seeking and the
distinguishing characteristics of individuals who seek that
information. This study profiles adults who use IMI to shed
light on the Internet’s potential as a tool for health promotion and disease prevention and on some of the determinants associated with IMI use. In contrast to previous
research, this study found much lower prevalence of IMIseeking behavior. Additionally, logistic regression models
suggested previously unrecognized relationships between
IMI-seeking behavior and sociodemographic characteristics,
computer and Internet use, and health care system use.
Discrepant Prevalence Estimates
Previous research, both peer-reviewed and self-published,
estimates a strong prevalence of Internet health information
use, from 408 –10,23–25 to more than 80% of adult Internet
users.1,2 This study, consistent with Bundorf et al.14 suggests
that Internet Medical Information use measured in a shorter
time frame (i.e., during the past 30 days) is much less
prevalent—in our findings, about 13.2% of all respondents
(n ⫽ 6,119) and 21.1% of respondents with Internet access
(n ⫽ 3,289).
This discrepancy in prevalence rates raises important concerns. One is the time frame referent used in questions about
information seeking. In most past research, respondents
were asked to recall periods ranging from 12 months to
“ever.” Though a longer time frame does provide details
about prevalence, it may also tax a respondent’s ability and
motivation to provide accurate information and therefore
result in strong measurement errors. Specifically, when
Weaver et al., Internet Medical Information Users
required to provide information from memory, respondents
frequently feel pressure to answer immediately, even if
providing accurate information requires longer recall and
calculation efforts. Typically, respondents estimate quickly
on the basis of general experience (e.g., “I’m on the Internet
every day, so I must have used health information some
time over the last year.”). However, these estimates are
commonly distorted by telescoping, a phenomenon where
respondents include more events in a time frame than actually
occurred.43 This may suggest that past estimates of Internet
information seeking are exaggerated. In contrast, this study
used a much narrower time frame: Respondents were asked to
consider their activities in the past 30 days.
Other studies, however, qualify this time frame explanation.
One report that used a 30-day time frame found that
approximately 53% of adults had “looked online for health
information”1. This projection is, of course, larger than the
one found in both the current study and the work of others
using shorter time reference periods.20 Unfortunately, sample characteristics and data collection techniques do not
provide ready explanations for this discrepancy since each
study used representative samples and well-established
interview methodologies. Instead, an alternative explanation—that prevalence rate variability results from the diverse
way in which the “health information” construct has been
operationalized across this body of research—appears most
plausible. Most past studies have used a combination of the
terms “health⫺”, “wellness⫺”, and/or “health care⫺” information seeking and/or use to assess the focal behavior. Research shows, however, that Internet information seekers
interpret these terms broadly to include concepts such as
symptoms and treatment; nutrition, exercise, and weight
loss; vitamins, supplements, and alternative medicines; and
as health insurance and health care providers.9 Unlike prior
research, the current study used a very specific operationalization: Respondents were asked if they had used “medical
services/information” from the Internet. This explanation is
consistent with other studies21,30,34 which show that specificity helps explain discrepancies in prevalence. Studies that
use a “health” and/or “wellness” operationalization1,9 yield
much higher prevalence estimates than studies that use
“health care” or “medical” information.10,26
One conclusion to be drawn is that prevalence estimates in much
past research are less compelling than previously thought. This
seems particularly relevant as we deliberate what role the Internet
might play in public health efforts and as we explore the relatively
new “online health information use” phenomena.
Determinants of Internet Medical Information Use
Although IMI use reported in this study appears modest
compared with past research, it still shows that many adults
use the Internet each month to seek medical information.
Building a coherent portrait of this audience is critical: Knowing what motivates segments of the population to seek IMI,
including what types of IMI (e.g., wellness, medical), allows
health professionals to design programs that target specific
audiences with messages tailored to yield particular outcomes.
The results of this investigation shed considerable light on
the determinants of IMI use, including previously unrecognized relationships between IMI-seeking behavior and sociodemographic characteristics, computer and Internet use,
Journal of the American Medical Informatics Association
Volume 16
and health care system use. The strongest relationships
involved health care system use. Past studies suggest that
individuals with disabilities or chronic conditions and those
with lower perceived health status are more likely to seek
health information online.10,21,29,30,32,34 The results of this
study show that respondents who use health care services,
whether for themselves or for a household member, are
more likely to engage in IMI use. The measures used in this
study were not indicators of self-reported health conditions;
however, the relationships found suggest that health services use predicts IMI use more strongly than sociodemographic factors and computer and Internet use.
This result is important, particularly because past research
suggests that many adults access health information online to
self-diagnose, to seek information about alternative treatments
or medicine, or to engage in health care strategies inconsistent
with medical recommendations.44,45 While only limited evidence shows that IMI use results in harmful health outcomes,46
research does raise concerns about the accuracy and credibility
of Internet health information.5,46 Furthermore, trends indicate
that IMI seekers are increasingly using peer-generated Internet
sources, such as blogs and Wikis, that may contain content that
is not credible or accurate.9 Therefore, it becomes even more
critical that we gain a strong understanding of how to reach
populations whose health care decisions are susceptible to
misinformation obtained online.
The sociodemographic and computer and Internet use determinants associated with IMI use reveal prominent differences between this study and past research, especially in
variables such as age, race/ethnicity, household income,
and education. Despite research that links race/ethnicity
with both Internet access11 and Internet health informationseeking behaviors,47 many studies do not consider race/
ethnicity.10,25 So while other reports find that race/ethnicity
is not significantly associated with seeking health information online,29 the data in this study reveal the opposite,
specifically that African Americans are less likely than their
white counterparts to seek IMI. More importantly, this
proved consistent across all four models. Even when controlling for possible confounders, such as sociodemographics, computer and Internet use, and health care system use,
the results still show that African American respondents
were much less likely than white respondents to use IMI.
Furthermore, in contrast to previous studies,10,25 IMI use
was not significantly associated with sociodemographic factors, such as education and household income. Although
these findings may reflect differences in the sampling and
study methodologies, future research to explore these associations is needed.
Past research also suggests that access to a broadband
connection is an important factor underlying health information seeking.2,48 Indeed, some researchers have proposed
reconceptualizing the “digital divide” (which focuses on
differences between individuals with and without Internet
access), arguing that it is more accurate now to speak of a
schism in access speed (i.e., dial-up connection v. broadband
connection).48,49 However, after controlling for sociodemographic factors, the data in this study show that respondents
who own a home computer and spend more time online are
more likely to report IMI use. Home broadband access did
not emerge as significant. Assessing computer and Internet
Number 5
September / October 2009
721
use, a measure lacking in most past studies,10,25 is potentially very important for future studies.
Limitations
While the findings of this study are meant to stimulate discussion about the role of the Internet in health promotion and
disease prevention, there are some limitations. The sample
included adults living in a large, geographic area around
Atlanta, GA, and consequently the results may have limited
generalizability to the broader United States population. Additionally, although the data were produced using a rigorous
methodology, they are from cross-sectional, self-report assessments and, as a result, may reflect certain biases. Furthermore,
the sampling frame, while randomly generated, was limited to
respondents in households with landline telephone service.
The exclusion of adults living in cell phone-only households
could have produced a bias, especially among younger respondents.50,51 Finally, the measure that was used to capture IMI
use is not without some limitations. For example, this brief
measure does not capture specifically what types of medical
information respondents were seeking, or if they found the
information that was desired. Clearly, these considerations
must be acknowledged when interpreting the findings.
Conclusions
Undoubtedly, the Internet is a promising tool for public
health and health care10 –12 and a potentially effective platform for health communication and education.13,14 As our
results show, it may be premature, however, to embrace the
Internet unilaterally as an effective asset for efforts that
target broad segments of the public. Because the sample in
this study was limited to adults from one major United
States metropolitan area, additional research is needed to
explore the relationships examined here among larger samples that are more representative of the United States adult
population. While the IMI use measure applied in this study
was more specific than the general “health information”
measures in prior research, future research is needed to
establish reliable, valid assessments of different dimensions
of health information seeking behaviors (e.g., medical, wellness, and other types of information). Further research is
also needed to explore the factors that motivate specific
populations to seek information online. Such research must
examine the constructs used to assess behaviors, including
an individual’s perceptions of these constructs, also how
measurement may impact prevalence estimates. Future research should also examine in-depth any determinants that
may impact online information seeking. Finally, more research is needed to identify how best to counteract any
adverse consequences that could result from health and
medical information gained from the Internet.
References y
1. Harris Interactive, Harris. Poll shows number of “cyberchondriacs”—adults who have ever gone online for health information—increases to an estimated 160 million nationwide, 2007;76:
[The Harris Poll]. Available at: http://www.harrisinteractive.com/
harris_poll/printerfriend/index.asp?PID⫽792. Accessed Aug 4,
2008.
2. Pew Internet and American Life Project. Online Activities: Total,
2008. Available at: http://www.pewinternet.org/trends.asp.
Accessed: Oct 9, 2008.
722
3. Horrigan JB. Homes Broadband Adoption. 2008. Available at:
http://www.pewinternet.org/PPF/r/257/report_display.asp.
Accessed: Aug 25, 2008.
4. Rose B, Lenski J. Internet and multimedia 2006: on-demand
media explodes, 2006. Available at: http://www/arbitron.
com/downloads/im2006study.pdf. Accessed: Jun 11, 2008.
5. Anderson JG. Consumers of e-health: Patterns of use and
barriers. Soc Sci Comput Rev 2004;22:242– 8.
6. Fox S. Health information online, 2005. Available at: http://
www.pewinternet.org/PPF/r/156/report_display.asp. Accessed:
Jun 11, 2008.
7. Fox S, Fallows D. Internet health resources, 2003. Available
at:http://www.pewinternet.org/PPF/3/95/report_display.asp.
Accessed: Jun 11, 2008.
8. Tu HT, Cohen G. Striking jump in consumers seeking health
care information, 2008. Available at: http://www.hschange.
org/CONTENT/1006. Accessed: Aug 21, 2008.
9. Elkin N. How America searches: Health and wellness, 2008.
Available at: http://www.icrossing.com/research/ho-americasearches-health-and-wellness.php. Accessed: Aug 1, 2008.
10. Baker L, Wagner TH, Singer S, Bundorf MK. Use of the internet
and e-mail for health care information: Results from a national
survey. J Am Med Assoc 2003;289:2400 – 6.
11. Bernhardt JM. Health education and the digital divide: Building
Bridges and filling chasms. Health Educ Res 2000;15:527–31.
12. Lintonen TP, Konu AI, Seedhouse D. Information technology in
health promotion. Health Educ Res 2008;23:560 – 6.
13. Berger M, Wagner TH, Baker LC. Internet use and stigmatized
illness. Soc Sci Med 2005;61:1821–7.
14. Bundorf MK, Singer S, Wagner TH, Baker L. Consumer’s use of the
internet for health insurance. Am J Manag Care 2004;10:609–16.
15. Joseph CLM, Peterson E, Havstad S, et al. A web-based, tailored
asthma management program for urban African-American high
school students. Am J Respir Crit Care Med 2007;175:560 – 6.
16. Rasmussen LM, Phanareth K, Nolte H, Backer V. Internet-based
monitoring of asthma: A long-term, randomized clinical study of
300 asthmatic subjects. J Allerg Clin Immunol 2005;115:1137– 42.
17. Marcus BH, Lewis BA, Williams DM, et al. A comparison of
internet and print-based physical activity interventions. Arch
Intern Med 2007;167:944 –9.
18. van den Berg MH, Ronday HK, Peeters AJ, et al. Engagement
and satisfaction with an internet-based physical activity intervention in patients with rheumatoid arthritis. Rheumatology
2007;46:545–52.
19. Blumenthal D. Stimulating the adoption of health information
technology. N Engl J Med 2009;360(15):1477–9.
20. Bundorf MK, Wagner TH, Singer SJ, Baker LC. Who searches the
internet for health information? Health Serv Res 2006;41:819 –36.
21. Goldner M. Using the internet and email for health purposes:
The impact of health status. Soc Sci Q 2006;87:690 –710.
22. Goldner M. How health status impacts the types of information
consumers seek online. Inf Commun Soc 2006;9:693–713.
23. Hesse BW, Nelson DE, Kreps GL, et al. Trust and sources of health
information. The impact of the internet and its implications for
health care providers: Findings from the health information national trends survey. Arch Intern Med 2005;165:2618 –24.
24. Rains SA. Health at high speed: Broadband internet access,
health communication, and the digital divide. Commun Res
2008;35:283–97.
25. Rains SA. Perceptions of traditional information sources and
use of the World Wide Web to seek health information: Findings
from the health information national trends survey. J Health
Commun 2007;12:667– 80.
26. National Cancer Institute. Health Information National Trends
Survey 2005 (HINTS 2005): Main study interview instrument—
English (Post-analysis version), 2005. Available at: http://www.
hints.cancer.gov/HINTS_2005_Instrument-English.pdf. Accessed:
Sept 29, 2008.
Weaver et al., Internet Medical Information Users
27. Weaver JB III, Thompson NJ, Weaver SS, Hopkins GL. Health
care non-adherence decisions and internet health information.
Comput Hum Behav. in Press.
28. Hanauer D, Dibble E, Fortin J, Col NF. Internet use among
community college students: Implications in designing healthcare interventions. J Am Coll Health 2004;52:197–202.
29. Ayers SL, Kronenfield JJ. Chronic illness and health-seeking
information on the internet. Health Interdiscip J Soc Study
Health Ill Med 2007;11:327– 47.
30. Lambert SD, Loiselle CG. Health information seeking behavior.
Qual Health Res 2007;17:1006 –19.
31. Pew Internet and American Life Project. Online Health Search,
2006. Available at: http://www.pewinternet.org/PPF/r/190/
report_display.asp. Accessed: Oct 21, 2008.
32. Rice RE. Influences, usage, and outcomes of internet health
information searching: Multivariate results from the pew surveys. Int J Med Inform 2006;75:8 –28.
33. Drentea P, Goldner M, Cotton S, Hale T. The association among
gender, computer use, and online health searching, and mental
health. Inf Commun Soc 2008;11:509 –25.
34. Pandey SK, Hart JJ, Tiwary S. Women’s health and the internet:
Understanding emerging trends and implications. Soc Sci Med
2003;56:179 –91.
35. Scarborough Research. Atlanta GA Market Report, Release 2, 2006
data, Atlanta DMA. New York: Scarborough Research; 2006.
36. Scarborough Research. Atlanta GA Market Report, Release 2, 2007
data, Atlanta DMA. New York: Scarborough Research; 2007.
37. U.S. Census Bureau. Current lists of metropolitan and micropolitan statistical areas and definitions, 2008. Available at:
http://www.census.gov/popuation/www/metroareas/metrodef.
html. Accessed: Oct 6, 2008.
38. Claritas. Target marketing and marketing research—Claritas, 2008.
Available at: http://www.claritas.com. Accessed: Oct 6, 2008.
39. Scarborough Research. Methodology. 2008. Available at: http://
www.scarborough.com/methodology.pdf. Accessed: Sept 22, 2008.
40. SAS Institute. Survey design specification, 2009. Available at:
http://support.sas.com/documentation/cdl/en/statug/59654/
HTML/default/statug_introsamp_sect008.htm. Accessed: Apr 27,
2009.
41. Tabachnick BG, Fidell LS. Using Multivariate Statistics, 5th edn,
Boston, MA: Allyn & Bacon, 2007.
42. Hosmer DW, Lemeshow S. Applied Logistic Regression, Hoboken, NJ: Wiley-Interscience; 2000.
43. DiIorio CK. Measurement in Health Behavior: Methods for
Research and Education, San Francisco, CA: Jossey-Bass, 2005.
44. Watson R, Springen K, Joseph N, et al. The new patient power.
Newsweek; 001(13)(26):54 – 60.
45. Weaver JB III, Thompson N, Weaver SS, Hopkins GL. Profiling
characteristics of individuals using internet health information
in health care non-adherence. Annual Meeting of the American
Public Health Association, San Diego, CA, p 136th, 2008.
46. Crocco AG, Villasis-Keever M, Jadad AR. Analysis of cases of
harm associated with use of the internet. J Am Med Assoc
2002;287:2869 –71.
47. Hsu J, Huang J, Kinsman J, et al. Use of e-health services
between 1999 and 2002: A growing digital divide. J Am Med
Inform Assoc 2005;12:164 –71.
48. Chang BL, Bakken S, Brown SS, et al. Bridging the digital divide:
Reaching vulnerable populations. J Am Med Inform Assoc
2004;11:448 –57.
49. Whaley KC. America’s digital divide: 2000 –2003. J Med Syst
2004;28:183–95.
50. American Association for Public Opinion Research. Do cell phones
affect survey research? 2008. Available at: http://www.aapor.org/
docellphonesaffectsurveyresearch. Accessed: Oct 10, 2008.
51. Blumberg SJ, Luke JV, Cynamon ML. Telephone coverage and
health survey estimates: Evaluating the need for concern about
wireless substitution. Am J Pub Health 2006;96:926 –31.