Scholarly article on topic 'Systematic review of the effectiveness of mass communication programs to change HIV/AIDS-related behaviors in developing countries'

Systematic review of the effectiveness of mass communication programs to change HIV/AIDS-related behaviors in developing countries Academic research paper on "Health sciences"

0
0
Share paper
Academic journal
Health Education Research
OECD Field of science
Keywords
{""}

Academic research paper on topic "Systematic review of the effectiveness of mass communication programs to change HIV/AIDS-related behaviors in developing countries"

HEALTH EDUCATION RESEARCH Theory & Practice

Vol.21 no.4 2006 Pages 567-597 Advance Access publication 17 July 2006

Systematic review of the effectiveness of mass communication programs to change HIV/AIDS-related behaviors in developing countries

1 2 3 4

Jane T. Bertrand *, Kevin O'Reilly , Julie Denison , Rebecca Anhang and

Michael Sweat3

Abstract

This review systematically examined the effectiveness of 24 mass media interventions on changing human immunodeficiency virus (HIV)-related knowledge, attitudes and behaviors. The intervention studies were published from 1990 through 2004, reported data from developing countries and compared outcomes using (i) pre- and post-intervention data, (ii) treatment versus control (comparison) groups or (iii) postintervention data across levels of exposure. The most frequently reported outcomes were condom use (17 studies) and knowledge of modes of HIV transmission (15), followed by reduction in high-risk sexual behavior (eight), perceived risk of contracting HIV/acquired immunodeficiency syndrome (AIDS) (six), interpersonal communication about AIDS or condom use (six), self-efficacy to negotiate condom use (four) and abstaining from sexual relations (three). The results yielded mixed results, and where statistically significant, the effect size was small to moderate (in some cases as low as 1-2% point increase). On two of the seven outcomes, at least

Johns Hopkins University, Bloomberg School of Public Health, Center for Communication Programs, Baltimore, MD 21202, USA, 2Department of HIV/AIDS, The World Health Organization, Geneva, Switzerland, 3Department of International Health, Johns Hopkins University, Bloomberg School of Public Health, Baltimore, MD 21202, USA and 4PhD Program in Health Policy, Harvard University, Cambridge, MA 02138, USA *Correspondence to: J. T. Bertrand. E-mail: jbertran@jhuccp.org

half of the studies did show a positive impact of the mass media: knowledge of HIV transmission and reduction in high-risk sexual behavior. Further rigorous evaluation on comprehensive programs is required to provide a more definitive answer to the question of media effects on HIV/ AIDS-related behavior in developing countries.

Introduction

The mass media have played a visible role in the human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) epidemic in developing countries since its onset in the early 1980s [1, 2]. Although many denied or minimized the importance of HIV/AIDS in the early days of the epidemic, almost all countries in the developing world used some form of mass communication to address the issue [2, 3]. Early on, these efforts focused on raising awareness of the existence of HIV/AIDS, the modes of transmission and the means of prevention. These efforts met with considerable success in raising awareness: in most countries, >90% of the population know the basic facts about HIV/AIDS [4]. The second generation of communication programs in the late 1980s and throughout the 1990s tended to focus more specifically on behavioral change related to abstinence, limiting one's number of sexual partners, and using condoms. In recent years, communication programs have expanded to address the full continuum from prevention to treatment to care and support [5]. Most of the mass media campaigns to date have focused on members of the general

© 2006 The Author(s). doi:10.1093/her/cyl036

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/ licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

public, or more narrowly on youth, but not on other high-risk populations.

The purpose of this paper is to review and synthesize the data from developing countries on the effectiveness of mass media interventions in changing HIV-related knowledge, attitudes and behaviors. Mass media interventions are any programs or other planned efforts that disseminate messages to produce awareness or behavior change among an intended population through channels that reach a broad audience. These channels include radio, television (TV), video, print and the Internet, and can take different forms such as radio variety shows, songs, spots, soap operas, music videos, films, pamphlets, billboards, posters and interactive Web sites. In the analysis, we have distinguished between 'broadcast' interventions, which include radio and/or TV, thus having the potential to reach a national audience, and 'small media' with more local reach (e.g. posters, pamphlets, audio programming, dramas and puppet shows). The latter tend to be face-to-face, interactive and community-based, with greater involvement of local stakeholders.

Some readers may seek the answer to a related question: why are some campaigns more effective than others? That is, what elements distinguish good campaigns from less effective ones? Unfortunately, this question goes beyond the scope of this paper, for reasons discussed in the final section.

Conceptual framework for the effects of mass media

Figure 1 illustrates how communication programs are expected to change HIV-risk behaviors.

Social, cultural, political, legal and economic factors often serve as obstacles to behavior change, though context can also facilitate change in certain circumstances. Within this context, the mass media are expected to affect a series of psychosocial factors, including knowledge, attitudes and self-efficacy. Changes in these factors are hypothesized to influence specific behaviors or practices, the most common of which are abstinence, reduction in number of sexual partners and condom use. In

countries in which the primary mode of HIV transmission is through sexual relations, the practice of these behaviors reduces the prevalence of HIV, the 'health outcome'. Program evaluation determines the degree to which the campaign reaches its objectives; it helps planners and scholars understand how or why a particular campaign worked and it provides information relevant for planning future activities [5].

Methodology

Inclusion and exclusion criteria

After establishing the written definition and theoretical framework for mass media, the synthesis team developed inclusion and exclusion criteria for study citations. To be eligible, papers had to present a mass media intervention as defined above; employ an evaluation design that compared outcomes using (i) pre- and post-intervention data, (ii) treatment versus control (comparison) groups or (iii) post-intervention data across levels of exposure; be published in a peer-reviewed journal from 1990 through 2004; and present data from a developing country, defined as a country bearing the World Bank designation of low-income, lower-middle income or upper-middle income economy [6]. Evaluation studies of condom social marketing campaigns were excluded, as these articles were reviewed in a separate analysis.

Search and acquisition

Trained staff used these criteria to search for eligible citations. A broad search was first initiated on computer-based search engines including the National Library of Medicine's Gateway system, PsycINFO, Sociological Abstracts, EMBASE and the Cumulative Index to Nursing & Allied Health Literature. We also searched the reference sections of papers that were selected for inclusion in the review. These new citations were acquired, screened, and if accepted, subjected to additional reference searches. The process was iterated until no new papers were identified. To supplement the computer database searches, we hand searched the

Legend

Determinant

Intervention Component

Intermediate Outcomes

Health Outcomes

Fig. 1. Conceptual framework for the effect of communication programs.

journals AIDS, AIDS and Behavior, AIDS Care, AIDS Education and Prevention and Journal of Health Communication for eligible citations. In addition, we contacted experts in the field to review our list of papers, and we solicited any missing references that they recommended. Additionally, we carefully reviewed the references from previous review papers and meta-analyses for possible citations. Staff downloaded the results from all searches into a database system.

The Project Principal Investigator (PI) and the Project Coordinator separately reviewed the pooled database generated by the search staff, and categorized the citations as (i) primary citations qualifying for inclusion in the synthesis; (ii) background citations not qualifying for inclusion but providing valuable information on mass media interventions;

(iii) citations to be acquired for further inquiry or

(iv) not relevant. The separate screened files from the PI and the Coordinator were then merged for comparison; citations with discordant screening were discussed to establish consensus (see Fig. 2).

Coding

Coding was conducted on all primary and background citations. Two independent coders extracted detailed information from each primary article using a structured coding form. Extracted data were transferred to an SPSS statistical database (SPSS™, Chicago, IL) for identification of intercoder discrepancies. Intercoder resolution was performed by a third person to correct for data entry error and to resolve different interpretations of the presentation of results.

The study rigor of each primary citation was also systematically assessed to determine whether the studies could provide an unbiased quantitative assessment of intervention effectiveness. We assessed the rigor of each primary study using an eight-point scale developed for the project. The scale was additive, with one point awarded for each item. The items were: prospective cohort, control or comparison group, pre-/post-intervention outcome data, random assignment to treatment groups, random selection of subjects for intervention and assessments, attrition, comparison group matching,

Fig. 2. Selection of articles for systematic review.

comparison group matching outcome measures and minimum requirements for inclusion in contextual coding (see Table I). Many studies used several research approaches of differing rigor; the scoring in Table I reflects the level of rigor with which data were collected and analyzed for the outcomes reported here, not necessarily for the overall study.

Outcomes

In an effort to synthesize the results from studies with diverse outcomes, we selected a subset of seven variables, three psychosocial factors and four behaviors, for purposes of this review. The seven outcomes are

(i) knowledge of HIV transmission;

(ii) perceived personal risk of contracting HIV/ AIDS;

(iii) self-efficacy to negotiate condom use or protect oneself;

(iv) discussion with others about HIV/AIDS or condom use;

(v) abstinence from sexual relations;

(vi) reduction in high-risk sexual behavior and

(vii) condom use.

Within each of the outcome areas outlined above, multiple measures were reported by the studies under review. To assess results, we combined different operational definitions of each outcome and classified results accordingly (e.g. 'condom use' included condom use at last sex, condom use at last sex with a casual partner, condom use with a sex worker, ever used a condom).

Results

The intensive search of the published literature revealed 24 articles that systematically evaluated the effects of mass communication programs on HIV/AIDS-related knowledge, attitudes and behaviors in developing countries, and that met the criteria for inclusion (see Table II). Of these 24, five used TV (with supporting media or alone), seven employed radio (with supporting media or alone) and the remaining 12 used 'small media' (with or without interpersonal communication, such as a group meeting or counseling). The majority of the evaluations of mass media programs published from 1990 to 1999 focused on small media (10 of 13 studies). The majority of the evaluations from 2000 onwards (8 of 11) examined programs that used radio or TV, with or without other supporting media. Thus, we observe an evolution in the types of programs evaluated toward communication programs designed to reach larger audiences using radio and/or TV.

This research represents experiences from around the globe, with the largest concentration of studies from Africa (n = 12), followed by Asia (n = 7) and Latin America and the Caribbean (n = 5). Of the 24, three were published during 1990-94, nine during 1995-99 and 12 during 2000-04.

For all but three of the studies, rigor scores range from 2 to 5 out of a possible 8 points; two studies scored a 1 out of 8, while another featured a highly

Table I. Quality assessment scores

Study Cohort Control or Pre-/post- Random Random Follow-up Comparison groups Comparison Final

comparison intervention assignment of selection of rate of >80% equivalent on groups equivalent quality

group, or data participants to participants for socio-demographics at baseline on score

outcomes by the intervention assessment outcome measure (out of 8)

level of exposure

TV plus other media

[30] No No Yes No NR NA NR NA 1

[20] No Yes No No Yes NA NA NA 2

[14] Yes Yes Yes No Yes NA No (cross-sectional groups differed on age) NA 4

Before/after analysis [25] No Yes Yes No No NA No NA 2

Post-only exposure analysis [25] No Yes No No No NA No NA 1

TV alone

[26] No Yes No No Yes NA NA NA 2

Radio plus other media

[19] No Yes No No Yes NA No NR 2

[10] Yes Yes Yes Yes Yes Yes Yes No 7

Radio alone

[18] No Yes No No No NA No NA 1

[15] No Yes Yes No Yes NA NR NR 3

[24] No Yes Yes No Yes NA No NR 3

[27] No No Yes No Yes NA NA NA 2

[12] No Yes Yes No Yes NA No NR 3

Small media with interpersonal communication

[17] No Yes Yes No Yes NA Yes NA 4

[11] Yes No Yes No No No NA NA 2

[9] Yes No Yes Yes Yes Yes NA NA 5

[16] No Yes Yes Yes Yes NA No NR 4

[23] Yes Yes No Yes Yes NA NR NR 4

[22] No No Yes No Yes NA NR NA 2

[21] Yes No Yes NR NR Yes NA NA 3

[33] Yes Yes Yes No NR No No NR 3

Small media alone

[13] No Yes Yes Yes Yes NA NR NR 4

[34] Yes No Yes No No NR NA NA 2

[8] No No Yes No Yes NA NA NA 2

[7] Yes Yes Yes No Yes NR No NA 4

Quality assessment scores were calculated for those outcomes reported in this synthesis. Other outcomes not reported in this synthesis may have been subject to higher or lower levels of rigor.

Table II. Description of interventions and evaluation study designs

Author(s) Description of intervention

Setting and target group

Entertainment Theoretical Primary education? basis objectives

Duration of Nationwide Reach and campaign campaign? frequency

Description of evaluation

Study design

Sample characteristics

Overview

TV plus other media

[30] Haiti, general population

[20] South Africa, junior secondary school learners

The AIDS prevention effort consisted of radio and TV messages, billboards, face-to-face contact and condom promotion

• Reach: NR

• Frequency: NR

Serial cross-sectional design comparing sero-prevalence rates among various rural and urban populations before and after the intervention.

• Sample size: NR

• Age: NR

• Sex: NR

Soul City Yes Soul City The Soul City NR Yes • Reach: more than Serial cross- • Sample size:

programs include: theory of weekly drama one-third of the sectional design, 3150

Soul City TVs social and disseminates learners had been however exposure • Age: mean

and radio programs behavioral basic exposed to four analysis performed 15.75

including a weekly change. information different Soul City with post- • Sex: 44.1%

drama that covers about the media sources >10 intervention cross- male, 55.9%

a range of health epidemic and its times, and about sectional data only. female

issues and consequences two-thirds six and

disseminates basic (the primary more time. Urban

information about objectives of the learners were

the epidemic and radio and exposed more often

its consequences newspaper to Soul City TV and

(an adult program); components Soul Buddyz than

Soul buddyz TV were not rural learners, and

(a children's reported). Soul City radio was

program), listened to more

newspaper often by rural

materials in learners.

the form of • Frequency: NR

health education

booklets and

a national life

skills program for

school children

in Grades 8-12.

Author(s) Description of intervention Description of evaluation

Setting and Overview Entertainment Theoretical Primary Duration of Nationwide Reach and Study design Sample

target group education? basis objectives campaign campaign? frequency characteristics

South Africa,

general

populations

Intervention components included (i) media programs including Soul City radio, TV and life skills program; (ii) community AIDS awareness forums; (iii) peer educators, including condom demonstrations and distribution; (iv) support groups for people living with HIV and (v) social care programs.

• Reach: NR

• Frequency: NR

Serial cross-sectional design comparing

(i) before to after data and

(ii) an exposure analysis using post-intervention data only.

intervention routes, radio, TV, large group demonstrations and small group activites. The film has been aired on national television and in conjunction with mobile video units throughout West and Central Africa.

• Sample size: baseline: 421, follow-up: 416

• Age: mean 20.8

• Sex: 44.9%, male 55.1% female

Bobo-Dioulasso, The Roulez Yes NR To encourage NR Yes • Reach: at follow- Serial cross- • Sample size:

Banfora and Protegee campaign the adoption up, two-thirds of sectional baseline: 764,

Niangoloko, is a regional mass of responsible the sample had study design. follow-up: 1032

Burkino Faso; media communication sexual been exposed to Differences • Age: 15-49 years

truck drivers strategy developed behaviors by at least two of between • Sex: 100% male

and their around a 30-min film the targeted the Roulez baseline and

assistants drama entitled 'Roulez Proteegee' and aimed at providing AIDS prevention messages to truckers and other mobile people in different target countries. The campaign uses billboards posted at major truckstops along population Proteegee activities. • Frequency: NR follow-up populations led the authors to conduct both pre- post-analyses and an exposure analysis, using post-intervention data only, to evaluate the mass media intervention.

Author(s) Description of intervention

Setting and target group

Entertainment education?

Theoretical

Primary objectives

Duration of campaign

Nationwide campaign?

Reach and frequency

Description of evaluation

Study design

Sample characteristics

TV alone [26]

Abijan, Boudepe and N'Douci Cote d'Ivoire, general population with electricity

Radio plus other media

Mutare,

Maphisa,

Nemanwa,

Nzvimbo,

Tongogara,

Kwekwe and

Mubaira,

Zimbabwe,

youths aged

10-24 years

and adults

'SIDA dans la Cite' is a weekly TV soap opera that describes the life of a family touched by HIV/AIDS. The series features popular music by Alpha Blondy and describes realistic scenarios that people who have multiple sexual partners can identify with.

Six-month multimedia campaign, directed at young people in five pilot sites. The campaign included posters, leaflets, newsletter, radio program, launch events, dramas, peer educators, a hotline and training FP providers in clinics to be youth friendly and the designation of youth friendly clinics.

Steps to behavior change framework

To educate the public about AIDS. Each episode is design to introduce at least one major theme (example, shows that wife of infected man can remain uninfected if she uses condoms).

To increase reproductive health and contraception knowledge.

Five months Yes

. Reach: 65% of the study sample had seen at least one episode of the soap opera. Frequency: once a week for 5 months.

. Reach: 97% reported exposure to at least one campaign component; 61% to >3 components. Exposure to individual campaign components: posters: 92%, launch events: 87%, leaflets: 70%, dramas: 46%, hotline: 7%.

Frequency: radio program: 26 episodes of the 1-hour weekly variety show were broadcast, theater troupes: daily performances for 2 months.

Cross-sectional study design comparing individuals exposed to the interventions versus those not exposed in terms of sexual-risk behavior and condom use.

Non-randomized pre-/post-intervention trial cross-sectional assessment comparing (i) five intervention and two comparison sites and (ii) combining data from all sites and analyzing outcomes by exposure to intervention components. Individual unit of analysis. Random selection of participants.

• Sample size: 2150

• Age: 15-49 years

• Sex: 47.4% male, 52.6% female

Sample size: baseline: 1426 (I: 973; C:453), follow-up: 1400 (I: 1000; C: 400) Age: I: 10-14 years: 33.0%; 15-19 years: 45.3%; 20-24 years: 21.7% C: 10-14 years: 19.7%; 15-19 years: 49.8%; 2024 years: 30.5% > Sex: ~50% female; 50% male

Six months

Author(s) Description of intervention Description of evaluation

Setting and Overview Entertainment Theoretical Primary Duration of Nationwide Reach and Study design Sample

target group education? basis objectives campaign campaign? frequency characteristics

Radio alone [18]

Kunshan The intervention NR NR To promote Twelve months No • Reach: NR Pre-/post- • Sample size:

county, consisted of ABCs (i.e. • Frequency: randomized baseline: 748 (NR

Jiangsu written materials, abstinence, reading materials: controlled trial by study arm),

province videos, radio monogamy, self-study, radio comparing two follow-up: 710

in eastern program, condom use program: once a intervention (I: 366; C: 344)

China, workshops, small or two areas week for 30 min. villages and two • Age: mean age:

young group discussions, within ABCs) control villages I: 24 years;

adults (18- home visits, sampled from two C: 25 years

30 years) personal counseling townships that • Sex: I: 51% female;

and free supply of were matched on 49% male; C: 52%

condoms. socio-economic female, 48% male

Intervention and demographic

emphasized sexual characteristics.

abstinence prior to Attempted to

marriage and the recruit all young

use of condoms adults in each of

for sexually the four randomly

active people, selected villages.

regardless of

marital status.

St Vincent A radio campaign. NR Although To encourage parents Two months Yes • Reach: NR Cross-sectional • Sample size:

and the The campaign not to talk • Frequency: NR survey comparing weighted

Grenadines, targeted primarily explicitly to teens about participants who sample 297

parent of parents of teenage stated, safer sex and were exposed to (I: 213; C: 84)

teens, teens children and the research condom use. the intervention • Age: I (15-19

and other central message was questions (intervention arm) years: 39.6%;

adults 'When you can't refer to to those who were 20-29 years:

protect them anymore constructs not exposed 25.4%; 30-44

... condoms can'. from the (comparison arm). years: 26.2%;

theory of Individual unit of 45-54 years:

reasoned analysis. Nation 8.8%), C (15-19

action/theory wide quota sample. years: 38.1%;

of planned Data weighted by 20-29 years:

behavior. age and gender. 25%; 30-44

years: 20.2%; 45-54 years:

Sex: I (50.7% female, 49.3% male), C (57.2% female, 42.8% male)

Ol -si

Author(s) Description of intervention

Description of evaluation

Setting and Overview Entertainment Theoretical Primary Duration of Nationwide Reach and Study design Sample

target group education? basis objectives campaign campaign? frequency characteristics

St Lucia, The entertainment Yes Bandura's social To promote the Phase 1: Yes • Reach: 12% of Pre-/post- • Sample size:

general education radio soap cognitive theory use of family February adult population intervention trial Baseline 753;

Population opera 'Apwe Plezi' which posits planning, the 1996 to May of St Lucia were serial cross- combined follow-

addressed 37 that individuals prevention of 1997, Interim: regular listeners sectional assessment up 1238 (first

educational issues learn new HIV and other June-July • Frequency: Phase comparing (i) follow-up 741,

including knowledge, behaviors by sexually 1997; Phase 2: 1: 260 15-min before to after second follow-up

attitudes and behaviors observing and transmitted July 1997 to episodes were data with the two 497); non-listener:

related to family imitating the diseases (STDs), September broadcast Tuesday follow-up surveys 799; casual listener:

planning, HIV behaviors of gender equity 1998; Phase 3: through Friday, combined for 288; regular

prevention, gender others who and other social until 2000 with repeat analysis and (ii) listener: 51

equity, relationship serve as role development episodes shown outcomes by • Age: mean age:

fidelity and domestic models. goals. on Monday. Phase listening status 29.3 years

violence. The radio 2: 105 15-min (non-listener, casual • Sex: 52% female;

soap opera had episodes aired listener and regular 48% male

positive, negative and three times a listener). Individual

transitional role models week. Phase 3: unit of analysis.

whose fates provided Street Theater: Probability selection

vicarious learning 21 performances of study participants.

experiences for in 1998

listeners to demonstrate

the consequences of

alternative behaviors.

Tanzania, Entertainment Yes Based on the To stimulate Seventy-nine Yes • Reach: 1994: Non-randomized • Sample size: 1993

General education radio soap work of Miguel interpersonal months 47% reported pre-/post- baseline: I 1793, C

public opera emphasizing Sabido, which communication exposure 1997: intervention trial 859; follow-up 1

four key HIV/AIDS draws heavily about AIDS by 58% serial cross-sectional (1994): I 1924, C

prevention themes: (i) on Bandura's showing role • Frequency: twice assessment comparing 861; follow-up 2

STD treatment, (ii) social cognitive modeling of per week for (i) I versus C; (ii) I (1995): I 1940, C

condom use; (iii) theory. characters 30 min versus C, controlling 861

AIDS is incurable and discussing for eight independent • Age: NR

transmitted through HIV/AIDS. variables and radio • Sex: NR

sexual contact and (iv) ownership and (iii)

and that various a multiple linear

rumors about AIDS regression (MLR)

are false. analysis to regress

ward-level change in the dependent variables against wardlevel listenership and 20 control variables. Individual and ward unit ofanalyses. Random selection of participants.

Author(s) Description of intervention

Description of evaluation

Setting and Overview Entertainment Theoretical Primary Duration of Nationwide Reach and Study design Sample

target group education? basis objectives campaign campaign? frequency characteristics

Bucaramanga, A radio campaign NR NR To create Three months No • Reach: 70% Pre-/post- • Sample size:

Colombia, emphasized condoms awareness of reported hearing intervention serial baseline: 944;

general use to prevent AIDS the role of a radio spot cross-sectional follow-up: 1440

population (two 10-s spots shown condoms in about AIDS and assessment • Age: NR

10 times daily preventing the condom in the comparing • Sex: ~60% female;

Monday-Friday). The AIDS. last 3 months responses before 40% male

first spot opens and • Frequency: two and after the

closes with the slogan: 10-s intervention.

'Prepared men do not advertisements Individual unit of

die of AIDS' and were broadcast an analysis. Random

PROFAMILIA average of 10 selection of

recommends the times a day, household. All

'Majestic' condoms Monday through eligible persons in

for safe sexual Friday. these households

relations. The second were interviewed.

spot advises 'If you

are having sexual

relations in the next

24 hours, think about

AIDS, and think about

a safe condom.

''Tahiti'', a safe

condom'.

Northern, A weekly radio drama Yes Modeling, To create Nine months Yes • Reach: NR Pre-/post- • Sample size:

Zambia Bemba performed in Bemba social learning awareness • Frequency: 39 intervention trial baseline: 1613 (949

speaking over a 9-month time hierarchy of regarding the 30-min drama comparing (i) the low access group,

Zambians period (August 1991 effects risks of getting episodes were sample before 664 high access

to June 1992). Each AIDS and the broadcast weekly and after the group); follow-up:

episode lasted 30 min importance of intervention and 1682 (997 low

and portrayed two preventing (ii) changes over access group, 685

families in Lusaka and transmission time among high access group)

their friends as they of the virus. participants most • Age: mean age

responded to the likely and least baseline: 25.9 years;

problems of rearing likely to have mean age follow-

teenaged children, listened to the up: 26.9 years

maintaining radio intervention. • Sex: ~50% female;

friendships, making (The intervention 50% male

ends meet, having arm is the high

sexual relations and access group

learning about AIDS. versus the comparison arm which is the low access group.)

Ol -si -si

■Nj 00

Table II. Continued

Author(s) Description of intervention

Description of evaluation

Setting and target group

Entertainment Theoretical education? basis

Primary objectives

Duration of Nationwide Reach and campaign campaign? frequency

Study design

Sample characteristics

Small media with interpersonal communication [17]

Khon Kaen A five-part (5-day) Yes Modeling, To reduce Five days No • Reach: NR Non-randomized • Sample size:

province of motivational social risk of HIV • Frequency: NR intervention trial baseline: 390; post:

Northeastern audio-drama was learning transmission comparing pre-/ 689 (I: 339 C: 350)

Thailand broadcast over principles of (in a context post-data from six • Age: mean age:

(rural), general village community in which men villages in the early 30s

population loudspeakers. Ten development engage in intervention arm • Sex: 56% female;

posters with each day's extramarital and post- 44% male

major issues acted as sexual activity, intervention only

daily advertisements. specifically with data from six

The drama topics commercial sex villages in a control

included married workers). To arm. Individual unit

men engaging make HIV/AIDS of analysis

commercial sex 'real' to the

workers, risk reduction villagers.

and dialogues among

women, spouses

and men. Meetings

were also held

with village

leaders, volunteers

were trained as

facilitators and

village discussions

on AIDS were held.

Wattala and Three dramas, Yes NR To create Twenty-four No • Reach: NR Pre-/post-cohort • Sample size:

Wadduwa, Sri performed by awareness months • Frequency: the intervention trial. baseline: 154;

Lanka, general volunteers and regarding the drama was Individual unit of follow-up: 97

population actors emphasized transmission performed 58 analysis. Non- • Age: mean age:

living on the awareness about and prevention times, with random selection 29.8 years

west coast. HIV/AIDS of HIV/AIDS. attendance of of study sample. • Sex: 18% female;

transmission and 4500 people at 76% male; 6%

prevention. Flyers the formative non-response

with prevention research site, 500

and transmission at evaluation Site

information were 1 and 900 at

distributed during evaluation Site 2

the drama

performances.

Author(s) Description of intervention

Description of evaluation

Setting and Overview Entertainment Theoretical Primary Duration of Nationwide Reach and Study design Sample

target group education? basis objectives campaign campaign? frequency characteristics

Libreville and High school Yes NR To demystify Comic book No • Reach: out of 964 Pre-/post-cohort • Sample size:

Lambarene, students received a the condom in distributed once. students given the intervention trial using baseline: 974;

Gabon High 15-min classroom a funny and Follow-up comic book at a 10% random sample post: 771

school students presentation on unusual way in assessment conducted baseline, 728 had of eligible classes in • Age: mean

AIDS from a order to induce 15-30 days after read it at follow-up 11 non-randomly age: 19 years

doctor with time behavior distribution. • Frequency: ofthe selected schools. • Sex: 45.3% female;

for questions change and 728 students who Individual unit of 54.7% male

afterward. They limit the reported reading analysis.

were then given number of new the comic book,

a comic book HIV infections 43.7% read it cover-

containing 36 one- to-cover once,

to three-page stories 29.5% read it cover-

to take home and to-cover multiple

read on their own. times, 17.2% read it

almost completely and 9.6% read it partially. The average amount of time spent reading was 90 min, with a mode and median of 30 min and a range of 1 min to 4 days.

Managua, Health education NR NR To create NR No • Reach: 44% aware Randomized pre-/ • Sample size:

Nicaragua Urban campaign awareness of of the visit of the post-controlled trial baseline: 2160

Nicaraguans emphasized HIV sexual health education with serial cross- (I:1294, C: 866);

aged 15-45 transmission and transmission of team sectional assessment. follow-up: 2277

years condom use. HIV and the • Frequency: NR Compares two (I: 1396, C:875)

Campaign ways to prevent intervention and two • Age: mean

components infection; control communities age: 28 years

consisted of house emphasis on matched on socio- • Sex: —54% women;

visits, leaflets on sexual economic status. 46% men

HIV and AIDS, transmission Individual unit of

stickers, posters, and protection analysis. Intervention

calendars, t-shirts using condoms. assigned by a coin

and condom toss. Random

distribution. selection of

participants from households.

Ol -si

Ol 00 o

Author(s) Description of intervention

Description of evaluation

Setting and target group

Entertainment education?

Theoretical basis

Primary objectives

Duration of campaign

Nationwide campaign?

Reach and frequency

Study design

Sample characteristics

Eighteen rural Information was Yes Behavioral To give knowledge NR No • Reach: 81% of Randomized • Sample size:

parishes in disseminated change for and correct individuals in the controlled trial baseline: NR;

Masaka district, through large and interventions misconceptions intervention arm, presenting post- follow-up 1:1 1677,

south west small group model about HIV/STI, and 9% in the intervention exposure C 1687; Follow-up

Uganda. meetings, village and promote safer comparison data combining data 2: I 1567, C 1695

Community drama shows, village sexual behavior communities from the intervention • Age: median at

members video shows and and practices. reported attending and comparison follow-up males: I

group or one-to-one at least one communities on 33.5, C 34; females:

discussions with Information sexual behavior and I 33; C 34

community educators. education HIV incidence. I the • Sex: follow-up 1

Information leaflets communication exposed participants -56% female 46%

were distributed at (IEC) activity in the and C the unexposed male

each of the IEC past year participants.

activities. The social • Frequency: NR. Individual unit of

marketing of condoms analysis

and voluntary HIV

counseling and

testing services

were implemented

in all communities.

East Moyo, An information NR NR To create awareness Twenty-four No • Reach: attended Pre-/post-intervention • Sample size:

Uganda (rural), pamphlet entitled regarding general months information session: trial serial cross- baseline: 1486;

general 'AIDS: be informed information about the 59%; received sectional assessment follow-up: 1744

population, and protected' was AIDS epidemic in pamphlet about comparing (i) before • Age: age range:

including produced. Community Uganda and Moyo AIDS: 42% to after data and (ii) 15-49 years; mean

Sudanese educators held district; length of the • Frequency: NR follow-up sub group age: NR

refugees information sessions incubation period; data based on • Sex: 50.7% female;

about AIDS need for correct use of exposure to (a) info 49.3% male

prevention and care, condoms in casual session and pamphlet,

including several sexual relationships; (b) info session only,

sessions for the STD symptoms and (c) pamphlet only and

general public and need for immediate (d) no exposure to any

particular groups such treatment; tolerance intervention element.

as military, police and and support for people Individual unit of

traditional healers. with AIDS. analysis. Random

Each session included selection of study

the distribution of participants.

locally produced

pamphlets in English

and Madi as well as

condom

demonstration and

distribution.

Author(s) Description of intervention

Setting and target group

Entertainment education?

Theoretical basis

Primary objectives

Duration of Nationwide campaign campaign?

Reach and frequency

Description of evaluation

Study design

Sample characteristics

Chittoor District, Andhra Pradesh, India, truck drivers

Sungai Kolok and Betong, Thailand, female commercial sex workers

The intervention consisted of films of local drama performances, folk media, group discussions and counseling by social workers provided at work and at the truckers' residences.

Music and messages shared through walkmans/ cassettes given to the sex workers. The STD clinics distributed leaflets, comic books and free condoms and showed video tapes. Sex establishments showed video tapes, had trained peer educators, held meetings with owners/ managers to support condom use, distributed condoms and were visited bi-weekly by a nurse. Leaflets with two condoms were placed in hotel rooms. Posters and stickers displayed in hotels and sex establishments.

To inform truck drivers about HIV/ AIDS, the dangers associated with it, use of condoms for safer sex and to bring about behavior change in sexual behavior.

AIDS risk To increase correct reduction knowledge regarding model HIV and its

prevention. To increase women's

One year

perceived vulnerability and social support from peers and managers.

• Reach: NR

• Frequency: films were shown and group discussions held once in every 2 months, counseling provided once

a month, and the frequency of the folk media component was not reported.

• Reach: NR

• Frequency: NR

Before/after intervention design

• Sample size: baseline: 300; follow-up: 300

• Age: NR

• Sex: 100% male

Non-randomized pre-/post-intervention trial with serial cross-sectional assessment comparing two study arms. Specific analyses compare (i) change within arm from pre-to post-intervention and (ii) differences between the two arms using data from those individuals who were interviewed both at baseline and follow-up (a smaller sub-sample of the total study population) is tested.

► Sample size: baseline: 751 (I: 408, C: 419); follow-up: 739 (I: 343, C: 320); participants who completed both pre- and post-surveys

(I: 159, C: 124)

Age: mean age: 25 years (intervention site); 24 years (comparison site)

► Sex: 100% female

Ol 00 ro

Author(s) Description of intervention

Description of evaluation

Setting and target group

Entertainment education?

Theoretical basis

Primary objectives

Duration of campaign

Nationwide campaign?

Reach and frequency

Study design

Sample characteristics

Small media alone [13]

Khon Kaen Province of Northeastern Thailand (rural), married women

Cape Town, South Africa, general population

Educational pamphlet No NR To create awareness Two months No • Reach: NR Randomized • Sample size:

distributed to every of HIV transmission, • Frequency: NR controlled pre-/post- baseline: 330; post-

household by village symptoms, and trial comparing (i) intervention: 654

health workers and consequences. post-intervention data (I and C: NR)

research team from the intervention • Age: age range: 16-

members. Pamphlets arm (12 villages that 50 years; mean age:

presented information received pamphlets); 31 years

on HIV transmission, (ii) post-intervention • Sex: 100% female

symptoms and data from a comparison

consequences arm (six villages) that

primarily with did not receive

pictures, including pamphlets and (iii) pre-

humanoid cartoon intervention data from

condoms dancing a before group (data

across the back. pooled from six villages that eventually received pamphlets and completed the follow-up survey and six villages that did not get pamphlets or participate in the follow-up survey). Individual unit of analysis, random selection of study subjects. Non-cohort sample.

Live puppet Yes NR To create awareness Three weeks No • Reach: NR Pre-/post-intervention • Sample size:

performance. The regarding HIV/AIDS • Frequency: NR trial comparing (i) baseline: NR;

story is about a main transmission and before to after data follow-up: 208;

character who is prevention. among attendees and comparison group

infected with HIV and (ii) intervention arm (post-intervention

passes the virus onto (pre-/post-group) only): 96

others through his versus a comparison • Age: median age:

sexual liaisons before arm (attendees who 24 years

eventually dying of were surveyed only • Sex: 52.7% female;

AIDS. During the once after the 47.3% male

show, the central performance) in order

messages concerning to assess the impact of

the prevention of HIV the pre-intervention

infection are survey on the study

explained outcomes. Individual unit of analysis. Convenience sample of study participants.

Author(s) Description of intervention

Setting and target group

Entertainment education?

Theoretical basis

Primary objectives

Duration of campaign

Nationwide campaign?

Reach and frequency

Description of evaluation

Study design

Sample characteristics

Choma district Zambia, rural villages

Madras, Tamil Nadu, India, inner city slums

Intervention consisted Yes NR To increase NR No • Reach: NR Pre-/post-intervention • Sample size:

of theater understanding • Frequency: theater trial cross-sectional baseline: 427;

performances, health the transmission group gave 55 assessment. Individual intervention: 494

talks by clinic staff/ of AIDS performances for unit of analysis. • Age: age range: 15-

community workers To promote the use 13 000 villagers. Random selection of 69 years; mean age:

and the distribution of of condoms Rural health centers households. 31.3 years; median

pamphlets/posters. (RHC) staff and age: 28 years

The campaign focused community health • Sex: 57% female;

on the transmission of workers gave —250 43% male

AIDS and condom use talks to 14 000

with condoms villagers.

distributed from

health centers. Health

workers,

schoolteachers and

traditional healers also

attended AIDS

seminars/anti-AIDS

clubs.

Three drama Yes NR To create Twenty-four No • Reach: NR Pre-/post-intervention • Sample size: I: pre-/

performances on the awareness of months • Frequency: 120 1- trial among attendees post-group 93; C

topic of HIV/AIDS HIV/AIDS to 2-hour shows at the drama (post-intervention

performed by information. were performed, performances with only): 99

a community theater with an average a comparison group • Age: age reported

group called attendance of 1000 surveyed post- by sub-groups only.

Nalamdana ('Are you people. Two of the intervention only. • Sex: pre-/post-

well' in Tamil). three HIV/AIDS Analyses include (i) group: 23.4%

shows were before to after female; 76.3%

performed four comparisons within male; comparison

times and one the intervention arm group: 44.9%

performed twice and (ii) I (before to female; 55.1% male

during the after) versus C.

evaluation period. Individual unit of

analysis. Random

selection of

participants.

IEC = information education communication; RHC = rural health centers.

Ol 00 CO

rigorous randomized control design, and therefore, scored a 7 (see Table I). Twenty-one of the 24 citations reported findings based on pre- and postintervention data; nine of the 24 citations compared results from treatment versus control or comparison groups and nine analyzed post-intervention only data comparing outcomes by level of exposure. (Some studies employed more than one of these approaches, and as such the total sums to >24.)

Of the seven outcomes examined, far more studies reported on condom use (17) and HIV knowledge (15) than on reduction in number of partners (8), interpersonal communication (6), perceived risk (6), self-efficacy (4) or abstinence/ age at sexual debut (3). A complete summary of the outcome measures and associated intervention effects generated from each citation appears in Table III. The results by outcome were as follows.

Knowledge of transmission

Fifteen studies measured knowledge of HIV transmission as an outcome. Roughly half of these reported positive effects on all or a plurality of knowledge measures, with effect sizes ranging from 2 to 100% improvements in the proportion of respondents with better knowledge; of the remaining studies, roughly half showed positive effects for some measures or population subgroups (e.g. women only). For example, in India, Valente and Bharath [7; articles that met the criteria for this review are indicated by * in the references] found significant differences between the intervention group and the control group on the percentage correct on 12 knowledge questions (97 versus 94%). After watching an educational theater performance, subjects of Trykker etal. [8] significantly increased their 'rejection' of incorrect modes of transmission, such as 'using secondhand clothes from a person having AIDS' (48-68%), 'drinking from the same cup as a person having AIDS' (4260%) and 'kissing a person having AIDS' (2637%). Similarly, Milleliri etal. [9] found significant increases in knowledge of various modes of transmission after high school students had been exposed to a comic book program in Gabon. In a study which scored 7 on the rigor scale, Xiaoming

et al. [10] showed large, significant increases in the intervention group regarding knowledge of modes of HIV transmission, including sexual intercourse (77-95%), multiple sexual partners (69-93%) and sharing needles for drug use (67-95%).

On the other hand, McGill and Joseph [11] did not detect significant differences in knowledge after drama performances in Sri Lanka, and Yoder et al. [12] did not find significant differences in knowledge of transmission between those with high access to a radio drama in Zambia and those with low access to it.

Perceived risk of contracting HIV/AIDS

The six studies that evaluated perceived HIV risk were evenly distributed over the categories of positive effects, no change or mixed results. Evaluation of a pamphlet campaign in Thailand by Elkins et al. [13] showed no significant changes in perceived personal risk of HIV. Similarly, Peltzer and Promtussananon [14] found no relationship between risk perception and any of four mass media components under study in South Africa. Vaughan et al. [15] reported that, after 2 years of radio soap opera broadcasts in Tanzania, those in the intervention group were significantly more likely to perceive that they were personally at risk than before the intervention (55-61%). This increase occurred despite strong contrary secular trends; the control group showed substantial reductions in perceived risk over the same time period (72-55%). Yoder et al. [12] also reported that Zambians exposed to a radio drama showed significantly higher belief that they could get AIDS than Zambians who were not exposed (30 versus 21%). Interventions evaluated by Pauw et al. [16] and Elkins et al. [13] showed stronger evidence for increased perception of HIV risk among female subjects; in fact, the Thai audio drama evaluated by Elkins et al. [17] showed significant decreases in perceived risk among men in the intervention group, an unintended consequence of the intervention.

Self-efficacy

Four studies evaluated self-efficacy to protect oneself or convince a sex partner to use a condom.

Table III. Data showing effects of communication interventions on eight HIV-AIDS-related outcomes, by outcome

Outcomes of interest by study

Results

Knowledge: modes of transmission [13]

Knowledge scale (mean ± 95% CI) Those who did NOT communicate about HIV

Those who did communicate about HIV

Mean response to a knowledge scale based on 37 questions on mode of transmission, availability of a cure, modes of prevention, diagnosis and appearance of a person infected with HIV. Correct responses were given a score of '1' and incorrect responses or responses which might produce ineffective prevention actions were scored '—1'. Respondents' scores ranged from —2 to 14. Mean knowledge score (women)

Mean knowledge score (men)

Mean knowledge score for male and female respondents at baseline and follow-up. Scores range from 0 to 14 and were calculated by summing the correct answers to a series of knowledge-based questions, such as: Is it possible to tell from someone's appearance whether they are carrying AIDS? Is there a cure for AIDS? How can AIDS be spread? Overall knowledge score (% correct) Overall knowledge score (mean score) Overall knowledge score was calculated by summing correct answers to the nine questions shown below. Specific knowledge scale items (% correct):

Before (4.91 ± 0.37), C (5.98 ± 0.63), I (6.12 ± 0.38)

Before (6.17 ± 0.42), C (7.44 ± 0.43), I (7.70 ± 0.31)

I (7.14-10.22, SIG), C (9.66), [I versus C, SIG] I (6.94-8.55, SIG), C (9.08), [I versus C, NS]

Before to after (74.5-77.7%, NS) Before to after (7.08-7.39, NS)

There is a cure for AIDS Before to after (81.9-85.5%, NS)

Mosquitoes can cause AIDS Before to after (78.5-84.6%, NS)

People with multiple sex partners may be more Before to after (89.0-84.3%, NS)

likely to get AIDS

You can tell a person has AIDS by looking at them Before to after (73.8-71.4%, NS)

If you use a water glass that has been used by a Before to after (74.7-76.1%, NS)

person who has AIDS you may get the disease

If you use a sharp object that has blood from Before to after (90.6-89.5%, NS)

an AIDS person, you may get the disease

If you have AIDS, symptoms will show in Before to after (46.9-54.5%, NS)

a few weeks

Homosexuals may be more susceptible to Before to after (62.3-76.1%, SIG)

getting the disease

Except for no sexual relations, condoms are Before to after (72.6-77.5%, NS)

the best method to protect against AIDS

HIV transmitted by blood Before to after (94.3-97.4%, SIG)

HIV transmitted sexually Before to after (96.8-99.1%, SIG)

HIV transmitted from mother to child Before to after (46.7-75.3%, SIG)

Cite false (incorrect) mode of transmission Before to after (6.2-4.3%, NS)

Table III. Continued

Outcomes of interest by study

Results

[16] HIV transmitted by sexual relations

HIV transmitted by common use of sharp instruments

HIV transmitted by kisses, sweat and saliva

HIV transmitted by blood transfusion

HIV transmitted by sharing needles

HIV transmitted during pregnancy

Do not know how HIV is transmitted

[14] Mean HIV/AIDS knowledge score (individual items below) People can protect themselves from HIV by using condoms correctly every time during sex? Can a person get HIV from mosquito bites? People protect themselves from HIV by having one uninfected faithful sex partner? People protect themselves from HIV by abstaining from sex?

Can a person get HIV by sharing a meal with someone who is infected?

Can a person get HIV by getting injections with a used needle?

Can a healthy-looking person be HIV infected? Can a pregnant woman infected with HIV transmit the virus to her unborn child? Can a woman with HIV transmit the virus to her newborn child through breastfeeding? What can a pregnant woman do to reduce the Risk Mother to child transmission (MTCT) HIV/AIDS knowledge score (dependent variable): higher score; lower score [20] HIV knowledge as it relates to each of the four

mass media components using Pearson's correlation

[34] Knowledge of modes of transmission

(no. of correct answers) [25] Knowledge of AIDS

Knowledge of modes of transmission Knowledge of prevention strategies

Females: I (82-87%, SIG), C (85-87%, NS), [I versus C, NS]; males: I (89-90%, NS), C (82-91%, SIG), [I versus C, NS] Females: I (7-12%, SIG), C (7-5%, NS), [I versus C, SIG]; males: I (9-15%, SIG), C (9-7%, NS), [I versus C, SIG] Females: I (7-6%, NS), C (11-8%, NS), [I versus C, NS]; males: I (11-6%, SIG), C (11-8%, NS), [I versus C, NS] Females: I (30-27%, NS), C (23-20%, NS), [I versus C, NS]; males: I (32-36%, NS), C (35-32%, NS), [I versus C, SIG] Females: I (24-31%; SIG), C (19-28%, SIG), [I versus C, NS]; males: I (28-42%, SIG), C (28-35%, SIG), [I versus C, SIG] Females: I (3-2%, NS), C (1-1%, NS), [I versus C, NS]; males: I (1-1%, NS), C (0.3-1%, NS), [I versus C, NS] Females: I (13-8%, SIG), C (10-7%, NS), [I versus C, NS]; males: I (7-4%, NS), C (6-4%, NS), [I versus C, NS] Before to after (7.11 versus 7.33, NS)

Before to after (91.1-85.5%, SIG)

Before to after (76.4-61.3%, SIG) Before to after (76-85%, SIG)

Before to after (68.8-81.2%, SIG)

Before to after (72.2-70.5%, NS)

Before to after (91.8-89.4%, NS)

Before to after (74-87.5%, SIG) Before to after (84.9-87.4%, NS)

Before to after (86.1-84.4%, NS)

Before to after (18.4-69.9%, SIG)

Peer educator exposure to HIV/AIDS message (b = 0.44) SIG; Soul Buddyz (b = 0.49) SIG Soul City TV (r = 0.08) SIG; radio (r = -0.03) NS; Soul buddyz (r = 0.02) NS; newspaper materials (r = 0.03) NS Before to after (% NR, SIG)

Before to after (55.5-63.4%, SIG) Before to after (39-61.7%, SIG) Before to after (53-77.3%, SIG)

Table III. Continued

Outcomes of interest by study

Results

[8] Modes of transmission (% correct)

Using second-hand clothes from a person having AIDS From a mosquito which has bitten a person having AIDS Drinking from same cup or bottle as a person having AIDS

Kissing a person having AIDS

Getting a blood transfusion from a person having AIDS Eating from the same plates as a person having AIDS Mean score on HIV/AIDS knowledge scale Knowledge scale (% correct out of 12 true/false questions)

Mean HIV/AIDS knowledge scale score during 1993-95

Routes by which HIV can be transmitted Sexual intercourse Multiple sexual partners Sharing needles for drug use Blood transfusion Mother to infant

Routes by which HIV cannot be transmitted

Using a public toilet

Shaking hand or touching body

Mosquito bites

Swimming pool

Kissing

[12] Routes of HIV transmission Perceived risk of HIV/AIDS

[13] Are you at risk?

[21] [7]

Are you personally at risk?

[15] [12]

Do you consider yourself at risk because of AIDS (females)?

Do you consider yourself at risk because of AIDS (males)?

Risk perception as it relates to each of the four mass media components using Pearson's correlation Perception of personally being at risk of HIV

Belief that she could get AIDS (women only)

Before to after (48.2-68.2%, SIG) Before to after (33.4-35.6%, NS) Before to after (41.9-60.3%, SIG)

Before to after (26.2-37.4%, SIG) Before to after (78.6-94.5%, SIG) Before to after (49.6-64.2%, SIG) Before to after (8.1-16.4%, SIG) Before to after (71-97%, SIG), I (97%), C (94%), [I versus C, SIG] I (base rate of 10.0 and increased 0.7 points), C (base rate of 10.09 and declined 0.5 points), analysis of variance (ANOVA) SIG; MLR, NS

I (77-95%, SIG), C (77-78%, NS), [I versus C, SIG] I (69-93%, SIG), C (68-74%, SIG), [I versus C, SIG] I (67-95%, SIG), C (68-67%, NS), [I versus C, SIG] I (59-92% SIG), C (58-65% SIG), [I versus C, SIG] I (76-97% SIG), C (80-79% NS), [I versus C, SIG]

I (46-92%, SIG), C (44-47%, NS), [I versus C, SIG] I (39-93%, SIG), C (40-45%, SIG), [I versus C, SIG] I (12-84%, SIG), C (13-10%, NS), [I versus C, SIG] I (26-88% SIG), C (30-34% NS), [I versus C, SIG] I (31-90%, SIG), C (34-33%, NS), [I versus C, SIG] I (NR) versus C (NR), [I versus C, NS]

Those who did NOT communicate about HIV: before (40%), C (36%), I (36%), [before versus C, NS], [I versus C, NS]; those who did communicate about HIV: before (44%), C (51%), I (45%), [before versus C, NS], [I versus C, NS] Women: no: I (61-43%), C (59%); yes: (36-54%), C (40%); uncertain: (3-3%), C (1%), [before to after, SIG], [I versus C, SIG]

Men: no: I (67-80%), C (82%); yes: I (22-18%), C (18%); uncertain: I (11-2%), C (0), [before to after, SIG], [I versus C, NS]

Very much: I (4-7%), C (4-5%); quite: I (8-15%), C (13-13%); a little: I (6-5%), C (7-7%); not at all: I (82-73%), C (77-76%), [I versus C, SIG] Very much: I (6-9%), C (6-9%); quite: I (11-11%), C (11-13%); a little: I (9-9%), C (6-7%); not at all: I (75-72%), C (77-71%), [I versus C, NS] Soul City TV (r = 0.00) NS; radio (r = —0.00) NS; Soul buddyz (r = 0.03) NS; newspaper materials (r = 0.03) NS

I (55-61%), C (72-55%); logit loglinear, logistic regression and MLR SIG I (29.7%), C (21.3%), [I versus C, SIG]

Table III. Continued

Outcomes of interest by study

Results

Self-efficacy [13]

Self-efficacy scales to protect oneself if husband suspected to be infected with HIV (mean ± 95% CI): those who did NOT communicate about HIV; those who did communicate about HIV The efficacy scale was calculated based on 10 potential responses to questions on whether and how a wife who suspected her husband might be infected could protect herself from infection. Correct responses were given a score of '1' and incorrect responses or responses which might produce ineffective prevention actions were scored '-1'. The scale contained two negatively scored items. Scores ranged from 0 to 6 If a married woman thinks she is at risk because of her husband, can she protect herself? Self-efficacy as it relates to each of the four mass media components using Pearson's correlation Have confidence to convince sex partner to use condom

Talked to others about HIV/AIDS

[13] Percent indicating that they had had a conversation about HIV/AIDS with: spouse; anyone

[19] [18] [25]

Have talked to others about condoms (women) Have talked to others about condoms (men) Have talked to others about AIDS (women) Have talked to others about AIDS (men) Women talked to husband about condoms Men talked to their wife about condoms Women talked to husband about AIDS Men talked to wife about AIDS Discussions with anyone about STIs/AIDS Suggested condom use to one's partner Likelihood of having spoken to a colleague about AIDS in the last year

Mentioned talking about AIDS with spouse (married only)

Mentioned talking about AIDS with children (those with children)

Talked about AIDS with spouse (prompted and unprompted responses)

Talked about AIDS with children (prompted and unprompted responses) Talked to spouse about AIDS (married only) Talked to spouse about AIDS (married only) among respondents with high access to radios Talked with child about AIDS (those with children only) Talked with child about AIDS (those with children only) among respondents with high access to radios

Before (2.32 ± 0.15), C (2.62 ± 0.24), I (2.55 ± 0.12), [I versus C, NS]; before (2.54 ± 0.22), C (2.79 ± 0.20), I (2.88 ± 0.14), [I versus C, NS]

Yes: I (86-85%, NS), C (79%), [I versus C, NS]

Soul City TV (r = 0.07) SIG; radio (r = 0.02) NS; Soul buddyz (r = 0.06) SIG; newspaper materials (r = 0.05) SIG I (83-92% SIG), C (78-84%, SIG) [I versus C, SIG]

Before to C (31-43%, SIG), I (44%), C (43%), [I versus C, NS]; before to C (27-61%, SIG), I (54%), C (61%), [I versus C. NS] I (32-54%, SIG), C (37%), [I versus C, SIG] I (56-71%, SIG), C (67%), [I versus C, NS] I (54-99%, SIG), C (61%), [I versus C, SIG] I (68-73%, SIG), C (78%), [I versus C, NS] I (43-68%, SIG), C (48%), [I versus C, SIG] I (62-65%, NS), C (47%), [I versus C, SIG] I (43-86%, SIG), C (61%), [I versus C, NS] I (66-78%, SIG), C (74%), [I versus C, NS] I (78%), C (67%), [I versus C, SIG] I (59.5%), C (56.5%), [I versus C, NS] Saw Roulez Protege billboards (OR = 1.37) NS, Heard Roulez Protege; radio shows (OR = 1.61) SIG, Saw Roulez Protege TV shows (OR = 1.01) NS, Participated in Roulez Protege; discussion group (OR = 1.79) SIG Before to after (22.7-34.9%, SIG), I (26.2-43.0%, SIG), C (20.4-29.6%, SIG), [I versus C, SIG] Before to after (5.8-8.4%, SIG), I (7.2-12.6%, NS), C (4.9-8.6%, NS), [I versus C, SIG] Before to after (38.1-52.7%, SIG), I (43.0-60.3%, SIG), C (34.9-47.7%, SIG), [I versus C, OR 1.18, NS] Before to after (13.8-20.4%, SIG), I (17.2-27.3%, SIG), C (11.7-16.1%, SIG), [I versus C, OR 1.18, NS] I (58.7%), C (50.8%), [I versus C, SIG] I (59.6%), C (60.8%), [I versus C, NS]

I (27.4%), C (18.9%), [I versus C, SIG] I (31.0%), C (24.5%), [I versus C, NS]

Table III. Continued

Outcomes of interest by study

Results

Abstains from sexual relations

[19] Adopting safer sexual behavior: said no to sex Adopting safer sexual behavior: continued abstinence

[16] Changed sexual practices (because of AIDS): abstinence or monogamy

[20] Delay of sexual activity among sexually inactive sample as it relates to each of the four mass media components using an F-test

Reduced number of sexual partners

[19] Sexually experienced participants who stuck to one partner due to the campaign Adopting safer sexual behavior: avoided sugar daddy

[16] Changed sexual practices (because of AIDS): abstinence or monogamy

[14] Number of non-commercial sex partners in the last 12 months

Number of commercial sex partners in the last 12 months

[22] Proportion of single women having casual sex in past year

Average number of casual partners (males) Proportion of men engaging in casual sex

[21] Percentage reporting pre-marital or extra-marital sex

[15] Number of sexual partners in previous year reported by sexually active men for 1993, 1994 and 1995 Number of sexual partners in previous year reported by sexually active women for 1993, 1994 and 1995

[10] Only one sexual partner in the past year

[12] Made another safer change to avoid AIDS (usually 'stick to one partner') among males

Made another safer change to avoid AIDS (usually 'stick to one partner') among females

Condom use

[13] Women who consider themselves to be at risk and use condoms to prevent transmission: those who did NOT communicate about HIV; those who

did communicate about HIV

[17] Among men, the frequency of condom use with prostitutes

Among men, the frequency of condom use with wife

[19] Sexually experienced participants who started condom use due to the campaign

[18] Ever used a condom Always uses condoms

I (52.7%), C (31.6%), [I versus C, SIG] I (31.5%), C (22.3%), [I versus C, SIG] Females: I(14-33%, SIG), C(13-45%, SIG), [I versus C, NS]; males: I(7-29%, SIG), C (5-28%, SIG), [I versus C, NS] Soul City TV (F = 8.00) SIG; radio (F = 0.53) NS; Soul buddyz (F = 0.11.54) SIG; newspaper materials (F = 0.89) SIG

I (20.4%), C (2.0%), [I versus C, SIG]

I (11.0%), C (9.1%), [I versus C, SIG] Females: I(14-33%, SIG), C(13-45%,SIG), [I versus C, NS]; males: I (7-29%, SIG), C (5-28%, SIG), [I versus C, NS] Before to after (0 = 3.4 versus 7.7%; 1 = 46.6 versus 76.3%; 2-3 = 34.2 versus 14.7%; 4+ = 16 versus 1.3%) SIG

Before to after (0 = 43.7 versus 94.4%; 1 = 37.9 versus 3.8%; 2-3 = 10.3 versus 1.7%; 4+ = 6.8 versus 0%) SIG Before to after (11-3%, SIG)

Before to after (0.29-0.19, SIG) Before to after (10.6-9.6%, NS) Before to after (40 versus 21%, SIG) I (2.3 to 1.6 to 1.6), C (2.2 to 1.5 to 1.9) ANOVA, MLR SIG

I (1.9, 1.3, 1.2), C (1.8, 1.2, 1.3) ANOVA, MLR SIG

I (93-96% NS), C (95-95%, NS), [I versus C, SIG] Before to after (54.4-54.8%, NS), I (56.0-58.1%, NS), C (53.2-51.9%, NS), [I versus C, NS] Before to after (35.8-42.6%, SIG), I (39.2-49.4%, NS), C (34.0-39.3%, NS), [I versus C, NS]

Before (2%), C (0%), I (7%), [before versus C, NS], [I versus C, NS]; before (13%), C (9%), I (11%), [before versus C, NS], [I versus C, NS]

Never: I (75-67%), C (80%); sometimes: I (9-15%),

C (7%); always: I (16-18%), C (13%);

[before to after, NS], [I versus C, SIG]

Never: I (76-70%), C (76%); sometimes: I (21-25%),

C (23%); always: I (3-5%), C (1%), [before

to after, NS], [I versus C, NS]

I (10.5%), C (2.0%), [I versus C, SIG]

I (69.5%), C(57.5%), [I versus C, NS] I (25.3%), C (26.0%), [I versus C, NS]

Table III. Continued

Outcomes of interest by study

Results

[16] Frequency of condom use among sexually experienced females

Frequency of condom use among sexually experienced males

Changed sexual practices (because of AIDS): use of condoms

[20] Condom use at last sex among sexually active

participants as it relates to each of the four mass media components using an F-test. [14] Male condom use at last sex with commercial

partner

Male condom use at last sex with non-commercial partner

Always used a condom over the last 12 months

with commercial partner(s)

Always used a condom over the last 12 months

with non-commercial partner(s)

Reasons for not using a condom when having

sex last with non-commercial partners

Past use of male condom if not used during the last 12 months with any partner Consistent condom use index Ever used a female condom

Consistent condom use index (dependent variable): higher score

[23] Likelihood of ever used condoms among males

Likelihood of ever used condoms among females

Likelihood use condom with last casual partner among males

Likelihood use condom with last casual partner among females

Never: I (91-83%), C (91-89%); sometimes: I (6-9%), C (6-6%); always: I (3-7%), C (3-5%), [I versus C, NS] Never: I (69-59%), C (70-64%); sometimes: I (25-28%), C (23-23%); always: I (6-13%), C (7-14%), [I versus C, NS] Females: I (42-44%, NS), C (40-29%, NS), [I versus C, NS]; males: I (33-44% SIG), C (34-40%, NS), [I versus C, NS]

Soul City TV (F = 0.00) NS; radio (F = 0.00) NS; Soul buddyz (F = 0.36) NS; newspaper materials (F = 0.00) NS

Before to after (NR-46%, NS)

Before to after (NR-33%, NS)

Before to after (25-23.1%, NS)

Before to after (14.9-11.1%, NS)

Before to after: not available (35.2-13.9%, SIG); too expensive (2.5-9.2%, NS); partner objected (17.6-33.9%,NS); don'tlike them (32.1-41.5%, NS); used other contraceptive (14.3-17.4%, NS); didn't think it was necessary (12.9-36.4%, SIG); didn't think of it (21.336.6%, SIG)

Before to after (55.9-49.7%, NS)

Before to after (NR versus NR, NS) Before to after (4.9 versus 5.4%, NS) HIV/AIDS messages from radio messages (b = -0.28) SIG; HIV/AIDS messages from Soul City radio drama (b = -0.22) SIG

No change (used one at both rounds): I (19%), C (22%) reference group; no change (never used one at both rounds): I (62%), C (62%), OR 0.72, NS; changed (from never to ever): I (13%), C (11%), OR 0.67, NS; changed (from ever to never): I (4%), C (5%), OR 0.39, NS No change (used one at both rounds): I (7%), C (7%) reference group; no change (never used one at both rounds): I (84%), C (87%), OR 1.43, NS; changed (from never to ever): I (5%), C (4%), OR 2.69, NS; changed (from ever to never): I (3%), C (2%), OR 2.70, NS No change (used one at both rounds): I (38%), C (36%) reference group; no change (never used one at both rounds): I (39%), C (42%), OR 0.64, NS; changed (from never to ever): I (18%), C (14%), OR 0.76, NS; changed (from ever to never): I (4%), C (8%), OR 0.13, NS No change (used one at both rounds): I (22%), C (33%) reference group; no change (never used one at both rounds): I (57%), C (58%), OR 1.62, NS; changed (from never to ever): I (6%), C (13%), OR 20.75, NS; changed (from ever to never): I (9%), C (3%), OR 1.76, NS

Table III. Continued

Outcomes of interest by study

Results

[22] Ever used condoms

Ever used condoms and had casual sex in the past year

Ever used condoms and had no casual sex in the past year

[26] Likelihood of condom use at last sexual encounter

[25] Condom use with occasional partner (among partners with occasional partners) Likelihood of condom use with an occasional partner

Ever use of condoms [21] Condom use

[15] Current condom use among respondents with more

than one sex partner for 1993, 1994 and 1995 [24] Currently use condoms among men in sexual unions

[27] Ever used condoms

Used condoms in past year [10] Frequency of condom use in the past year: never; sometimes; always

Condom use in last sexual intercourse [12] Ever used a condom among males

Ever used a condom among females

Ever used condoms (unmarried men or married men with non-marital partners) Ever used condoms (unmarried women or married women with non-marital partners)

All respondents: before to after (6-6%, NS); women: before to after (3-3%, NS); men: before to after (9-9%, NS)

All respondents: before to after (23-46%, SIG); women: before to after (6.3-33.3%, NS); men: before to after (27-48.2%, NS)

All respondents: before to after (4.4-3.7%, NS); women: before to after (2.8-2.4%, NS); men: before to after (6.1-5.0%, NS) Female: non-viewers (reference group); saw one to four episodes: OR 0.97 NS; saw five to nine episodes: OR 0.92 NS; saw 10+ episodes: OR 1.31 SIG; Male: non-viewers (reference group); saw one to four episodes: OR 0.93 NS; saw five to nine episodes: OR 1.71 SIG; saw 10+ episodes: OR 2.38 SIG Before to after (69.4 versus 89.6%, SIG)

Saw Roulez Protege billboards (OR = 0.69) NS; heard

Roulez Protégé; radio shows (OR = 0.89) NS; saw

Roulez Protege; TV shows (OR = 1.81) NS; participated

in Roulez Protege discussion group (OR = 0.65) NS

Before to after (28.7 versus 31.4%) NS

Before to after (40.8 versus 86.7%, SIG)

I (6%, 9%, 13%), C (15%, 10%, 2%) logit loglinear

and logistic regression SIG, MLR NS

Before to after (21-24%, NS); by listening status: non-

listener (24%); casual listener (20%); regular listener

(31%) [NS]

Before to after (25.2-33.6%) [SIG] Before to after (8.2-12.1%) [SIG]

I (81-80% NS), C (79-79%, NS), [I versus C, SIG];

I (14-12% NS), C (19-18%, NS), [I versus C, SIG];

I (5-8% NS), C (2-3%, NS), [I versus C, SIG]

I (9-14% SIG), C(13-12%, NS), [I versus C, SIG]

Before to after (38.0-43.1%, SIG) I (49.7-50.5%, NS),

C (29.0-36.4%, SIG), [I versus C, SIG]

Before to after (14.3-22.5%, SIG) I (20.8-27.0%, NS),

C (10.9-20.3%, SIG), [I versus C, SIG]

Before to after (44.0-51.7%, NS) I (52.9-59.4%, NS),

C (36.7-43.9%, NS), [I versus C, SIG]

Before to after (17.4-29.1%, SIG) I (20.3-33.3%, NS),

C (16.1-27.3%, SIG), [I versus C, NS]

Any value with a significance of P < .05 is in bold. MCT = mother to child transmission.

The findings were evenly split between positive effects and no effects. In Thailand, Elkins et al. [13] reported no significant changes in self-efficacy to protect oneself if one's husband is suspected to be

infected with HIV. In a later study, when Elkins et al. [17] asked Thai villagers, 'if a married woman thinks she is at risk because of her husband, can she protect herself', no differences were observed

between intervention and control groups. In China, both the intervention and the control group showed increases in confidence to convince sex partners to use condoms, but the increase for the intervention group was significantly larger (83-92% versus 78-84%) [8]. Peltzer and Promtussananon [14] found significant associations between self-efficacy and exposure to newspaper materials, the Soul City campaign's TV programming and Soul Buddyz, a Soul City spin-off campaign targeted at children.

Talked to others about HIV/AIDS or condom use

The six studies measuring this outcome differed in terms of the person with whom the discussion occurred (e.g. spouse, children, 'someone', colleague). Three studies reported on communication with a spouse or partner regarding AIDS, while two studies measured communication with a spouse or partner regarding condoms. Results were split among positive, mixed and no effects, regardless of who the discussions were with and whether they were about AIDS or condoms. For example, the evaluation by Middlestadt et al. [18] of a radio-only campaign in St Vincent and the Grenadines (1995) did not find a difference between those exposed to the campaign and those unexposed on communication about condoms; however, Elkins et al. [17] found substantial, significant differences between those exposed to the Thai audio drama and those who were not (68 versus 48% for women, 65 versus 47% for men). In the same study, significantly more women and men talked with their spouses about AIDS after the campaign than before (43-86% for women, 66-78% for men); however, this difference was not significantly different from the control group. The same finding was reported in Elkins's earlier study in Thailand, suggesting that a secular trend toward increased communication about HIV may have been at play [13].

or mixed. Kim et al. [19] reported that those in the intervention group of a multimedia campaign in Zimbabwe were significantly more likely than controls to have continued abstinence (32 versus 22%) and to have 'said no to sex' (53 versus 32%). Pauw et al. [16] reported significant increases in both intervention and control groups for changing sexual practices toward abstinence or monogamy due to AIDS, but found no significant differences between them. In South Africa, Peltzer and Prom-tussananon [14] found significant associations between delay of sexual activity and exposure to newspaper materials, the Soul City campaign's TV programming and Soul Buddyz.

Reduction in high-risk sexual behaviors

The clear majority of eight studies in this category yielded positive effects. The studies measured different aspects of the phenomenon, including number of sexual partners in the past year, percentage of men engaging in casual sex, percent avoiding a sugar daddy and percent avoiding commercial sex workers. Peltzer et al. [20] found that those exposed to the Soul City campaign in South Africa had significantly fewer non-commercial and commercial sex partners in the past year. Ubaidullah [21] reported that after receiving an intervention, only half as many truck drivers reported pre-marital or extra-marital sex. Kim et al. [19] reported a substantial difference between intervention and control groups on the 'sticking to one partner' variable (20 versus 2%). The proportion of single women having casual sex decreased significantly from before to after a campaign evaluated by Schopper et al. [22] in Uganda (11-3%); although the proportion of single men having casual sex did not change, the average number of casual partners did significantly decline from 0.29 to 0.19. Small but significant changes were found by Xiaoming et al. [10] and Vaughan et al. [15] with regard to number of partners in the previous year.

Abstinence from sexual relations

Only three studies measured this outcome, each using different measures; the results were positive

Condom use

More studies evaluated effectiveness based on condom use (17 of 24) than any other outcome.

Of the 12 evaluations of programs using radio and/ or TV (with or without other media), all but one included this outcome. The modal response over the 17 studies was 'no effect' followed by 'positive' or 'mixed'. In Uganda, Quigley et al. [23] did not show significant change on the measures 'ever use condoms' or 'use condom with last casual partner'. The radio campaign in St Vincent and the Grenadines [18] also failed to show changes on the variables 'ever used a condom' and 'always used a condom'. The evaluation of the educational radio soap opera in St Lucia also did not report changes in condom use among men in sexual unions [24]. Pauw et al. [16] showed no significant increases in the frequency of condom use due to the house-to-house campaign in Nicaragua, and Schopper et al. [22] reported no significant increases in ever use of condoms after a pamphlet and community education campaign in Uganda. Increases in condom use were not found to be associated with exposure to elements of the Roulez Protege campaign in Burkina Faso [25] or a variety of HIV-related media in South Africa [14].

However, Kim et al. [19] reported 5-fold higher condom use among sexually experienced campaign participants than among controls (11 versus 2%), and Shapiro et al. [26] found that likelihood of condom use at last sexual encounter was significantly higher among those who had seen more episodes of 'SIDA dans la Cite' TV drama. Ever use of condoms and use of condoms in the past year both rose sharply in Colombia after a radio advertising campaign aimed at increasing condom use (25-34% and 8-12%, respectively) [27]. Significant 1-2% differences between intervention and control groups were observed by Xiaoming et al. [10] in China on the variables frequency of condom use and condom use in last sexual intercourse.

In addition to examining these outcomes for the group of 24 studies, we attempted to identify patterns by type of intervention, distinguishing between those that used radio and/or TV (with or without other media) and those employing small media (with or without interpersonal communication). Given the small number of studies (3-8) that reported five of the outcomes, we opted only to ex-

amine this question for knowledge (with 15 studies) and condom use (17 studies). In terms of knowledge, the interventions using small media—with interpersonal communication or alone—showed similar effects, both in significance and size, to programs using radio and/or TV. With regard to condom use, evaluations of interventions using small media were less likely to measure this outcome and those that did were less likely to show positive effects. Of six studies that did show a positive outcome for condom use, five used radio and/or TV, alone or with other media.

Discussion

How effective have communication programs been in changing HIV-related behavior? A number of critics have questioned the effectiveness of communication interventions as conducted to date [28, 29].

The current review yielded mixed results on the effectiveness of the mass media to change HIV-related behaviors in developing countries. On most of the outcomes examined across studies, we found no statistically significant impact. Among those that did show significant impacts, the effect sizes— while often statistically significant—were typically small to moderate in size. However, on two of the seven outcomes, at least half of the studies did show a positive impact of the mass media: knowledge of HIV transmission and reduction in high-risk sexual behavior. By contrast, the predominance of evidence for the remaining five outcomes—perceived risk of HIV, self-efficacy, interpersonal communication with partner/spouse, abstinence and condom use—showed mixed results or no effect.

This paper falls short of providing a definitive answer to the question 'what is the impact of mass media on HIV/AIDS-related behaviors in developing countries' for two reasons. First, although we identified 24 articles that met the inclusion criteria, many of these studies had weak designs. For example, one study originally established as a randomized trial for other purposes did not use the

'arms' of the study in the analysis of communication effects [23]. Another based the conclusion of 'no effect' on the increase in HIV prevalence between Time 1 and Time 2, during which time the media carried HIV prevention messages through multiple channels; yet no attempt was made to link exposure to the campaign and HIV status [30]. In another case, the study design called for a baseline and follow-up survey; yet the time elapsed between baseline and follow-up as well as the non-comparability of the two samples on socio-demographic factors resulted in basing the evaluation largely on the post-intervention data [25].

Second, the studies included in this review— representing all published evaluations through 2004 that met the inclusion criteria—do not capture the current state-of-the art for mass media campaigns for HIV/AIDS prevention. The 'evolution' in types of programs studied—from those involving small media to those using TV and/or radio—is reflective of the trend among governments, donor agencies and in-country implementing organizations to go to scale. The current analysis did not include a single study that evaluated what communication experts would consider a comprehensive behavior change program: one that uses the full gamut of media— TV, radio, billboards, posters, pamphlets and other media linked with community-level activities (e.g. mobile vans, outreach events) to reach multiple segments of the general public with messages on HIV/AIDS. (In fact, one study did report on what may have been a comprehensive program, but made no linkage between exposure to the campaign and its effects.) This 'ideal' exists in a number of developing countries today, yet possibly because such comprehensive programs are still relatively new, no results were available in the published literature as of the end of 2004.

Policy makers, donors and practitioners are often frustrated at evaluators' inability to answer the question: 'what makes some campaigns more effective than others?' The experience from commercial advertising and marketing has defined many of the best practices that are now used in promoting social and behavior change. Marketing researchers in developed countries have honed techniques for

establishing 'what works' by tracking sales using a variety of techniques: scanning bar codes in stores in the target area, tracking number of orders placed by phone or over the Internet for catalogue sales. However, such techniques are not readily applicable to evaluating programs designed to change social norms or behaviors in developing countries, without a means of tracking sales on an hourly or daily basis.

Even if international agencies were willing to fund such research to identify what makes an effective program, methodological problems exist. First, most evaluation studies focus on a single campaign, making 'systematic comparisons' across campaigns impossible. Second, it is difficult to disaggregate the effects of different components of a given campaign. One can stagger the introduction of different components into a campaign and track the point at which change occurs or accelerates; yet change that occurs after introducing a specific component may reflect lagged response to previously disseminated components. Third, experienced practitioners are loath to 'experiment' with time-tested techniques (e.g. audience segmentation) for the academic purpose of 'proving' that these techniques are effective. Fourth, relatively few campaigns undergo evaluation to determine effectiveness, let alone the factors behind their success. For these reasons, the published literature contains relatively little empirical testing or experimentation to determine what factors or characteristics make for an effective behavior change communication program in developing countries.

Ideally, we would have analyzed the data by sex of the respondent. However, only eight of the 24 studies disaggregated the data by gender. Thus, we did not attempt to incorporate this variable into the current review. Future research on effects would greatly benefit from disaggregation of results by gender.

In keeping with Hornik's findings in his edited volume of studies entitled Public Health Communication, Evidence for Behavior Change [31], this review underscores the need for alternative study designs to randomized trials as the optimal means for evaluating full coverage mass media programs.

Only five of the 24 studies in the current review randomly allocated subjects to a treatment group. In four of the five cases, the intervention was limited to small media, making it possible to expose one group to the communication intervention, while withholding it from the other. The only exception was Xiaoming et al. [10] which used a pre-/post-randomized controlled trial comparing two intervention villages and two control villages sampled from two townships that were matched on socio-economic and demographic characteristics. Thus, there were no studies of full coverage media programs with random allocation of subjects to treatment areas, nor were there any studies that involved the randomization of a large number of communities. The fact remains that it is not viable to assign subjects randomly to treatment groups when the intervention consists of full coverage programs aiming to reach the largest possible audience, which is the case with national AIDS prevention programs in most countries.

One frequently used alternative for measuring effects is to compare outcomes by level of exposure, also known as 'dose response' analysis. This approach can yield highly biased results if no attempt is made to control for socio-demographic factors or access to media. For example, a strong association between levels of exposure and behavior change may merely reflect the effect of education and urbanization on both variables. To address this bias, researchers often control for socio-economic status and access to media. However, this does not resolve the issue of reverse causality (i.e. that people already doing the behavior may be more attentive to the messages about it). A more statistically advanced approach to measuring communication effects involves the use of propensity score analysis [32, 33]. This methodology relies on postintervention only (cross-sectional) data with no control or comparison group. In the ideal case, eval-uators would continue to collect pre- and post-data to demonstrate the expected change on key outcome indicators, but would use propensity score analysis on the post-intervention data to establish the link between exposure and the desired outcome, controlling for socio-demographic factors and

access to media. Testing for endogeneity further strengthens the causal inference drawn from propensity scoring.

Even though international donor agencies and governments have invested millions of dollars in different types of communication interventions in developing countries, relatively few have been subjected to any type of rigorous evaluation to date. In addition, few studies address the costs and o

cost-effectiveness of mass communication pro- n

gramming, leaving funders and policy makers g

without the data necessary to determine which S.

intervention strategies offer the greatest 'bang for o

the buck' (i.e.—lowest cost per person reached or g

outcome influenced). /

This review addresses an important question for .

HIV/AIDS program managers, communication re- f

searchers, donor agency staff and others: to what jj

extent do communication programs impact HIV/ |

AIDS-related behaviors? It presents a systematic &

review and analysis of the relatively limited number J

of studies on this topic, and underscores the need ag

for researchers working in this area to ensure that c

their work finds its way into the published literature S

to help us better understand (i) the outcomes on n

which communication programs have the greatest C

effect, (ii) the magnitude of these effects, (iii) the SS

elements of a communication program that contrib- b

ute to its effectiveness and (iv) the cost effective- i

ness of communication programs in HIV/AIDS A

prevention. Given the emergence of communica- M

tion programs with national scope in many de- S

veloping countries, we need further evaluation of J

programs that go to scale and refinements in the g

methodologies for evaluating such programs when 2

randomization of subjects is not an option. Such 5

research will be of greatest benefit to program managers if it includes detailed descriptions of the interventions under study, including media channels, main messages, duration, reach, frequency and underlying theoretical principles. Researchers will look for greater methodological rigor and convergence toward a common list of psychosocial outcomes and behaviors, allowing greater comparability across studies. Communication programs continue to be at the heart of the

HIV/AIDS response, yet much work remains to be done in building the evidence base for their effectiveness.

Acknowledgements

The authors wish to thank Priya Emmart, Jennifer Gonyea, Amy Gregowski, Andrea Ippel, Sarah Kessler, Juliana Kohler, Devaki Nambiar, Anne Palaia and Emma Williams for their coding work, and Caitlin Kennedy and Amy Medley for their coordination of the project. This research was supported by the World Health Organization, Department of HIV/AIDS, The US National Institute of Mental Health, grant number 1R01 MH071204, and The Horizons Program. The Horizons Program is funded by The US Agency for International Development under the terms of HRN-A-00-97-00012-00.

Conflicts of interest

None declared.

References

1. Liskin L. Using mass media for HIV/AIDS prevention. AIDS Care 1990; 2: 419-20.

2. Myhre SL, Flora JA. HIV/AIDS communication campaigns: progress and prospects. J Health Commun 2000; 5(Suppl.): 29-45.

3. Oakley A, Fullerton D, Holland J. Behavioural interventions for HIV/AIDS prevention. AIDS 1995; 9: 479-86.

4. Measure DHS+. HIV/AIDS Survey Indicators Database. Calverton, MD: Macro International. Available at: http:// www.measuredhs.com/hivdata. Accessed: May 19, 2006.

5. McKee N, Bertrand JT, Becker-Benton A. Strategic Communication in the HIV//AIDS Epidemic. New Delhi: SAGE Publications, 2004.

6. World Bank. Countries and Regions. Available at: http:// www.worldbank.org/. Accessed: 16 July 2004.

7. *Valente TW, Bharath U. An evaluation of the use of drama to communicate HIV/AIDS information. /AIDS Educ Prev 1999; 11: 203-11.

8. *Trykker H, Kalumba K, Hamming A et al. Changes in public knowledge and awareness of AIDS in Zambia. /AIDS 1992; 6: 1408-9.

9. *Milleliri JM, Krentel A, Rey JL. Sensitisation about condom use in Gabon (1999): evaluation of the impact of a comic book. Cah Sante 2003; 13: 253-64.

10. *Xiaoming S, Yong W, Choi WK et al. Integrating HIV prevention education into existing family planning services: results of a controlled trial of a community-level intervention for young adults in rural China. Aids Behav 2000; 4: 103-10.

*McGill D, Joseph WD. An HIV/AIDS awareness prevention project in Sri Lanka: evaluation of drama and flyer distribution interventions. Int Q Commun Health Educ 1996; 16: 237-55.

*Yoder PS, Hornik R, Chirwa BC. Evaluating the program effects of a radio drama about AIDS in Zambia. Stud Fam Plann 1996; 27: 188-203.

*Elkins D, Maticka-Tyndale E, Kuyyakanond T et al. Evaluation of HIV/AIDS education initiatives among women in northeastern Thai villages. Southeast Asian J Trop Med Public Health 1996; 27: 430-42.

14. *Peltzer K, Promtussananon S. Evaluation of Soul City school and mass media life skills education among junior secondary school learners in South Africa. Soc Behav Pers 2003; 31: 825-34.

15. *Vaughan PW, Rogers EM, Singhal A et al. Entertainment-education and HIV/AIDS prevention: a field experiment in Tanzania. J Health Commun 2000; 5(Suppl.): 81-100.

16. *Pauw J, Ferrie J, Rivera VR et al. A controlled HIV/AIDS-related health education programme in Managua, Nicaragua. AIDS 1996; 10: 537-44.

17. *Elkins D, Maticka-Tyndale E, Kuyyakanond T. Toward reducing the spread of HIV in northeastern Thai villages: evaluation of a village-based intervention. AIDS Educ Prev 1997; 9: 49-69.

*Middlestadt S, Fishbein M, Albarracin D et al. Evaluating the impact of a national AIDS prevention radio campaign in St. Vincent and the Grenadines. J Appl Soc Psychol 1995; 25: 21-34.

*Kim YM, Kols A, Nyakauru R et al. Promoting sexual responsibility among young people in Zimbabwe. Int Fam Plan Perspect 2001; 27: 11-9.

*Peltzer K, Philip S. Evaluation of HIV/AIDS prevention intervention messages on a rural sample of South African youth's knowledge, attitudes, beliefs and behaviours over a period of 15 months. J Child Adolesc Ment Health 2004; 16: 93-102.

*Ubaidullah M. Social vaccine for HIV prevention: a study on truck drivers in South India. Soc Work Health Care 2004; 39: 399-414.

22. *Schopper D, Doussantousse S, Ayiga N et al. Village-based AIDS prevention in a rural district in Uganda. Health PolicyPlan 1995; 10: 171-80.

23. *Quigley MA, Kamali A, Kinsman J et al. The impact of attending a behavioural intervention on HIV incidence in Masaka, Uganda. AIDS 2004; 18: 2055-63.

24. *Vaughan PW, Regis A, St Catherine E. Effects of an entertainment-education radio soap opera on family planning and HIV prevention in St. Lucia. Int Fam Plan Perspect 2000; 26: 148-57.

*Tambashe BO, Speizer IS et al. Evaluation of the PSAMAO ''Roulez Protege'' mass media campaign in Burkina Faso. AIDS Educ Prev 2003; 15: 33-48.

26. *Shapiro D, Meekers D, Tambashe B. Exposure to the 'SIDA dans la Cite' AIDS prevention television series in Cote d'Ivoire, sexual risk behaviour and condom use. AIDS Care 2003; 15: 303-14.

27. *Vernon R, Ojeda G, Murad R. Incorporating AIDS prevention activities into family planning organization in Columbia. Stud Fam Plann 1990; 21: 335-43.

28. Airhihenbuwa CO, Makinwa B, Obregon R. Toward a new communications framework for HIV/AIDS. J Health Commun 2000; 5(Suppl.): 101-11.

29. Scalway T. Missing the Message? 20 Years of Learning from HIV/AIDS. London: Panos Institute, 2003. *Pape JW. AIDS: Results of current prevention efforts in Haiti. AIDS Res Hum Retroviruses 1993; 9(Suppl. 1): S143-5. Hornik RC (ed). Public Health Communication: Evidence for Behavior Change. New Jersey: Lawrence Erlbaum Associates, Publishers, 2002.

32. Babalola S, Vondrasek C. Communication, ideation and contraceptive use in Burkina Faso: an application of the propensity score matching method. J Fam Plan Reprod Health Care 2005; 31: 207-12.

33. *van Griensven GJ, Limanonda B, Ngaokeow S et al. Evaluation of a targeted HIV prevention programme among female commercial sex workers in the south of Thailand. Sex Transm Infect 1998; 74: 54-8.

34. *Skinner D, Metcalf CA, Seager JR et al. An evaluation of an education programme on HIV infection using puppetry and street theatre. AIDS Care 1991; 3: 317-29.

35. Valente TW. Evaluating communication campaigns. In: Rice RE, Atkin CF (eds). Public Communication Campaigns, 3rd edn. Thousand Oaks, CA: Sage Publications, 2001, 105-24.

Received on December 16, 2004; accepted on April 14, 2006