Cancer Survivorship: Embracing the FutureSupplement to Cancer

Information Support for Cancer Survivors

Bradford W. Hesse, PhD1

Neeraj K. Arora, PhD1

Ellen Burke Beckjord, PhD2

Lila J. Finney Rutten, PhD1

1 Division of Cancer Control and PopulationSciences, National Cancer Institute, Bethesda,Maryland.

2 RAND Corporation, Pittsburgh, Pennsylvania.

Survivors’ information-seeking behavior has traditionally been documented

through analysis of inquiries to hotlines and cancer information services. Data

from these self-selected inquiries tend to be restricted to a time around diagno-

sis, and to those populations possessing the wherewithal and motivation to

seek information actively. The current study used data from a general popula-

tion survey to assess 1) the prevalence of information-seeking behavior among

survivors in the general population, 2) characteristics of seekers versus non-

seekers, 3) ratings of information-seeking experience, and 4) actual versus

preferred sources of information. Data were analyzed from the 2005 adminis-

tration of the Health Information National Trends Survey (HINTS). HINTS is a

cross-sectional, random digit dial telephone survey, weighted to provide esti-

mates for the general population. Nearly half of all Americans (48.7%) indicated

that they have looked for cancer information from 1 source or another. Percen-

tages were highest for those who have been touched by cancer (63.1% of cancer

survivors and 54.6% of those with family histories) and lowest for those with

no cancer history (27.6% of those with no history). Quality concerns topped the

list of information-seeking experiences for those recalling the last time they

looked. Patterns of information-seeking revealed a discrepancy between pre-

ferred and actual source when tracked over years since diagnosis. Information-

seeking is prevalent among cancer survivors and does not diminish over time.

Prescriptions are given for reengineering the information environment to

improve long-term outcomes for survivors. Cancer 2008;112(11 suppl):2529–

40. � 2008 American Cancer Society.

KEYWORDS: cancer survivors, consumer health information, national survey, in-formation seeking.

I n a feature article appearing in the Health section of the New

York Times on August 14, 2005, journalist Jan Hoffman summar-

ized the modern cancer patient’s experience in one, seemingly para-

doxical, headline: ‘‘Awash in information, patients face a lonely,

uncertain road.’’1 Describing the experience of Meg Gaines, diag-

nosed with ovarian cancer and pictured holding mounds of com-

puter printouts and clipped articles, the headline captured an

inherent contradiction for cancer patients and survivors in the midst

of the information age. Never before has so much information about

cancer been made so immediately available to patients and practi-

tioners alike, yet never has it appeared so difficult to integrate the

raw data of evidence-based oncology into a coherent lifetime guide

for healthy survivorship. As suggested in a subsequent article in the

same newspaper by journalist Denise Grady,2 a diagnosis of cancer

carries with it a lifetime of ‘‘making sense’’3 in an environment of

uneven care4 and fragmented communications.5

Address for reprints: Bradford W. Hesse, PhD, Di-vision of Cancer Control and Population Sciences,National Cancer Institute, 6130 Executive Blvd.,MSC 7365, EPN 4068, Bethesda, MD 20892-7365; Fax: (301) 480-2198; E-mail: hesseb@mail.nih.gov

Supplement sponsored by the American CancerSociety’s Behavioral Research Center and theNational Cancer Institute’s Office of CancerSurvivorship.

Supported by National Cancer Institute Contract#N02-PC-15003.

Received December 5, 2007; revision receivedJanuary 2, 2008; accepted February 20, 2008.

ª 2008 American Cancer SocietyDOI 10.1002/cncr.23445Published online 21 April 2008 in Wiley InterScience (www.interscience.wiley.com).

2529

WHY INFORMATION NEEDS ARECRITICAL INDICATORSIn 2000 the Institute of Medicine (IOM) released its

landmark report on patient safety titled To Err is

Human.6 The gist of the report was straightforward

and direct. Systemic fractures in the organizational

practice of medicine have created an epidemic of

avoidable medical errors and lost opportunities for

treatment that, when factored together, result in an

untold number of deaths each year. The fractures are

not unique to medicine, they occur in every human-

based system in which variation occurs as a natural

consequence of normal cognitive and physical limi-

tation.7 What is different is that the culture of health-

care has not yet evolved the safety mechanisms—

redundancies in communications, technologic back-

ups, etc—that have been instituted in other high-risk

industries.

One of the key fracture points, according to

authors of the IOM report, is in the area of commu-

nication. At their best, effective communications can

ensure that information in the sociotechnical envir-

onment8 of modern healthcare can be used to opti-

mize positive outcomes for the patient.9,10 Patient-

centered communications can serve to 1) foster heal-

ing relationships between providers and patients

over time, 2) bring patients into the informational

loop for their own care, 3) address emotional con-

cerns, 4) support decision-making, 5) manage uncer-

tainty, and 6) enable patient self-management.11,12

When communications fail, when noise enters the

system or critical conversations fail to happen, the

result can be a catastrophic error or a missed oppor-

tunity for intervention.6,7

The influence of these fractures in cancer survi-

vorship was made apparent by a study of 119 breast

cancer patients who failed to receive guideline-

recommended adjuvant therapy.4 In deconstructing

events leading to the systemic exclusion of treat-

ments, authors of the study concluded that a seri-

ous ‘‘communication breakdown’’ occurred in the

delivery of care; that ‘‘the loop from recommenda-

tion to treatment’’ was not closed. According to the

hospital staff interviewed, far too few mechanisms

were built into the system of care that would allow

surgical staff to follow-up with their patients, to

ensure that the patients attended necessary refer-

rals, or to monitor adherence to recommended

treatment.4 Similar reviews of patients presenting

with late-stage cervical cancer (a cancer that is

wholly preventable in the population through early

detection) revealed that a lack of adherence to rou-

tine screening examinations accounted for a major-

ity of late-stage cases.13

ASSESSING CANCER SURVIVORS’INFORMATION-SEEKING EXPERIENCESAssessing patients’ needs can, and should be, a

starting point for improving the delivery of infor-

mation services to cancer survivors. One way of

assessing information needs is to evaluate the data

collected systematically by 1 of the publicly or pri-

vately funded cancer information services. Indeed,

much has already been learned from these centers

regarding the nature of callers’ questions, the

composition of information seekers, and the psy-

chologic motivations associated with looking for

answers to self-generated questions.14–17 Going

online through Web-based messaging and email

has opened up other avenues for collecting data

from public inquiries.18–20

Nevertheless, data from the call centers do not

represent the full breadth of potential callers’ needs.

The self-initiated transactions logged at these centers

come from those members within the public space

who 1) learned about the existence of the call line

or Web site, 2) were sufficiently motivated to access

the resource, and 3) had the wherewithal to make

the call or access the Web.15,21,22 Researchers have

argued that these self-selected inquiries, although

useful, are not generalizable to the full population of

cancer information seekers but tend to be limited to

callers of higher socioeconomic status. Calls into the

hotlines also tend to cluster temporally around diag-

nosis and treatment decisions, and do not represent

the full continuum of cancer survivorship from diag-

nosis to a lifetime of vigilance and coping. One

review found that only 10% of catalogued studies (9

of 92) assessed the current needs of posttreatment

survivors.23

The purpose of this study, then, was to gain

insight into the information-seeking experiences of

cancer survivors as distributed through the general

population. To do this, data were analyzed from a

national probability sample of adults in the U.S. with

sufficient breadth to cover the information needs of

people reporting themselves as cancer survivors,

those with family members touched by cancer, and

those with no personal or family experience of can-

cer. With such a sample frame in place, the following

research questions were identified as important:

� Research Question 1. How prevalent is cancer in-

formation-seeking in the general population and

how do cancer survivors compare with other

groups in terms of motivated usage of cancer in-

formation resources?

� Research Question 2. Who among cancer survi-

vors, as sampled from the general population,

2530 CANCER Supplement June 1, 2008 / Volume 112 / Number 11

reported taking an active part in information-seek-

ing activities?

� Research Question 3. Where do information-seek-

ing survivors turn when looking for quality cancer

information? Is there a difference between where

survivors would prefer to go in contrast to where

they actually sought information?

� Research Question 4. How do information-seeking

survivors rate their satisfaction with the existing

cancer information environment? Are there points

of concern about the information environment

that stand out against others?

MATERIALS AND METHODSData SourceData for this study are from the 2005 iteration of the

Health Information National Trends Survey (HINTS

2005), a nationally representative survey designed to

assess the impact of the cancer information environ-

ment and the public’s knowledge of, attitudes to-

ward, and behaviors related to cancer and cancer

prevention. Comprehensive reports on the concep-

tual framework of HINTS and sample designs are

published elsewhere.24,25

Data for HINTS 2005 were collected from Febru-

ary through August 2005. The cross-sectional survey

was administered by trained interviewers to repre-

sentative samples of American households drawn

from all telephone exchanges in the U.S. One adult

(aged �18 years) was selected from each household

to participate in the full survey during a household

screening. In 2005, response rates were 34% at the

household screening level (ie, the initial contact with

the household used for sampling purposes), and 61%

at the sampled person interview level (ie, completion

of the interview by the sampled household member).

Every sampled adult who completed a questionnaire

was assigned a final sampling weight and a set of 50

replicate sampling weights. These sampling weights

were used for the purpose of computing nationally

representative estimates, to adjust for nonresponse,

and to reduce the sampling variance of estimators

through utilization of information from more robust

sources than the corresponding HINTS estimates (eg,

estimates obtained through the Current Population

Survey, which has much larger sample sizes than

HINTS). Data were available from 5586 extended

interviews for use in analyses.

MeasuresInformation-seekingHINTS 2005 respondents were asked ‘‘Have you ever

looked for cancer information from any source?’’

(yes/no). Respondents also indicated the timing of

their most recent information search (within the last

year/ > 1 year ago).

Information source preferencesInformation source preference was assessed by ask-

ing respondents ‘‘Imagine that you had a strong

need to get information about cancer. Where would

you go first?’’ Responses were combined into 3

groups: Health Care Provider (HCP), Internet, and

Other (includes print media, family/friends/cowork-

ers, cancer organization/cancer survivor, and 1–800

telephone numbers).

Information source useFor HINTS respondents who had personally searched

for cancer information, source use was assessed by

asking ‘‘The most recent time you wanted informa-

tion on cancer, where did you go first?’’ Responses

were coded into the same 3 groups used to assess

source preference: Health Care Provider (HCP), Inter-

net, and Other.

Information-seeking experiencesExperiences seeking cancer information were

assessed using the Information SEeking Experiences

Scale (ISEE).26,27 HINTS respondents who had per-

sonally searched for cancer information indicated

their agreement on a 4-point scale to statements

about their most recent search: ‘‘It took a lot of effort

to get the information that you needed’’; ‘‘You felt

frustrated during your search for the information’’;

‘‘The information you found was too hard to under-

stand’’; and ‘‘You were concerned about the quality

of information.’’ These items were examined indivi-

dually and as a continuous scale, in which higher

scores indicate better information search experiences

(a 5 .76).

Sociodemographic variablesSex, age, education, annual household income, race/

ethnicity, and whether a respondent had health in-

surance were included in the current study. Three

discrete categories of cancer history were also

included: 1) no personal or family history of cancer,

2) no personal but a family history of cancer, and 3)

personal history of cancer. For analyses of cancer

survivors, time since diagnosis (�1 year postdiagno-

sis, 2–5 years postdiagnosis, 6–10 years postdiagno-

sis, and 111 years postdiagnosis) was included as

well.

Information Support/Hesse et al. 2531

Formative Testing of ItemsAll measures in the survey were evaluated in a cogni-

tive laboratory using concurrent protocol analysis

techniques.28,29 Items were reviewed up to 3 times,

depending on the degree of modification needed to

stabilize interpretation and use. The full survey was

also pilot-tested before the main field period, and a

translated version of the instrument was created in

Spanish for non-English-speaking, Hispanic respon-

dents.30

Data AnalysisAnalyses were conducted using SAS-callable

SUDAAN software (version 9.0)31 to account for the

complex survey design of HINTS 2005 and to obtain

appropriate standard errors and 95% confidence

intervals (95% CIs) for point estimates. Responses of

‘‘refused’’ or ‘‘don’t know’’ were counted as missing.

Weighted descriptive statistics are presented, and

weighted data were used in all inferential statistical

analyses. Bivariate analyses (chi-square) estimated

associations between sociodemographic characteris-

tics and cancer history as well as information out-

comes and cancer history. For cancer survivors,

binomial logistic regression estimated associations

between sociodemographic characteristics and infor-

mation-seeking; multinomial logistic regression esti-

mated associations between sociodemographic

characteristics and information source preferences.

For survivors who had ever searched for cancer infor-

mation, multinomial logistic regression estimated

associations between sociodemographic characteris-

tics and source use; multivariate linear regression

estimated associations between survivor characteris-

tics and information-seeking experiences.

Informed ConsentData were collected as part of HINTS, a national sur-

vey of the general population. HINTS went through

an expedited review with the National Cancer Insti-

tute’s Institutional Review Board in 2001 that gave it

an ‘‘exempt’’ status; hence, a formal informed con-

sent was not needed. Clearance for conducting the

survey was also obtained from the U.S. government’s

Office of Management and Budget (OMB #0925-

0538).

RESULTSOf the 5586 observations available for analyses, 5553

could be identified in terms of survivorship status. Of

these, 865 reported having a personal history of can-

cer, 3397 reported having a family history of cancer

only, and 1291 reported having no personal or family

history with cancer. Table 1 depicts the weighted

population characteristics along with the accompa-

nying chi-square statistics within each of these

groupings. The majority of the demographic variables

appeared to be distributed in a way that is reflective

of the population at large for adults aged �18 years

(eg, median income at around the < $50,000 mark;

available at www.census.gov. Accessed on September

4, 2007). However, differences across groups in de-

mographic characteristics were observed (Table 1).

Compared with other groups, respondents with a

personal history of cancer were more frequently

female and non-Hispanic white. Survey respondents

with a history of cancer were also somewhat older

and more frequently reported having health insur-

ance than other groups. Among respondents with a

personal history of cancer, the most frequently iden-

tified types of cancer included skin cancer (29.5%),

breast cancer (16.9%), and prostate cancer (10.1%).

The distribution of reported cancers among the

HINTS sample is consistent with cancer incidence

estimates derived during a similar time period, sug-

gesting that the HINTS distribution of survivors is

fairly representative of the national population.32

Research Question 1. Information-seeking Prevalenceand Survivorship StatusIn Figure 1 we investigate the prevalence of cancer

information-seeking, defined as having ever searched

for cancer information, as estimated for various seg-

ments of the population. For comparison purposes,

we begin with an overall estimate for the entire sam-

ple frame; that is, for all noninstitutionalized Ameri-

cans aged �18 years. That estimate suggests that a

little less than half (46.7%) of the adult population in

the U.S. has actively looked for cancer information

from any source; among these respondents, 71.2%

reported that they had sought said information

within the last year. Next to the overall estimate, we

overlay estimates for subpopulations based on

responses to questions regarding survivorship. The

first comparison separates the estimates by survivor-

ship status, with only 27.6% of those who have never

been ‘‘touched by cancer’’33 indicating that they had

engaged in active searching. The percentage rises

dramatically for the 2 groups most closely influenced

by cancer, with up to 54.6% of those with family his-

tories and up to 63.1% of cancer survivors looking

for information. The trend shows an overall increase

in engagement with the information environment

among people who have been touched by cancer

2532 CANCER Supplement June 1, 2008 / Volume 112 / Number 11

either through a family or personal history

(v2 5 185.67; P < .01).

In Figure 2, we switch our frame of analysis to

just individuals who have searched for cancer infor-

mation, and from among those we indicate percen-

tages of searchers who have looked for cancer

information during the past year. Of note is that in-

formation seekers, regardless of cancer status, appear

to be highly engaged with the cancer information

environment: Overall, 71.2% of information seekers

had conducted their most recent search within the

past year. Not surprisingly, nearly all cancer survivors

diagnosed within the past year had searched for can-

cer information during that same time period

(97.2%). Although smaller percentages of survivors

FIGURE 1. Information-seeking (ever searched for cancer information).HINTS indicates Health Information National Trends Survey.

TABLE 1Respondent Characteristics Grouped by Survivor Status (n 5 5553)

Respondent characteristic

Weighted % within survivor status group (95% CI)

Personal history

(n 5 865)

Family history only

(n 5 3397)

No history

(n 5 1291)

Gender (chi-square 5 10.53; P < .01)

Male 44.2 (42.0–46.4) 46.8 (46.0–47.6) 52.8 (51.2–53.4)

Female 55.8 (53.6–58.0) 53.2 (52.4–54.0) 47.2 (45.6–48.8)

Age (chi-square 5 439.69; P < .01)

18–34 y 4.8 (3.6–6.0) 31.8 (31.1–32.5) 39.6 (38.0–41.2)

35–49 y 17.7 (15.5–19.9) 33.7 (33.1–35.3) 27.6 (26.4–28.8)

50–64 y 32.4 (30.4–34.4) 22.2 (21.6–22.8) 20.4 (19.0–21.8)

65–74 y 22.6 (21.0–24.2) 7.6 (7.2–8.0) 7.7 (6.9–8.5)

�75 y 22.4 (20.6–24.2) 4.8 (4.5–5.1) 4.7 (4.0–5.4)

Education (chi-square 5 22.51; P < .01)

Less than high school 14.1 (12.7–15.5) 11.8 (11.1–12.5) 21.1 (19.6–22.6)

High school 31.5 (29.6–33.4) 30.1 (29.3–30.9) 28.5 (26.8–30.2)

Some college 30.7 (28.5–32.9) 34.0 (33.1–34.9) 27.9 (25.9–29.9)

College graduate 23.7 (22.0–25.4) 23.9 (23.3–24.5) 22.5 (21.3–23.7)

Annual income (chi-square 5 29.30; P < .01)

< $25K 25.9 (24.0–27.8) 19.3 (18.2–20.4) 26.6 (24.9–28.3)

$25K to < $50K 22.4 (20.5–24.3) 21.4 (20.2–22.6) 20.3 (18.9–21.7)

$50K to < $75K 16.2 (14.7–17.7) 19.4 (18.3–20.5) 17.2 (15.9–18.5)

�$75K 18.8 (17.2–20.4) 26.3 (25.0–27.6) 19.4 (18.0–20.8)

Refused/DK/missing 16.7 (15.0–18.4) 13.7 (12.7–14.7) 16.6 (14.9–18.3)

Race/ethnicity (chi-square 5 172.61; P < .01)

Hispanic/Latino 5.1 (4.1–6.1) 9.2 (8.5–9.9) 26.1 (24.5–27.7)

NH white 84.7 (83.1–86.3) 74.0 (65.0–83.0) 54.3 (52.8–55.8)

NH African American 5.1 (4.2–6.0) 10.6 (9.8–11.4) 10.7 (9.6–11.8)

NH other 5.2 (4.1–6.3) 6.7 (6.0–7.4) 8.9 (7.6–10.2)

Health insurance (chi-square 5 74.37; P < .01)

Have health insurance 93.8 (92.5–95.1) 86.5 (85.6–87.4) 71.5 (69.2–73.8)

Do not have health insurance 6.2 (4.9–7.5) 13.5 (12.6–14.4) 28.5 (26.2–30.8)

Time since diagnosis

�1 y 15.8 (12.3–19.4)

2–5 y 31.3 (26.0–36.7)

6–10 y 16.1 (12.9–19.4)

�11 y 36.7 (32.1–41.3)

95% CI indicates 95% confidence interval; DK, did not know; NH, non-Hispanic.

Information Support/Hesse et al. 2533

had recently searched further from time of diagnosis,

the percentage of survivors who had sought cancer

information in the past year was relatively constant

across the remaining groups and remained high at

approximately half the sample. Finally, compared

with survivors overall, information seekers with a

family history or no history of cancer were more

likely to have searched for cancer information during

the past year (v2 5 17.13; P < .01).

Table 2 displays differences in our other informa-

tion-related study outcomes in relation to survivor-

ship status. Although the majority of all groups

reported preferring an HCP as a first source of cancer

information, this preference was stronger among

HINTS respondents with a personal or no history of

cancer (v2 5 73.70; P < .01). Conversely, although a

majority would prefer to use their HCP to get infor-

mation concerning cancer, only a minority of seekers

actually did: significantly more respondents with a

personal history of cancer had used an HCP as a first

source of cancer information (41.9%) compared with

those with a family (21.6%) or no history (16.2%) of

cancer (v2 5 47.0; P < .01). Finally, information

search experiences varied by survivorship status.

Although suboptimal information search experiences

were fairly prevalent among all HINTS respondents,

cancer survivors were least likely to report that their

most recent search took a lot of effort (v2 5 7.20;

P 5 .03), and respondents with a family history of

cancer were most likely to have been frustrated by

their most recent search (v2 5 7.19; P 5 .03).

Research Question 2: Characteristics ofInformation-seeking SurvivorsWe next sought to understand what differentiated in-

formation seekers from nonseekers within the survi-

vorship population. Table 3 shows the results of a

multivariate regression model estimating the odds of

information-seeking. Information-seeking survivors

tended to be younger and to be better educated.

Compared with survivors aged �75 years, survivors

ages 35 to 49 years were more than twice as likely to

have ever searched for cancer information (odds ra-

tio [OR] of2.71), and survivors ages 50 to 64 years

were nearly 4 times more likely to have searched (OR

of 3.89). The probability of information-seeking

increased steadily with years of education; compared

with survivors who did not complete high school,

those who did had >3 times the odds of having

searched for cancer information (OR of 3.76), and

this ratio nearly doubled for survivors with some (OR

of 6.11) or completed (OR of 6.66) college education.

Research Question 3. Survivors’ Actual and PreferredSources of InformationIn HINTS 2005, respondents were asked to imagine a

hypothetical situation in which they had a strong

need to look for cancer information. Given that need,

respondents were asked where they would likely go

FIGURE 2. Information-seeking (searched within the past year). Hx indi-cates history; HINTS, Health Information National Trends Survey.

TABLE 2Information Outcomes by Survivor Status

Information outcome

Weighted % within survivor status group (SE)

Personal history Family history No history

Ever searched for cancer information? (chi-square 5 185.67; P < .01)

Yes 63.11 (2.23) 54.64 (1.14) 27.56 (1.74)

No 36.89 (2.23) 45.36 (1.14) 72.44 (1.74)

Most recent search for cancer information (chi-square 5 17.13; P < .01)

Within the last y 60.47 (2.78) 72.75 (1.18) 74.97 (3.78)

>1 y ago 39.53 (2.78) 27.25 (1.18) 25.03 (3.78)

Information source preference (chi-square 5 73.70; P < .01)

Health care provider 68.49 (2.13) 50.53 (1.06) 61.25 (2.03)

Internet 19.99 (1.81) 34.14 (1.10) 26.36 (1.52)

Other 11.52 (1.61) 15.32 (0.90) 12.39 (1.29)

Source used first in most recent search (chi-square 5 47.01; P < .01)

Health care provider 41.88 (2.80) 21.59 (1.32) 16.21 (2.37)

Internet 36.04 (2.67) 50.07 (1.37) 48.87 (3.78)

Other 22.08 (2.09) 28.34 (1.22) 34.92 (3.69)

It took a lot of effort to get the information you needed (chi-square 5 7.20; P 5 .03)

Strongly agree/agree 30.47 (2.50) 38.69 (2.00) 37.56 (3.21)

Strongly disagree/disagree 69.53 (2.50) 61.31 (2.00) 62.44 (3.21)

You felt frustrated during your information search (chi-square 5 7.19; P 5 .03)

Strongly agree/agree 23.15 (2.48) 28.29 (1.46) 22.27 (2.18)

Strongly disagree/disagree 76.85 (2.48) 71.71 (1.46) 77.73 (2.18)

The information you found was too hard to understand (chi-square 5 0.39; P 5 .82)

Strongly agree/agree 24.32 (2.51) 23.67 (1.03) 22.11 (2.72)

Strongly disagree/disagree 75.68 (2.51) 76.33 (1.03) 77.89 (2.72)

You were concerned about the quality of the information (chi-square 5 2.18; P 5 .34)

Strongly agree/agree 44.48 (2.96) 48.73 (1.55) 44.84 (3.20)

Strongly disagree/disagree 55.52 (2.96) 51.27 (1.55) 55.16 (3.20)

SE indicates standard error.

2534 CANCER Supplement June 1, 2008 / Volume 112 / Number 11

first for information. Responses for the full survey

frame were placed into 1 of 12 categories according to

prompts on the Computer Assisted Telephone Inter-

view screen. Categories included books (1.9%), bro-

chures (0.2%), cancer organization (3%), family (4.1%),

friend/coworker (1%), healthcare provider (54.3%),

Internet (29.9%), library (2.4%), magazines (0.3%),

newspapers (0.3%), someone with cancer (0.2%), tele-

phone information services (0.1%), other specified

(0.5%), and a residual category of not ascertained or do

not know (1.6%). Overwhelmingly, responses clustered

into 2 main categories, with HCP leading as the most

common preferred source of cancer information (an

estimated 54.3% of the entire U.S. adult population)

followed by the Internet (an estimated 29.9%). None of

the other single categories came close to matching fre-

quencies in these 2 individual categories.

Those in the sample who had reported looking

for cancer information were asked to indicate where

they had gone first. In this case, the pattern

switched for the main categories with the majority

indicating that they had gone to the Internet first

(46.7%), and a smaller, but still substantive, percent-

age reporting that they had gone to their HCP first

(23%). This overall finding of preferring to go to

HCP first, but actually using the Internet first repli-

cates a pattern uncovered previously in the HINTS

2003 data.34

TABLE 3Multivariate Regression Models of Information Outcomes for Survivors

Survivor characteristic

Information seeking*

(OR; 95% CI)

Information source

preferencey (OR; 95% CI)

Information source

use{ (OR; 95% CI) ISEE§ (ß; SE ß)

Gender P 5 .21 P 5 .76 P < .01 P 5 .27

Male 1.00 1.00 1.00 0.00

Female 1.32 (0.85–2.07) 1.07 (0.63–1.83) 1.27 (0.69–2.33) 3.18 (2.89)

Age P < .01 P < .01 P < .01 P 5 .39

18–34 y 2.00 (0.55–7.26) 13.31 (1.97–89.70) 12.62 (1.02–156.43) 7.35 (12.24)

35–49 y 2.71 (1.34–5.45) 26.42 (6.67–104.60) 10.63 (2.80–40.40) 11.58 (4.85)

50–64 y 3.89 (1.95–7.75) 25.87 (7.34–91.14) 6.71 (1.76–25.62) 5.91 (4.95)

65–74 y 1.33 (0.75–2.36) 11.07 (2.83–43.25) 6.78 (1.87–24.58) 5.09 (4.47)

�75 y 1.00 1.00 1.00 0.00

Education P < .01 P 5 .08 P 5 .01 P 5 .70

Less than high school 1.00 1.00 1.00 0.00

High school 3.76 (1.82–7.77) 0.87 (0.23–3.24) 1.00 1.81 (6.09)

Some college 6.11 (2.81–13.28) 1.36 (0.42–4.41) 1.56 (0.64–3.77) 2.18 (5.25)

College graduate 6.66 (2.99–14.87) 2.21 (0.61–8.03) 2.93 (1.23–6.97) 21.97 (6.60)

Annual income P 5 .32 P 5 .52 P 5 .95 P 5 .09

<$25K 1.00 1.00 1.00 0.00

$25K to <$50K 1.30 (0.67–2.52) 0.98 (0.43–2.24) 1.00 (0.31–3.18) 8.65 (4.72)

$50K to <$75K 2.10 (1.01–4.35) 1.35 (0.54–3.38) 1.19 (0.30–4.76) 9.59 (5.40)

�$75K 1.35 (0.62–2.91) 1.67 (0.67–4.19) 1.34 (0.32–5.65) 12.27 (5.84)

Refused/DK/missing 1.00 (0.54–1.84) 1.70 (0.64–4.51) 1.06 (0.36–3.10) 1.14 (4.83)

Race/ethnicity P 5 .89 P 5 .22 P 5 .42 P 5 .22

Hispanic/Latino 1.28 (0.39–4.15) 1.36 (0.27–6.86) 0.74 (0.13–4.33) 25.77 (6.02)

NH white 1.00 1.00 1.00 0.00

NH African American 0.64 (0.18–2.31) 0.31 (0.07–1.41) 1.99 (0.63–6.27) 22.58 (6.35)

NH other 1.03 (0.32–3.24) 0.36 (0.09–1.47) 1.15 (0.28–4.72) 214.15 (8.91)

Health insurance P 5 .70 P 5 .61 P 5 .99 P 5 .69

Have health insurance 1.00 1.00 1.00 0.00

Do not have health insurance 0.83 (0.32–2.18) 0.47 (0.08–2.87) 1.04 (0.23–4.77) 3.00 (7.53)

Time since diagnosis P 5 .78 P 5 .60 P 5 .03 P 5 .49

�1 y 1.00 1.00 1.00 0.00

2–5 y 1.02 (0.48–2.18) 1.65 (0.74–3.69) 2.56 (0.81–8.14) 0.78 (4.31)

6–10 y 0.78 (0.41–1.46) 1.60 (0.61–4.21) 2.05 (0.71–5.91) 22.47 (3.93)

�11 y 0.91 (0.46–1.79) 1.35 (0.53–3.47) 0.93 (0.34–2.53) 24.18 (4.17)

OR indicates odds ratio; 95% CI, 95% confidence interval; ISEE, Information SEeking Experiences Scale; SE, standard error; DK, does not know; NH, non-Hispanic.

* Binary logistic model; OR 5 search/no search.y Multinomial logistic model; OR 5 prefer Internet/prefer healthcare provider.{ Multinomial logistic model for information seekers only; OR 5 used Internet/used healthcare provider.§ Linear model for information seekers only; higher scores indicate better search experiences.

Information Support/Hesse et al. 2535

To understand source use and source preference

in more detail, we used multinomial logistic regres-

sion to compare the odds of having used or prefer-

ring the Internet to the odds of having used or

preferring an HCP for information-seeking. The

results are displayed in Table 3. Of the sociodemo-

graphic characteristics examined in this study, only

survivors’ age was found to be significantly asso-

ciated with preferring the Internet over an HCP as a

first source of cancer information. All survivors

aged < 75 years were more likely than those aged

�75 years to prefer the Internet to an HCP as an in-

formation source (P < .01). Similarly, among survi-

vors who had searched for information, those

aged < 75 years were more likely than those aged

�75 years to have used the Internet before an HCP

in their most recent search (P < .01). Furthermore,

survivors who had completed college were more

likely than those who had completed high school or

less to have used the Internet before an HCP in their

most recent search (P < .05), and post hoc compari-

sons revealed that survivors 2 to 5 years from diag-

nosis were more likely to have used the Internet

before an HCP compared with those �11 years after

diagnosis (P < .05).

The pattern of source preference versus source

usage is more nuanced when evaluating survivors’

data than it is for the general population. Figure 3

compares preference against usage across years since

diagnosis, using estimated percentages adjusted for

the influence of the demographic variables (age, sex,

education, race/ethnicity, income, and health insur-

ance). As with the overall population, survivors pre-

ferred contacting their HCPs regardless of how long

it had been since they were diagnosed. The usage

data show a different pattern. During the first year

after diagnosis, cancer patients actually reported

going to their HCPs as a source of first resort. That

pattern changes, however, when examining years 2

through 10. Here Internet usage exceeds provider

usage as a first source. The trend reverses again for

those living with their cancers for > 11 years (a delay

effect that might be confounded with generational

effects on Internet use34).

Research Question 4. Survivors’ Information-seekingExperiencesSurvivors who reported looking specifically for cancer

information in the past were asked to recall their last

searching experience. These respondents were then

asked to provide ratings on the ISEE scale. Figure 4

portrays the weighted proportions within this

response group of those answering ‘‘strongly agree’’

or ‘‘somewhat agree’’ to each of the scales’ constitu-

ent items. From a scan of the items in the scale, con-

cerns over quality (‘‘I was concerned about the

quality of the information’’) stood out as a dimension

of particular concern. Nearly half of survivors

(44.6%), among those who recounted their experi-

ence of looking for information, expressed worry

over the quality of the information they retrieved.

A continuous index of search experiences was

examined using multivariate linear regression, in

which a higher ISEE scores indicated better informa-

tion search experiences. None of the sociodemo-

graphic variables considered in this study were

found to be significantly associated with search

experiences; there was marginal evidence that

increasing annual income was associated with more

positive information-seeking experiences (P 5 .09).

FIGURE 3. Cancer survivors’ information source preferences and informa-tion source use. Weighted percentages are adjusted for gender, age, educa-

tion, annual income, race/ethnicity, and health insurance status. FIGURE 4. Cancer survivors’ (n 5 865) information-seeking experiences:

weighted.

2536 CANCER Supplement June 1, 2008 / Volume 112 / Number 11

DISCUSSIONFrom the HINTS 2005 data it is apparent that

‘‘looking for information about cancer’’ is not an ex-

perience restricted to a privileged segment of the

population but appears to be quite common in the

public at large. From the data presented in this arti-

cle, 46.7% of the general U.S. adult population, or a

little over a million Americans, have looked at 1 time

or another for information regarding the disease, its

causes, its prevention, its treatment, or its sequelae

(Fig. 1). These data are consistent with analyses con-

ducted on the HINTS 2003 data.35

Not surprisingly, a motivated use of information

resources appears to be the hallmark of many who

have been touched by cancer. Nearly two-thirds of

cancer survivors and a little more than half of those

with family histories of cancer reported having

looked for information about cancer from any

source. By contrast, barely more than 25% of those

with no personal or family experience of cancer

reported searching for information about the disease.

For cancer survivors, the experience is nearly univer-

sal during the first year since diagnosis, with 97% of

those with a recent diagnosis having looked for can-

cer information during the previous 12 months.

Interestingly, the need to search for information

about cancer appears to continue, with roughly half

of those survivors diagnosed in the past (2 to 11 or

more years) having reported looking for cancer infor-

mation within the previous 12 months. These data

match analyses of the HINTS 2003 data, suggesting

that many (although not all) cancer survivors look

for information from a variety of sources, including

the Web.36

Information-seeking Beyond the ‘Transition’Supportive, effective communications should be the

hallmark of a world-class healthcare system for can-

cer survivors. As the authors of the IOM 2006 report

From Cancer Patient to Cancer Survivor: Lost in Tran-

sition reasoned, a patient-centered approach is

essential to survivorship care. This includes a respon-

siveness to patient needs, and effective communica-

tion and information sharing.37 Nevertheless, the

authors of the report expressed concern that as

patients transitioned to the status of ‘survivor,’ they

might find themselves ‘lost’ in an information envir-

onment that is fragmented and that has abdicated

responsibility for long-term support of survivors’

needs.

The data reported in the current study provide at

least a preliminary indication of information-seeking

patterns at this phase of transition. From the multi-

nomial model we constructed to explain preference

for Internet versus Provider, we know that age was

the most significant predictor of Internet usage. The

finding is consistent with other research on demo-

graphics and Internet use.34 In Figure 3, we mapped

preference and usage patterns onto years since diag-

nosis, while controlling statistically for the influence

of age and other demographic variables. From the

graph, we do not observe many differences because

of preference: all survivors showed a strong interest

in going to their provider teams first before going to

other sources. However, the usage patterns suggest a

different story. Those who have just been diagnosed

report going to their providers first, as might be

expected. Presumably, access to a treatment team

would be frequent during this first year and cancer

patients would find it easy to pose their questions to

medical staff. During Years 2 to 10 after diagnosis,

the point of transition, the pattern reverses and the

profile resembles that of the general population. Sur-

vivors prefer going to their HCP for quality informa-

tion, but in reporting actual usage revealed going to

the Internet as a source of first resort. The Internet,

always on and always available, became the first

place to look whenever a ‘strong need’ arose to

obtain information on cancer and cancer survivor-

ship. Given the uneven and unstructured environ-

ment of the World Wide Web,38 it is little wonder

that concerns regarding quality topped the list of

reactions to information-seeking. Moreover, our mul-

tivariate model, in which we regressed a continuous

score from our information-seeking experiences scale

(ISEE) onto demographics, did not yield any signifi-

cant relationships: Everyone experienced some

degree of confusion when looking for cancer infor-

mation, a finding that duplicates research done with

the scale on the general population.26

Several interesting differences in information-

seeking by survivorship status were observed. Prefer-

ence for an HCP as a first source of cancer informa-

tion was stronger among respondents with a

personal or no history of cancer compared with

those with a family history. Furthermore, although a

majority of respondents indicated a preference to

use their HCP to get information about cancer, sig-

nificantly more respondents with a personal history

of cancer had used an HCP as a first source of cancer

information compared with those with a family or no

history of cancer. These findings, coupled with the

greater frustration in information-seeking reported

by respondents with a family history of cancer, are

consistent with previous research documenting dis-

satisfaction with regard to information-seeking

among family members of cancer patients.39

Information Support/Hesse et al. 2537

Although physicians and other healthcare profes-

sionals are consistently named as the preferred

source of cancer information among family members

of cancer patients, in previous research, family mem-

bers report difficulty obtaining desired information

and limited contact with physicians.39

Engineering ‘Deep Support’ Into the Health InformationEnvironmentOne implication of these data is to consider ways

of providing ‘deep support’ for cancer survivors’ in-

formation needs well beyond the point of diagnosis

and treatment.40 Historically, the business systems

that support medicine have been built around a

transactional model41: patients present with symp-

toms, often too late for preemptive care42; they are

treated; and then they are dismissed. Transactional

systems strain when dealing with chronic disease,

preventive care, and personalized service.43 Yet the

new systems must evolve to meet the demands of

21st century medicine.44 Medicine in the 21st cen-

tury must use advances in information technology to

create a service that is predictive of upcoming need,

preemptive of early disease process, and persona-

lized with individually tailored plans for treatment

and self-management.44 The transformation is espe-

cially important in oncology, in which continuous

monitoring over the lifespan will yield population

gains from prevention, early detection, and control

over metastasis.42 Following recommendations from

IOM, these new systems must become ‘patient cen-

tric,’ using information technology to create ‘healing

relationships’ with patients and their caregivers over

time.

It is easy to imagine from these data how the

environment of medicine can be reengineered to

support the move to a relationship model. Electronic

medical records (EMRs) are being put in place to

guarantee patients that their medical records are ac-

cessible to all members of the care team within and

across life phases.45 Personal Health Records, teth-

ered to the continuity of the EMR, are being used to

support self-management by offering patients full

and complete access to their own health information

and to offer ‘information therapy’ for the different

questions that emerge over the lifespan.38 Appoint-

ment management systems, with reminders and sta-

tus checks available to patients as well as providers,

can be used to ensure continuity of care well beyond

the transition period.

All these implementations are under priority

consideration by the Department of Health and

Human Services, the IOM, and countless Health In-

formation Technology companies throughout the

country.44 Its importance to healthy survivorship is

paramount. In fact, some have argued that solving

the information dilemma for patients and providers

will offer one of the most important contributions to

population health in the coming decade.5,46–48

Limitations of the DataThe data presented in this article come from the

2005 administration of the HINTS, a cross-sectional

view of cancer information attitudes and behavior in

the general population. Caution should be taken not

to overinterpret the observed associations between

study variables, given the cross-sectional nature of

the survey. The associations should be interpreted in

light of converging evidence from other sources.

Some of the correlations, such as the association

between age and Internet use or sex and informa-

tion-seeking, have been well documented. Others,

such as the curvilinear association between Internet

reliance and time since diagnosis, are new. These

correlations merit continued study in other venues

to substantiate and confirm. Future studies could

also offer variations in wording for key items in the

survey and could explore issues of convergent, con-

struct, and predictive validity in greater depth.

Another limitation of the data has to do with the

falling response rates in random digit dial (RDD) sur-

veys. The HINTS program was designed after the rig-

orous and exacting procedures for coverage and

selection used in other federal surveillance mechan-

isms. Nevertheless, RDD response rates are falling

universally in all federal surveys because of a variety

of pressures in the communication environment (tel-

emarketing, the ‘do not call’ list, etc). The implica-

tions of these trends for nonresponse bias are not

completely known. Fortunately, psychometric reviews

suggest that population estimates for the types of

items included in HINTS (eg, self-reports of health

behaviors such as smoking) have not been affected

dramatically by declining response rates.49

To maximize sample size in the multivariate

analyses, information-seeking was defined according

to having ever sought cancer information, rather

than having sought information during the past year.

Analysis of more recent cancer information may be

more robust and reveal stronger associations with

variables of interest.

ConclusionsSome have observed that a diagnosis of cancer sig-

nifies a lifetime of ‘making sense’ in an information

environment that appears fragmented, uneven, and

even disparate in quality to survivors. The data pre-

sented in this study confirm those expectations, but

2538 CANCER Supplement June 1, 2008 / Volume 112 / Number 11

they also offer a perspective on survivors’ informa-

tion-seeking behavior across years since their initial

diagnosis. The perspective provides a glimpse into

how the health information environment can be

reengineered to shore up support for the goals of

self-vigilance, self-monitoring, and healthy living that

characterize life after treatment. Health information

technology can play a salubrious role in this reengi-

neering effort, but only if it enhances the ‘healing

relationship’ of patients, their families, and their

healthcare providers in safeguarding health.

REFERENCES1. Hoffman J. Awash in information, patients face a lonely,

uncertain road. New York Times August 14, 2005. Available

at URL: http://www.nytimes.com. Accessed September 4,

2007.

2. Grady D. Cancer patients, lost in a maze of uneven care.

New York Times. July 29, 2007. Available at URL: http://

www.nytimes.com. Accessed September 4, 2007.

3. McCaughan E, McKenna H. Never-ending making sense:

towards a substantive theory of the information-seeking

behaviour of newly diagnosed cancer patients. J Clin Nurs.

2007;16:2105–2113.

4. Bickell NA, LePar F, Wang JJ, Leventhal H. Lost opportu-

nities: physicians’ reasons and disparities in breast cancer

treatment. J Clin Oncol. 2007;25:2516–2521.

5. Viswanath K. Science and society: the communications

revolution and cancer control. Nat Rev Cancer. 2005;5:828–

835.

6. Kohn LT, Corrigan J, Donaldson MS. To Err is Human:

Building a Safer Health System. Washington, DC: National

Academy Press; 2000.

7. Norman DA. The Design of Everyday Things. 1st Basic

paperback ed. New York: Basic Books; 2002.

8. Coiera E. Four rules for the reinvention of health care.

BMJ. 2004;328:1197–1199.

9. Kreps GL, O’Hair D. Communication and Health Out-

comes. Cresskill, NJ: Hampton Press; 1995.

10. O’Hair D, Kreps GL, Sparks LD. Handbook of Communica-

tion and Cancer Care. Cresskill, NJ: Hampton Press; 2006.

11. Arora NK. Interacting with cancer patients: the significance

of physicians’ communication behavior. Soc Sci Med. 2003;

57:791–806.

12. Epstein RM, Street RL. Patient-Centered Communication in

Cancer Care: Promoting Healing and Reducing Suffering.

Bethesda, MD: National Cancer Institute; 2007.

13. Leyden WA, Manos MM, Geiger AM, et al. Cervical cancer

in women with comprehensive health care access: attribut-

able factors in the screening process. J Natl Cancer Inst.

2005;97:675–683.

14. Bright MA. The National Cancer Institute’s Cancer Informa-

tion Service: a premiere cancer information and education

resource for the nation. J Cancer Educ 2007;22(1 suppl):S2–

S7.

15. Johnson JD. Cancer-related Information Seeking. Cresskill,

NJ: Hampton Press; 1997.

16. Marcus AC, Woodworth MA, Strickland CJ. The Cancer In-

formation Service as a laboratory for research: the first 15

years. J Natl Cancer Inst Monogr. 1993;14:67–79.

17. Finney Rutten LJ, Squiers L, Hesse B. Cancer-related infor-

mation sought by the general public: evidence from the

National Cancer Institute’s Cancer Information Service,

2002–2003. J Cancer Educ. 2007;22:91–98.

18. La Porta M, Hagood H, Patt J, Hodorowski JK, Treiman K.

The NCI’s Cancer Information Service: meeting the public’s

cancer information needs via the internet. J Cancer Educ.

2007;22(1 suppl):S26–S34.

19. Smith CA, Stavri PZ, Chapman WW. In their own words? A

terminological analysis of e-mail to a cancer information

service. Proc AMIA Symp. 2002:697–701.

20. Bright MA. Web sites: the National Cancer Institute’s Can-

cer Information Service. J Cancer Educ. 2007;22:5.

21. Wilkinson GS, Wilson J. An evaluation of demographic dif-

ferences in the utilization of a cancer information service.

Soc Sci Med. 1983;17:169–175.

22. Freimuth VS. Narrowing the cancer knowledge gap

between whites and African Americans. J Natl Cancer Inst

Monogr. 1993:81–91.

23. Rutten LJ, Arora NK, Bakos AD, Aziz N, Rowland J. Infor-

mation needs and sources of information among cancer

patients: a systematic review of research (1980–2003).

Patient Educ Couns. 2005;57:250–261.

24. Hesse BW, Moser RP, Rutten LJ, Kreps GL. The health infor-

mation national trends survey: research from the baseline.

J Health Commun. 2006;11(suppl 1):vii–xvi.

25. Nelson DE, Kreps GL, Hesse BW, et al. The Health Informa-

tion National Trends Survey (HINTS): development, design,

and dissemination. J Health Commun. 2004;9:443–460; dis-

cussion 481–444.

26. Arora NK, Hesse BW, Rimer BK, Viswanath K, Clayman ML,

Croyle RT. Frustrated and confused: the American publicrates its cancer-related information-seeking experiences

[Electronic publication]. J Gen Intern Med. 2008;23:223–228.

27. Arora NK, Johnson P, Gustafson DH, McTavish F, Hawkins

RP, Pingree S. Barriers to information access, perceived

health competence, and psychosocial health outcomes:

test of a mediation model in a breast cancer sample.

Patient Educ Couns. 2002;47:37–46.

28. Willis GB. Cognitive Interviewing: A Tool for Improving

Questionnaire Design. Thousand Oaks, CA: Sage Publica-

tions; 2005.

29. Sudman S, Bradburn NM, Schwarz N. Thinking About

Answers: The Application of Cognitive Processes to Survey

Methodology. 1st ed. San Francisco: Jossey-Bass; 1996.

30. Davis T, Park I, Covell J, Rizzo L, Cantor D. Health Informa-

tion National Trends Survey: Final Report. Rockville, MD:

Westat; 2005.

31. Research Triangle Institute. SUDAAN Users Manual. 9th ed.

Research Triangle Park, NC: Research Triangle Institute; 2006.

32. Espey DK, Wu XC, Swan J, et al. Annual report to the

nation on the status of cancer, 1975–2004, featuring cancer

in American Indians and Alaska Natives. Cancer. 2007;110:

2119–2152.

33. Johnson J, Blanchard J, Harvey C. Americans’ view of can-

cer. Support Care Cancer. 2000;8:24–27.

34. 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

first Health Information National Trends Survey. Arch In-

tern Med. 2005;165:2618–2624.

35. Rutten LJ, Squiers L, Hesse B. Cancer-related information

seeking: hints from the 2003 Health Information National

Trends Survey (HINTS). J Health Commun. 2006;11(suppl

1):147–156.

Information Support/Hesse et al. 2539

36. Mayer DK, Terrin NC, Kreps GL, et al. Cancer survivors in-

formation seeking behaviors: a comparison of survivors

who do and do not seek information about cancer. Patient

Educ Couns. 2007;65:342–350.

37. Hewitt ME, Ganz PA,Institute of Medicine (U.S.). From

Cancer Patient to Cancer Survivor: Lost in Transition.

Washington, DC: National Academies Press; 2006.

38. Kemper DW, Mettler M. Information Therapy: Prescribed

Information as a Reimbursable Medical Service. 1st ed.

Boise, ID: Healthwise; 2002.

39. Rutten LJ, Squiers L, Treiman K. Requests for information

by family and friends of cancer patients calling the

National Cancer Institute’s Cancer Information Service.

Psychooncology. 2006;15:664–672.

40. Shneiderman B. Leonardo’s Laptop: Human Needs and the

New Computing Technologies. Cambridge, MA: MIT Press;

2002.

41. Zuboff S, Maxmin J. The Support Economy: Why Corpora-

tions Are Failing Individuals and the Next Episode of Capit-

alism. New York: Viking; 2002.

42. von Eschenbach AC. A vision for the National Cancer Pro-

gram in the United States. Nat Rev Cancer. 2004;4:820–828.

43. Wagner EH. Chronic disease care. BMJ. 2004;328:177–178.

44. Culliton BJ. Extracting knowledge from science: a conver-

sation with Elias Zerhouni. Health Aff (Millwood). 2006;

25:w94–w103.

45. Brailer D. Action through collaboration: a conversation

with David Brailer. The national coordinator of HIT

believes that facilitation, not mandates are the way to

move the agenda forward. Interview by Robert Cunning-

ham. Health Aff (Millwood). 2005;24:1150–1157.

46. Institute for the Future. Health and Health Care 2010: The

Forecase, the Challenge. 2nd ed. San Francisco, CA: Jossey-

Bass; 2003.

47. Cayton H. The flat-pack patient? Creating health together.

Patient Educ Couns. 2006;62:288–290.

48. Institute of Medicine (U.S.). Committee on Quality of

Health Care in America. Crossing the Quality Chasm: A

New Health System for the 21st Century. Washington, DC:

National Academy Press; 2001.

49. Biener L, Garrett CA, Gilpin EA, Roman AM, Currivan DB.

Consequences of declining survey response rates for

smoking prevalence estimates. Am J Prev Med. 2004;27:

254–257.

2540 CANCER Supplement June 1, 2008 / Volume 112 / Number 11