Ask question

Research Methods: Quantitative Research

Comparing Strategies for Health Information Dissemination: Messengers That Can Help or Hinder

Jessica Fishman, PhD 1 , Patricia Greenberg, MS

2 , Margy Barbieri Bagga, MS

3 ,

David Casarett, MD, MA 4 , and Kathleen Propert, PhD

5

Abstract

Purpose: To test the effects of different messengers on the dissemination of health information.

Design: An experimental study exposed participants to 12 news articles pertaining to 1 of 3 health topics framed from the perspective of 4 generic messengers: religious figures, doctors, celebrity patients, or ordinary patients. Participants select as many of the 12 articles as desired.

Setting: A cancer clinic within a large, urban hospital serving a sociodemographically diverse patient population.

Participants: Eighty-nine patients with a history of cancer.

Measures: The primary outcome was the frequency with which each news story was selected.

Analysis: Summary statistics and a general estimating equation model.

Results: For each health topic, news articles using celebrity messengers were the least likely to be selected; almost half of the participants (36 [41.4%] of 87) rejected all such articles. Articles linked to religious figures were equally unpopular (P ¼ .59). Articles that used doctors or ordinary patients as the messenger were very likely to be selected: Nearly all women (84 [96.6%] of 87) selected at least one of these. Furthermore, the odds of choosing articles linked to celebrities or religious leaders were statistically significantly lower than the odds of choosing those linked to ordinary patients or doctors (P < .01).

Conclusion: Commonly used generic messengers had large effects on the dissemination of information. Health materials linked to celebrities or religious figures were consistently less likely to be selected than those linked to ordinary patients, or doctors.

Keywords health communication, cancer communication, mass media effects, information dissemination, persuasion, influence, behavior change, messenger, source, implementation science

Purpose

Health information is made widely available to millions

through news reporting, where various types of messengers

(or sources) are used to convey information. For example,

news coverage of health issues often begins with a

celebrity-focused introductory “hook.” 1-3

In other cases,

ordinary patients serve as the messenger. Sometimes, news

reports rely on doctors as the messenger, and in yet other

cases, they use religious leaders.

Despite the various types of messengers that appear in the

news media, the value of using one type versus another is

unclear because the effects have yet to be compared empiri-

cally. Studies of news coverage have examined the effects of

celebrity messengers, but they do not compare them to the

effects of using noncelebrity messengers. 1,4-8

The comparison

between different types of messengers is particularly

meaningful for mass-mediated health communication where

it is highly feasible to change the messenger and use any of

these potential messengers. In response, the current study

1 Department of Psychiatry, Perelman School of Medicine, University of

Pennsylvania, Philadelphia, PA, USA 2

Hackensack Meridian Health, Jersey Shore University Medical Center,

Neptune, NJ, USA 3 University of Pennsylvania Health System, Philadelphia, PA, USA 4 Department of Medicine, School of Medicine, Duke University, Durham, NC,

USA 5

Department of Biostatistics, Perelman School of Medicine, University of

Pennsylvania, Philadelphia, PA, USA

Corresponding Author:

Jessica Fishman, University of Pennsylvania, Room 3103, 3535 Market Street,

Philadelphia, PA 19104, USA.

Email: [email protected]

American Journal of Health Promotion 2018, Vol. 32(4) 932-938 ª The Author(s) 2017 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/0890117117733780 journals.sagepub.com/home/ahp

builds on earlier literature by comparing different types of

messengers while controlling for other factors, such as the

amount of health information offered, which can have poten-

tially confounding effects. 9-11

The goal of the current study is

to test which type of messenger is particularly effective at

attracting attention to health information.

This study is conducted with patients with breast cancer

because, although research on cancer communication has been

growing rapidly, there is little evidence about best practices,

especially with women. 9-13

Cancer is a leading cause of mor-

bidity and mortality, with breast cancer being one of the most

common types, and it is important to test strategies for sharing

health information with those affected. 14,15

Patients with can-

cer, and in particular, those with breast cancer, have a docu-

mented and substantial need for health information, 16-18

and

the dissemination of information about cancer through the

news can play a pivotal role in reducing the disease

burden. 19-21

In addition, breast cancer receives more funding

than other diseases, including other cancers, meaning that the

current findings will be relevant to future interventions funded

to improve communication with patients having breast

cancer. 22

It is important to identify the most effective

messengers for this large population of women, since the ideal

messenger may differ for other populations, and a stronger

evidence base can potentially improve women’s health.

Methods

Design

In this study, patients met individually and privately with a

research assistant who offered them a varied selection of health

materials. Specifically, the patient encountered copies of 12

different news articles and were told they could select any news

articles that were of interest to take home and read. The

research assistant briefly described each article as pertaining

to 1 of 3 different health topics from the perspective of reli-

gious figures, doctors, celebrity patients, or ordinary patients.

This design allowed 3 tests of whether these common yet dis-

tinct types of messengers (the independent variable) influenced

whether the material would be selected or rejected (the depen-

dent variable).

The dependent variable of interest is whether patients

select a news article because this outcome is a rate-limiting

one in studies of mass communication effects. Whether the

information influences patient knowledge, beliefs, health-

care decisions, or other health-related behaviors, depends on

the patient first deciding to select a news article. Therefore, it

is important to initially test the effects messengers have on

whether patients will select a news article. The current study

is also designed to test whether these effects depend on the

health topic presented.

This study was designed to compare the effects of the gen-

eric messenger, which is frequently used by the news media in

headlines and “teasers.” Recent examples of the celebrity gen-

eric messenger include the following news articles: “Celebrity

Cancer Patients and Survivors,” “Stars Who Have Battled

Cancer,” “14 Celebrities with Breast Cancer,” and “11

Celebrities Affected by Breast Cancer.” 23-26

Similarly, health

reporting also refers generically to doctors as messengers, as

these news headlines illustrate: “Doctors excited by new cancer

treatment” and “Doctors object to high cancer-drug

prices.” 27,28

In yet other headlines, the generic messenger may

include religious leaders or ordinary patients. 29,30

Despite the common journalistic practice of using generic

messengers, we are not aware of studies empirically comparing

their effects. Past studies have examined the effects of a celeb-

rity messenger but, in addition to lacking a comparison to

another messenger, they differ from the current study because

they have studied cancer news coverage of a specific celebrity

(eg, Kylie Minogue or Nancy Reagan, when diagnosed with

breast cancer). 4-8

The effects attributed to a specific celebrity

may not be generalizable to another famous individual or to the

generic celebrity messenger. These prior studies have also

measured outcomes that differ from those of interest to the

current study.

Participants

To strengthen the generalizability of the results among

patients with breast cancer, the current study was designed

to attract a socioeconomically diverse sample. Specifically,

using relatively inclusive eligibility criteria, the study

enrolled patients with breast cancer who were at least 21 years

old and spoke English. There were no other requirements.

These eligibility criteria were also designed to include rela-

tively healthy patients (with lower morbidity and mortality

risks), as well as sicker ones, and therefore represent patients

with breast cancer across the spectrum.

Setting

Participants were recruited from an outpatient breast cancer

clinic in a large, urban hospital (located in Philadelphia). Those

who were eligible, based on clinic records, were approached

while waiting at the clinic for a scheduled appointment. The

clinic setting is of interest because it may offer many neglected

opportunities for disseminating health-related materials. A

majority of Americans visit health-care providers at least annu-

ally, and those with life-threatening diagnoses, such as cancer,

do so more frequently. 31

Furthermore, clinics often have long

wait times, creating a “captive” audience for any health infor-

mation provided. For these reasons, health promotion research-

ers have urged clinics to increase efforts to provide health

materials for their patient population. 32,33

Experimental Procedures and Measures

During recruitment, a research assistant approached eligible

patients in the clinic and assessed their interest in study partic-

ipation. Interested patients provided informed consent for this

study (which was approved by the institutional review board of

Fishman et al. 933

the University of Pennsylvania and the Abramson Cancer Cen-

ter). To avoid bias during recruitment, the study was described

in very general terms. Potential participants were told that the

study was designed to summarize patients’ daily activities and

that there were no correct or incorrect responses.

Study participants first answered a brief set of questions

about their daily activities, which were not actually germane

to the current tests. Subsequently, the research assistant

thanked them for their participation and, as a token of appre-

ciation, offered participants an array of informational materials

that they were invited to select from, depending on which (if

any) were of interest. As confirmed by exit questions (data not

shown), participants were not aware that the study was testing

the effects of messengers on their selection behavior. Partici-

pants only perceived that they were offered a wide-ranging

selection of health information simply as a courtesy.

This study was designed to test the effects of different mes-

senger types when patients first encounter the headline or

“teaser,” as this is a point when they decide whether to forgo

that news story. Therefore, when offering the news articles, the

research assistant briefly summarized each article (in 10 words

or less), reflecting the small amount of information typically

presented by an Internet or print news headline or a “teaser”

before the start of a television or radio news segment. To con-

trol for potential order effects, the sequence in which articles

were offered to each patient was randomly determined.

The 12 articles presented to the patients focused on 1 of 3

different health topics: (1) surviving cancer, (2) coping with a

terminal illness, or (3) health-care reform controversies. These

topics are commonly covered by the news media, and they are

widely considered important public health topics. 34

For each

article addressing a particular topic, the research assistant’s

description of an article was the same except for the messenger

referenced. For example, one article was described as one in

which doctors discuss a particular topic, whereas another was

described as one in which religious figures discuss the (same)

particular topic, and so forth, for each of the 4 messengers. (For

reasons discussed above, instead of personally identifying indi-

vidual religious figures, doctors, patients, or celebrity patients,

each messenger type was referenced generically.)

The frequency with which each news story was selected

determined the study’s primary outcome, but the study also

included a secondary outcome, proving another means of test-

ing messenger effects with the same participant. For this sec-

ondary outcome, the research assistant asked the participants to

rate their level of interest in each article. In turn, patients

selected a number from a Likert scale ranging from 0 to 10,

where 0 ¼ not at all interested and 10 ¼ extremely interested. (Prior to data collection, the reason for collecting this informa-

tion was again disguised from participants to avoid response

bias, and as mentioned above, exit interviews confirmed that

participants were not aware of the effects being tested.)

Other measures collected included standard sociodemo-

graphic factors (eg., age, race, education levels, and marital

status) that were self-reported. Participants also self-reported

their health status on a 5-point scale ranging from “poor” to

“excellent.” In addition, a patient’s health status was mea-

sured objectively using the cancer staging noted in their

clinic records. These sociodemographic and clinical charac-

teristics were used to describe the sample and test for possi-

ble predictors.

Analyses

The primary analytic goal was to compare the likelihood that

news articles with different messengers of health information

would be selected. Summary tables and box plots created a

numerical and visual representation of the findings. To for-

mally test the effects of each of the 4 messengers, a general

estimating equation (GEE) model with an exchangeable corre-

lation structure clustered on the individual participant, and a

logit link was used to predict the likelihood of article selection,

with the messenger as the sole predictor. 35

The reference level

was the celebrity messenger, which is commonly used by the

news media and of interest to a growing field of research. 1,4-8

After analyzing the actual selection frequencies, the study

also analyzed the mean levels of interest for each news story by

messenger type. To more formally test the patients’ level of

reported interest in each messenger, another GEE model sim-

ilar to the above, but now with an identity link, was used to

model patients’ average level of desire for information from

each messenger on the 0 to 10 scales. (The reference level was

again celebrities.) All statistical analyses were performed using

SAS/STAT software version 9.3.

Results

In this study, staff approached 122 patients with breast cancer

and 89 (73%) agreed to participate. All but 2 were included in the final analyses, resulting in a relatively diverse sample of

women with a history of breast cancer, as measured by age,

race, education, and other characteristics (Table 1). Women

ranged from 30 to 80 years old, with a median age of 56 years

(data not shown). About one-third of the sample self-

identified as black/African American, while the rest self-

identified as white/Caucasian. The patients reported varying

degrees of education: 84% graduated from high school and half of the participants had received some college education.

(According to the US Census, those enrolled had slightly less

education, on average, than the general US population.) The

sample also included patients with different levels of health,

as subjectively assessed by the study participant and objec-

tively by clinic records. For example, about half reported

having very good or excellent health.

Patients were likely to select information about the health

topics offered. More specifically, 79% (69/87) of patients selected at least 1 news article about a particular health topic.

Additionally, analyses showed that patients typically selected

at least 5 of the 12 available news articles, and all but 2 parti-

cipants selected at least 1 news article.

Although there was a high chance that articles would be

selected, the likelihood that a news article was selected

934 American Journal of Health Promotion 32(4)

depended on the messenger. Table 2 summarizes the number of

articles selected pertaining to each messenger type, regardless

of the topic, because the effect of each messenger did not vary

significantly between topics (data not shown). In other words,

the messenger effects were consistent across each of the 3 tests

using different health topics.

As shown in Table 2, the news articles using celebrity mes-

sengers were the most likely to be rejected; almost half of the

study participants (36 [41.4%] of 87) rejected all 3 celebrity- focused news articles available. Participants were also likely to

reject information linked to the religious messenger. In con-

trast, news articles that used ordinary patients or doctors as the

messenger were likely to be selected—nearly all participants

(84 [96.6%] of 87) selected at least one of these news stories. More than half of the participants (51 [58.6%] of 87) selected all 3 articles sharing information from the perspective of ordi-

nary patients. Most participants (49 [56.3%] of 87) also selected all 3 available articles sharing the perspective of doc-

tors. By contrast, only about a quarter of participants selected

all 3 of the celebrity-focused articles.

Consistent with the results presented in Table 2, a GEE

model that was fit to the data found that the odds of choosing

celebrity-linked health information were statistically signifi-

cantly lower than the odds of choosing information linked to

ordinary patients or doctors (P < .01). In addition, the odds

were not statistically significantly different from 0 that a

celebrity-linked news article would be chosen. The materials

linked to celebrity patients and religious figures were found to

be equally unpopular (P ¼ .59). As represented in Figure 1, when individuals rated their

level of interest in each article, the articles including celebrity

messengers received the lowest values. The information linked

to ordinary patients and doctors repeatedly received the highest

ratings and were each twice as high as the ratings for the

celebrity-linked articles, resulting in statistically significantly

differences (P < .01). Specifically, the mean interest levels

(with 95% confidence interval) were 2.7 (2.1-3.4), 3.2 (2.5- 4.0), 5.5 (4.9-6.1), and 5.8 (5.3-6.4), respectively, for articles

using celebrities, religious leaders, ordinary patients, and doc-

tors as the messenger.

This study also examined whether possible predictors

included available sociodemographics (such as race, age, edu-

cation level, and other factors reported in Table 1) or clinical

characteristics (such as cancer staging or self-assessments of

health). Among this sample with diverse sociodemographics

and wide-ranging levels of health, no results were significantly

predictive when analyzing whether articles were selected or

rejected. The same was true when analyzing the degree to

which patients rated being interested in each article.

Conclusion

In this study of patients with breast cancer, women frequently

selected the materials available, reflecting a substantial inter-

est in receiving health information. Patients wanted the mass-

mediated health information even though they were at a clinic

to meet with clinicians personally, suggesting that patients

may have high informational needs regardless of any direct

access to experts. Because this study was conducted in a

clinic, the findings suggest that clinic settings may present

feasible, valued opportunities for disseminating health infor-

mation, as others have advocated. 32,33

The findings also document large effects suggesting that,

for this study population, the messenger matters. By a sub-

stantial margin, the health materials using celebrity messen-

gers were the least likely to be selected. When religious

figures were the messenger, the health materials were also

unlikely to be selected. When the materials used ordinary

patients or doctors as the messenger, study participants were

very likely to select them.

Articles using doctors or ordinary patients as the messenger

may have been selected most often because the study was

conducted in a clinical setting, where the doctors and patients

physically present may have “primed” participants to value

doctors and patients as informational sources. Possibly, these

messengers would not be preferred in nonclinic settings, such

as work or home settings. However, it is also possible that if

this study was conducted in a nonclinic setting, that preferences

for doctor and patient messengers may be even stronger than

documented. In settings where doctors and fellow patients are

rare, and not immediately accessible for interpersonal

Table 1. Study Sample Characteristics.

Characteristics Frequency (N ¼ 87)a Percentage

Education Some high school 14 16.1 Some college 44 50.6 Some post-graduate 29 33.3

Race Black/African American 27 31.0 White/Caucasian 57 65.5 Other 3 3.5

Marital status Married/living with partner 52 59.8 Other 25 28.7 Missing 10 11.5

Employment status Working part-time 48 55.2 Working full-time 11 12.6 Unemployed 28 32.2

Cancer stage b

Stage 0 to 1 33 37.9 Stages 2 and 3 28 32.2 Stage 4 26 29.9

Self-reported health Poor 12 13.8 Good 35 40.2 Very good 29 33.3 Excellent 10 11.5 Missing 1 1.20

a Two of the 89 patients enrolled were excluded from this table and all analyses.

One patient was unable to complete the study due to time constraints. The other patient excluded was male and all others were female. bData based on clinic charts.

Fishman et al. 935

communication, their value as sources may actually increase.

Indeed, the clinic setting may have suppressed the full extent to

which these messengers are actually preferred.

It is also possible that the study setting does not have a large

moderating effect. Given the large effects found in the current

study, which showed an overwhelming preference for messen-

gers who are doctors or ordinary patients, the potentially mod-

erating effect of the study setting would have to be very strong

to reverse these preferences. Nonetheless, to test whether the

setting has a large moderating effect, additional studies should

be conducted in different settings.

Yet other study limitations should be emphasized. The evi-

dence base can also be expanded by testing messenger effects

for additional health topics, such as nutrition, vaccine safety, or

mental health treatment. Although ordinary patients and doc-

tors were the preferred messenger, consistently for each topic

studied, future studies with additional topics can provide more

tests of generalizability.

Although enrollment in the current study was high and

resulted in a diverse sample (as measured by various socio-

demographic and clinical characteristics), and although the

effects were strong, it would be worthwhile to test the extent

to which results are generalizable to all women with a history

of breast cancer. For example, one may question whether the

effects depend on one’s levels of education. We were able to

consider this question because the sample had a diverse range

of educational levels, and we did not find this to be the case.

The study population also included relatively healthy individ-

uals, as well as sicker ones, and when examining these possible

clinical predictors, nothing was significant. Still, since there are

no other known tests of these effects, the findings may or may

not apply beyond the study population, which was recruited

from a single clinic. Future tests could include different study

populations to further test generalizability.

Future research could also explore the extent to which pre-

ferences for messengers of mass-mediated health information

(which is the focus of the current study) reflect preferences for

interpersonal communication. Traditionally, the medical model

has considered doctors to be the ideal interpersonal messenger

for health-related information, due to their professional exper-

tise. More recently, however, models for lay health advisors

have implied that fellow patients can also serve as valued mes-

sengers precisely because they are lay.

Because this is the first study of its kind, the findings need

not be interpreted as conflicting with earlier research findings.

For example, although the current findings do not suggest that

the celebrity messenger should be used to disseminate infor-

mation to this study population, the findings need not be

viewed as conflicting those from earlier studies of a celebrity

messenger. 4-8

Some have argued that celebrity involvement is

ideal and have referred to research on what has been labeled the

“Katie Couric” effect, which refers to a celebrity’s work pro-

moting news coverage of cancer. 12

However, earlier studies

have examined news coverage of a single celebrity (without

a comparison to a noncelebrity messengers) and therefore

address a different question than the one of interest to the

current study. In fact, compared to the current study, earlier

studies had different independent and dependent variables.

The study did not find that religious leaders are ideal mes-

sengers, but the findings need not be viewed as conflicting

those from earlier studies of religious leaders, who have been

described as “uniquely positioned” with high credibility among

religious communities. 36

As a review of the literature sum-

marizes, great promise has been placed on these messengers:

“Health promotion researchers and practitioners agree on the

importance of clergy involvement, participation, and

Table 2. Selection of News Stories by Messenger Type.

The Number (%) of Participants Who Selected News Articles When the Messenger was . . .

Doctors Patients Religious Leaders Celebrities

Number of articles selected 0 4 (4.6%) 5 (5.7%) 33 (37.9%) 36 (41.4%) 1 7 (8.1%) 10 (11.5%) 8 (9.2%) 14 (16.1%) 2 27 (31.0%) 21 (24.2%) 15 (17.3%) 14 (16.1%) 3 49 (56.3%) 51 (58.6%) 31 (35.6%) 23 (26.4%) Selection of at least 1 article 83 (95.4%) 82 (94.3%) 54 (62.1%) 51 (58.6%)

Figure 1. Levels of interest in news stories by messenger type. Note: symbols designate the mean, and the line inside the bar designates the median. The bar for the celebrity median overlaps with the shaded area’s lower border. The confidence interval is represented by the shaded area.

936 American Journal of Health Promotion 32(4)

endorsement as an essential element for intervention

success.” 37

Conceivably, clergy may be more effective in other

contexts not examined by the current study. For example, they

may be more effective than other messengers when the inter-

vention relies largely on interpersonal communication of health

information. In contrast, this study compared messengers for

mass-mediated health communication.

To summarize, the current study has several limitations that

should be considered when interpreting results. Nonetheless,

for several health topics tested among a diverse population of

women with breast cancer, a consistent pattern emerged. Two

types of messengers—namely, ordinary patients and doctors—

substantially increased the likelihood that individuals would

select the available health information.

The findings also suggest that, for this study population,

the commonly used generic celebrity messenger may be least

effective at disseminating health information. Health advo-

cates have argued that “People with a personal history of

cancer—especially well-known or celebrity survivors—and

multiple mass media channels are key resources for

dissemination.” 38

Reflecting a similar assumption, a news

editor at the Huffington Post focused on the (generic) celeb-

rity messenger for an article entitled “13 of the Biggest Celeb-

rity Health Battles,” while claiming “Their stories and

courage will inspire you.” 39

Given that the current study find-

ings do not support this industry practice, the findings may

have practical implications.

Ideally, future attempts to disseminate health information

will embrace the most effective strategies, as determined by

additional studies. Fortunately, the messenger used to promote

health information can be easily switched with little cost.

Because mass media can influence public health by widely

disseminating information, there may be broad, immediate

opportunities to improve health communication.

Acknowledgments

The authors would like to thank the patients who participated in this

study.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to

the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for

the research, authorship, and/or publication of this article: American

Cancer Society, MRSGT grant.

References

1. Corbett J, Mori M. Medicine, media, and celebrities: news cov-

erage of breast cancer, 1960-1995. J Mass Commun Q. 1999;

76(2):229-249.

2. Wright K, Sparks L, O’Hair D, eds. Health Communication in the

21st Century. Oxford, United Kingdom: Wiley-Blackwell; 2013.

3. Obregon R, Waisbord S, eds. The Handbook of Global Health

Communication. Oxford, United Kingdom: Wiley-Blackwell;

2012.

4. Chapman S, McLeod K, Wakefield M, Holding S. Impact of news

of celebrity illness on breast cancer screening: Kylie Minogue’s

breast cancer diagnosis. Med J Aust. 2005;183(5):247-250.

5. Metcalfe D, Price C, Powell J. Media coverage and public reac-

tion to a celebrity cancer diagnosis. J Public Health. 2011;33(1):

80-85.

6. MacArthur GJ, Wright M, Beer H, Paranjothy S. Impact of media

reporting of cervical cancer in a UK celebrity on a population-

based cervical screening programme. J Med Screen. 2011;18(4):

204-209.

7. Noar SM, Ribisl KM, Althouse BM, Ayers JW. Using digital

surveillance to examine the impact of public figure pancreatic

cancer announcements on media and search query outcomes.

J Natl Cancer Inst Monogr. 2013;(47):188-194.

8. Nattinger AB, Hoffmann RG, Howell-Pelz A, Goodwin JS.

Effects of Nancy Reagan’s mastectomy on choice of surgery for

breast cancer by US women. JAMA. 1998;279(10):762-766.

SO WHAT? Implications for Health Promotion Practioners and Researchers

What is already known on this topic?

If disseminated, health information can have many posi- tive effects on populations. When attempting to pro- mote health information, different types of messengers can be used to disseminate the information, but little is known about their relative effectiveness.

What does this article add?

The findings from this study suggest that some different but commonly used types of messengers may not be equally effective at disseminating health information among patients with breast cancer—a large and impor- tant population of women with substantial informational needs. In separate tests on various health topics, infor- mational materials were unlikely to be selected by these patients if celebrity figures or religious leaders were the messengers, but likely to be selected if ordinary patients or physicians were the messengers.

What are the implications for health promotion practice or research?

This evidence suggests that the type of messenger used can influence the likelihood that mass-mediated health information disseminates. Those attempting to improve health communication with this patient population may consider these results. Fortunately, the decision to change a particular type of messenger is highly feasible and low cost.

Fishman et al. 937

9. Hornik R, ed. Public Health Communication: Evidence

for Behavior Change. New York, NY: Lawrence Erlbaum

Associates; 2002.

10. Rice R, Atkin C, eds. Public Communication Campaigns.

London, United Kingdom: Sage; 2013.

11. Lawton B, Rose S, McLeod D, Dowell A. Changes in use of

hormone replacement therapy after the report from the Women’s

Health Initiative: cross sectional survey of users. BMJ. 2002;

327(7419):845-846.

12. Viswanath K. Science and society: the communications revo-

lution and cancer control. Nat Rev Cancer. 2005;5(10):

828-835.

13. Gathirua-Mwangi WG, Monahan PO, Stump T, Rawl SM, Skin-

ner CS, Champion VL. Mammography adherence in African-

American women: results of a randomized controlled trial. Ann

Behav Med. 2015;50(1):70-78.

14. American Cancer Society. Cancer Facts and Figures 2016.

Atlanta, GA: American Cancer Society; 2016.

15. United Nations World Health Organization. Cancer Fact Sheet for

2015. http://www.who.int/mediacentre/factsheets/fs297/en/.

Updated 2016. Accessed July 9, 2016.

16. Moldovan-Johnson M, Tan A, Hornik R. Navigating the cancer

information environment: the reciprocal relationship between

patient-clinician information engagement and information seek-

ing from nonmedical sources. Health Commun. 2014;29(10):

974-983.

17. Basch E, Thaler H, Shi W, Yakren S, Schrag D. Use of informa-

tion resources by patients with cancer and their companions. Can-

cer. 2004;100(11):2476-2483.

18. Castleton K, Fong T, Wang-Gillam A, et al. A survey of Internet

utilization among patients with cancer. Support Care Cancer.

2001;1(8):1183-1190.

19. Stryker JE, Emmons KM, Viswanath K. Uncovering differences

across the cancer control continuum: a comparison of ethnic and

mainstream cancer newspaper stories, Prev Med. 2007;44(1):

20-25.

20. Yanovitzky I, Blitz CL. Effect of media coverage and physician

advice on utilization of breast cancer screening by women 40

years and older. J Health Commun. 2000;5(2):117-134.

21. Meissner HI, Potosky AL, Convissor R. How sources of health

information relate to knowledge and use of cancer screening

exams. J. Commun Health. 1992;17(3):153-165.

22. National Institutes of Health. Estimates of Funding for Various

Research, Condition, and Disease Categories (RCDC) March 7,

2014. http://report.nih.gov/categorical_spending.aspx. Accessed

August 20, 2014.

23. Celebrity cancer patients and survivors. Newsday. 2014. https://

newsday.com. Accessed September 1, 2016.

24. Stars who have battled cancer. New York Daily News. https://

NYdailynews.com. Accessed September 1, 2016.

25. Stewart K. 18 Celebrities with breast cancer. Everyday Health.

2014. https://Everydayhealth.com. Accessed September 1, 2016.

26. Sangweni Y. 11 Celebrities affected by breast cancer. Essence

Magazine. https://Essence.com. Accessed September 1, 2016.

27. Terranella S. Doctors excited by new cancer treatment. ABC

News. http://abcnews.go.com/Health/DrJo hnson/sto ry?

id¼117508. Accessed June 2, 2017. 28. Whalen J. Doctors object to high cancer-drug prices. The Wall Street

Journal. 2015. https://www.wsj.com/articles/doctors-object-to-

high-cancer-drug-prices-1437624060. Accessed June 2, 2017.

29. Welch A. How couples cope with cancer on Valentine’s Day, and

every day. 2016. CBS News. http://www.cbsnews.com/news/how-

couples-cope-with-cancer-on-valentines-day-and-every-day/.

Accessed June 2, 2017.

30. Andrews E. Religious leaders prepare to fight patent on genes.

New York Times. 1995. http://www.nytimes.com/1995/05/13/us/

company-news-religious-leaders-prepare-to-fight-patents-on-gen

es.html? mcubz¼0. Accessed June 2, 2017. 31. Ambulatory care use and physician office visits, National Hospi-

tal Ambulatory Medical Care Survey. 2014. Center for Disease

Control and Prevention, National Center for Health Statistics.

http://www.cdc.gov/nchs/fastats/physician-visits.htm. Accessed

March 6, 2017.

32. Oermann M. Effects of educational intervention in waiting room

on patient satisfaction. J Ambul Care Manage. 2003;26(2):

150-158.

33. Prochaska J, Zabinski M, Calfas K, Sallis J, Patrick K. PACEþ: interactive communication technology for behavior change in

clinical settings. Am J Prev Med. 2000;19(2):127-131.

34. Fishman JM, Tenhave T, Casarett D. Cancer and the media: how

does the news report on treatment and outcomes? Arch Intern

Med. 2010;170(6):515-518.

35. Diggle P, Heagerty P, Liang K, Zeger S. Analysis of Longitudinal

Data. 2nd ed. Oxford, United Kingdom: Oxford University Press;

2002.

36. Anshel M, Smith M. The role of religious leaders in promoting

healthy habits in religious institutions. J Relig Health. 2014;53(4):

1046-1059.

37. Bopp M, Baruth M. Leading their flocks to health? Clergy health

and the role of clergy in faith-based health promotion interven-

tions. Fam Commun Health. 2013;36(3):182-192.

38. Daher M. Cultural beliefs and values in cancer patients. Ann

Oncol. 2012;23(suppl 3):66-69.

39. Schocker L. 13 of the biggest celebrity health battles. Huffington

Post. 2013. http://www.huffingtonpost.com/2013/12/31/celeb

rity-health-2013_n_4518887.html. Accessed August 2, 2014.

938 American Journal of Health Promotion 32(4)