Assessing the Relationship of Attitude toward the Ad to Intentions to use Direct-to-Consumer Drugs

MISSING IMAGE

Material Information

Title:
Assessing the Relationship of Attitude toward the Ad to Intentions to use Direct-to-Consumer Drugs A Systematic Quantitative Meta-Analysis
Physical Description:
1 online resource (104 p.)
Language:
english
Creator:
Jung,Wan Seop
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Mass Communication, Journalism and Communications
Committee Chair:
Treise, Deborah M
Committee Members:
Goodman, Jennifer R
Weigold, Michael F
Dodd, Virginia J

Subjects

Subjects / Keywords:
direct -- dtc -- meta -- prescription
Journalism and Communications -- Dissertations, Academic -- UF
Genre:
Mass Communication thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Attitude toward the ad (Aad) is widely known to be an essential predictor of behavioral intentions. Therefore, a number of studies have addressed Aad in the DTCA literature. Despite this interest in Aad, there has not been a comprehensive attempt to investigate general findings across independent DTCA studies. Such an investigation is useful in understanding the general strength and variability of the relationships and the study conditions that moderate those relationships. For example, while some studies have reported no evidence of a significant effect of Aad on behavioral intentions, others have reported a significant effect. Furthermore, different studies have found widely varying magnitudes of the Aad effect on behavioral intentions. In order to assess the strength and variability of the Aad-intention relationship, the current research meta-analyzed Aad effects aggregated across all available research in the extant DTCA literature. In addition to the assessment of the relationship between Aad and intentions, this study also investigated the relationships between Aad and its antecedents and the potential moderating variables. The results of this meta-analysis provide considerable insight into the effects of Aad in the contexts of DTCA and the state of DTCA research. As with any meta-analysis, the data provide a quantitative summary. In the current meta-analysis, the data provided a summary of 278 samples reported in the 36 articles for which the author could obtain usable data. Variables were classified into three levels. The first level included demographic characteristics, ad awareness, involvement, health status, and drug usage. The second level contained consumers? attitudes toward the ad. The third included behavioral intentions. The first level directly and/or indirectly affected the second and third. The second level directly influenced the third. As shown in Tables 4-1 through 4-5 and analyzed above, the aggregated study effects suggested a significant relationship between Aad and a number of important constructs, including both antecedents (education, r = -.12 and Zr = -.12 and income, r = -.08, Zr = -.08) and consequences (behavioral intention, r = .19 and Zr = .20 and pharmacist intention, r = .15, Zr = .15). In addition, the results also found that consumers? intentions were influenced by personal characteristics, including gender (r = .02 and Zr = .08), health status (r = -.12 and Zr = -.12), drug usage (r = .14 and Zr = .15), and ad exposure (r = .23 and Zr = .24). The results showed that consumers who (a) were less educated, (b) had a low income (c) were female, (d) were in bad health, (e) took a lot of drugs, or (f) were exposed to advertising frequently tended to have more favorable attitudes toward DTCA than those who (a) were more educated, (b) had a high income (c) were male, (d) were in good health, (e) took few drugs, or (f) were exposed to advertising less frequently. However, in general, the strength of each of these relationships was small or small to moderate.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Wan Seop Jung.
Thesis:
Thesis (Ph.D.)--University of Florida, 2011.
Local:
Adviser: Treise, Deborah M.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043239:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 ASSESSING THE RELATIONSHIP OF ATTITUDE TOWARD THE AD TO INTENTIONS TO USE DIRECT TO CONSUMER DRUGS : A SYSTEMATIC QUANTITATIVE META ANALYSIS By WAN SEOP JUNG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Wan Seop Jung

PAGE 3

3 To m y family

PAGE 4

4 ACKNOWLEDGMENTS Above all, I thank my wife who supported me throughout the entire process that led to this dissertation Without her, I would not be Dr. Jung. Additionally, this project could not have been completed without the guidance and help of Dr. Debbie Treise. She provides the perfect blend of criticism and encouragement, revisio n and hope. Moreover, I want to thank my committee members Dr. Mike Weigold, Dr. Robin Goodman, and Dr. Virginia Dodd for providing feedback and support throughout this process.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 ABSTRACT ................................ ................................ ................................ ................................ ..... 9 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 11 Direct to Consumer Advertising ................................ ................................ ............................ 11 Purpose of the Study ................................ ................................ ................................ ............... 16 2 LITERATURE REVIEW ................................ ................................ ................................ ....... 19 Effect of the Attitude Toward the Ad ................................ ................................ ..................... 2 1 Factors Affecting Attitude Toward the Ad ................................ ................................ ............. 2 4 Ad Awareness and Ad Exposure ................................ ................................ ..................... 2 4 Demographic Factors ................................ ................................ ................................ ....... 2 6 Health Status and I nvolvement ................................ ................................ ........................ 2 6 Other Fact ors A ffecting Attitude Toward the Ad ................................ ............................ 28 Potentional Moderating Variables ................................ ................................ .......................... 28 Sample ................................ ................................ ................................ ............................. 28 Type of Research ................................ ................................ ................................ ............. 29 Measurement of Constructs ................................ ................................ ............................. 29 Theoretical Basis ................................ ................................ ................................ ............. 3 0 Other Research Ch aracteristics ................................ ................................ ....................... 3 0 Chapter Summary ................................ ................................ ................................ ................... 3 1 3 METHOD ................................ ................................ ................................ ............................... 35 Meta Analysis ................................ ................................ ................................ ......................... 35 Meta Analysis Process ................................ ................................ ................................ ............ 36 Step 1: Database Development ................................ ................................ ........................ 38 Step 2: The Conversion ................................ ................................ ................................ ... 40 Step 3: Method of An alysis ................................ ................................ ............................. 41 Rater Reliability ................................ ................................ ................................ ...................... 42 Fixed Versus Ra ndom Effect Models ................................ ................................ ..................... 43 File Drawer Problems ................................ ................................ ................................ ............. 45 Ch apter Summary ................................ ................................ ................................ ................... 4 6

PAGE 6

6 4 RESULTS ................................ ................................ ................................ ............................... 49 Antecedents of Attitude Toward the Ad ................................ ................................ ................. 5 0 Attitude Toward the Ad and Outcomes ................................ ................................ .................. 5 1 Ad Awareness ................................ ................................ ................................ ......................... 5 3 Other Relationships ................................ ................................ ................................ ................ 5 4 Moderator Analyses ................................ ................................ ................................ ................ 5 6 5 DI SCUSSION ................................ ................................ ................................ ......................... 67 Discussions and Implications ................................ ................................ ................................ 6 8 Role of Att itude Toward the Ad ................................ ................................ ...................... 6 8 Role of Ad Awareness ................................ ................................ ................................ ..... 73 Role of Antecedents of Attit ude Toward the Ad ................................ ............................. 7 4 Limitations ................................ ................................ ................................ .............................. 76 Chapter Summary ................................ ................................ ................................ ................... 7 8 APPENDIX A CODED CHARACTERISTICS OF INCLUDED STUDIES ................................ ................ 79 B OTHER CHARACTERISTICS OF INCLUDED STUDIES ................................ ................. 82 C CODING MANUAL AND FORMS ................................ ................................ ...................... 8 8 LIST OF REFERENCES ................................ ................................ ................................ ............... 95 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 104

PAGE 7

7 LIST OF TABLES Table page 1 1 Summary of DTCA Effects ................................ ................................ ............................... 18 2 1 Summary of H ypotheses ................................ ................................ ................................ .... 32 2 2 Summary of Research Questions ................................ ................................ ....................... 33 3 1 The Conversion Statistical Equations ................................ ................................ ................ 4 8 3 2 Interrater Reliability for Coded Categories ................................ ................................ ........ 4 8 3 3 Analysis of the File Drawer Problem ................................ ................................ ................. 4 8 4 1 Analysis of the Re lationships between Antecdents and A ad ................................ .............. 59 4 2 Analysis of the Relationships between A ad and Intentions ................................ ................ 59 4 3 Analysis of the Relationships between A ad and Other Outcomes ................................ ...... 6 0 4 4 Analysis of the Non Hypothesized Relationships ................................ ............................. 6 1 4 5 Subgroup Means by Moderator Variables ................................ ................................ ......... 6 3

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Suggested Model of DTCA Effects ................................ ................................ .................. 34 4 1 Summary of the Re sults ................................ ................................ ................................ ..... 58

PAGE 9

9 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ASSESSING THE RELATIONSHIP OF ATTITUDE TOWARD THE AD TO INTENTIONS TO USE DIRECT TO CONSUMER DRUGS : A SYSTEMATIC QUANTITATIVE META ANALYSIS By Wan Seop Jung August 2011 Chair: Debbie Treise Major: Mass Communication Attitude toward the ad ( A ad ) is widely known to be an essential predictor of behavioral intentions. Therefore, a number of studies have addressed A ad in the DTCA literature. Despite this interest in A ad there has not been a comprehensive attempt to investigate general findings across independent DTCA studies. Such an investigation is useful in understanding the general strength and variability of the relationships and the study conditions that moderate those relationships. For example, while some studies have reported no evidence of a significant effect of A ad on behavioral intentions, other s have reported a significant effect. Furthermore, different studies have found widely varying magnitudes of the A ad effect on behavioral intentions. In order to assess the strength and variability of the A ad intention relationship, the current research me ta analyzed A ad effects aggregated across all available research in the extant DTCA literature. In addition to the assessment of the relationship between A ad and intentions, this study also investigated the relationships between A ad and its antecedents and the potential moderating variables. The results of this meta analysis provide considerable insight into the effects of A ad in the contexts of DTCA and the state of DTCA research.

PAGE 10

10 As with any meta analysis, the data provide a quantitative summary. In t he current meta analysis, the data provided a summary of 278 samples reported in the 36 articles for which the author could obtain usable data. Variables were classified into three levels. The first level included demographic characteristics, ad awareness, involvement, health status, and drug usage. intentions. The first level directly and/or indirectly affected the second and third. The second level directly influe nced the third. A s shown in Tables 4 1 through 4 5 and analyzed above, the aggregated study effects suggested a significant relationship between A ad and a number of important constructs, including both antecedents (education, r = .12 and Z r = .12 and income, r = .08, Z r = .08) and consequences (behavioral intention, r = .19 and Z r = .20 and pharmacist intention, r = .15, Z r = characteristics, including g ender ( r = .02 and Z r = .08), health status ( r = .12 and Z r = .12), drug usage ( r = .14 and Z r = .15), and ad exposure ( r = .23 and Z r = .24). The results showed that consumers who (a) were less educated, (b) had a low income (c) were female, (d) were in bad health, (e) took a lot of drugs, or (f) were exposed to advertising frequently tended to have more favorable attitudes toward DTCA than those who (a) were more educated, (b) had a high income (c) were male, (d) were in good health, (e) took few drugs, or (f) were exposed to advertising less frequently. However, in general, the strength of each of these relationships was small or small to moderate

PAGE 11

11 CHAPTER 1 INTRODUCTION Direct to Consumer Advertising Direct to consumer prescription drug advertising (DTCA) is becoming increasingly common in the USA. DTCA refers to any promotional effort by a pharmaceutical firm to present prescription drug information to consumers v ia the mass media (Wilkes, Bell and Kra vitz 2000). Due to the controversy surrounding D TCA, there are only two developed countries that permit it: the Unite d States and New Zealand (Hoek and Gendall 2002). Among the controversial issues is that DTCA encourages the inappropriate use of medications and drives up dr ug spending (Donohue, Cevasco and Rosenthal 2007). While DTCA was a radical idea little less than 30 years ago, it has grown significantly over the last 3 decades and extended into a variety of health conditions. in an American newspaper (Young 1969), inspiring others to promote their patented m edicines as well (Wilkes et al. 2000). For the next 200 years, patented drugs such as Pectoral Drops, Balsam of Life, and Cordial appeared in newspapers, magazines, and medi cine shows. By the early 1800s, the pharmaceutical industry and the press had developed a mutually beneficial relationship. For example, newspapers received the biggest proportion of their income from drug advertising, and the drug industry spent more to p romote their products than other industries. With the enactment of the Federal Food, Drug, and Cosmetic Act in 1938, the US Food and Drug Administration (FDA) was given the authority to oversee the safety of food, drugs (both prescription and over the c ounter), and cosmetics. However, the Federal Trade Commission (FTC) still had the authority to control drug advertising and medical devices. In 1962, the Kefauver Harris Amendment, also known as the Drug Efficacy Amendment,

PAGE 12

12 introduced requirements for phar maceutical advertising to include, for example, information about the effectiveness, safety, and side effects of drugs, the size of warning/side effect statements (e.g., font size) and the benefits and risk s of taking the drugs (Peltzman 1973). Furthermore the Kefauver Harris Amendment transferred responsibility for prescription drug advertising from the FTC (which still had the authority to regulate advertising for over the counter dr ugs) to the FDA (Calfee 2002). Even though the legal groundwork for drug advertising was laid in 1962, there is no specific point in time when the pharmaceutical industry began to open it to patients/consumers. However, of particular importance, during the late 1960s and early 1970s the FDA utilized the patient package insert (PPI). PPIs, first developed in 1968 ultimately led to DTCA (Pines 1999). The first PPI was designed to promote isoproterenol inhala tion products (Federal Register 1968). Then the concept of the PPI was extended to birth control pills and estrogen replace ment therapy. To provide medical information to patients, during the 1970s, the FDA increasingly and consistently required information about prescription medicine to be provided directly to consumers (Federal Register 1979). Until 1981, the pharmaceutical industry had promoted drugs and medical equipment exclusively to physicians and other health care professionals because only physicians are drugs (Weissman, Blu menthal, Silk, Zapert, Newman and Leitman 2003). In addition, pharmaceutical marketers believed that promoting drugs directly to consumers would be suicidal due to the fear that physicians would never prescribe drugs that bypassed them. Thus, promoting prescription drugs to non medical professionals was inconceivable. Before DTCA was implemented, health care decisions were dominated by physicians because they monopolized the

PAGE 13

13 medical/medicine information. The traditional pharmaceutical marketing program c onsisted of: latest offerings; and (c) the provision to doctors by mail of drug information. In the late 1970s and early 1980s, as the perception that consumers h ad the right to know about drug information for better medical services emerged, pharmaceutical companies sought to promote their products directly to consumers. As part of this tr end a book about prescription drugs the (PDR) became available in consumer bookstores. The PDR was written exclusively for physicians and health care professionals, and from it, consumers were able to gain access to drug information and learn about the effects of drugs. The breakthrough of implemen ting drug advertising came from British based Boots Pharmaceuticals, which advertised its ibuprofen product, Rufen. A second DTCA was undertaken by Merck Sharp & Dohme, which advertised its pneumonia vaccine, Pneumovax in in 1981. The adver tisements for Rufen and Pneumovax raised questions in terms of the potential negative effects of DTCA. For example, physicians at the FDA believed that even though there was a public health benefit in advertising a prescription drug, DTCA could provide inc orrect inform ation regarding drug uses (Hoen 1998). In 1982, the FDA Commissioner Arthur Hull Hayes, Jr. spoke to the Pharmaceutical Advertising Council, summarizing the status of DTCA as the FDA saw it and predicting exponential growth in DTCA. Even thou gh the Commissioner did not intend to advocate for and they began to advertise their products directly to consumers. In September of 1982, the FDA officially requested that the pharmaceutical industry voluntarily avoid advertising its products directly to consumers, and the industry halted the promotion of its products.

PAGE 14

14 Over the next two years, the industry and the FDA investigated the impact of DTCA on consumers, whi ch is considered the early research on DTCA. During the investigation period, only price comparison advertisements were allowed. There were two noteworthy studies that t ended to acquire more information about the positive aspects of products than abou t the negative aspects (Morris and Millstein 1984). Another study showed that consumers preferred prescription drug information in detail and would consider DTCA favorably (M orri s, Brinberg, Klimberg, Rivers and Millstein 1986). The studies of Morris and Millstein (1984) and Morris et al (1986) indicated that the FDA had opened the door for the pharmaceutical industry. However, the FDA felt that DTCA should be controlled and t he pharmaceutical industry needed specific guidance regarding DTCA. On September 9, 1985, the FDA announced guidelines stipulating that DTCAs should be fairly balanced in terms of benef it and risk information (Calfee 2002). Subsequently, pharmaceutical com panies began to more broadly promote their products directly to consumers in print in industry could not utilize television to promote its products because it was i mpossible to meet the approved pr escription information) (Calfee 2002). Specifically, prescription d rug marketers were unable to provide all of the risk information to patients in a limited amount of time. Nevertheless, DTCA consistently increased from $12 million in 1989 to $55 million in 1991, $164 million in 1993, $340 million in 1995, and al most $ 1 billion in 1997 (Wilke 1997a; Wilke 1997b). The number of drugs that utilized

PAGE 15

15 DTCA gradually increased from six in both 1994 and 1995, to nine in 1996, and fourteen in 1997 (Pines 1999). The pharmaceutical industry reached a turning point in the late 199 0s. In January of 1997, then Commissioner Dr. David A. Kessler left the FDA. Since he had opposed television advertising of prescription drugs, his departure from the FDA provided momentum for a new television advertising policy. In August of 1997, the FDA Industry on Consumer t elevision drug advertising (FDA free number to call for more information; (b) concurrent print advertisements containing a brief summary of the risks; (c) Web address where more information about the drug would be available; or (d) specific drug information that could be obtained from pu blicly accessible locations such as hospitals or pharmacies (FDA 1997). During the next two years, the FDA reviewed the policy and further investigated the effects of DTCA on patients, and in August of 1999, the FDA (1999) issued its last guidance on DTCA. Since the FDA had allowed advertising for prescription drugs to be broadcast directly to consumers in 1997, the amount of DTCA for prescription drugs as well as the research on DTCA increased rapidly (An 2008; Jung, Kim and Rhee, 2010). For example, in 19 96, total spending on DTCA in the United States was approximately $1 billion, but that amount increas ed to $4 billion in 2005 (Brett 2007) and $4.3 billion in 2009 (Kaise r Family Foundation 2010). This rate of increase in DTCA spending was nearly 330% from 1996 to 2005, and its average rate of growth was 14.3% from 2002 to 2005 (Donohue, Cevasco and Rosenthal, 2007). And, advertising of prescription drugs (DTCA) represents 60% of the total spending on drug advertising (General Accounting Office 2002).

PAGE 16

16 Purp ose of the Study The dramatic growth of DTCA spending has accompanied a great deal of studies dealing with various aspects of DTCA since the early 1980s. These aspects include the following: policy and r egulation of DTCA (e.g., Calfee 2002; Sheehan 2003; H oek and Gendall 2002), media environment (e.g., Brownf ield, Bernhardt, Phan, Williams and Parker 2004), consumer behavior (e. g., Deshpande, Menon, Perri III and Zinkhan 2004; Yuan 2008; Hau sman attitudes t oward DTCA (e.g., Paul, Handlin and Stanton 2002), economic impact of DTCA (Kopp and Sheffet 1997) and advertising content (e.g., Kaphingst, Dejong, Rudd and Daltroy, 2004). Among other DTCA topics, the most controversial and most frequently researched issue in the DTCA literature is the effects of DTCA, especially with respect to the outcomes we can expect via DTCA and some constructs that affect the effects of DTCA. Table 1 1 displays the summary of the literature on DTCA effects. Despite the large volume of research in the area of DTC A, the findings in terms of ad effects have been inconsistent. For example, while some studies have reported no evidence of a significant effect of DTCA on con sumer behavior (e.g., Williams and Hensel 1995), others have reported a significant effect (e.g., An 2007). Furthermore, different studies have found widely varying magnitudes of DTCA effect on consumer behavior. This suggests that DTCA research should first develop an understanding of the nature of the underlying relationship between advertising outc omes and antecedents such as A ad in order to determine whether the patterns of these relationships are consistent or inconsistent across other independent studies. Furthermore, there should be an attempt to systematically review DTCA related research. To u nderstand thoroughly what previous research has revealed regarding the role of DTCA, both qualitative and quantitative research can be used. However, a qualitative review is problematic in that it can neither account for the quality of the study nor for th e issues of the

PAGE 17

17 eff ect size of each study (Capella 2005). Thus, a meta analysis that statistically cumulates empirical findings is a better method of reviewing the research results. The current research employs a meta analytic methodology using quantitativ e statistical results from the DTCA Given the amount of speculation concerning the role of A ad and the amount of research that has been devoted to the topic, it is useful to consider what is known about the concept and to identify the most fruitful avenues for future research. Accordingly, the present study aims to address several issues. First, it examines studies on the antecedents of A ad in DTCA (e.g., gender, age, ethnici ty, media exposure, health status / involvement, ad awareness). Second, it examines studies on the consequences of A ad in DTCA (e.g., intention to ask doctors to prescribe the advertised drug, intention to find more information regarding the advertised dru g, intention to ad Fourth, the study examines the potential variables that moderate the relationships between A ad and its antecedents and consequences. Finally, it offers s uggestions for future research.

PAGE 18

18 Table 1 1 Summary of DTCA Effects 1. DTCA Effects on Patients Relevant Literature a. intentions to request that physicians prescribe the advertised drug An 2007; Mehta and Pruvis 2003; Hausman 2008 b. intentio ns to ask physicians for more information about the advertised drug An 2007; Herzenstein, Misra and Posavac 2005 c. intentions to search for information about the advertised drug Herzenstein, Misra and Posavac 2005; Morris et al 1986 d. intentions to di scuss symptoms with physicians Miller and Blum, 1993; Yuan 2008 e. advertising recall Baca et al. 2005 f. attitude toward the brand Hausman 2008 g. attitude toward the DTCA regulations/policies Huh et al. 2004 2. DTCA Effects on Physicians a. phy sicians' attitudes toward DTCA Paul et al. 2002; Petroshius, Titus and Hatch, 1995 b. drug prescription writing Bell, Wilkes and Kravitz 1999; Wilkes et al. 2000 c. education Myers et al. 2006 3. DTCA Effects on Physician Patient Relationship a. hori zontal relationship Berger et al. 2001 b. lengthened clinical encounter Lipsky and Taylor 1997, Robinson et al. 2004 4. DTCA Effects on the Pharmaceutical Industry a. market expansion Iizuka 2004

PAGE 19

19 CHAPTER 2 LITERATURE REVIEW The pharmaceutical indus try is one of the most advertisin g intensive industries (Brekke and Kuhn 2006). However, advertising pharmaceutical products has generated a controversial debate since its inception. For example, DTCA proponents claim there is an educational value to DTCA available medical treatment options. Advocates also assert that DTCA may enable patients to notice a disease in the early stages. By contrast, opponents argue that DTCA encour ages inappropriate use of medications and drives up drug spending and price. Since most DTCA fails to inform consumers of the potential (side) effects of drug mis and over use or to provide directions for adequate usage, it can be considered as disseminat ion of improper information about the potential and foreseeable risks connected to prescription drugs; it also overstates the efficacy of th e advertised drugs (Lee, Salmon and Paek 2007). Opponents are also concerned about whether FDA guidelines for DTCA a re strict enough to ensure that consumers are informed about all relevant drug information. Ironically, such criticism is evidence that DTCA has a significant effect on consumer and physician behavior. Furthermore, more and more empirical evidence has been collected of the effect that DTCA has on consumer knowledge, awareness, and attitudes toward DTCA, as well as behavior related to health care treatment. The significant effect of DTCA has led many researchers to contribute to the debate surrounding DTCA. A number of empirical studies have addressed various aspects of DTCA. Much of the previous research has focused on governmental reg ulation and policy (e.g., Green 1995; Reichertz 1996; Statman and Tyebjee 1984 ), DTCA industry and manag ement issues (e.g., Leffler 1981; Rheinstein 1982), or DTCA effects (e.g., Alperstein and Peyrot 1993), prior to

PAGE 20

20 1997 when the FDA allowed pharmaceutical companies to advertise prescription drugs on television (Jung, Kim and Rhee 2010). In addition, questions related to what DTCA does or what its effects are have often been investigated in extant DTCA literature. As DTCA is a common phenomenon, greater attention has been paid to identifying the variables that predict the desired behavioral outcomes, which is the ultimate goal of the marketing efforts for pharmaceutical products. As a result of prior research on DTCA, some variables have been identified that anticipate the likelihood of such behavior in the context of DTCA, such as A ad (e.g., Mehta and Purvis 2003). Not surpris ingly, A ad is the most frequently employed variable in DTCA related research to pred ict consumer behaviors (Wilson and Till 2007) because A ad is considered the best indicator of ad vertising effectiveness (Haley and Baldinger 1991). Despite its importance, there has not been a comprehensive attempt to evaluate the general findings across independent studies. Moreover, several research findings on the relationship between A ad and its outcome variables vary in terms of the strength and direction of the relatio nships. Therefore, evaluating the general findings across independent studies will be useful in understanding the general variability and strength of the relationships and the research conditions (e.g., methodological and research environment differences) that moderate those relationships. This is because A ad related studies have been conducted in various methodological contexts, but there has been no attempt to assess the robustness of A ad effects across different methodological conditions (Rao and Monroe 1989). The current study begins with a review of the literature regarding A ad in the context of DTCA and then a review of the preceding variables and outcome variables and other moderators, using a meta analysis technique. The results of this study help de termine the strength and direction of

PAGE 21

21 relationships between A ad the preceding variables, and the outcome variables. Figure 1 Table 2 1 and 2 2 provide a summary of the hypotheses and research questions. Effect of the Attitude t oward the Ad The concept of attitude toward the ad (A ad ) has been subject to a great deal of empirical study in the context of DTCA, both as an antecedent and a consequence of other advertising related variables of interests. In spite of the importance of A ad no one has comprehensi vely and systematically attempted to assess the empirical findings across other independent studies related to DTCA and A ad As other researchers have emphasized meta of interest in a meta analysis concern the robustness of the relationships studied and the specification of conditions that limit these relations 1992, p. 35). This study is the first to review and analyze previous DTCA related research findings in terms of the relation ad and its antecedents and consequences. In the current study, the variability and strength of the relationship between A ad and its antecedent and outcome constructs will be investigated using a meta analysis techni que. To understand the relationship between A ad and advertising effects, it is necessary to explore the concept of attitude and identify the roles of attitude in general. The concept of attitude has played a critical role in the fields of psychology and e ducation to understand human thought and behavior. Since 1974, more than 34,000 published studies have addressed attitudes in some way (Kraus 1995). The roles of attitude that researchers have identified are that they, in some way, influence, direct, guide or predict actual behavior, and researchers have taken much interest in this attitude behavior relationship. In the fields of advertising and marketing, researchers have applied the concept of attitude to advertising, created the concept of attitude tow ard the ad (A ad ), and tested whether the

PAGE 22

22 role of A ad is similar to attitude in general. A ad one of the most important constructs in advertising research, favorability/unfavorability tow ard 1986 p. 130) The general importance and construct of A ad was introduced by Shimp (1981), who viewed A ad as an important mediator of brand choice (also see Mitchell and Olson 1981). Furthermore, a great deal of advertising research has investigated the roles of A ad in determining advertising outcomes, because brand attitudes and behavioral intentions are functions of A ad in general (e.g., MacKenzie et al. 1986; MacKenzie and Lutz 1989). Shimp (1981) proposed the Three Alternatives Brand Choice Mechanisms, suggesting attitude transfer from advertising to brand which culminates in brand choice. MacKenzie et al. (1986) also proposed the Four Alternative Structural Specifications of the mediating role of A ad that are causal models derived from A ad research as a mediator: affect transfer model, dual mediating model, reciprocal mediation model, and independent influences model. Although there are several models to explain the role of A ad they have something in common i n that A ad is viewed as an affective construct and an influence on intentions (Homer 1990). DTCA researchers have incorporated the concept of A ad in terms of how A ad affects tionship between A ad and intentions. However, in the context of DTCA, it should be noted that because it is impossible that a consumer will purchase certain drugs wi thout a prescription. Instead of measuring purchase intentions directly, DTCA research to date has investigated various types of behavioral intentions as outcomes of A ad such as intentions to request that physicians prescrib e the advertised drug (e.g., An 2007 ; Mehta and Purvis 2003; Hausman 2008),

PAGE 23

23 intentions to ask physicians for more information abou t the advertised drug (e.g., An 2007; Herzenstein, Misra and Posavac 2005), intentions to discuss symptoms with physicians (e.g., Miller and Blum 1993; Yuan 2008), and intentions to visit thei r physicians(e.g., Gonul, Cater and Wind 2000). In addition to the effects of A ad on intentions, another marketing variable, brand attitudes, was inve stigated as well (e.g., Hausman 2008). In sum, the earlier foundationa l theoretical models in ad attitude research would lead one advertising, as well as attitudes toward certain branded products. In other words, it could be argued that if consumers have positive attitudes toward DTCA, they are more likely to adopt the specific advertised brand and vice versa. Based on the majority of past research and attitude/A ad brand attitude behavior models, it is hypothesized that A ad is a sign ificant predictor of behavioral intentions and attitude toward the brand in the context of DTCA. Accordingly, H 1 and H 2 are used to demonstrate this relationship in the model from Figure 2 1. H 1 : In extant DTCA literature, attitude toward the ad is positi vely related to behavioral intentions (e.g. drug request intention, drug inquiry intention, drug information search intention, and physician visit intention). H 2 : In extant DTCA literature, attitude toward the ad is positively related to attitude toward th e brand. Factors Affecting Attitude toward the Ad The previous section reviewed the effects of A ad on behavioral intentions in the context of DTCA. The extant literature on A ad posits that advertising may influence consumer behavior by influencing the con sumers attitudes toward the ad. The question that naturally arises is What determines A ad ? T he results of hypotheses and research questions will be useful for pharmaceutical marketers to develop market segmentation strategies. F or example, several

PAGE 24

24 studi es investigated the relationships among demographic characteristics, A ad and behavior. Pharmaceutical marketers can use the findings of those studies to decide on the target audiences that are suited to their marketing purposes. I n the next section, the l iterature on the factors affecting A ad will be reviewed. Ad Awareness and Ad Exposure The effect of ad awareness on A ad and other outcomes can be explained by early experimental research on the mere exposure effect, referring to the idea that the repetit ion of ads leads individuals to give a more positive evaluation of an advertised product/brand and an advertisement itself (Tellis 1988). R esearch on the mere exposure effect has emphasized the power of familiarity (i.e., exposure to advertisements increas es familiarity with them). Zajonc (1980) argues that familiar objects tend to be preferred over unfamiliar ones. A number of researchers have found support for the familiarity effect such that familiar brands/ads were more likely to produce favorable attit udes toward the ad / brand than unfamiliar ads / brands ( Campbell and Keller 2003; Dahlen 2001 ; Janiszewski 1993 ; Pechmann and Stewart 1990 ). An interesting finding in the literature of mere exposure effect is that mere exposure even to nonsense photos or words generates positive attitudes (Zajonc 1980). Using the elaboration likelihood model, Petty and Caci o ppo (1979) also insist that consumers in some situations are often reliant upon simple heuristics (e.g., brand familiarity) when making purchase decisi ons, and it is well known that there is a high multicollinearity between ad exposure and ad awareness. T herefore, it is hypothesized that ad awareness and ad exposure are predictors of A ad in the context of DTCA. Several studies have addressed various pre ceding variables of ad awareness such as medication condition, drug use, gender and age (e.g., Perri and Nelson 1987; Bell et al. 1999;

PAGE 25

25 Huh and Becker 2005). Those studies found that higher levels of ad awareness were related to older people, regular drug users, females, and those who are concerned about their health. H owever, other studies revealed somewhat different findings (e.g., Alperstein and Peyrot 1993). F or instance, they found that young people are more aware of the advertised drugs. In addition t o the problem of inconsistent results regarding the preceding variables of ad awareness, theoretical studies of ad awareness in the DTCA literature are scant. T hus, antecedents of ad awareness and their strength of the relationships will be investigated. A s such, the relationships are depicted as H 3 and RQ 1 a and b in the model from Figure 2 1. H 3 : Ad awareness and ad exposure are positively related to A ad in extant DTCA literature. RQ 1 a: Which preceding variables of ad awareness have been investigated in extant DTCA literature? RQ 1 b: How strong are the relationships between the preceding variables and ad awareness in extant DTCA literature? Demographic Factors D emographic factors such as age, gender, and ethnicity have been investigated to determine whet her they affect A ad For example, age has been tested in terms of its impact on the degree to which consumers have favorable/unfavorable attitudes toward DTCA. Several studies have tested the role of age and found that older consumers tend to have more fav orable attitudes toward DTCA (Williams and Hensel 1995; Menon, Deshpande, Zinkhan and Perry 2004; Baca, Holguin and Stratemeyer 2005). Gender difference is one of the main research topics in the field of women s studies. Metha and Purvis (2003) investigate d the role of gender and found that women valued DTCA more than men. T he impact of ethnicity has been examined in the context of DTCA as well (e.g., Huh and Becker 2004). In addition to the role of demographic factors as

PAGE 26

26 preceding variables of A ad some st udies (e.g., Peyrot et al. 1998) have argued that demographic factors bypass A ad and directly affect consumer behavior. Although the relationships between A ad and demographic factors have been examined in several studies, little is known about the effects of demographic factors on A ad and behavior. This is because most researchers have included demographic variables in DTCA studies more as descriptive statistics than as explanatory variables. That is, there has been relatively little theoretical work aimed at explaining why demographic variables should be related to A ad in the DTCA literature. To find the common patterns of the relationships among demographic factors and Aad, the present study investigates the effect sizes of each study regarding the relati onships. These relationships are shown as RQ 2 and RQ 3 in the model from Figure 2 1. RQ 2 : In extant literature how strong is the relationship between demographic factors (e.g., age, gender, and ethnicity) and A ad ? RQ 3 : In extant DTCA literature how strong is the relationship between demographic factors and intentions to request the advertised drugs, intentions to ask for more information about the advertised drugs, intentions to search more information, and intentions to visit physicians? Health Status and Involvement If consumers are concerned about their health, they may process arguments in DTCA carefully because they are highly involved with the ad arguments. According to Petty and an olvement level determines his or her use of either the central route or the peripheral route. Involvement is and Olson 1988). The ELM consists of a central route and a peripheral route. The central route is

PAGE 27

27 characterized by considerable cognitive elaboration. This occurs when individuals focus in depth on the central features of an issue, person, or message. When people process information centrally, they carefully eval uate message arguments, ponder the implications of the communicator s ideas, and relate information to their own knowledge and values. H owever, the peripheral route is entirely different. Rather than examining issue relevant arguments, people examine the m essage quickly or focus on simple cues to help them decide whether to accept the argument. T he simple cues can include a communicator s physical attractiveness or speaking / writing style or a pleasant association between the message and the music playing in the background. The ELM posits that under high involvement conditions, people use a central route to process information, whereas under low involvement conditions, they use a peripheral route to process information. Based on the ELM, it is hypothesized that a consumer s health condition and involvement with a medical condition are significant predictors of A ad in the context of DTCA. In addition, it will be examined whether the health status factor directly affects consumer behavior. Accordingly, H 4 and RQ 4 are used to demonstrate this relationship in the model from Figure 2 1. H 4 : In extant DTCA literature, the consumer s health status and involvement with medical conditions are related to the consumer s A ad ; specifically, consumers who have (perceived) adverse health conditions and who are highly involved with a medical condition, have more favorable attitudes toward the ad than are consumers with good health conditions and low involvement.

PAGE 28

28 RQ 4 : In extant DTCA literature, how strong is the relationshi p between the health status and intentions to request the advertised drugs, intentions to ask for more information about the advertised drugs, intentions to search more information, and intentions to visit physicians? Other Factors Affecting Attitude towar d the Ad Other factors, such as education and income, have rarely been examined to determine whether they influence A ad or behavior in the context of DTCA. The current study will attempt to explore other factors affecting A ad or behavior. RQ 5 : Which preced ing variables of A ad or behavior besides demographic factors, health status/involvement, and ad exposure have been investigated in extant DTCA literature? RQ 6 : How strong are the relationships between the preceding variables and A ad or behavior in extant DTCA literature? Potential Moderating Variables As noted before, the primary purposes of a meta analytic study are to assess the strength of the relationships and specific conditions that limit the generalizability of these relationships. The current stud y assesses the robustness of the relationships between A ad and its preceding and outcome variables in extant DTCA literature. H owever, in terms of research methods and environments, this particular research stream encompasses diverse studies. This suggests that the methodological decisions might influence the robustness of pairwise relationships. M any meta analysis researchers have provided useful guidelines on how to code study characteristics for moderator analyses (e.g., Hedges and Olkin 1985; Rosenthal 1984). A s is typical in meta analytic studies, research characteristics will be investigated as to whether they moderate the advertising effects in extant DTCA literature. Coding for the research characteristics will include

PAGE 29

29 the following: (a) type of samp le, (b) type of research methodology (c) measurement of constructs, and (d) theoretical basis (if any). Sample The use of student samples has been a subject of debate in quantitative research (Calder, Phillips, & Tybout, 1981). The type of study subject ( student or not) often functions as a moderator because the homogeneity of the student sample may produce strong bias effects that are not typically found in the general population and which culminate in a bias toward stronger effects. H 5 : The type of samp le moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by student samples than by non student samples. Type of Research Due to the vulnerability of the use of student sample and the reliability issues of measuring instruments, the survey method tends to be more variable across independent studies with regard to research findings as compared to the experiment method. H 6 : The type of research moderates the stre ngth of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by survey studies than by experimental studies. Measurement of Constructs To measure A ad ad awareness, and intentions, some studies have used multi item scales, whereas others have used a one item scale. The analysis of the number of scale items has been suggested in a meta analytic method because multi item scales tend to be more reliable and

PAGE 30

30 sensitive in general. T he cur rent study expects that multi item scales may lead to greater effect sizes due to less attenuation from measurement error (Johnson and Eagly 1989). H 7 : The number of scale items (one item scale vs. multi scale item) moderates the strength of pairwise relat ionships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by studies using multi item scales than by studies using one item scales. Theoretical Basis According to Farley and Lehman (1986), theory driv en studies tend not to have spurious effects. Therefore, it is necessary to examine the theoretical foundations of each independent study to analyze the study quality. H 8 : The theoretical foundation moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by theory driven studies than by non theory driven studies. Other Research Characteristics In addition to the research characteristics addressed above, other c haracteristics will be investigated to determine whether they moderate the strength of pairwise relationships studied in the context of DTCA. RQ 7 : Which research characteristics moderate the strength of pairwise relationships studied in extant DTCA litera ture? To assess the impact of each research characteristic on the strength of pairwise relationships, the studies will be separated into two subsets. An overall and a subset mean effect size will be calculated for each characteristic.

PAGE 31

31 Chapter Summary The chapter 2 literature review provided an overview of the background knowledge and critical information regarding the research constructs (A ad antecedents, and consequences ) of the current study. A theoretical background for the A ad outcome relationships w as presented first with MacKensie et al. s model (1986). Then a discussion of the theoretical and empirical background of A ad antecedents relations followed. From the literature on DTCA, this study tests the strength of the relationships between A ad and ( a) the intention to ask doctors to provide more information about the advertised drug, (b) intention to ask doctors to prescribe the advertised drug, (c) intention to consult doctors to discuss their symptoms, (d) intention to visit physicians, and (e) att itude toward the brand. The DTCA literature suggests that a variety of factors have an impact on A ad This study also investigates the strength of the relationship between A ad and 1) demographic variables (age, gender, and ethnicity), 2) health status/invo lvement, 3) ad exposure/media consumption, 4) ad awareness and 4) other potential preceding variables. Some studies (e.g., Peyrot et al. 1998) have argued that preceding variables such as demographic factors, health status/involvement, media exposure, and ad awareness bypass A ad and directly affect behavioral intentions. The current study analyzes the strength of the relationships between preceding variables and outcome variables (e.g., intentions). The overview of the literature review led to the develo pment of the hypotheses and research questions. This study reports the findings in two sections. First, it reports the results for consistencies and the strength of the relationships with A ad If the results are consistent, moderators do not exist, and if they are inconsistent, moderators exist. Second, moderator analyses are conducted, and the study reports the results of the analyses.

PAGE 32

32 Table 2 1 Summary of Hypotheses Hypotheses H 1 : In extant DTCA literature, a ttitude toward the ad is positively relate d to behavioral intentions (e.g. drug request intention, drug inquiry intention, drug information search intention, and physician visit intention ). H 2 : In extant DTCA literature, attitude toward the ad is positively related to attitude toward the brand. H 3 : Ad awarenes s and ad exposure are positively related to A ad in extant DTCA literature. H 4 A ad ; specifically, consumers who ha ve (perceived) adverse health conditions and who are highly involved with a medical H 5 : The type of sample moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be dete cted by student samples than by non student samples. H 6 : The type of research moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by survey studies than by experimental studies. H 7 : The number of scale items (one item scale vs. multi scale item) moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by studies using multi item scales than by studies using one item scales. H 8 : The theoretical foundation moderates the strength of pairwise relationships studied in extant DTCA literature; specifically, a stronger relationship is more likely to be detected by the ory driven studies than by non theory driven studies.

PAGE 33

33 Table 2 2 Summary of Research Questions Research Questions RQ 1 a : Which preceding variables of ad awareness have been investigated in extant DTCA literature? RQ 1 b: How strong are the rela tionships between the preceding variables and ad awareness i n extant DTCA literature? RQ 2 : In extant literature how strong is the relationship between demographic factors (e.g., age, gender, and ethnicity) and A ad ? RQ 3 : In extant DTCA literature how str ong is the relationship between demographic factors and intentions to request the advertised drugs, intentions to ask for more information about the advertised drugs intention s to search more information, and intentions to visit physicians ? RQ 4 : In extan t DTCA literature, how strong is the relationship between the health status and intentions to request the advertised drugs, intentions to ask for more information about the advertised drugs intention s to search more information, and intentions to visit ph ysicians ? RQ 5 : Which preceding variables of A ad or behavior besides demographic factors, health status/involvement, and ad exposure have been investigated in extant DTCA literature? RQ 6 : How strong are the relationships between the preceding variables and A ad or behavior i n extant DTCA literature? RQ 7 : Which research characteristics moderate the strength of pairwise relationships studied in extant DTCA literature?

PAGE 34

34 Ad Attitude Brand Attitude Behavioral Intention H2 H1 Ad Awareness H3 I nvolvement / Health Condition H4 Consumer Characteristics RQ1 RQ2, 5, 6 RQ3 and 6 RQ4 Figure 2 1. Hypothesized Model of DTCA Effects Coded Study Charac teristics : Sample (H 5 ), Type of Research (H 6 ), Measurement of Constructs (H 7 ), Theoretical Basis (H 8 ), Other Research Characteristics (RQ 7 )

PAGE 35

35 CHAPTER 3 METHOD Meta Analysis Many researchers (e.g., Glass 1977; Rosenthal 1984) have argued that research activi ties in the social sciences, including mass communication are in crisis While the n atural sciences are based on standardized and commonly accepted techniques and methods this is not the case in the social sciences E xplaining and understanding human beha vior is a very difficult task due to it s complexity, research environments are difficult to control, and sampling characteristics, methods, and techniques differ across studies (Wolf 1986). For example, many studies have examined advertising effects in the context of DTCA. These studies not only use different definitions, variables, procedures, methods, samples, and so on, but their conclusions are often at odds with each other. It is possible that research environments in the social sciences lead to an en ormous waste of scholarly effort (Kulik 1983) because those scholarly efforts cannot be cumulated. I f we view science as the accumulation of information and knowledge (Hunter et al. 1982), to be considered a science, it is important to cumulate previous st udies. I t then becomes critical to establish guidelines for reliable and valid reviews, integrations, and syntheses of studies examining similar research questions (Jackson 1980). The principles of meta analysis provide the guidelines for cumulating previo us studies. In addition to the environmental problems of the social sciences, there is an important perspective to adopt when reviewing previous literature According to Kirk (2001) social scientists are interested in answering three basic questions when examining the relationships/differences between variables/groups: (a ) statistical significance (b ) effect size and (c ) practical significance A statistical ly significant outcome indicates the likelihood that there is

PAGE 36

36 a real relationship between variabl es. In other words, the level of significance ( p value) indicates the probability that an outcome could happen. However, a very small p value does not imply an important finding in any practical sense. If the sample size is very large, a small p value can occur even though the difference between groups is small. In this case, the difference may not be significant in practical terms. The use of an effect size is a good way to overcome the limitation of the statistical significance test Meta analysis is the principal method that I will use for my dissertation because the main purpose of my dissertation is to review previous research on direct to consumer advertising of prescription drugs. To thoroughly understand what previous research has revealed, both qua litative and quantitative review s can be used. However, a qualitative review, in counter distinction to a meta analysis, is problematic in that it cannot account for the quality of the study or for issues of effect size for each study. Therefore, a better method of summarizing research results is to conduct a met a analysis, which statistically, or quantitatively, ac cumulates research findings because the procedures employed in meta analysis permit quantitative reviews and synthes e s of the research literatur e that address es these issues. This cumulative view of previous research provides an opportunity to draw a big picture in a research enterprise. Meta analysis techniques have been also used to test certain hypotheses regarding the concept of A ad I n this f ashion, meta analysis is used not only as a summary device, but also as a hypothesis testing technique. Meta Analysis Process Meta analytic studies are rapidly growing in social sciences, and they follow three steps: database creation, data conversion, and the analysis method.

PAGE 37

37 T he first step in meta analytic research is to collect relevant studies and to extract information to build a database of research findings across independent studies related to a research topic. T he focus is usually the bivariate re lationships between the constructs of interest. According to Lipsey and Wilson (2001), there are several eligibility criteria for studies, that can be included in meta analysis. Studies should have distinguishing features. Since meta analysis is a techniqu e for testing statistical results across independent research findings related to the same topic, any papers lacking an element of study interest should be eliminated. A meta analytic review combines the research findings of studies to evaluate the magnitu de and significance of diverse measures of effect sizes (Fern and Monroe 1996). To be meta analyzed, a study should contain statistical information (e.g., t F p value, chi square, correlation coefficient) sufficient to calculate an effect size, because meta analysis cannot be used to summarize qualitative studies such as theoretical papers If a study lacks statistical information, the author can either calculate or estimate the required statistics using the information presented in the studies or conta ct the authors of the studies to obtain the missing statistical information. If neither way is available, these studies will be eliminated. Field (2001) stated, In meta analysis, the basic principle is to calculate effect sizes for individual studies, con vert them to common metric, and then combine them to obtain in average effect size (p. 162). The second step of meta analytic research is to transform the collected statistical information into a standardized form, if needed. The third step in meta analys is is to analyze the obtained statistical information. Saxton (2006) argued that the goal of meta analysis is to test whether the obtained results from independent studies confirm/disconfirm the strength and direction of the relationship between variables. In other words, meta analysis treats data from independent studies as a part of one study. F urthermore, the analysis of data from multiple studies is conducted by calculating effect sizes.

PAGE 38

38 Step 1: Database Development All extant literature on the effects of DTCA was reviewed and synthesized in this dissertation. For the purpose of this review and synthesis, t he current study meta analyze d the range of articles that deal with DTC A and appeared in U.S. and international journals from 1981, when Leffler s (1 981) DTCA study appeared to issues current at the time when the present analysis was conducted (2011). Advertising is an applied field; thus both marketing and communication have contributed to advertising research. In addition to marketing and communicati on, other fields like medicine, medical science, and/or law have investigated the role that DTCA plays in each field because of its inter disciplinary nature. Thus, this study analyzes DTCA related papers for a wide variety of fields (e.g., advertising, ma rketing, communication, health communication, health care, and law). Studies for inclusion in the review were identified through computerized searches using the following: ( a ) Journal of American Medical Association & Archives ( JAMA & Archives ), ( b ) Ebsco Source Premier ( c ) PubMed Central ( d ) Science Direct ( e ) Springer Link ( f ) JSTOR ( g ) ProQuest ( h ) Wiley InterScience (i) Business Source Premier (j) Academic Search Premier (k) Social Sciences Citation Index, and (l) PSYCHLIT Additionally, Googl e Scholar was used to search articles using the following search term s: advertising of prescription drugs, DTC, DTCA, pharmaceutical advertising pharmaceutical promotion, prescription drug advertising, promotion of prescrip tion drugs, drug promotion, advertising of prescription drugs, and drug advertising. The search terms were based on the keywords that each author listed for his or her article. For example, Bell et al. (1999) used prescription drugs and drug prom otion as keywords for their study. After looking into all search terms listed in the literature on DTCA, the above search terms were developed.

PAGE 39

39 In addition to the computer search, a manual search was conducted in two fashions. First, a fter reading the s elected articles, the reference lists of previous reviews of DTCA literature were also screened to ensure a complete review Second, an article by article search of 21 academic journals was performed. The following journals were included for the search: Jo urnal of Medical Marketing, Journal of Clinical Oncology, Journal of Consumer Marketing, Journal of General Internal Medicine, Health Care Management Science, Journal of Advertising, International Journal of Advertising, Journal of Health Communication, Pu blic Health Care Report, Health Marketing Quarterly, Journal of Advertising Research, Journal of Health Care Marketing, American Journal of Health Behavior, Journal of Business Ethics, Research in Social and Administrative Pharmacy, Journal of Health and H uman Services Administration, Journal of Product & Brand Management, Annals of Family Medicine, Health Affair, Health Communication, and Communication Research All referenced papers in these published articles were included in the study. The search proces s yielded 36 articles that presented empirical findings that could be used in the meta analysis. N o restrictions were place on the inclusion of the studies other than that they must have analyzed DTCA effects. The studies included contained correlations be tween A ad and a wide array of variables. All relevant studies were coded into three separate databases: A ad its antecedents relationships, A ad its consequences relationships, and other relationships. The separate coding processes were based on the researc h agenda of the current study. First, this study hypothesized that A ad is a significant predictor of intentions and brand attitude. Second, research questions were developed about the relationship between A ad and its antecedents. Lastly, in addition to loo king at A ad related pairwise relationships, the current study also examined at the relationships

PAGE 40

40 between antecedents of A ad and consequences of A ad For the purpose of analyzing data, the databases were conveniently divided into three main sections. (1 ) A ad its antecedents relationships The total sample size across independent studies was 51 reported correlation coefficients or other statistics that could be transformed into correlation coefficients. The correlations were collected from studies publish ed in peer reviewed journals between 1991 and 2009. T hese studies reported many preceding variables including age, gender, ethnicity, education, health status, involvement, ad exposure, ad awareness, drug usage, and income. (2) A ad its consequences relatio nships The total sample size across independent studies was 51 reported correlation coefficients or other statistics that could be transformed into correlation coefficients. The correlations were collected from studies published in peer reviewed journals between 1995 and 2007. These studies reported a number of outcome variables including behavioral intentions. (3) Other relationships The total sample size across independent studies was 107 reported correlation coefficients or other statistics that could be transformed into correlation coefficients. The correlations were collected from studies published in peer reviewed journals between 1986 and 2010. These studies reported not hypothesized variables. Step 2: The Conversion The term effect size refers to t he magnitude of effect observed in a study. The effect size can be represented in two ways: standardized mean difference (d) and correlation coeffi cient (r) (Hedges and Olkin 198 1). The standardized mean difference effect sizes ( d ) are devised to calculate the degree of difference between group means. In other words, it is often used extensively for expressing and combining the results of studies that assess the effectiveness of an

PAGE 41

41 experimental treatment. The correlation coefficient effect sizes ( r ) are de signed to calculate the size of a relationship between variables. In meta analysis, an estimate of effect magnitude is obtained for each study. I ndividual effect magnitude estimates can then be averaged to obtain an overall estimate of effect magnitude. Before conducting a meta analytic study, a researcher must determine the effect size metric to use (e.g., d or r ). Next, statistical information has to be converted into the common standardized form (e.g., d or r ). The effect size used in the current resea rch was the correlation coefficient (r), which is the square root of the variance explained by a given variable or combination of variables (Rosenthal 1994). The correlation coefficient was chosen as the measure of effect size because it is easy to compute from a t or F statistic and easy to interpret (Janiszewski, Noel and Sawyer 2003). This study utilized correlation coefficient values for data analyses. H owever, it is well known that correlation coefficients are not normally distributed. I t is, therefor e, conventional in meta analysis to convert correlations to z scores using Fisher s r to z transformation Z r = .5 [ ln (1 + r ) ln (1 r ) ], where ln( x ) is the natural logarithm function. This study reports both r and z and assesses whether there is a difference between the two values. In order to synthesize the empirical findings across independent studies, it was needed to convert all test statistical information to a standardized form, r Moreover, statistical tests such as t tests, F tests, chi squ are statistics, and p value are not effect sizes because for any given effect, their value increases as the sample size increases (Rothstein, McDaniel and Borenstein 2002). Therefore, when necessary, the statistical information in primary studies was conve rted into the correlation coefficient effect size (Arthur, Jr., Bennett, Jr. and Huffcutt 2001). Examples of equations for transformation to r are illustrated in Table 3 1

PAGE 42

42 Step 3: Method of Analysis A number of relationships pertinent to the hypotheses a nd research questions were examined. Once transformed into the common effect size metric ( r ), the individual effect sizes for independent studies can be synthesized to obtain an average or pooled effect size To calculate the pooled effect size, the effect where the results of studies that have large sample sizes receive more weight. To test the calculation of the aggregated effect size, significance in meta analysis is generally gauged by If the confidence interval does not include zero, the effect size is significant. After combi ning the effect sizes, proposed Hunter and Schmidt (1990) moderator analyses are necessary. Moderator analyses are helpful to gain additional insight into the research pairwise relationships and to refine the strength/direction of the relationships. In the current study, moderator analyses were performed by categorizing the total sample into subgroups based on the following research characteristics: (a) type of sample, (b) type of research methodology, (c) measurement of construct, (d) theoretical basis (if any), and (e) other coding categories. (1) Type of sample. The samples of each study will be coded in terms of student vs. non student. (2) Type of research. Meta analysis cannot be used to summarize qualitative studies such as theoretical papers. First, to sort out meta analyzable studies, research papers will be coded in terms of qualitative vs. quantitative. Second, quantitative studies will further be categorized into two methodologies: experiment and survey. (3) Measurement of construct. The measureme nts employed in each study will be coded in terms of single item scale vs. multi item scale. Each relationship contains two constructs. If both

PAGE 43

43 constructs employed a multi item scale, the relationship was coded as a multi item scale relationship. Other tha n that, all relationships were coded as single item scale relationships. (4) Theoretical basis. The theoretical foundation, or lack thereof, will be coded if any theory is employed in a study. (5) Other research characteristics. In addition to the researc h characteristics addressed above, other characteristics such as the year of publication, effect size estimation (no estimation vs. estimation), product category (inductively categorized), funding source (government and private), and field of publication ( marketing, communication, advertising, medical science, and other). Rater Reliability In meta analytic studies, the assessment of rater reliability is of great importance to critics, reviewers, and general readers. O bservations and coding of the study c haracteristics should be reliable and objective. Accurate coding is undoubtedly crucial to the conclusions generated from any synthesis of study effect sizes. Thus, it was vital to perform an interrater reliability check on the accuracy of the coding proce ss. Interrater reliability refers to the degree of agreement or consistency that exists between two or more raters (Klein 2000). The reliability of raters is assessed to determine whether the coding process is free from bias or error. In most meta analyt ic studies, two or more raters code most (or all) of the available studies on the basis of coding schemes. For the current research, another rater was recruited to rate a subset of the articles in the database, and this rater was trained on the coding cate gories that were investigated in the current research. The subset of ratings was performed on six articles chosen by the author from the complete database of coded articles. This subset represents 16.7% of the final database. I nterrater reliability was def ined as the frequency of agreement on codes divided by the total number of coded categories per section, expressed as a percentage.

PAGE 44

44 Reliability was calculated on the four coding categories listed in Table 3 2 The levels of agreement between the primary an d second coders were generally high, with a mean overall agreement of 98.2%. Fixed Versus Random Effect Models Meta analysis is used to obtain the true effect sizes in a population by combining effect sizes from independent studies. T here are two ways to a scertain the true effect sizes: fixed effects and random effects models (Hedges 1992; Hedges and Vevea 1998). In the fixed effect model (also called the homogeneous case), the true effect sizes in the population are fixed but have unknown constants. It is assumed that the effect size in the population is the same for all studies included in a meta analysis. Another model for the effect size synthesis is the random effect model. T he assumption of the random effect model is that the population effect sizes va ry randomly from study to study because every study in a meta analysis comes from a population that is likely to have a different effect size to any other study in the meta analysis. This is also referred to as the heterogeneous case. P opulation effect siz es are assumed to be sampled from a universe of possible effect, which is called a super population (Hedges 1992, Hunter and Schumidt 2000). In statistical terms, the choice of one model over another makes a difference in the calculation of the standard errors associated with the combined effect size. O ne of the main differences between the two models is that the fixed effect model is appropriate only for conditional inferences, whereas the random effect model facilitates unconditional inferences. Socia l science researchers typically want to make unconditional inferences; thus, the random effect model is often more appropriate. However, the standard errors in the random effect model are larger than in the fixed model when effect sizes are heterogeneous ( Field 2001). M oreover, transformed correlation coefficients, which are used in

PAGE 45

45 fixed effect models, can eliminate a bias in the untransformed correlation coefficients, which are used in random effect models (Silver and Dunlap 1987). T he transformation (e.g ., Fisher s r to z transformation) corrects for a skew in the sampling distribution of correlation coefficients. F or the present study, therefore, the fixed effect model was used to combine effect sizes from primary studies. File Drawer Problem A common co ncern surrounding the meta analytic research method is that researchers can never be certain that their review contains all studies pertaining to the research domain. As published studies are more easily obtained, it is more likely that a meta analysis con tains the highest quality studies of a given subject area, which are also often those containing statistically significant outcomes. Moreover, the situation that any number of unpublished works could influence overall findings is a persistent problem for m eta analysis. Such studies remain in the This problem was given the name file drawer problem by Rosenthal (1979). The file drawer problem appears to have two causes: the reluctance of researchers to report their null r esults and the reluctance of professional journal editors to include studies whose results fail to reach statistical significance. Taking this into account, Rosenthal developed a fail safe N statistic. As Brown (1992) states, The fail safe N statistic is a follow up test used with meta analysis to estimate the number of new, unpublished, or unretrieved non significant (null result) studies what would, on the average, change the significance of a meta analysis study to non significance (p. 179). To calcu late the fail safe N, Rosenthal (1979) provided a formula that used the combined Z scores from the articles included in the meta analysis to determine the number of non significant (or null effect) studies. The formula is as follows:

PAGE 46

46 X = [(SUM Z) 2 / G] k where X = the number of studies needed to reverse the statistically significant findings, k = the number of studies combined in the meta analysis, (SUM Z) = the sum of the Z scores for the individual studies, and G = the Z score that falls at the p crit ical value being evaluated. The current study set the p critical value at .05. Therefore, a Z score of 1.645 was the denominator of the formula. For example, in Table 5, the fail safe N analysis yielded an answer of 47,112 for the A ad intention relation. T his means that in order to bring the meta analytic review s level of statistical significance down to the .05 level, 47,112 non significant studies would be needed. Rosenthal (1969) wrote one could regard as resistant to the file drawer problem any combin ed results for which the tolerance level X reaches 5k + 10 (p/ 640), where k equals the number of studies in the meta analysis This means that if a meta analysis contained 45 studies (the number of studies for the A ad intention relation), if the fail saf e N value were 235 or bigger then results could be considered stable. The 47,112 markedly exceeds the 235. This result would be ideal for the current study. T his confidence is due to the unlikely chance that 47,112 studies that found non significant result s were not included in the current study because they failed to be published or because they were simply overlooked by the author in the literature review. Other fail safe Ns for each pairwise relationship were also pr esented in Table 3 3 Chapter Summary Chapter 3 provided a description of the methodology used in conducting the current meta analysis examining the effect of ad attitudes on behavioral intentions and other pairwise relationships. The usefulness of the meta analytic technique and the three st eps of conducting meta analysis were presented. T he three steps included database development, effect size transformation, and the method of analysis. Studies for the meta analysis were found through the

PAGE 47

47 use of electronic databases and bibliographies of ot her literature. Study selection was based on the selection criteria. After all studies were selected and coded, coding reliability and publication bias were assessed to ensure that the results of the analysis are appropriated drawn.

PAGE 48

48 T able 3 1 The Conversion Statistical Equations t to r = t t 2 + N 2 ) t distribution, df (degrees of freedom) d to r = d d 2 + 4 ) N (sample size) F to r F ) / ( F + df error ) 2 (chi square) 2 to r 2 / N ) F distribution Z t o r = Z N 1 ) d (effect size) Table 3 2. Interrater Reliability for Coded Categories Section Interrater Agreement Bibliographic information 100% Study participants 100% Methodology 96.1% Effect size 97.3% Overall 98.2% Table 3 3. Analysis of the File Drawer Problem Relationship k Fail Safe N at .05 level Relationship k Fail Safe N at .05 level Age A ad 6 86 Age Intention 21 104 Gender A ad 5 16 Gender Intention 12 185 Ethnicity A ad 6 228 Ethnicity Intention 10 5 Education A ad 7 712 Education Intention 16 13 Health Status A ad 9 8 Health Status Intention 9 845 Involvement A ad 3 55 Involvement Intention 4 332 Ad exposure A ad 5 2 Drug usage Intention 10 1176 Ad awareness A ad 2 1 Income Intention 8 6 Drug use A ad 4 0 Media consumption Intention 7 8 Income A ad 4 109 Ad exposure Intention 10 2608 Ad awareness Intention 8 10 Aad Intention 45 47112 A ad Pharmacist Intention 4 56 A ad Friend Intention 2 3 Note. k = number of studies in the meta analysis *Fail Safe N ( X = [(SUM Z) 2 / G] k) > Tolerance level X (5k + 10)

PAGE 49

49 CHAPTER 4 RESULTS The overall objective of the current study is to provide a quantitative review of antecedents A ad consequences constructs and to investig ate their relationships in the context of DTCA. The variables included in these relati onships are depicted in Figure 2 1 This study employed the meta analytic technique to statistically identify the strength and direction of the pairwise relationships. U p on completion of the coding process, it was determined that 36 studies would contribute data for the current meta analytic database. Thirty six independent studies provided 278 effect sizes. The sample sizes ranged from 80 to 3001 ( M = 941.00; SD = 802.40) yielding a total sample of 261,597. Thirty studies were published in the 2000s, 4 in the 1990s, and 2 in the 1980s. A ppendix A and B summarized other characteristics of collected studies such as authorships, author disciplines, published fi elds and sites characteristics of samples, and methodology. The results of this study will be reported in four different areas. These areas address the following: (a) antecedents of A ad (b) consequences of A ad (c) other relationships, and (d) the potential moderati ng influence of study characteristics on the pairwise relationships. Tables 5 through 9 contain the results of the meta analysis for the hypotheses and research questions. The tables display the key results from each studied topic. T he information found in these tables includes the number of participants in each analysis ( N ), the number of independent effect sizes (i.e., correlation coefficients) in each analysis ( k ), the mean observed correlation ( r ), Fisher s Z (Z r ), the standard deviation of Fisher s Z ( SD Z r ), and 95% confidence intervals around Fisher s Z (CI Z r ). The interpretation of effect size magnitude is guided by Cohen s (1988) definitions of small ( r = .10), moderate ( r = .30), and large ( r = .50) effect sizes. Cohen (1988) established the medium effect size as one that was large enough so that people would naturally recognize it in

PAGE 50

50 everyday life, the small effect size to be one that was noticeably smaller, but not trivial, and the large effect size to be the same distance above the medium effect s ize as small was below it. Figure 4 1 provide s a summary of the results of the current study. Antecedents of Attitude toward the Ad Demographic factors, ad exposure, drug usage, health status, involvement, ad awareness, and ad exposure were examined as an tecedents of A ad The results of the conducted antecedents A ad meta a nalysis are presented in Table 4 1 The total sample size across the collected empirical studies was 51,753 with 51 reported or converted correlations. The number of independent effect sizes ( k ), the means of the observed correlations ( r ) and the Fisher s Z r and 95% confidence intervals around the Fisher s Z r (CI Zr ) between antecedents and A ad are as follows: Age A ad ( k = 6, r = .03, Z r = .03, CI Zr 5% = .15, and CI Zr 95% = .09), Gen der A ad ( k = 5, r = .02, Z r = .02, CI Zr 5% = .08, and CI Zr 95% = .10), Ethnicity A ad ( k = 6, r = .12, Z r = .12, CI Zr 5% = .27, and CI Zr 95% = .02), Education A ad ( k = 7, r = .12, Z r = .12, CI Zr 5% = .23, and CI Zr 95% = .01), Health status A ad ( k = 9, r = .02, Z r = .02, CI Zr 5% = .04, and CI Zr 95% = .08), Involvement A ad ( k = 3, r = .27, Z r = .32, CI Zr 5% = .77, and CI Zr 95% = 1.41), Ad exposure A ad ( k = 5, r = .02, Z r = .02, CI Zr 5% = .24, and CI Zr 95% = .0.29), Ad awareness A ad ( k = 2, r = .04, Z r = .04, CI Zr 5% = 2.00, and CI Zr 95% = 2.08), Drug usage A ad ( k = 4, r = .05, Z r = .05, CI Zr 5% = .08, and CI Zr 95% = .17), and Income A ad ( k = 4, r = .08, Z r = .08, CI Zr 5% = .13, and CI Zr 95% = .03). The findings of a sta tistical significance at the 95% confidence level show that education A ad and income A ad relationships do not include 0, which indicates that education and income are significant predictors of A ad Although the education and income A ad relationships a re statistically significant, the strength of the relationships is small. Such other factors as age,

PAGE 51

51 gender, ethnicity health status, involvement, ad exposure, ad awareness, and drug usage do not predict consumers attitudes toward the ad in the extant DT CA literature. The results reject H 3 and H 4 which predicted that A ad was a function of ad awareness/ad exposure and health status/involvement. Attitude t oward the Ad and Outcomes The current research investigated the relationship between A ad an d outcomes variables such as behavioral intenti ons and brand attitudes. Table 4 2 presents the results of these analyses. The total sample size across the collected studies was 54,282 with 45 reported or converted correlations. Consistent with H1, A ad is a significant predictor of behavioral intentions. The mean correlation between A ad and behavior intentions was 0.19, which was statistically significant. The A ad intention relationship was consistently positive as indicated by the confidence interval, whic h did not include zero. The findings of a statistical significance at the 95% confidence level indicated that the relationship falls within a 0.14 0.24 interval. In the DTC studies, behavioral intentions have been operationalized in four different ways such as intention to request physicians to prescribe the advertised drugs (Intention 1), intentions to ask physicians for more information about the advertised drugs (Intention 2), intentions to search more information about the advertised drugs (Intentio n 3), and intentions to visit their physicians (Intention 4). To clarify the relationship, the current research further analyzed the relationship A ad and four differently operationalized intentions. The means and confidence intervals of correlations are k = 8, r = .29, Z r = .29, CI Zr 5% = .18, and CI Zr 95% = .40 for the A ad and Intention 1 relation, k = 14, r = .16, Z r = .16, CI Zr 5% = .08, and CI Zr 95% = .24 for the A ad Intention 2 relation, k = 4, r = .15, Z r = .15, CI Zr 5% = .01, and CI Zr 95% = .31 for

PAGE 52

52 the A ad Intention 3 relation, and k = 19, r = .18, Z r = .18, CI Zr 5% = .11, and CI Zr 95% = .25 for the A ad Intention 4 relation. Regardless of the operationalization of the behavioral intentions, all relationships are at least marginally significant. More specifically, A ad is a statistically significant predictor of consumers intentions (a) to request physicians to prescribe the advertised drug, (b) to ask physicians for more information about the advertised drugs, and (c) to visit their physicians. T he re lationship between A ad and intentions to search more information about the advertised drugs is marginally significant. Most studies investigating the A ad intention relation have focused on patient physician relationships. H owever, some studies have been interested in looking at the patient and non physician relationships such as patient pharmacist, patient patient s friend, and patient nurse relationships. T hese studies have assessed the role of A ad in terms of predicting (a) intentions to ask pharmacists for more information about the advertised drugs (Pharmacist Intention), (b) intentions to ask nurses for more information about the advertised drugs (Nurse Intention), and (c) intentions to ask patient friends/family for more information about the adver tised drugs (Friend Intention). The means and confidence intervals of the correlations are k = 4, r = .15, Z r = .15, CI Zr 5% = .11, and CI Zr 95% = .19 for the A ad and pharmacist intention relation and k = 2, r = .13, Z r = .13, CI Zr 5% = 1.43, and CI Zr 95% = 1.69 for the A ad and friend intention relation. The relationship between A ad and nurse intention has not appeared in multiple studies. Since meta analytic research needs at least two research findings on the same topic, the relationship was not included in the current study. S tatistical significance at the 95% confidence level indicated that consumers attitudes toward the ad are positively related to their intentions to ask pharmacists for more information about the advertised prescription drugs. The re sults of the patient and non physician relations are presented in Table 4 3 Furthermore, in H 2 it was

PAGE 53

53 predicted that brand attitude would be a function of A ad However, the relationship could not be tested because only one study had addressed the role of A ad in terms of predicting brand attitudes. Ad Awareness The comprehensive search for previous research on DTCA found that a number of studies have investigated the construct of ad awareness. H owever, only two studies addressed ad awareness as a predicto r of A ad The number of independent effect sizes ( k ), the means of the observed correlations ( r ) and the Fisher s Z r and 95% confidence intervals around the Fisher s Z r (CI Zr ) between ad awareness and A ad are as follows: Ad awareness A ad ( k = 2, r = .04, Z r = .04, CI Zr 5% = 2.00, and CI Zr 95% = 2.08). Thus, the third hypothesis is rejected. In addition to the relationships between ad awareness and A ad this study also analyzed the relationship between ad awareness and behavioral intentions. H owever, the results revealed that the relationship is small and insignificant ( k = 8, r = .09, Z r = .09, CI Zr 5% = .08, and CI Zr 95% = .26). Besides the relationship, such constructs as age, education, and health status have been tested as antecedents of ad awareness The number of independent effect sizes ( k ), the means of the observed correlations ( r ), and the Fisher s Z r and 95% confidence intervals around Fisher s Z r (CI Zr ) for antecedents and ad awareness relationships are as follows: Age Ad awareness ( k = 3, r = .08, Z r = .08, CI Zr 5% = .27, and CI Zr 95% = .43), Education Ad awareness ( k = 2, r = .02, Z r = .02, CI Zr 5% = 1.69, and CI Zr 95% = 1.65), and Health status Ad awareness ( k = 2, r = .03, Z r = .03, CI Zr 5% = 1.51, and CI Zr 95% = 1.58). The resul ts of the meta analyses could not find significant antecedents and outcomes of ad awareness.

PAGE 54

54 Other Relationships The current study also investigated A ad not involved relationships (e.g., personal difference variables behavioral intention relation). Ta ble 4 4 presents the results of these analyses. DTCA researchers have investigated the effects of such personal differences as demographic factors on behavioral intentions. The number of independent effect sizes ( k ), the means of the observed correlations ( r ) and Fisher s Z r and 95% confidence intervals around Fisher s Z r (CI Zr ) for the non hypothesized relationships are as follows; Age Intention ( k = 21, r = .02, Z r = .02, CI Zr 5% = .02, and CI Zr 95% = .07), Gender Intention ( k = 12, r = .05, Z r = .05 CI Zr 5% = .02, and CI Zr 95% = .08), Ethnicity Intention ( k = 10, r = .01, Z r = .01, CI Zr 5% = .05, and CI Zr 95% = .03), Education Intention ( k = 16, r = .02, Z r = .02, CI Zr 5% = .07, and CI Zr 95% = .04), Health status Intention ( k = 9, r = 12, Z r = .12, CI Zr 5% = .15, and CI Zr 95% = .09), Involvement Intention ( k = 4, r = .23, Z r = .24, CI Zr 5% = .03, and CI Zr 95% = .51), Drug usage Intention ( k = 10, r = .14, Z r = .15, CI Zr 5% = .09, and CI Zr 95% = .20), Income Intention ( k = 8, r = .02, Z r = .03, CI Zr 5% = .11, and CI Zr 95% = .17), Media consumption Intention ( k = 7, r = .06, Z r = .06, CI Zr 5% = .07, and CI Zr 95% = .19), and Ad exposure Intention ( k = 10, r = .23, Z r = .24, CI Zr 5% = .16, and CI Zr 95% = .32). Statistical sig nificance at the 95% confidence level shows that gender intention, health status intention, drug usage intention, and ad exposure intention relationships do not include 0, which indicates that gender, health status, drug usage, and ad exposure are significant predictors of behavioral intentions in the extant DTCA literature. Such other factors as age, ethnicity education, involvement, income, and media consumption do not predict consumers behavioral intentions.

PAGE 55

55 As noted previously, DTCA researchers have operationalized behavioral intentions in four different ways (e.g., intention to request physicians to prescribe the advertised drugs Intention 1, intentions to ask physicians for more information about the advertised drugs Intention 2 intentions to discuss symptoms/the advertised drugs with physicians Intention 3, and intentions to visit their physicians Intention 4). To clarify the relationship between personal difference variables and behavioral intentions, this study further in vestigated the relationships between personal difference variables and four differently operationalized intentions. Statistical significance at the 95% confidence level shows that drug usage is the only significant predictor of intention 1 ( k = 3, r = .10, Z r = .10, CI Zr 5% = .07, and CI Zr 95% = .14), health status, drug usage, and ad exposure are the significant predictors of intention 2 (for health status k = 3, r = .10, Z r = .10, CI Zr 5% = .21, and CI Zr 95% = .00, for drug usage k = 4, r = .13, Z r = .13, CI Zr 5% = .10, and CI Zr 95% = .16, and for ad exposure k = 4, r = .34, Z r = .36, CI Zr 5% = .26, and CI Zr 95% = .45), and gender, health status, and ad exposure are the significant predictors of intention 3 (for gender k = 6, r = .05, Z r = .05, CI Zr 5% = .02, and CI Zr 95% = .08, for health status k = 5, r = .14, Z r = .14, CI Zr 5% = .19, and CI Zr 95% = .09, for ad exposure k = 6, r = .16, Z r = .16, CI Zr 5% = .13, and CI Zr 95% = .19). No study has investigated the relationship between personal differe nce variables and Intention 4. In addition to the findings of the relationships between personal difference variables and behavioral intentions in the extant DTCA literature, the researcher has studied such other relationships as gender drug usage, ethnic ity drug usage, education drug usage, age health status, education health status, involvement price perception, involvement health status, drug knowledge behavioral intention, ad message clarity A ad age ad exposure, and gender ad exposure. H owever, those relationships have not appeared in multiple studies. Since meta

PAGE 56

56 analytic research needs at least two research findings on the same topic, the relationships were not included in the current study. Moderator Analyses Moderator analyses were conducted to f urther clarify the strength of each pairwise relationship. Hunter et al. (1982) suggested that a moderator will show itself in the way: the average correlation coefficient will vary from subset to subset (e.g., between student sample and non student sample ). The differences between the subset results were tested statistically using a t test. Table 10 reports the results of the moderator analyses including the means and significance of each moderator for both observed correlation coefficients and Fisher s Z transformations of the correlation coefficients. The first conclusion drawn from Table 10 is that the results of the analyses of the observed correlation coefficients and Fisher s Z transformations are almost identical. Thus, the discussion below does not distinguish between the observed correlations and Fisher s Z transformation results (the statistics for the subgroup analyses are for Fisher s Z transformations). The analyses were mainly conducted on five factors: study sample characteristic (student vs. non student and local vs. nationwide sample), number of scale items (single vs. multiple item scale), theoretical foundation (theoretical vs. atheoretical study), research method (experiment vs. survey), and degree of effect size estimation (no estimation vs. estimation). I t was expected that studies would use a student sample and multi item scale, and theories would have stronger relationships than those using a non student sample, single item measures, and non theories. H owever, no relationships were affe cted by the type of sample, the number of scale items, and theoretical foundation.

PAGE 57

57 This study also tested the moderator effects of the degree of effect size estimation (no estimation vs. estimation), study sampling location (local sample vs. nationwide sa mple), and research method (experiment vs. survey). The analyses revealed that a stronger relationship between drug usage and behavioral intentions was detected in the studies with reported effect sizes than those with estimated effect sizes (.07 no esti mation vs. .03 estimation). The study sampling location appeared to have a more consistent effect than did other moderators across all relationships identified. The use of a local sample resulted in stronger relationships than the use of a nationwide sam ple in age A ad (.16 local sample vs. .07 nationwide sample), ethnicity intention ( .11 vs. .00), and income intention (.24 vs. .05) relationships. The research method was tested as a potential moderator. H owever, no relationships were affected by the research method.

PAGE 58

58 Ad Attitude Brand Attitude Behavioral Intention H2 H1 Ad Awarenes s H3 Involvement / Health Condition H4 Consumer Characteristics RQ1 RQ2, 5, 6 RQ3 and 6 r = .04 r = .08 ~ .27 r = .27 Figure 4 1. Summary of Results

PAGE 59

59 Table 4 1. Analysis of the Relationship between Antecedents and Aad IV DV N k R Z r SE Zr CI Zr 5% CI Zr 95% Age A ad 7634 6 .03 .03 .05 .15 .09 Gender 2141 5 .02 .02 .03 .7 0 .1 0 Ethnicity 7573 6 .12 .12 .06 .27 .02 Education 9067 7 .12 .12 .05 .23 .01 Health Status 11147 9 .02 .02 .03 .04 .08 Involvement 480 3 .27 .32 .25 .77 1.41 Ad exposure 2295 5 .02 .02 .09 24 .29 Ad awareness 468 2 .04 .04 .16 2 .00 2.08 Drug use 4207 4 .05 .05 .08 .08 .17 Income 6741 4 .08 .08 .02 .13 0.03 Note. IV = Independent Variable, DV = Dependent Variable, k = number of correlation coefficients, r = mean observed correlat Zr = Zr 5% = lower bound of the confidence interval for Zr Table 4 2. Analysis o f the Relationship between Aad and Intentions IV DV N k r Z r SD Zr CI Zr 5% CI Zr 95% A ad Overall Intention 54,282 45 .19 .2 0 .03 .14 .25 A ad Intention 1 7618 8 .29 .29 .06 .18 .4 0* Intention 2 18479 14 .16 .16 .04 .08 .24 Intention 3 2342 4 .15 .15 .08 .01 .31 Intention 4 25843 19 .18 .18 .04 .11 .25 Note. Intention 1 = Intention to request physicians to prescribe the advertised drugs, Intention 2 = Intention to ask physicians for more information about the advertised drugs, Intention 3 = Intentions to discuss symptoms/the advertised drugs with physician, Intention 4 = Intention to visit a physician, k = number of correlation coefficients, r = mean observed correlation, Z r = Fisher s Z between Aad and intention, SD Zr = estimated strand ard deviation of Fisher s Z CI Zr 5% = lower bound of the confidence interval for Fisher s Z CI Zr 95% = upper bound of the confidence interval for Fisher s Z

PAGE 60

60 Table 4 3. Analysis of the Relationship between Aad and Other Outcomes IV DV N k r Z r SE Zr CI Zr 5% CI Zr 95% Aad Pharmacist Intention 1164 4 0.15 0.15 0.01 0.11 0.19 Friend Intention 450 2 0.02 0.02 0.03 0.7 0.1 Note. Aad = Attitude toward the ad, Pharmacist Intention = intention to ask pharmacist more information about the advertised dru gs, Friend Intention = intention to ask patients' friends more information about the advertisied drugs, k = number of correlation coefficients, r = mean Zr = estimated strandard deviation o Zr Zr

PAGE 61

61 Table 4 4. Analysis of the Non Hypothesized Relationships IV DV N k R Z r SE Zr CI Zr 5% CI Zr 9 5% Age Overall Intention 21,208 21 .02 .02 .02 .02 .07 Age Intention 1 3,334 4 .01 .01 .09 .26 .29 Intention 2 10,049 10 .03 .03 .03 .04 .11 Intention 3 7,825 7 .02 .03 .02 .04 .08 Intention 4 0 0 N/A N/A N/A N/A N/A Gender Overall Intention 1 0,515 12 .05 .05 .01 .02 .08* Gender Intention 1 80 1 .08 .08 N/A N/A N/A Intention 2 3,897 5 .05 .05 .02 .01 .11 Intention 3 6,538 6 .05 .05 .02 .02 .08* Intention 4 0 0 N/A N/A N/A N/A N/A Eth nicity Overall Intention 11,965 10 .01 .01 .02 .05 .03 Eth nicity Intention 1 1,647 1 .05 .05 N/A N/A N/A Intention 2 4,152 5 .02 .02 .03 .09 .05 Intention 3 6,166 4 .02 .02 .03 .1 0 .07 Intention 4 0 0 N/A N/A N/A N/A N/A Education Overall Intention 15,200 16 .02 .02 .03 .07 .04 Education Intention 1 315 2 .06 .07 .17 2.25 2.12 Intention 2 8,454 9 .0 0 .0 0 .02 .05 .05 Intention 3 6,431 5 .03 .03 .06 .19 .14 Intention 4 0 0 N/A N/A N/A N/A N/A Health Status Overall Intention 11,822 9 .12 .12 .01 .15 .09 Health Status Inte ntion 1 80 1 .09 .09 N/A N/A N/A Intention 2 4,511 3 .1 0 .1 0 .02 .21 .0 0* Intention 3 7,231 5 .14 .14 .02 .19 .09 Intention 4 0 0 N/A N/A N/A N/A N/A Involvement Overall Intention 4,904 4 .23 .24 .08 .03 .51 Involvement Intention 1 186 1 .14 .14 N/A N/A N/A Intention 2 4,718 3 .27 .28 .11 .19 .75 Intention 3 0 0 N/A N/A N/A N/A N/A Intention 4 0 0 N/A N/A N/A N/A N/A

PAGE 62

62 Table 4 4. Continued IV DV N k R Z r SE Zr CI Zr 5% CI Zr 95% Drug Usage Overall Intention 10,070 10 .14 .15 .02 09 .2 0* Drug Usage Intention 1 4,082 3 .1 0 .1 0 .01 .07 .14 Intention 2 3,737 4 .13 .13 .01 .1 0 .16 Intention 3 2,251 3 .21 .22 .07 .08 .52 Intention 4 0 0 N/A N/A N/A N/A N/A Income Overall Intention 7,879 8 .02 .03 .06 .11 .17 Income Intentio n 1 80 1 .37 .38 N/A N/A N/A Intention 2 3,181 3 .05 .05 .04 .13 .24 Intention 3 4,618 4 .08 .08 .02 .16 .0 0 Intention 4 0 0 N/A N/A N/A N/A N/A Media Consumptio n Overall Intention 833 7 .06 .06 .05 .07 .19 Media Consumptio n Intention 1 240 3 .04 .04 .12 .47 .55 Intention 2 240 3 .06 .06 .08 .27 .39 Intention 3 353 1 .12 .12 N/A N/A N/A Intention 4 0 0 N/A N/A N/A N/A N/A Ad Exposure Overall Intention 8,769 10 .23 .24 .03 .16 .32 Ad Exposure Intention 1 0 0 N/A N/A N/A N/A N/A Int ention 2 3,015 4 .34 .36 .03 .26 .45 Intention 3 5,754 6 .16 .16 .01 .13 .19 Intention 4 0 0 N/A N/A N/A N/A N/A Note. Intention 1 = Intention to request physicians to prescribe the advertised drugs, Intention 2 = Intention to ask physicia ns for more information about the advertised drugs, Intention 3 = Intentions to search more information about the advertised drugs, Intention 4 = Intention to visit their physicians, k = number of correlation coefficients, r = mean observed correlation, Zr = Zr Zr Zr 95% = upper bound of the

PAGE 63

63 Table 4 5. Subgroup Means b y Moderator Variables Student versus non student single versus multi item scale Theoretical versus atheoretical study No estimation versus estimation Good versus bad estimation Experiment versus survey Local sample versus nationwide sample Age Aad r 08 vs. 0.8 .10 vs. .03 .03 vs. .03 .08 vs. .06 N/A N/A .16 vs. .07* Zr .08 vs. 0.9 .10 vs. .03 .03 vs. .03 .08 vs. .06 .16 vs. .07* k = 6 2 vs. 4 3 vs. 3 4 vs. 2 4 vs. 2 1 vs. 5 Gender Aad r .03 vs. .01 .05 vs. .03 .01 vs. .03 N/A N/A N/A N/A Zr .03 vs. .01 .05 vs. .03 .01 vs. .03 k = 5 1 vs. 4 3 vs. 2 4 vs. 1 Ethnicity Aad r .07 vs. .16 .10 vs. .15 .10 vs. .15 .06 vs. .26 N/A N/A .26 vs. .06 Zr .07 vs. .17 .10 vs. .16 .10 vs. .16 .06 vs. .27 .27 vs. .0 6 k = 6 1 vs. 5 3 vs. 3 3 vs. 3 4 vs. 2 2 vs. 4 Education Aad r .11 vs. .12 .18 vs. .04 .07 vs. .18 .08 vs. .22 N/A N/A .12 vs. .16* Zr .11 vs. .12 .18 vs. .04 .07 vs. .19 .08 vs. .22 .12 vs. .16* k =7 1 vs. 6 4 vs. 3 4 vs. 3 5 v s. 2 1 vs. 6 Health Status Aad r .04 vs. .01 .01 vs. .04 .05 vs. .05 .03 vs. .01 .02 vs. .00 N/A .13 vs. .01 Zr .04 vs. .01 .01 vs. .04 .05 vs. .05 .03 vs. .01 .02 vs. .00 .13 vs. .01 k = 9 2 vs. 7 4 vs. 5 6 vs. 3 6 vs. 3 8 vs. 1 2 vs. 6 Inv olvement Aad r N/A N/A N/A N/A .40 vs. .00 N/A N/A Zr .48 vs. .00 k = 3 2 vs. 1

PAGE 64

64 Table 4 5. Continued Student versus non student single versus multi item scale Theoretical versus atheoretical study No estimation versus estimation Good versus bad estimation Experiment versus survey Local sample versus nationwide sample Ad Exposure Aad r .01 vs. .05 N/A .32 vs. .05 .01 vs. .05 N/A .32 vs. .05 .23 vs. .01 Zr .01 vs. .05 .33 vs. .05 .01 vs. .05 .33 vs. .05 .23 vs. .01 k = 5 3 vs 2 1 vs. 4 3 vs. 2 1 vs. 4 1 vs. 3 Ad Awareness Aad r N/A N/A N/A N/A N/A N/A N/A Zr k = 2 Drug Usage Aad r .05 vs. .05 .00 vs. .06 .13 vs. .02 .04 vs. .07 .06 vs. .00 N/A .07 vs. .03 Zr .05 vs. .05 .00 vs. .06 .13 vs. .02 .04 vs. .07 .06 vs. .00 .07 vs. .03 k = 4 2 vs. 2 1 vs. 3 1 vs. 3 1 vs. 3 3 vs. 1 2 vs. 2 Income Aad r N/A .09 vs. .07 .09 vs. .07 .09 vs. .07 N/A N/A N/A Zr .09 vs. .07 .09 vs. .07 .09 vs. .07 k = 4 3 vs. 1 4 vs. 1 5 vs. 1 Ad Awarenes s Intention r N/A .11 vs. .06 .20 vs. .05 N/A N/A N/A .18 vs. .07 Zr .11 vs. .06 .22 vs. .05 .19 vs. .07 k = 8 7 vs. 1 2 vs. 6 5 vs. 3 Aad Intention r .07 vs. .21* .17 vs. .20 .22 vs. .16 .17 vs. .21 .19 vs. .21 .19 vs. .16 .19 vs. .19 Zr .07 vs. .22* .17 vs. .25 .24 vs. .16 .17 vs. .22 .20 vs. .22 .20 vs. .16 .20 vs. .20 k = 45 6 vs. 39 29 vs. 16 22 vs. 23 19 vs. 26 42 vs. 3 41 vs. 4 10 vs. 35

PAGE 65

65 Table 4 5. Continued Student versus non student single versus multi item scale Theoretic al versus atheoretical study No estimation versus estimation Good versus bad estimation Experiment versus survey Local sample versus nationwide sample Aad Pharmacist Intention r .11 vs. .16 .14 vs. .15 .14 vs. .15 .14 vs. .15 N/A N/A .13 vs. .17 Zr .11 vs. .17 .14 vs. 16 .14 vs. 16 .14 vs. .15 .13 vs. .17 k = 4 1 vs. 3 2 vs. 2 2 vs. 2 1 vs. 3 2 vs. 2 Age Intention r .03 vs. .04 N/A .02 vs. .03 .01 vs. .03 .01 vs. .04 .02 vs. .05 .07 vs. .01 Zr .03 vs. .04 .02 vs. .03 .01 vs. .03 .01 vs. .04 .0 2 vs. .05 .08 vs. .01 k = 21 3 vs. 17 5 vs. 16 7 vs. 14 12 vs. 9 19 vs. 2 4 vs. 17 Gender Intention r .07 vs. .04 .06 vs. .00* .04 vs. .05 .07 vs. .03* .06 vs. .00 N/A .08 vs. .04 Zr .07 vs. .04 .06 vs. .00* .04 vs. .05 .07 vs. .03* .06 vs. .00 .08 vs. .04 k = 12 3 vs. 9 10 vs. 2 5 vs. 7 6 vs. 6 10 vs. 2 2 vs. 10 Ethnicity Intention r .03 vs. .00 .02 vs. .01 .01 vs. .01 .01 vs. .01 .02 vs. .05 N/A .11 vs. .00* Zr .03 vs. .00 .02 vs. .01 .01 vs. .01 .01 vs. .01 .02 vs. .05 .11 vs. .00* k = 10 4 vs. 6 8 vs. 2 5 vs. 5 6 vs. 4 9 vs. 1 1 vs. 9 Education Intention r .02 vs. 0.03 .02 vs. .01 .03 vs. .01 .01 vs. .04 .02 vs. .00 N/A .07 vs. .01 Zr .02 vs. 0.04 .02 vs. .01 .03 vs. .01 .01 vs. .04 .02 vs. .00 .07 vs. .01 k = 16 5 vs. 11 13 vs. 3 5 vs. 11 8 vs. 8 15 vs. 1 5 vs. 11 Health Status Intention r .12 vs. .12 .12 vs. .13 .14 vs. .11 .12 vs. .12 N/A N/A .09 vs. .13 Zr .12 vs. .12 .12 vs. .13 .14 vs. .11 .12 vs. .13 .09 vs. .13 k = 9 2 vs. 7 7 vs. 2 3 vs. 6 6 vs. 3 2 vs. 7

PAGE 66

66 Table 4 5. Continued Student versus non student single versus multi item scale Theoretical versus atheoretical study No estimation versus estimation Good versus bad estimation Experiment versus survey Local sam ple versus nationwide sample Involvement Intention r N/A N/A .30 vs. .17 .17 vs. .30 N/A .30 vs. .17 N/A Zr .32 vs. .17 .17 vs. .32 .32 vs. .17 N = 4 2 vs. 2 2 vs. 2 2 vs. 2 Drug Usage Intention r .18 vs. .11 N/A N/A .21 vs. .12* .17 vs. .10 N/A .13 vs. .15 Zr .19 vs. .11 .22 vs. .12* .17 vs. .10 .13 vs. .15 k = 10 5 vs. 5 3 vs. 7 7 vs. 3 2 vs. 8 Income Intention r N/A N/A .04 vs. .06 .04 vs. .06 .07 vs. .19* N/A .24 vs. .05* Zr .04 vs. .06 .04 vs. .06 .07 vs. .19* .25 vs. .05* k = 8 3 vs. 5 3 vs. 5 5 vs. 3 2 vs. 6 Media Consumption Intention r .12 vs. .05 N/A N/A N/A .12 vs. .05 N/A N/A Zr .12 vs. .05 .12 vs. .05 k = 7 1 vs. 6 1 vs. 6 Ad Exposure Intention r .22 vs. .32 .22 vs. .32 .32 vs. .22 .22 vs. 32 N/A .32 vs. .22 N/A Zr .23 vs. .33 .23 vs. .33 .33 vs. .23 .23 vs. .33 .33 vs. .23 k = 10 9 vs. 1 9 vs. 1 1 vs. 9 9 vs. 1 1 vs. 9 Note. All data are means. k = number of correlation coefficients, r = mean observed correlation, Zr ansformation *Significant at p < .05

PAGE 67

67 CHAPTER 5 DISCUSSION A ad is widely known to be an essential predictor of behavioral intentions. Therefore, a number of studies have addressed A ad in the DTCA literature. Despite this interest in A ad there has not been a comprehensive attempt to investigate general findings across independent DTCA studies Such an investigation is useful in understanding the general strength and variability of the relationships and the study conditions that moderate those relations hips (Brown and Stayman 1992 p. 34). For example, while some studies have reported no evidence of a significant effect of A ad on behavioral intentions (e.g., Williams and Hensel 1995), others have reported a significant effect (e.g., An 2007). Furthermore different studies have found widely varying magnitudes of the A ad effect on behavioral intentions. In order to assess the strength and variability of the A ad intention relationship, the current research meta analyzed A ad effects aggregated across all ava ilable research in the extant DTCA literature. In addition to the assessment of the relationship between A ad and intentions, this study also investigated the relationships between A ad and its antecedents and the potential moderating variables. The results of this meta analysis provide considerable insight into the effects of A ad in the contexts of DTCA and the state of DTCA research. As with any meta analysis, the data provide a quantitative summary. In the current meta analysis, the data provided a sum mary of 278 samples reported in the 36 articles for which the author could obtain usable data. Variables were classified into three levels. The first level included demographic characteristics, ad awareness, involvement, health status, and drug usage. The second level contained consumers attitudes toward the ad. The third included behavioral intentions. The first level directly and/or indirectly affected the second and third. The second level directly influenced the third.

PAGE 68

68 A s shown in Tables 4 1 throug h 4 5 and analyzed above, the aggregated study effects suggested a significant relationship between A ad and a number of important constructs, including both antecedents (education, r = .12 and Z r = .12 and income, r = .08, Z r = .08) and consequences (b ehavioral intention, r = .19 and Z r = .20 and pharmacist intention, r = .15, Z r = .15). In addition, the results also found that consumers intentions were influenced by personal characteristics, including gender ( r = .02 and Z r = .08), health status ( r = .12 and Z r = .12), drug usage ( r = .14 and Z r = .15), and ad exposure ( r = .23 and Z r = .24). The results showed that consumers who (a) were less educated, (b) had a low income (c) were female, (d) were in bad health, (e) took a lot of drugs, or (f) were exposed to advertising frequently tended to have more favorable attitudes toward DTCA than those who (a) were more educated, (b) had a high income (c) were male, (d) were in good health, (e) took few drugs, or (f) were exposed to advertising less frequent ly. H owever, in general, the strength of each of these relationships was small or small to moderate. Discussions and Implications Role of Attitude toward the Ad T he results of this dissertation challenge the effect of A ad on consumers behavioral intentio ns in terms of DTCA. Although A ad is a statistically significant predictor of consumers intentions, it has a small to moderate effect according to Cohen s effect size interpretation guide (e.g., .1 = small, .3 = moderate, and .5 = strong). More specifical ly, since consumers need prescriptions to buy an advertised drug, DTCA studies have operationalized intentions in different ways (e.g., i ntention to request physicians to prescribe the advertised drugs, i ntention to ask physicians for more information abou t the advertised drugs, i ntentions to search more information about the advertised drugs, and i ntention s to visit a physician ). Among the four

PAGE 69

69 different types of intentions, patients intentions to request prescriptions for the advertised drugs are the str ongest outcome of A ad a moderate effect. A ad has a small effect on the other types of intentions. The results of the meta analysis of the A ad intention relation in the context of DTCA have important theoretical and practical implications. Since Lutz et a l. (1983) proposed the dual mediation model, which explains the A ad intention relation, many studies have tested this relation in the fields of advertising and marketing. I t is worth comparing the results of the current study with those in the previous stu dy that meta analyzed the dual mediation model. Brown and Stayman (1992) combined the effect sizes of A ad intention relationships across independent studies. T he research found that A ad has a strong effect on consumers intentions ( r = .43), which is much stronger than that found in the current study ( r = .19). From the marketing and DTCA literature, four explanations for the limited effect of A ad on intentions is the present study seemed plausible. First, it is possible that the operationalization of the behavioral intentions culminate d in the effect size difference. Brown and Stayman s study aggregated studies that used consumers purchase intentions in the A ad intention relation. H owever, since consumers do not have the final right of product purchase in prescription drug sales, drug request intention drug inquiry intention, drug information search intention and visit intention are usually considered by DTCA researchers as more proper variables for the outcomes of DTCA. The present study synthes ized the studies that employed drug related behavioral intentions, not purchase intentions. The second possible explanation of the limited DTCA effect hinges on the traditional relationships between patients and doctors. Advocates of DTCA contend that the advent of DTCA has given consumers opportunities that they have never had before. T hey claim that

PAGE 70

70 consumers can take an active role in the treatment of their medical conditions via the knowledge consumers acquire from DTCA. H owever, there is a disparity b etween reality and expectation. For example, even though many patients have the desire to question the appropriateness of physician prescribed decisions, some of them are unwilling to ask about the advertised prescription drugs. Patients believe that physi cians may view patient inquiries or prescription requests as a sign of distrust or even disrespect (Petroshius et al. 1995). Although pharmaceutical industry advocates point to the educational value of DTCA, patients do not obtain enough medical informatio n, either because the amount of information delivered via DTCA is limited or because the content of the medical information is difficult. Patients limited access to medical information pertaining to various prescription drugs has culminated in patients re lying heavily on the advice of their physicians to select appropriate medications. This reliance causes the patients unwillingness to question or request prescription drugs. Third as mentioned previously, A ad is an affective construct representing consu mers feelings of favorability/unfavorability toward the ad itself (MacKenzie et al. 1986, p. 130). When consumers are about to choose their medical treatment options, their involvement level is usually high. This implies that consumers rely more on caref ul evaluation of advertising information than on their feelings about drug advertising for the better medical decision. In other words, consumers tend to value trustworthiness/believability of the advertisement or the advertiser more than their attitudes t oward the ad when they process health related information. In addition, many have insisted for a long time that if consumers do not believe what is being said, the probability of evoking a desired response is greatly weak ened (Atkin and Beltramini 2007). T hus, although the attitude behavioral intention relation in the context of DTCA has been statistically supported, the strength of the relation is small to moderate.

PAGE 71

71 Lastly, potentially the effects of A ad for different product categories are different. T h e results of this study revealed that the A ad effect for health products is lower than that for normal consumer goods. It is possible that the lower A ad effect is due to consumers decision making processes, which are different for prescription products in general than they are for other product categories. This implies that future research should address how and why those processes are different. In addition, future research also needs to investigate to what extent A ad effects are different for different p roduct categories. This result suggests that future research needs to develop other constructs that predict consumer behavior better than A ad in the context of health related communications and identify other possibilities that limit the A ad and intention relationship. As noted previously, A ad is an affective construct. A cognitive construct would be a better predictor of consumers medical decision making behavior. T his fact implies that future researchers need to find or develop a new construct that can replace A ad In addition to developing a cognitive construct, it is also important to investigate other factors that can cause the limited effect of A ad For example, patients accessibility to medical services is directly related to their intentions to vi sit a doctor s office, which is a pre step for DTC drug prescriptions. Such a small to moderate effect of A ad on patients intentions is the reason that drug manufacturers are questioning the efficiency of their consumer advertising (Narayanan Desiraju and Chintagunta 2004). H owever, a brand s DTCA expenditure can influence product category sales (Narayanan et al. 2004). First, since the FDA allowed the DTCA of prescription drugs on television in 1997, the prescription drug market has increased by 330% f rom 1996 to 2005 (Donohue, Cevasco and Rosenthal 2007). Rosenthal et al. (2002) also reported noteworthy findings; they found that up to 22% of category growth could be attributed to advertising

PAGE 72

72 expenditures. DTCA has been criticized for encouraging the in appropriate use of medications and for driving up drug spending. However, the criticism is evidence of the increased product category size. In addition to the influence of DTCA on product category sales, it is well known that a brand s advertising expendit ure can affect the brand s share of category sales and reduce price sensitivity, which enables pharmaceutical companies to raise drug prices. M oreover, pharmaceutical marketers have often seen sales of advertised prescription drugs increase rapidly after the beginning of their advertisements targeting the general public (Davis 2000). A f ter airing DTCA, Allegra and Zyrtec, for instance, experienced sales increases of 100% and 56%, respectively. Claritin, also a heavy advertiser, was the leading DTC advertis er in the first half of 1998, and during that time, its sales increased 32%. When Davis conducted his study, Claritin, Allegra, and Zyrtec were DTC drugs. A fter being on the market as a prescription drug for a long time and being used by a large number of people, allergy drugs were changed from DTC to OTC (over the counter) status. Consumers tendency to request advertised drugs more than non advertised drugs increased the sales of advertised drugs dramatically after the initiation of drug advertising. Ther efore, even though a strong effect size was not detected in the A ad intention relation, pharmaceutical advertising is essential to increasing the product category size and to increasing a brand s share in the category It is also noteworthy that consumer s positive attitudes toward the ad lead to their intentions to ask their friends/family or pharmacists about the advertised drugs. T hose relationships can be undervalued because even pharmacists cannot be involved in prescribing DTC drugs. H owever, if we view the relationships as an information exchange process, it is important because the communication between consumers and friends/pharmacists results in increased awareness in terms of the effects of the advertised drugs and/or common symptoms

PAGE 73

7 3 related to the drugs. In other words, the advent of DTCA affords patients the opportunity to be better informed about their prescribed medication options. Informed patients are more likely to request prescriptions for the advertised drugs and to question the appropri ateness of a physician s prescription decisions. Role of Ad Awareness It has been empirically proven that people have a tendency to prefer familiar objects over unfamiliar ones ( e.g., Zajonc 1980). I n addition to the empirical evidence, the effect of awar eness can be explained by the mere exposure effect, contending that the simple repetition of ads culminates in consumers favorable attitudes toward the ad and the advertised product. B ased on the previous research and theory, it was expected that ad aware ness would be an important predictor of consumers attitudes toward the ad. In spite of the importance of the awareness effect in advertising and marketing, the comprehensive literature search found only two studies addressing the effect of ad awareness in the context of DTCA. Thus, the result of the hypothesis about the ad awareness effect on A ad cannot be generalized. Not only the ad awareness effect on A ad but also its effect on consumer behavior has been investigated in the DTCA literature M ost of th e studies on ad awareness effects on consumer behavior stem from Lavidge and Steiner s (1961) hierarchy of effect model. In the model, there are five pre steps of purchase: awareness knowledge liking preference conviction purchase. Many researche rs have since proposed similar models that explain the consumer purchase process (e.g., Colley 1961). These hierarchy models commonly contend that ad awareness is a pre step of the advertising goal. Based on the hierarchy models, eight studies have address ed ad awareness in terms of predicting consumer behavior. T he results of the current

PAGE 74

74 analyses revealed that the relationship between ad awareness and behavioral intentions is small and insignificant ( k = 8, r = .09, Z r = .09, CI Zr 5% = .08, and CI Zr 95% = .26). The insignificant effect of ad awareness suggests that increasing ad awareness, one of the common advertising goals, cannot lead to achieving communication breakthroughs. I n advertising circles, some believe that the proliferation of products and ad vertisers competing for the space in the consumer s mind (e.g., ad clutter ) may explain this limited effect of ad awareness. H owever, it is important to view DTCA in terms of health communication. A s mentioned above, many factors can affect consumer respon se behavior to DTC drugs (e.g., rational decision making process, physician s authority, and long term relationship between physician and patient). Thus, it is unlikely that the previously discussed effect on the consumer s intention to ask friends/family or pharmacists about the advertised drugs has an actual influence on DTC drug sales. Role of Antecedents of Attitude toward the Ad According to empirical findings and theories in the DTCA literature, it was expected that consumer characteristics would be related to DTCA effectiveness. C onsumer characteristics consist of demographics, involvement/health status, ad awareness/ad exposure, and health characteristics (such as health conditions and prescription drug utilization). DTCA effectiveness, including A a d and behavioral intentions, were analyzed as outcomes of DTCA. The current meta analysis identified the variables that affect consumers attitudes toward the ad; education and income. The results revealed that gender and ethnicity were marginally signific ant predictors of A ad M ore specifically, consumers who were less educated, poor, female, and non white were more likely to have favorable attitudes toward the ad than those who were more educated, rich, male, and white.

PAGE 75

75 A nother interesting finding was th at consumer characteristics were directly related to behavioral intentions. F or instance, consumers higher behavioral intentions were a function of gender (female), health status (poor health condition), prescription drug utilization (high drug consumptio n), and frequent ad exposure. There are two implications of the results of the relationships between antecedents and intentions. F irst, it is noteworthy that frequent ad exposure is related to patients behavioral intentions. I t implies that advertising m edia planners need to focus more on frequency than on reach or other criteria related to the effect measurements of the media. Second, p harmaceutical companies can use the findings of the present study on antecedents of DTCA effects to develop market segme ntation strategies. For instance, although DTCA is an effective communication tool for some consumer groups (e.g., less educated), there are also many consumers whose attitudes toward the ad are not favorable and who are not willing to request drug prescri ptions or information. T his means that to increase the sales of pharmaceutical products, marketers have to utilize other customized marketing tools such as drug price off coupons for those who have unfavorable attitudes toward DTCA and low behavioral inten tions. How to spend limited advertising budgets effectively is critical for advertisers. How to save the budget is a more important task for them. T he results of the analysis about the relationships among antecedents, A ad and behavioral intentions can he lp save the budget because the results provide guidance for audience targeting. F or example, highly educated people are less reliant on DTCA information when they make a medical treatment decision. T herefore, DTCA for those consumers would be less effectiv e than for less educated consumers. In sum, several studies have investigated the relationships between consumer characteristics and behavioral

PAGE 76

76 intentions, and the findings of the studies are valuable when developing advertising targeting strategies. In addition to the targeting strategies, the results of the analysis about antecedents of Aad can be used to develop more effective advertisements. For instance, consumers who are in bad health tend to evaluate drug ad arguments carefully, whereas those who are in good health examine the ad messages quickly or focus on simples cues. T he consumer s tendency implies that the use of message framing is helpful to create persuasive messages for different consumers (e.g., good health consumers vs. bad health consum ers). Message framing c an be conceptualized as the perspective that people react differently to different but objectively equivalent messages (K hberger 1998). The analogy of gambling (e.g., 50 % of probability of losing money vs. 50 % of probability of earn ing money ) is a common example of message framing. The features of health related behavior can be framed in terms of the benefits of engaging in the behavior (called a positive /gain frame), or in terms of the costs of failing to engage in the behavior (a n egative /loss frame). Message framing insists that consumers prefer a positively framed message when they examine drug ad messages quickly or focus on simples cues, while they prefer a negatively framed message when they evaluate the ad arguments carefully. Thus, the results of consumer characteristics can be used to create more persuasive messages. Limitations In synthesizing previous studies on A ad in the DTCA field, Cooper s (1989) suggestion of validity issues guided the current study. Cooper emphasized the importance of defining the objectives and scope of the meta analytic review, searching for studies, reporting analysis procedures and results, and interpreting the results. The results and conclusions of the current study should be evaluated with Coope r s criteria of validity issues. Although the findings of the

PAGE 77

77 meta analysis provide good implications for both advertising practitioners and researchers, some limitations were unavoidable. First, the scope of the analysis assessing issues pertaining to the advertising effect was broad. A comprehensive and thorough search for relevant studies found 36 articles for which usable quantitative data were available. H owever, in spite of the author s efforts, data for four additional studies identified as pertinent to the topic of this study were unavailable. Thus, some differences might have resulted if the not included studies had strong effects. Second, the number of included empirical studies, based on the set of inclusion criteria, was limited. Even though adv ocates of meta analysis proclaim it is appropriate to test effects from a limited number of studies, particular caution was used to interpret the findings of the current study. More specifically, some cells in the moderator analyses and some relationships had a very small number of study effects (as few as a single observation for the moderator analyses and two observations for relationship analyses). L imited data availability did not allow for meaningful interpretation of some relationships and the effects of moderators on pairwise relationships. Another main cause of the small number of studies included in the meta analysis was the lack of necessary statistics for calculating effect sizes. F or instance, some studies reported only statistical significance o f their results without p value, sample size, and other statistical values. Therefore, researchers should mindful of reporting necessary statistics for other researchers in primary studies. F inally, this study was limited in representing all research doma ins in extant pharmaceutical advertising literature because the focus was on consumer responsiveness to DTCA. I n other words, it is recommended that future research address the impact of DTCA on physicians in terms of their prescription writing habits and responsiveness to patient drug

PAGE 78

78 inquiries and requests. In addition to the effect on patients, although advertising of p rescription drugs (DTCA) represent s 60% of the total spending on drug advertising (General Accounting Office 2002) for better understand ing of the landscape of drug advertising in general, future studies need to address the role of over the counter drug advertising (OTCA). The unique feature of DTC drugs is that consumers cannot directly purchase prescription drugs. H owever, the purchase p rocess for OTC drugs is different from that of DTC drugs. Thus, it would be interesting to compare the effect of OTCA on behavioral intentions with that of DTCA or the results of Brown and Stayman s study. Summary Researchers have devoted considerable att ention to the investigation of consumers attitudes toward the ad and their effect on consumer behavior. The impact of this well known predisposition has been tested in the context of DTCA, and it appears to be mixed. T he current study is the first attempt to investigate the effect of DTCA across independent DTCA studies with a meta analytic approach. Two primary conclusions can be drawn from this meta analysis. The results from the current meta analysis suggest that A ad is a statistically significant predi ctor of behavioral intentions, but that A ad has a small to moderate effect size in terms of affecting consumers intentions. T his study also found that some demographic factors are related to consumers attitudes toward the ad. Although moderator analyses were conducted to clarify further the strength of each pairwise relationship studied, consistent patterns were not detected. Gaining a greater understanding of the relationships surrounding A ad has implications fo r researchers and practitioners.

PAGE 79

79 APPENDIX A CODED CHARACTERISTIC S OF INCLUDED STUDIE S Study Primary author's discipline Student or Nonstudent sample Single or mu l tiple scales Theoretical foundation yes or no Sample size Effect size estimation Study design Sample collection Lee et al. (2007) Advert ising Nonstudent Single Yes 2141 No estimation Survey Nationwide sample Choi & Lee (2007) Advertising Nonstudent Single No 1301 Estimation Survey Nationwide sample Everett (1991) Communication Nonstudent Single No 238 Estimation Survey Local sample Meht a & Purvis (2003) Other Nonstudent Single No 1475 Estimation Survey Nationwide sample Wilson & Till (2007) Communication Nonstudent Mu l tiple No 2290 No estimation Survey Nationwide sample Deshpande et al. (2004) Medical Nonstudent Mu l tiple Yes 382 No est imation Survey Nationwide sample Huh et al. (2005) Communication Student Single No 353 Estimation Survey Local sample Huh & Becker (2005) Communication Student Mutiple No 688 No estimation Survey Nationwide sample Atkin & Beltramini (2007) Marketing Non student Mutiple No 93 Estimation Survey Local sample Hausman (2008) Marketing Nonstudent Single No 455 No estimation Survey Local sample An (2007) Communication Nonstudent Single Yes 189 Estimation Survey Local sample Huh & Lanteau (2009) Communication Nonstudent Single Yes 644 Estimation Survey Nationwide sample Kavadas et al. (2007) Marketing Nonstudent Single Yes 156 Estimation Experi ment Local sample

PAGE 80

80 Study Primary author's discipline Student or Nonstudent sample Single or mu l tiple scales Theoretica l foundation yes or no Sample size Effect size estimation Study design Sample collection Beltramini (2006) Marketing Nonstudent Single No 2653 No estimation Survey Nationwide sample Joseph et al. (2008) Marketing Nonstudent Single No 168 Estimation Surve y Nationwide sample Bell et al. (1999) Communication Nonstudent Single No 329 No estimation Survey Local sample Spake & Joseph (2007) Marketing Nonstudent Single No 154 Estimation Survey Local sample Singh & Smith (2007) Marketing Nonstudent Single Yes 288 Estimation Survey Nationwide sample Limbu & Torres (2009) Marketing Nonstudent Mutiple Yes 186 Estimation Experi ment Local sample Perri & Nelson (1987) Marketing Nonstudent Mutiple No 139 Estimation Survey Local sample Williams & Hensel (1995) Mark eting Nonstudent Mutiple Yes 132 No estimation Survey Local sample Baca et al. (2005) Marketing Student Mutiple Yes 205 Estimation Survey Local sample Gonul et al. (2000) Marketing Nonstudent Mutiple No 318 Estimation Survey Nationwide sample Liu et al. (2005) Medical Nonstudent Mutiple Yes 375 Estimation Survey Nationwide sample Parnes et al. (2009) Medical Nonstudent Single No 1647 Estimation Survey Local sample Rehne & Moldrup (2008) Medical Nonstudent Single No 3001 Estimation Survey Local sample Morris et al. (1986) Other Nonstudent Single No 1507 Estimation Survey Local sample

PAGE 81

81 Study Primary author's discipline Student or Nonstudent sample Single or mu l tiple scales Theoretical foundation yes or no Sample size Effect size estimation Study design S ample collection Sumpradit et al. (2002) Medical Nonstudent Single Yes 342 Estimation Survey Local sample Schommer et al. (2005) Medical Nonstudent Single No 80 Estimation Survey Nationwide sample Weissman (2003) Medical Nonstudent Mutiple No 1022 Estim ation Survey Local sample Maddox (1999) Marketing Nonstudent Single No 132 Estimation Survey Local sample Herzenstein et al. (2004) Marketing Nonstudent Single No 960 Estimation Survey Local sample Abel et al. (2009) Medical Nonstudent Single No 348 Est imation Survey Nationwide sample Bhutada et al. (2009) Medical Nonstudent Mutiple Yes 138 Estimation Experi ment Nationwide sample Delorme et al. (2010) Communication Nonstudent Mutiple No 235 Estimation Survey Nationwide sample Yang et al. (2010) Medica l Nonstudent Single Yes 150 Estimation Survey Nationwide sample

PAGE 82

82 APPENDIX B OTHER CHARACTERISTIC S OF INCLUDED STUDIE S Study Name of Journal # of Autho rs # of Relations Reported Dominant Ethnic Group of Sample Dominant Gender of Sample Relati onships Investigated Lee et al. (2007) Journal of Advertising 3 46 White Female 6, 7, 8, 9, 10, 15, 24, 26, 42, 43, 60, 61, 44, 62, 63, 45, 64, 50, 67 Choi & Lee (2007) Journal of Advertising 2 6 White Female 24, 26, 42, 60 Everett (1991) Journal of Adv ertising Research 1 6 N /A N/A 16, 18, 21, 36, 63 Mehta & Purvis (2003) Journal of Advertising Research 2 6 N/A Female 16, 17, 33, 39, 42, 49 Wilson & Till (2007) Journal of Advertising Research 2 6 White Female 24, 42, 45, 46, 47 Deshpande et al. (2004) Journal of Health Communication 3 8 White Female 7, 10, 23, 26, 69, 70, 71 Huh et al. (2005) Journal of Health Communication 3 10 White Female 25, 26, 28, 54, 55, 57, 66, 92, 93 Huh & Becker (2005) International Journal of Advertising 2 65 White Female 6, 7, 8, 9, 10, 12, 14, 25, 26, 52, 53, 54, 55, 60, 61, 62, 68, 72, 73, 74, 75, 76, 77, 78, 79, 80, 87, 88, 89, 90, 91 Atkin & Beltramini (2007) Journal of Marketing Communications 2 1 N/A Male 12 Hausman (2008) Journal of Advertising Research 1 1 White Female 22 An (2007) Journal of Health Communication 1 6 N/A Female 23, 24, 94, 95, 96, 97 Huh & Lanteau (2009) Communication Research 2 1 N/A Female 95 Kavadas et al. (2007) Journal of Consumer Marketing 3 4 N/A Female 5, 11, 31, 81

PAGE 83

83 Study Name of Journ al # of Autho rs # of Relations Reported Dominant Ethnic Group of Sample Dominant Gender of Sample Relationships Investigated Beltramini (2006) Journal of Business Ethics 1 6 N/A N/A 23, 24, 26 Joseph et al. (2008) International Journal of Pharmaceutical and Healthcare Marketing 3 6 N /A Female 60, 61, 67 Bell et al. (1999) Journal of General Internal Medicine 3 4 White Female 1, 2, 4, 13 Spake & Joseph (2007) Journal of Consumer Marketing 2 2 N/A Male 23, 26 Singh & Smith (2007) Journal of Consumer Mar keting 2 2 N/A Female 26 Limbu & Torres (2009) Journal of Health and Human Services Administration 2 4 N/A N/A 11, 31, 38, 86 Perri & Nelson (1987) Journal of Health Care Marketing 2 1 N/A Female 13 Williams & Hensel (1995) Journal of Health Care Market ing 2 5 N/A N/A 9, 10, 24, 28, 29 Baca et al. (2005) Journal of Consumer Marketing 3 4 N/A N/A 6, 10, 14, 24 Gonul et al. (2000) Health Care Management Science 3 9 N/A N/A 6, 9, 10, 14, 15, 26, 28, 29, 30 Liu et al. (2005) Research in Social Administrat ive Pharmacy 4 15 White Female 24, 25, 28, 42, 43, 44, 45, 52, 53, 54, 55, 82, 83, 84, 85 Parnes et al. (2009) Annals of Family Medicine 7 3 N o dominant group Female 33, 35, 39 Rehne & Moldrup (2008) Journal of Medical Marketing 2 4 N/A N/A 1, 2, 10, 14 Morris et al. (1986) Public Health Reports 5 1 White Female 9

PAGE 84

84 Study Name of Journal # of Autho rs # of Relations Reported Dominant Ethnic Group of Sample Dominant Gender of Sample Relationships Investigated Sumpradit et al. (2002) American Journal of Hea lth Behavior 3 2 N/A Female 23 Schommer et al. (2005) Research in Social and Administrative Pharmacy 3 18 N/A Male 23, 24, 33, 34, 36, 37, 41, 42, 43, 46, 50, 51 Weissman (2003) Health Affair 1 2 White N/A 18, 64 Maddox ( 1999) Journal of Product and Br and Management 1 2 N/A Male 33, 42 Herzenstein et al. (2004) Marketing Letters 3 7 N/A N/A 24, 25, 39, 49, 52, 55, 58 Abel et al. (2009) Journal of Clinical Oncology 4 5 White N/A 1, 2, 3, 4, 5 Bhutada et al. (2009) Health Marketing Quarterly 3 6 White Male 11, 12, 31, 32, 47, 48 Delorme et al. (2010) Journal of Health Communication 3 2 White Female 8 Yang et al. (2010) Journal of the National Medical Association 7 3 Black Female 16, 18, 20 Note. Numbers in relationships investigated represents the fo llowing relationships: 1. age Ad awareness 2. gender Ad awareness 3. ethnicity Ad awareness 4. education Ad awareness 5. health status Ad awareness 6. age A ad 7. gender A ad 8. ethnicity A ad 9. education A ad 10. health status A ad 11. involvement A a d 12. ad exposure A ad 13. ad awareness A ad 14. drug use A ad 15. income A ad 16. Ad awareness Intention to request the advertised drugs 17. Ad awareness Intention to ask more info about the advertised drugs 18. Ad awareness Intention to discuss with p hysicians 19. Ad awareness Intention to tell their doctor they had seen the ad 20. Ad awareness screening test participation

PAGE 85

85 21. Ad awareness intention to change doctors if request refused 22. A ad brand attitude 23. A ad Intention to request the advert ised drugs 24. A ad Intention to ask doctors more info about the advertised drugs 25. A ad Intentions to find more info 26. A ad Intention to discuss with physicians 27. A ad Intention to tell their doctor they had seen the ad 28. A ad Intention to ask pha rmacists more info about the advertised drugs 29. A ad Intention to ask friends more info about the advertised drugs 30. A ad Intention to ask purses more info about the advertised drugs 31. involvement brand attitude 32. ad exposure brand attitude 33. age Intention to request the advertised drugs 34. gender Intention to request the advertised drugs 35. ethnicity Intention to request the advertised drugs 36. education Intention to request the advertised drugs 37. health status Intention to request th e advertised drugs 38. involvement Intention to request the advertised drugs 39. drug use Intention to request the advertised drugs 40. income Intention to request the advertised drugs 41. media consumption Intention to request the advertised drugs 42 age Intention to ask more info about the advertised drugs 43. gender Intention to ask more info about the advertised drugs 44. ethnicity Intention to ask more info about the advertised drugs 45. education Intention to ask more info about the advertise d drugs 46. health status Intention to ask more info about the advertised drugs 47. involvement Intention to ask more info about the advertised drugs 48. ad exposure Intention to ask more info about the advertised drugs 49. drug use Intention to ask mo re info about the advertised drugs 50. income Intention to ask more info about the advertised drugs 51. media consumption Intention to ask more info about the advertised drugs

PAGE 86

86 52. age Intentions to find more info 53. gender Intentions to find more info 54. ethnicity Intentions to find more info 55.education Intentions to find more info 56. health status Intentions to find more info 57. drug use Intentions to find more info 58. income Intentions to find more info 59. ad exposure Intentions to find more info 60. age Intention to discuss with physicians 61. gender Intention to discuss with physicians 62. ethnicity Intention to discuss with physicians 63. education Intention to discuss with physicians 64. health status Intention to discuss with phy sicians 65. media consumption Intention to discuss with physicians 66. drug use Intention to discuss with physicians 67. income Intention to discuss with physicians 68. ad exposure Intention to discuss with physicians 69. ad clarity Aad 70. quality o f ad info Aad 71. quality of risk info Aad 72. age ad exposure 73. gender ad exposure 74. ethnicity ad exposure 75. education ad exposure 76. health status ad exposure 77. drug use ad exposure 78. gender drug use 79. ethnicity drug use 80. educat ion drug use 81. health status involvement 82. age Intention to ask pharmacists more info about the advertised drugs 83. gender Intention to ask pharmacists more info about the advertised drugs 84. ethnicity Intention to ask pharmacists more info about the advertised drugs 85. education Intention to ask pharmacists more info about the advertised drugs 86. involvement price perception 87. age health status 88. gender health status 89. ethnicity health status 90. education health status 91. age dru g use 92. gender Intention to ask friends more info about the advertised drugs 93. drug use Intention to ask friends more info about the advertised drugs 94. interest in health info Intention to request the advertised drugs

PAGE 87

87 95. drug knowledge Intention to request the advertised drugs 96. interest in heatlh info Intention to ask doctors more info about the advertised drugs 97. drug knowledge Intention to ask doctors more info about the advertised drugs

PAGE 88

88 APPENDIX C CODING MENUAL AND FO RMS 1. Study ID number Assign a unique identification number to each study. I f a report presents two independent studies, i.e., two independent outcome studies with different participants, then add a decimal to the study ID number to distinguish each study within a report and code each independent study. [ID] 2. Coder [CODER] 1. Primary coder 2. Secondary coder 3. Primary author (LN, FI) [AUTHOR] 4. Primary author affiliation: [AUTHAFF] 5. Primary author s discipline [AUTHDISC] 1. A dvertising 2. M arketing/Business 3. C ommunication (including health communication) 4. M edical 5. H ealth marketing 6. Other (specify): 9. Cannot tell 6. Bibliographic info in APA format: [REF] ______________________________________________________________ ___ 7. Year of publication _ _ (9999 if unknown) [PUBYR] 8. Name of journal: [NAMEJOU] 9. Journal ranking: Impact factor of the Journal if readily available. Impact factors were taken from Journal Citation reports Social Sciences & Genera l Science Editions (2004). [JOURANKING] 10. Field of publication [PUBFIELD] 1. Advertising 4. Medical 2. Marketing 5. Health marketing 3. Communication 6. Other (specify):

PAGE 89

89 11. Where was this study conducted? [SITE] 1. U.S 3. O ther: 2. Ne w Zealand 9. C annot tell Sample Descriptors 1. Mean age of sample. Specify the approximate or exact mean age. Code the best information available; estimate mean age from grade levels if necessary. I f mean age cannot be determined, enter 99.99. [AGE] 2 Predominant race. Select the code that best describes the racial makeup of the sample. 1. Greater than 60% White 5. M ixed, none more than 60% 2. Greater than 60% Black 6. Mixed, cannot estimate proportion 3. Greater than 60% Hispanic 9. C annot tel l 4. Greater than 60% other minority [RACE] 3. Predominant sex of sample. Select the code that best describes the proportion of males in the sample. [GENDER1] 1. Less than 5% male 4. B etween 51% and 95% male 2. Between 5% and 49% male 5. G reater than 95% male 3. 50% male 9. C annot tell 4. Actual proportion of males in the sample: If the proportion cannot be determined, enter 99.99 [GENDER2] 5. Use of student sample. Select the code that best describes the sample characteristic 1. Student sample 9. Cannot tell 2. Non student sample [STUDENT] 6. Predominant education level. Select the code that best describes the proportion of college and hi gher level in the sample. [EDU] 1. Less than 5% 4. B etween 51% and 95 % 2. Between 5% and 49% 5. G reater than 95% 3. 50% 9. C annot tell 7. Predominant income level. S elect the code that best describe the proportions of $50,000 and higher income bracket in the sample. [INCOME] 1. Less than 5% 4. B etween 51% and 95%

PAGE 90

90 2. Between 5% and 49% 5. G reater than 95% 3. 50% 9. C annot tell 8. Other participant characteristics and value. Describe other sample descriptor not coded above. [OTHERCHAR] Effect Size For each effect size, code all of the foll owing items. N ote that studies will have different numbers of effect sizes, and hence, different numbers of effect size level data coding forms. 1. Study ID: [ID] 2. Effect size number As sign each ES within a study a unique number and name. Multi ple effect sizes are numbered sequentially, e.g., 1, 2, 3, 4, etc. [ESID] Code for Constructs 3 Dependent Variable (Construct measured ) [DV] 1. Ad awareness 2. Attitude toward the ad 3. A ttitude toward the brand 4. I ntention to request the ad vertised drugs 5. I ntention to ask doctors more info about the advertised drugs 6. Intentions to find more info 7. Intention to discuss with physicians 8. Intention to visit their physicians 9 I ntention to ask pharmacists more info about the advertis ed drugs 10 I ntention to ask friends more info about the advertised drugs 11 I ntention to ask nurses more info about the advertised drugs 12. D rug use 13. H ealth status 14. Ad exposure 15. Other 4. Number of scale item to measure DV [SCALEDV] 1. Single item 2. Multi item 9. Cannot determine or not reported

PAGE 91

91 5. Independent Variable (Construct measured) [IV] 1. Age 9. Ad awareness 2. Gender 10. Attitude toward the ad 3. E thnicity 11. Attitude toward the brand 4. E ducation 12. Other: 5. H ealth sta tus 13. Drug use 6. I nvolvement 14. Income 7. A d exposure 15. D rug knowledge 8. M edia time consumption 6. Number of scale item to measure IV [SCALEIV] 1. Single item 2. Multi item 9. Cannot determine or not reported 7. Reliability and validity provid ed? [RELVAL] 1. Yes 2. No 9. Cannot tell 8. Type of measure. What are the constructs measured? [MEASURE] 1. B ehavioral observation/observational measure 2. R ating scale/checklist/questionnaire 3. S tandardized test 4. D irect assessment 5. O ther: __ ___________________________ 9. C annot tell 9. Focus of measure What is the focus of the article that the author(s) is(are) interested in? 1. A ttitude toward the ad 2. A ttitude toward the brand 3. B ehavioral intention 4. O ther: _______________________ _____ 9. C annot tell [FOCUS] 10 Theoretical foundation. Is the article based on a specific theory? [ THEOFOUND ] 1. Y es 2. No 9. Cannot tell

PAGE 92

92 11. I f yes, specify the theory employed: [THEORY] 12. Types of relationship. Variable types in the p airwise relationships. [RELATION] 1. Two dichotomous variables 2. A dichotomous and continuous variables 3. Two continuous variables Codes for Effect size 13. Statistical value for the pairwise relationship: [STATVALUE] 14. Types of statistical value for the rel ationship. [STATUSED] 1. C hi square 5. t 2. F 6. P value 3. Z 7. other 4. r 15 Effect size Calculate effect size using the excel effect size determination program or by hand using the procedures outlined in Table 3 2 : [ES] 16. Statistical sig nificance of the pairwise relationship [SIGNIFICANCE] 1. Y es 2. N o 9. Cannot tell 17 Confidence rating in effect size computation. [ESTIMATION] 1. highly estimated (have N and crude p value only, such as p<.01, and must reconstruct via rough t te st equivalence) 2. moderate estimation (have complex but relatively complete statistics, such as multifactor ANOVA, as basis for estimation) 3. some estimation (have unconventional statistics and must convert to equivalent t values or have conventional sta tistics but incomplete, such as exact p value). 4. slight estimation (must use significance testing statistics rather than descriptive statistics but have complete statistics of conventional sort) 5. no estimation (have descriptive data such as means, sta ndard deviations, frequencies, proportions, etc. and can calculate the effect size directly)

PAGE 93

93 Methodological Feature 1. Type of design [DESIGN] 1. Experiment 4. Econometric 2. Quasi experiment 5. Other: 3. Survey 2. If a study is a survey. Survey sam ple collecting method [SURVEY] 1. P hone 2. M ail 3. P aper pencil 4. online 3. Survey sampling [SURSAMP] 1. S tratified random 2. R andom 3. C onvenience sample 4. P robability sample 4. Sampling place [SAMPPLACE] 1. L ocal 2. N ationwide 3. F oreign 5. Response rate [RESRA TE] C annot, enter 1000 6 Method of assignment to experimental groups [ASSIGNMETH] 1. Random 2. Non random 3. Other: _______________ 9. Cannot determine 7 Study funding [FUNDING] 1. Government funding 2. Private funding 3. No funding 4. P artial funding

PAGE 94

94 9. Cannot determi ne 8 Study setting [STUDYSETTING] 1. Lab 2. Field 3. Hybrid 9. Cannot determine 9. Product name and category (inductively identified): __________ [PRODUCTCATE] 10 Experimental study stimulus [STIMULUS] 1. Print 2. Audio 3. Visual 4. Audio+visual 5. Other 9. C annot determine or no reported

PAGE 95

95 LIST OF REFERENCES Aaker, David A. and George S. Day (1974), A Dynamic Model of Relationships among Advertising, Consumer Awareness, Attitude, and Behavior, Journal of Applied Psychology 59 (3), 281 286 Alperstein, Neil M. and Mark Peyrot (1993), Consumer Awareness of Prescription Drug Advertising, Journal of Advertising Research July/August, 50 56 An, S oontae (2008) D irect to C onsumer A dvertising and S ocial P erception of the P revalence of D epressio n: A pplication of the A vailability H euristic Health Communication, 23 (6), 499 505. _______ (2007), Attitude toward Direct to Consumer Advertising and Drug Inquiry Intention: The Moderating Role of Perceived Knowledge, Journal of Health Communication 12(6), 567 580 Anderson, C raig A., and Brad J. Bushman (2001) V iolent V ideo G ames on A ggressive B ehavior, A ggressive C ognition, A ggressive A ffect, P hysiological A rousal, and P rosocial B ehavior: A M eta A nalytic R eview of the S cientific L itera ture Psychological Science, 12, 353 359 Ankem, Library and Information Science Research, 27 (2), 164 176 Atkin, Joann L. and Richard F. Beltramini (2007), Exploring the Percei ved Believability of DTC Advertising in the US, Journal of Marketing Communications 13 (3), 169 180 Baca, Erin E., Juan Holguin Jr., and Andreas W. Stratemeyer (2005), Direct to Consumer Advertising and Young Consumers: Building Brand Value, Journal o f Consumer Marketing 22 (7), 379 387 Bansal, Harvir S. and Shirley F. Taylor (2002), Investigating Interaction Effects in the Theory of Planned Behavior in a Service Provider Switching Context, Psychology & Marketing, 19 (5), 407 425 Baron, Reuben M. and David A. Kenny (1986), The Moderator Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations, Journal of Personality and Social Psychology 51 (6), 1173 1183 Bell, Robert A., Michael S. W ilkes, Richard L. Kravitz (1999), Advertisement Induced Prescription Drug Requests Patients Anticipated Reactions to a Physician who Refuses, Journal of Family Practice 48 (6), 446 452 Berger, Jeffrey T., Pieter Kark, Fred Rosner, Samuel Packer and Al len J. Bennett (2001), Direct to Consumer Drug Marketing, Mount Sinai Journal of Medicine 68 (3), 197 202

PAGE 96

96 Berndt, Ernest (2001), The U.S. Pharmaceutical Industry: Why Significant Growth in Times of Cost Containment? Health Affairs 20 (2), 100 114 Bo rnstein, Robert F. (1989), Exposure and Affect: Overview and M e ta Analysis of Research, 1968 1987, Psychological Bulletin 106 (2), 265 289 Bradley, Lynette, R. and Julie Magno Zito (1997), Direct to Consumer Prescription Drug Advertising, Medical Car e 35 (1), 86 92 Brett, A llen S. (2007) to consumer advertising of prescription drugs Journal of Watch General Medicine, 357, 673 681 Brown, Steven P. and Douglas M. Stayman (1992), Antecedents and Consequences of Attitude toward the Ad: A Me ta analysis, Journal of Consumer Research 19 (June), 34 51 Brownfield, Erica D., Jay M. Bernhardt, Jennifer L. Phan, Mark V. Williams, Ruth M. Parker (2004), Direct to Consumer Drug Advertisements on Network Television: An Exploration of Quantity, Freq uency, and Placement, Journal of Health Communication 9, 491 497 Bush, Alan J., Rachel Smith, and Craig Martin (1999), The Influence of Consumer Socialization Variables on Attitudes Toward Advertising: A Comparison of African Americans and Caucasians, Journal of Advertising 28 (3), 13 24 Bushman, B rad J., and Craig A. Anderson (2001) V iolence and the American P ublic: Scientific F act versus M edia M isinformation American Psychologist, 56, 477 489 Calder, Bobby J., Lynn W. Phillips, and Alice M. Tybout (1981), Designing Research for Application, Journal of Consumer Research 8 (September), 197 207 Calfee, John E. (2002), Public Policy Issues in Direct to Consumer Advertising of Prescription Drugs, Journal of Public Policy & Marketing 21 (2), 174 193 Journal of Consumer Research, 30, 292 304 Capella, Michael L. (2005), A Review of the Effect of Advertising on Cigarette Initiation, Continuation and Brand Behavior: A Mixed Method Approach, a dissertation for the degree of doctor of philosophy in marketing, Mississipp i State University, Mississippi The Role of Involvement in Attention an d The Journal of Consumer Research, 15(2), 210 224 Journal of Personality and Social Psycholog y 45 (20), 241 256

PAGE 97

97 Choi, Sejung M. and Wei to Consumer (DTC) Pharmaceutical Advertising on Patient Journal of Advertising 36 (3), 137 149 Colley, Russel H. (1961), Defining Adver tising Goals for Measured Advertising Results, New York, NY: Asso ciation of National Advertisers Deshpande, Aparna, Ajit Menon, Matthew Perri III, and George Zinkhan (2004), Direct to Consumer Advertising and its Utility in Health Care Decision Making: A Consumer Perspective, Journal of Health Communication 9, 499 513 Donohue, J. M., Berndt, E. R. Rosenthal, M. B., Epstein, A. M., and Frank, R. G. (2004). Effects of pharmaceutical promotion on adherence to the treatment guidelines for depression. Medi cal Care 42 (12), 1176 1185 ______, Marisa Cevasco, and Meredith B. Rosenthal (2007), A Decade of Direct to Consumer Advertising of Prescription Drugs, New England Journal of Medicine 357 (7), 673 681 Everett, Stephen E. (1991), Lay Audience Response to Prescription Drug Advertising, Journal of Advertising Research 31 (2), 43 49 Farley, John U., and Donald R. Lehmann (1986), Meta Analysis in Marketing: Generalization of Response Models Lexington, MA: Lexington Books Federal Register 33 (1968): 88 12 Federal Register 44 (1979): 40016 Fern, Edward F. and Kent B. Monroe (1996), Effect Size Estimates: Issues and Problems in Interpretation, Journal of Consumer Research 23 (September), 89 105 Field, Andy P. (2001), Meta Analysis of Correlation C oefficients: A Monte Carlo Comparison of Fixed and Random Effects Methods, Psychological Methods, 6 (2), 161 180 Food and Drug Administration (1997), Center for Drug Evaluation and Research, Draft Guidance for Industry on Consumer directed Broadcast Ad vertisements, (July 1997), Federal Register 62 (155), 43171 73 ______ (1999), Center for Drug Evaluation and Research, Guidance for Industry: Consumer directed Broadcast Advertisements, (August 1999), Federal Register, 64 (152), 43197 98 General Acc ounting Office. (2002). Prescription Drugs: FDA Oversight of Direct to Consumer Advertising Has Limitations Washington, DC: United States General Accounting Office Glass, Gene V. (1976), Primary, Secondary, and Meta Analysis of Research, Educational Re searcher 5 (10), 3 8

PAGE 98

98 Green, William (1995), Consumer Directed Advertising of Contraceptive Drugs: The FDA, Depo Provera, and Product Liability, Food and Drug Law Journal 50, 553 568 Haley, Russel I. and Allan L. Baldinger (1991), The ARF Copy Resear ch Validity Project, Journal of Advertising Research 31 (April/May), 11 32 Harris, Richard J. (1983), Information Processing Research in Advertising Hillsdale, NJ: Lawrence Erlbaum Associates Hausman, Angela (2008), Direct to Consumer Advertising and Its Effect on Prescription Request, Journal of Advertising Research 48 (1), 42 56 Hedges, Lawrence V. and Ingram Olkin (1985), Statistical Methods for Meta analysis Orlando, FL: Academic Press ______ (1992), Meta Analysis, Journal of Educational St atistics 17 (4), 279 296 ______ & Vevea, J. L. (1998) and R andom E ffects M odels in M eta A nalysis Psychological Methods 3, 486 504 Herzenstein, Michal, Sanjog Misra, and Steven S. Posavac (2004), How Consumers Attitude Toward Direct to Cons umer Advertising of Prescription Drugs Influence Ad Effectiveness, and Consumer and Physician Behavior, Marketing Letters 15 (4), 201 212 Hoek, Janet and Philip Gendall (2002), To Have or Not to Have? Ethics and Regulation of Direct to Consumer Adverti sing of Prescription Medicines, Journal of Marketing Communications 8, 1 15 Hoen, E. (1998), Direct to Consumer Advertising: For Better Profits or For Better Health? American Journal of Health System Pharmacy, 55, 594 597 Homer, Pamela M. (1990), Th e Mediating Role of Attitude Toward the Ad: Some Additional Evidence, Journal of Marketing Research 27 (February), 78 86 Huh, Jisu and Lee B. Becker (2004), to consumer prescription drug advertising: Internationa l Journal of Advertising 24 (4), 441 466 ______, Denise E. Delorme, Leonard N. Reid (2004), The Third Person Effect and Its Influence on Behavioral Outcomes in a Product Advertising Context: The Case of Direct to Consumer Prescription Drug Advertising, Communication Research 31 (5), 568 599 Hunter, John E. and Frank L. Schmidt (1990), Methods of Meta analysis: Correcting Error and Bias in Research Findings Newbury Park, CA: Sage ______, ______, and G. E. Jackson (1982), Meta Analysis: Cumulating Res earch Findings across Studies Beverly Hills, CA: Sage Publications.

PAGE 99

99 Iizuka, Toshiaki (2004), What Explains the Use of Direct to Consumer Advertising of Prescription Drugs? Journal of Industrial Economics 52 (3), 349 379 Journal of Consumer Research, 20(3), 376 392 ______, Hayden Noel, and Alan G. Sawyer (2003), A Meta analysis of the Spacing Effect in Verbal Learning: Implications for Research on Advertising Repetition and Consumer Memo ry, Journal of Consumer Research, 30 (June), 138 149 Johnson, Blair T. and Alice H. Eagly (1989), Effects of Involvement on Persuasion: A Meta analysis, Psychological Bulletin 106 (2), 290 314 Jung, Wan S., Thre e Decades of Direct to Consumer Advertising of Prescription Drugs, Will be presented at to the annual conference of AEJMC, Denver, CO, August, 2010 Kaiser Family Foundation (2010), Prescription Drug Fact Sheet, (assessed April 23, 2011), available at http://www.kff .org/rxdrugs/upload/3057 08.pdf Kaphingst, Kimberly A., William Dejong, Rima E. Rudd, and Lawren H. Darltroy (2004), A Content Analysis of Direct to Consumer Television Prescript ion Drug Advertisements, Journal of Health Communication 9, 515 528 Keller, Kevin L. (2002), Strategic Brand Management: Building, Measuring, and Managing Brand Equity (2 ed.). Upper Saddle, River, NJ: Prentice Hall Kraus, Stephen J. (1995), Attitudes and the Prediction of Behavior: A Meta Analysis of the Empirical Literature, Personality and Social Psychology Bulletin 21 (1), 58 75 Kravitz, Richard L., Ronald M. Epstein, Mitchell D. Feldman, Carol E. Franz, Rahman Azari, Michael S. Wilkes, Ladson H inton, and Peter Franks (2005), Influence of Patients Requests for Direct to Consumer Advertised Antidepressants: A Randomized Controlled Trial, Journal of American Medical Association 293(16), 1995 2002 Lavidge, Robert J. and Gary A. Steiner (1961), A Model for Predictive Measurements of Advertising Effectiveness, Journal of Marketing 25 (6), 59 62 Leffler, Keith B. (1981), Advertising Journal of Law and Economics 24 (1), 45 74 Li psey, Mark W. and David B. Wilson (2001), Practical Meta Analysis (1 ed.). CA: Sage Publications, Inc. Lipsky, MS and CA Taylor (1997), The Opinions and Experiences of Family Physicians Regarding Direct to Consumer Advertising, Journal of Family Practic e 45 (6), 495 499

PAGE 100

100 MacKenzie, Scott B. and Richard J. Lutz (1989), A Empirical Examination of the Structural Antecedents of Attitude toward the Ad in an Advertising Pretesting Context, Journal of Marketing 53 (April), 48 65 ______, Richard J. Lutz, an d George E. Belch (1986), The Role of Attitude toward the Ad as a Mediator of Advertising Effectiveness: A Test of Competing Explanations, Journal of Marketing Research 23 (May), 130 143 f Advertising: A Meta Journal of Public Policy & Marketing, 16 (2), 205 216 McGuire, William J. (1978), An Information Processing Model of Advertising Effectiveness. I n: Davis, H.L. and Silk, A.J. (eds) Behavioral and Management Science in Marke ting New York, NY: Ronald Press Mehta, Abhilasha and Scott C. Purvis (2003), Consumer Response to Print Prescription Drug Advertising, Journal of Advertising Research 43, 194 206 Menon, Ajit M., Aparna D. Deshpande, George M. Zinkhan, and Matthew Pe rry (2004), A Model Assessing the Effectiveness of Direct to Consumer Advertising: Integration of Concepts and Measures from Marketing and Healthcare, International Journal of Advertising 23 (1), 91 112 Miller, Lucinda G. and Alan Blum (1993), Physici an Awareness of Prescription Drug Costs: A Missing Element of Drug Advertising and Promotion, Journal of Family Practice 36 (1), 33 37 Mitchell, Andrew A. and Jerry C. Olson (1981), Are Product Attribute Beliefs the Only Mediator of Advertising Effects on Brand Attitudes? Journal of Marketing Research 18 (August), 318 322 Morris, Louis A., and Lloyd G. Millstein (1984), Drug Advertising to Consumers: Effects of Formats for Magazine and Television Advertisements, Food Drug Cosmetic Law Journal 39, 497 503 ______, David Brinberg, Ron Klimberg, Carole Rivers, and Lloyd G. Millstein (1986), The Attitudes of Consumers Toward Direct Advertising of Prescription Drugs, Public Health Reports 101, 82 89 Myers, Melanie F., Man Huei Chang, Cynthia Jorgens en, William Whitworth, Sidibe Kassim, James A. Litch, Lori Armstrong, Barbara Bernhardt, W. Andrew Faucett, Debra Irwin, Jody Mouchawar, and Linda A. Bradley (2006), Genetic Testing for Susceptibility to Breast and Ovarian Cancer: Evaluating the Impact of a Direct to Consumer Marketing Campaign on Physicians Knowledge and Practices, Genetics in Medicine 8 (6), 361 370

PAGE 101

101 Narayanan, Sridhar, Ramarao Desiraju, and Pradeep K. Chintagunta (2004), Return of Investment Implications for Pharmaceutical Promotion al Expenditures: The Role of Marketing Mix Interations, Journal of Marketing 68 (October), 90 105 Park, Jin Seong and Jean M. Grow (2008), The social Reality of Depression: DTC Advertising of Antidepressants and Perceptions of the Prevalence and Lifeti me Risk of Depression, Journal of Business Ethics, 79 (4), 379 393 Paul, David P., Amy Handlin, and Angela D Auria Stanton (2002), Primary Care Physicians Attitudes toward Direct to Consumer Advertising Prescription Drugs: Still Crazy after All These Y ears, Journal of Consumer Marketing 19 (7), 564 574 Journal of Consumer Research, 17(2), 180 191 Peltzman, Sam (1973 ), An Evaluation of Consumer Protection Legislation: The 1962 Drug Amendments, Journal of Political Economy 81 (5), 1049 1091 Perri, Matthew and Author A. Nelson Jr. (1987), An Exploratory Analysis of Consumer Recognition of Direct to Consumer Adverti sing of Prescription Medications, Journal of Health Care Marketing 7 (1), 9 17 Perri, Matthew and W. Michael Dickson (1988), Consumer Reaction to a Direct to Consumer Prescription Drug Advertising Campaign, Journal of Health Care Marketing 8 (2), 66 69 Peyrot, Mark, Neil M. Alperstein, Doris Van Doren, and Laurence G. Poll (1998), Direct to Consumer Ads can Influence Behavior, Marketing Health Services 18 (2), 26 32 Petroshius, Susan M., Philip A. Titus and Kathryn J. Hatch (1995), Physician At titudes Toward Pharmaceutical Drug Advertising, Journal of Advertising Research 35 (6), 41 51 Petty, Richard E., and John T. Cacioppo (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Cha nge, New York: Springer Verlag Pin to and Drug Law Journal, 54, 489 518 Rao, Akshay R. and Kent B. Monroe (1989), The Effect of Price, Brand Name, and Store Name on Buyers Perceptions of Product Quality: An Integrative Review, Journal of Marketing Research 26 (August), 351 357 365

PAGE 102

102 Rheinstein, Pe ter H. (1982), The New Frontiers of Prescription Drug Promotion Food and Drug Law Journal 37, 330 335 Roth, Martin S. (1996), Patterns in Direct to Consumer Prescription Drug Print Adve rtising and Their Policy Implications, Journal of Public Policy and Marketing 15 (1), 63 75 Saxton, Matthew L. (2006), Meta Analysis in Library and Information Science: Method, History, and Recommendations for Reporting Research, Library Trends, 55 (1 ), 158 170 Sheehan, Kim Bartel (2003), Balancing Acts: An Analysis of Food and Drug Administration Letters about Direct to Consumer Advertising Violations, Journal of Public Policy & Marketing 22 (2), 159 169 Shimp, Terence A. (1981), Attitude toward the Ad as a Mediator of Consumer Brand Choice, Journal of Advertising 10(2), 9 15 Direct to Consumer Prescription Drug Journal of Consume r Marketing 22 (7), 369 378 Statman, Meir and Tyzoon T. Tyebjee (1984), The Case of the Pharmaceutical Industry and Drug Substitution Laws Journal of Public Policy & Marketing 99 112 Tellis, Gerard J. (1988), Advertising Exposure, Loyalty, and Brand Purchase: A Two Stage Model of Choice, Journal of Marketing Research 25 (May), 134 144 Weissman, Joel S., David Blumenthal, Alvin J. Silk, Kinga Zapert, Michael Newman, and Robert Leitman (2003), Consu mers Reports on the Health Effects of Direct to Consumer Drug Advertising, Health Affairs 22 (1), 82 95 Wilke, Michael (1997a), Ad Fever Sweeps Health Care Industry Advertising Age (January 13) Wilke, Michael (1997b), FDA Close to Easing Limits on Te levision Ads Advertising Age (June 30) Wilkes, Michael S., Robert A. Bell, and R. Kravitz (2000), Direct to Consumer Prescription Drug Advertising: Trends, Impact, and Implications, Health Affairs 19 (2), 110 128 Williams, James R. and Paul J. Hensel (1995), Direct to Consumer Advertising of Prescription Drugs: Attitudes toward This Type of Advertising Influence the Information Seeking Behavior of Older Adults, Journal of Health Care Marketing 15 (1), 35 41

PAGE 103

103 Wilson, Rick T. and Brian D. Till (2007) Direct to Consumer Pharmaceutical Advertising: Building and Testing a Model for Advertising Effectiveness, Journal of Advertising Research 47 (3), 270 282 Wolf, Fredric M. (1986). Meta analysis: Quantitative methods for research synthesis Beverly Hi lls, CA: Sage Yaros, Ann Marie (2007). Attitudes and Behaviors of Human Services Students Associated with Direct To Consumer Advertising (DTCA) of Prescription Drugs. Dissertation Abstracts International (UMI No. 3250898) Young, James H. (1969), The Medi cal Messiahs: A Social History of Health Quackery in Twentieth Century America Princeton University Press, Princeton, NJ Yuan, Sheng (2008), Public Responses to Direct to Consumer Advertising of Prescription Drugs, Journal of Advertising Research 48 ( 1), 30 41 Zajonc, Robert B. (1980), Feeling and Thinking: Preferences Need No Inferences, American Psychologist 35, 151 175

PAGE 104

104 BIOGRAPHICAL SKETCH Wan Seop Jung grew up in South Korea and attended Chung Ang University before arriving at the University of Florida. His compositions have been performed across the United States and Korea, and his work can be found in Health Marketing Quarterly and Journal of Communication in Healthcare Jung s interests include direct to consumer advertising, anti smoking campaign, and social marketing.