AIDS Education and Prevention, 20(6), 486–503, 2008
© 2008 The Guilford Press
PAPPAS-DELUCA ET AL.
Entertainment-Education and HIV Testing
sEriaL draMa and outcoMEs rELatEd
to Hiv tEsting in Botswana
Katina A. Pappas-DeLuca, Joan Marie Kraft, Christine Galavotti,
Lee Warner, Maungo Mooki, Phil Hastings, Todd Koppenhaver,
Thierry H. Roels, and Peter H. Kilmarx
Makgabaneng is an entertainment–education radio serial drama written and
produced in Botswana to promote prevention of HIV. This effort is part of
the national response to HIV/AIDS. Broadcast of the serial drama began in
August 2001, and two new 15-minute episodes air each week. We exam-
ined associations between exposure to Makgabaneng and outcomes related
to HIV testing, including stigmatizing attitudes, intention to be tested, talk-
ing with a partner about testing, and testing for HIV, among 555 sexually
active respondents. The four measures of exposure to Makgabaneng were
frequency of listening, duration of listening, talking about the program,
and attentiveness to and identification with relevant characters. Data were
collected approximately 18 months after the drama began airing. We found
positive associations between exposure to the program and intermediate
outcomes, including lower level of stigmatizing attitudes, stronger intention
to have HIV testing, and talking to a partner about testing. Although asso-
ciations were identified with all four measures of exposure, increased dura-
tion of listening was associated with more positive outcomes than the other
measures. This finding suggests that longer term exposure to entertainment–
education programming may be important for behavior change.
Katina A. Pappas-DeLuca, Joan Marie Kraft, Christine Galavotti, and Lee Warner are with the Centers for
Disease Control and Prevention (CDC), Atlanta, GA. Maungo Mooki is with Nesswana, Gaborone, Bo-
tswana. Phil Hastings is with Far Harbor, LLC, Austin, TX. Todd Koppenhaver is with Axiom Resource
Management, Falls Church, VA, and the BOTUSA Project, Gaborone, Botswana. Thierry H. Roels, and
Peter H. Kilmarx are with the CDC, Atlanta, GA, and the BOTUSA Project, Gaborone, Botswana.
Address correspondence to Katina Pappas-DeLuca, Centers for Disease Control and Prevention, Division
of Reproductive Health, Women’s Health and Fertility Branch, 4770 Buford Hwy. NE, Mailstop K-34,
Atlanta, GA; e-mail: firstname.lastname@example.org.
The authors thank the staff of the Makgabaneng radio serial drama and the government of Botswana for
support for the performance of this research. They also thank the team of colleagues who contributed to
conceptualization and implementation of the strategy for Modeling and Reinforcement to Combat HIV.
The findings and conclusions in this article are those of the authors and do not necessarily represent the
views of the U.S. Centers for Disease Control and Prevention.
ENTERTAINMENT-EDUCATION AND HIV TESTING 487
Botswana ranks among the countries hardest hit by HIV. In 2004 HIV prevalence
among persons 15–49 years old was approximately 25% (National AIDS Coordi-
nating Agency [NACA], 2005). Although prevalence is high, knowledge of infection
status is not; the majority of an estimated 300,000 citizens of Botswana in this age
group are unaware that they are infected with HIV (NACA, 2005; Rakgoasi, 2005).
In 2001, as a part of its national response to HIV/AIDS, the government of Bo-
tswana partnered with the BOTUSA Project of the U.S. Centers for Disease Control
and Prevention (CDC) to adopt a communications strategy for behavior change,
Modeling and Reinforcement to Combat HIV (MARCH). Grounded in behavioral
science theory, the MARCH strategy builds on two fundamental principles of be-
havior change: showing people how to change through role models (modeling) and
supporting efforts to change primarily through interpersonal communication and
normative change (reinforcement) (Galavotti, Pappas-DeLuca, & Lansky, 2001).
MARCH’s modeling component in Botswana consists of an entertainment–
education serial drama that has been broadcast on the radio since 2001. Activities
to reinforce awareness of HIV include groups that listen to and discuss this serial
drama (Makgabaneng, a rocky place), a magazine (Go Sharpo!), and community-
wide activities, such as facilitated community meetings and road shows. Botswana
reports near universal (96%) access to radio (Thapisa & Megwa, 2002). All the
program components address multiple objectives related to psychosocial factors and
behavior change. These objectives include increasing (a) positive attitudes toward
and behaviors of abstinence, monogamy, and condom use; (b) knowledge and use
of services to help prevent mother-to-child transmission of HIV; and (c) favorable
attitudes toward HIV testing and motivation for testing.
An interim evaluation of MARCH’s modeling component, the Makgabaneng
radio serial drama, was performed by assessment of psychosocial and behavioral
outcomes related to HIV testing. We focused on the modeling component exclu-
sively, because the reinforcement activities were not fully developed or operational at
the time of the assessment. HIV testing was selected as the behavior change process
to evaluate for two reasons. First, learning one’s HIV status through counseling and
testing has been identified as an important component of effective HIV prevention
strategies (Allen et al., 1992; Allen et al., 2003; DeCock, Marum, & Mbori-Ngacha,
2003; Kamenga et al., 1991), particularly in countries with high HIV prevalence.
Second, at the time of the assessment, Makgabeneng story lines that focused on HIV
testing had progressed farther in behavior change modeling than other measurable
MarcH’s ModELing coMponEnt
In contrast to learning through personal experience, observational learning occurs
by witnessing another person perform (model) a behavior. Behavioral modeling may
influence the observer’s behavior by (a) providing information on how to perform the
behavior (knowledge), (b) influencing perceptions that certain outcomes are likely
to result from performing the behavior (outcome expectations), and (c) influencing
awareness of personal ability to perform the behavior (self-efficacy) (Bandura, 1997,
488 PAPPAS-DELUCA ET AL.
2004). The key elements of MARCH’s modeling component are (a) progression of
change over time, (b) use of entertainment education as a vehicle for modeling, (c)
use of contrast modeling in program content, and (d) creation of character models
similar to the target audience.
MODELING THE BEHAVIOR CHANGE PROCESS AT A REALISTIC PACE
New behaviors are often acquired gradually, by accomplishment of small steps in
a process of change (Bertrand, 2004; Prochaska & DiClemente, 1983; Prochaska,
Redding, & Evers, 1997; Rogers, 2003; Valente & Fosados, 2006). Steps to behav-
ior change include psychosocial changes (e.g., knowledge, attitudes, and intention),
as well as preliminary behavioral changes related to the ultimate behavioral goal.
MARCH programs are designed to be in place for a long time; the first several
months are used to capture and retain a loyal audience by developing a compelling
and entertaining program. Subsequently, behavioral models change their behavior
over time, experiencing setbacks and facilitators along the way, changing attitudes
and adopting steps toward behavior change until they change behavior.
USING ENTERTAINMENT-EDUCATION AS THE MEDIUM
Entertainment-education is a communication strategy designed to both entertain and
educate an audience to increase knowledge and inspire changes in attitudes, norms,
and behaviors (Singhal & Rogers, 2004). Long-running, serialized entertainment–
education programs use fictional characters to model a realistic change process over
a series of episodes. Audience members have an opportunity to follow the evolution
of characters and relationships over time, learn about the backgrounds and skills of
characters, and observe the consequences of the decisions they make. In addition, a
long-running format provides several opportunities for the audience to become ac-
quainted and emotionally involved with characters (Bandura, 1997, 2004; Piotrow
& de Fossard, 2004).
USING CONTRAST MODELING TO FACILITATE COMPARATIVE
WEIGHING OF BEHAVIORAL CHOICES AND OUTCOMES
Contrast modeling occurs when the positive outcomes of “good” behaviors are jux-
taposed to the negative outcomes of “bad” behaviors (Bandura, 1997, 2004). Like
Mexican television producer Miguel Sabido, MARCH uses three main character
types to illustrate contrast modeling: positive, negative, and transitional (Bandura,
1997; Singhal & Rogers, 1999). Contrasting the rewards of the good behavior of
positive characters with the undesirable consequences of the negative character’s
bad behavior strengthens the perception of likely behavioral outcomes associated
with different choices. The transitional character, intended to be most like the target
audience, struggles between good and bad influences. Gradually, these characters
make their way through the process of behavior change, experiencing appropriate
consequences along the way.
USING CHARACTER ROLE MODELS DESIGNED TO BE
ENTERTAINMENT-EDUCATION AND HIV TESTING 489
SIMILAR TO THE AUDIENCE TO INCREASE
Researchers have long noted the role of perceived audience–character similarity as
a contributing factor in the influence of media personalities on audiences (Bandura,
1997; Bertrand, 2004; Liebes & Katz, 1990; Maccoby & Wilson, 1957). Audience
perceptions of similarity to characters (homophily) may be maximized when story
lines and characters are based on formative research that identifies attributes of the
target audience and common barriers and facilitators of the target behavior. Tran-
sitional characters, as the agents of change, are designed to be most like the target
MakgaBanEng storY LinEs rELatEd to Hiv tEsting
Makgabaneng is written, produced, and acted by residents of Botswana, and two
new episodes have been aired weekly since 2001. Program content is reviewed by
a technical advisory committee of local experts on content; additional guidance on
behavioral science theory is provided by the CDC and technical support on drama
production is provided by Media Support Solutions. The program story lines incor-
porate a village, a city, and a cattle post and depict current realities in Botswana. By
February 2003, when we began data collection for this assessment, Makgabaneng
had begun to model the behavior change process for HIV testing in several story
In one story line, after becoming engaged to be married, Duke (39-year-old male
truck driver) and Ausi Babinang (40-year-old female shopkeeper), both transitional
characters, repeatedly discuss the importance of being tested together for HIV and
their intention to do so as they plan their marriage. Although they eventually have
the HIV test, they had not done so by the time of the assessment.
Another story line involved Moses (45-year-old bank manager, transitional
character), his wife, Thandi (35-year-old journalist, positive character), and his mis-
tress, Mary (27-year-old bank clerk, negative character). During a time when his
job requires him to live in the village away from his wife and children in the city,
Moses has an affair with Mary, a materialistic and promiscuous young woman. On
discovering the affair, Thandi gives Moses the cold shoulder until he confesses to and
ends the affair with Mary and, at Thandi’s insistence, agrees to HIV testing with her.
When Moses and Thandi reveal the test results to one another, the audience learns
that Thandi is HIV-negative and Moses is HIV-positive. Despite their discordant
status, the couple remain reunited and Thandi supports Moses as he wrestles with
the stigma of HIV, copes with the knowledge of his infection, starts and continues
therapy to prevent tuberculosis, and attempts to live positively. At the time of the
assessment, Mary had not yet been faced with the reality of her own serostatus or
experienced any negative health consequences of her own risky behaviors.
The aims of the evaluation were to provide program feedback on audience reach and
acceptability of the program to the audience and to assess preliminary indications of
the effectiveness of the program. We focus on intermediate psychosocial and behav-
490 PAPPAS-DELUCA ET AL.
ioral outcomes in addition to actual reports of HIV testing because the assessment
was an interim evaluation conducted before the program was complete and because
attention to both intermediate steps and final outcomes is the most appropriate ap-
proach to evaluating programs that model behavior change as a process (Bertrand,
2004; Piotrow, Kincaid et al., 1997). The study consisted of a single, cross-sectional
assessment that began approximately 18 months after the initial broadcast of Mak-
gabaneng, when story lines related to HIV testing were under way and one key tran-
sitional character had been tested for HIV. Another transitional character, Masego,
had also been tested for HIV as part of a story line on prevention of mother-to-child
transmission (Sebert-Kuhlmann et al., 2008). Data were collected between February
and May 2003 in face-to-face interviews by using a structured questionnaire in a
multistage sampling approach (Figure 1).
We selected for assessment 7 of 22 health districts from among the most populous
districts. These districts are spread across a wide geographic area including urban
and rural areas, and they accounted for almost one half (49%) of the citizens of
Botswana in the target age range, 15–49 years. We systematically selected 60 cen-
sus enumeration areas (EAs) from the 7 districts by using a random starting num-
ber and an equal-interval sampling approach. Each EA was segmented into smaller
clusters of approximately 30 plots. One cluster was randomly selected within each
EA, and 30 households were systematically chosen by randomly selecting a starting
household and continuing in a numerical sequence by house number. In the first dis-
trict, Gaborone, the number of households with no eligible respondents was larger
than expected, so the number of households per cluster in the remaining districts
was increased from 25 to 30. We selected 1,730 households; 38 (2.2%) were aban-
doned, 203 (12%) were unoccupied at the time of the fieldwork, and the residents
in 23 (1.3%) refused to participate in the enumeration process to identify an eligible
household member. Of the 1,466 households that were enumerated, 961 (65%) had
eligible household members, and interviews were completed for 807 (84%). In addi-
tion to the criterion of living in the household during the time of fieldwork, eligibility
was limited to persons aged 15–49 years who could speak and understand Setswana.
This population was the primary target audience for Makgabaneng.
One person was randomly selected from each household and asked to partici-
pate in the interview. Each respondent gave verbal consent for participation. Inter-
viewers and respondents were matched by sex. Men and boys were sampled at 1.3
times the rate for women and girls, because fewer men and boys were expected to
be available during the time of fieldwork. The study was reviewed and approved by
institutional review boards in Botswana and at the CDC.
We developed the survey instrument in collaboration with the Technical Work-
ing Group for HIV-Related Behavioral Surveys in Botswana. The instrument was
translated forward and backward between Setswana and English, and discrepancies
were resolved. Each interviewer conducted multiple practice interviews during three
pilot tests, and the instrument was revised after each pilot test. The mean duration
of the interviews was 57 minutes (median = 55 minutes).
ENTERTAINMENT-EDUCATION AND HIV TESTING 491
We selected intermediate psychosocial and behavioral outcomes that could
suggest progress toward the final behavioral objective of testing for HIV, including
stigma related to HIV, intention to test in the near future, and talking about test-
ing with a partner. HIV-related stigma is an important psychosocial outcome that
has been associated with HIV testing (Hutchinson & Mahlalela, 2006; MacPhail,
Pettifor, Coates, & Rees, 2006). Intention to test was selected because it is a key
antecedent to behavior change (Ajzen & Fishbein, 1980) and an important indicator
of progress in behavior change (Prochaska et al., 1997). “Talking to a partner about
testing” was selected as an intermediate behavioral variable, because it was modeled
in the program as a preparatory behavior. We were interested in exploring the role of
other psychosocial variables, including perceived self-efficacy, perceived norms, and
outcome expectations related to HIV testing. However, preliminary analysis of the
psychometric properties of the scales to assess these variables suggested that further
use of these scales was not warranted.
Stigma. Six statements adapted from other work (Joint United Nations Programme
on HIV/AIDS, 2000) addressed stigmatizing attitudes (Appendix 1). Respondents in-
dicated their level of agreement on a 5-point scale ranging from “strongly agree” to
“strongly disagree.” Reliability analysis was performed, and one item was dropped
because of low interitem correlation. Thus, a five-item scale for stigma, created by
averaging responses across all items (mean = 2.11, SD = .73, alpha = .74), was used
to assess stigmatizing attitudes; higher scores indicated stigmatizing attitudes.
Intention. Respondents who reported never being testing for HIV were asked to rate
how likely they were to have an HIV test in the next 3 months. Possible responses
were as follows: (1) “not at all,” (2) “a little,” (3) “moderately,” (4) “very,” and (5)
Talking About HIV Testing. In relation to the most recent sexual partner, we asked
respondents, “Have the two of you ever discussed getting tested for the AIDS virus?”
Data on persons who had talked about testing were compared with data on those
who had not talked about testing (referent).
Testing for HIV. Respondents were asked if they had ever been tested for HIV and
if so, the month and year of the most recent HIV test. Data on persons tested dur-
ing or after December 2001 (3 months after first potential exposure to the program)
were compared with data on those who either were tested before December 2001 or
within each EA
FIGURE 1. Sampling Process.
492 PAPPAS-DELUCA ET AL.
were never tested (referent). We selected 3 months after Makgabaneng first aired to
conservatively estimate a point in time before which influence of the program on the
audience was unlikely. Persons previously tested were included in the comparison
group for two reasons: First, repeated HIV testing is important in a country where
HIV prevalence among adults is 25% or higher. Second, it was unknown whether
persons tested since December 2001 had also been tested previously but decided to
test again. To ensure that any relationships between program exposure and testing
were not obscured by our decision to combine those who tested before but not after
December 2001 with those who never tested in the reference group, we conducted
additional analysis omitting those who tested prior to December 2001 and com-
paring only those who tested after December 2001 to those who had never tested.
Similar results were obtained.
Multiple measures of program exposure may best capture the complexity of
audience involvement and identification with program characters (Noar, 2006; Papa
et al., 2000; Rubin & McHugh, 1987). To explore different aspects of audience in-
volvement in the radio program and identification with the characters, we assessed
four measures of exposure to the program: (a) how often respondents listened (“fre-
quency”), (b) how long they had listened (“duration”), (c) whether they had spoken
with anyone about Makgabaneng in the previous 3 months (“discussion”), and (d)
whether they were attentive to and engaged with key characters in story lines related
to HIV testing (“identification”).
Frequency. We dichotomized frequency of listening into two groups: persons who
reported listening to the program one or more times each week and those who re-
ported listening less than weekly or not at all. Although two new episodes are aired
each week, back-to-back airing enables a person to have full exposure to program
content by listening once a week.
Duration. When fieldwork began, the program had been on the air for 18 months.
Respondents were asked, “Since you started listening to Makgabaneng, how much
time has passed?” We dichotomized the duration of time respondents reported lis-
tening into periods of 1 year or longer compared with less than 1 year or not at all
(referent). This threshold of 1 year represents exposure to at least two thirds of the
program content aired by that time. Because early episodes were designed to draw
a loyal audience base before the start of an extended modeling process, the “longer
term” group needed a sufficient duration (e.g., 1 year or longer) for exposure to a
fuller modeling process including intermediate and final outcomes. The 39 persons
who reported not remembering or not knowing how long they had been listening
were coded as missing.
Discussion. Talking to someone about Makgabaneng was defined as a report of talk-
ing to anyone about the program in the past 3 months and was compared with not
speaking to anyone or not listening to the program (referent).
Identification. Because of the variability with which audience–character identifi-
cation has been conceptualized and measured (Cohen, 2001) and our need for a
concise measure in the survey instrument, we assessed two basic elements of audi-
ence–character identification: attention to and likability of characters. This variable
ENTERTAINMENT-EDUCATION AND HIV TESTING 493
was created by combining responses from two questions: “Please mention which
Makgabaneng characters you are aware of” and “Who is your favorite character?”
Respondents who mentioned one or more key characters from an HIV-testing story
line but did not name such a character as a favorite were defined as “attending but
not identifying with.” Respondents who mentioned relevant characters in the first
question and named a relevant character as a favorite were defined as “attending
and identifying.” Respondents who did not listen to the program or did not name a
relevant character were the referent (“neither attending nor identifying”). Key char-
acters were defined as positive, negative, or transitional characters whose personal
story lines were related to modeling behaviors related to HIV testing.
Analyses were restricted to respondents who were recently sexually active because
sexual transmission is the most common mode of HIV transmission in Botswana
and the radio serial drama promotes testing for sexually active people. Recently
sexually active was defined as having had sexual intercourse in the 12 months before
Data were weighted to account for the undersampling of Gaborone households,
the oversampling of boys and men, the number of eligible residents per household,
and nonresponse. SUDAAN was used for analysis to properly account for our com-
plex sampling design. Individual respondents were the unit of analysis. We conducted
bivariate analyses to identify and test associations between each dimension of expo-
sure to the radio program and each outcome. Multivariable analysis was conducted
to explore potential confounding effects of measured sociodemographic factors. We
used unconditional logistic regression for analysis of the outcomes of “testing” and
“talking about testing,” multinomial regression for the “intention to test” outcome,
and ordinary least squares regression for the “stigma” outcome. Each multivariable
model was adjusted for sex, age, education, marital status, age at first sex, and rural
or urban residence (Central Statistics Office of Botswana) because we hypothesized
that the variable may have influenced the likelihood of program exposure or because
it was identified in the literature as related to HIV-testing behavior (Bond, Lauby,
& Batson, 2005; Houston, Archibald, Strike, & Sutherland, 1998; Pronyk et al.,
CHARACTERISTICS OF STUDY SAMPLE
Overall, 69% (555) of the 807 respondents were currently or recently sexually ac-
tive. Of these 555, 45% were male, 63% were 25 years of age or older, 36% had
no formal education or had primary schooling only, 38% were currently married
or cohabiting with a partner, and 73% were living in an urban area. Age at sexual
initiation was 12–32 years (mean = 18).
1. Weighted data presented
494 PAPPAS-DELUCA ET AL.
MAKGABANENG PROGRAM REACH1
Nearly one half (46%, 243) of respondents reported listening to Makgabaneng one
or more times each week, 15% (76) reported listening for 1 year or longer, and 19%
(100) reported talking with someone about Makgabaneng in the past 3 months. In
addition, 23% (122) of the respondents attended to the story line (i.e., spontane-
ously named any relevant character) and 9% (50) attended to and identified with
relevant characters (i.e., spontaneously named one or more relevant characters and
identified a relevant character as a favorite). Approximately 90% (68) of the 76 re-
spondents who reported listening to Makgabaneng for a year or longer also reported
listening one or more times each week.
UNIVARIATE AND BIVARIATE ANALYSES
Table 1 presents the proportion of respondents who reported each outcome related
to HIV testing, overall and by each measure of program exposure. Overall, the mean
score for stigma was 2.11 on a scale of 5 which reflects low levels of stigmatizing
attitudes. Of the 326 respondents who had never been tested, 27% reported no in-
tention to test in the next 3 months, and 29% reported being extremely likely to test.
More than one half (57%, 301) of the 524 persons who responded to the question
on talking with their partner about HIV testing gave a positive response, and 116
(22%) of the 505 who responded to the question on HIV testing since December
2001 gave a positive response.
Bivariate associations between exposure to the radio program and HIV-related
outcomes revealed statistically significant associations between the length of time
listening to the drama (duration) and the intermediate outcomes of HIV-related
stigma, intending to be tested, and talking about testing with a partner. In addi-
tion, talking about Makgabaneng (discussion) was associated with two intermediate
outcomes: stigma and talking about testing. The remaining measures, frequency of
listening and identification, were significantly associated with stigma related to HIV.
No significant bivariate associations existed between any dimension of exposure and
whether participants had been tested for HIV since December 2001.
To test the relationship between exposure to the radio serial drama Makgabaneng
and outcomes related to HIV testing, four multivariable models were constructed
for each outcome. These models corresponded to each of the four measures of pro-
gram exposure (frequency, duration, discussion, and identification). We controlled
across all models for a consistent set of demographic factors, and each model in-
cluded only one measure of exposure to the program. HIV testing was not associ-
ated with any measure of program exposure in bivariate or multivariate models
or in additional analysis comparing data on persons tested since 2001 with data
on those who were never tested. Thus further data on HIV testing as an outcome
are not presented here. Likewise, we do not present data from multivariate models
considering the exposure dimension of discussion, because it was not associated
with any of the outcomes. Relationships between the three remaining measures of
exposure and stigma related to HIV, intending to test, and talking about testing
with a partner are presented here.
ENTERTAINMENT-EDUCATION AND HIV TESTING 495
TABLE 1. Proportion of 555 Sexually Active Respondentsa Reporting Outcomes Related to Testing for Human Immunodeficiency Virus, by Measures of Exposure to
the Radio Serial Drama Makgabaneng, Botswana, 2003
Measures of Exposure (%)
n = 243
n = 283
n = 76
n = 428
n = 100
n = 426
n = 50
n = 122
n = 355
HIV stigmac (547)
Mean = 2.11
Intention to be testedd
(326 never tested)
Not at all likely
A little likely
Talked with partner
about testing (524)
HIV testing since
aSexually active defined as having sexual intercourse in the 12 months before the survey. bValues are weighted and rounded to nearest whole number. cThe stigma scale averages responses
across all items; higher scores indicated stigmatizing attitudes. dStatistically significant differences reflect differences across all levels of intention only among persons who were never tested.
*p < .05, **p < .01.
496 PAPPAS-DELUCA ET AL.
Stigma Related to HIV. Frequency, duration, and identification all had a statistically
significant association with HIV-related stigma (Table 2). Greater levels of exposure
to Makgabaneng, including more frequent listening, longer-term listening, and at-
tentiveness to and identification with the characters, were associated with lower
levels of stigmatizing attitudes.
Intention to Test for HIV. Both frequency and duration of listening were signifi-
cantly related to a respondent’s intention to test in the next 3 months (Table 3).
Respondents who reported listening to Makgabaneng one or more times each week
were almost twice as likely (adjusted odds ratio [AOR] = 1.8) as less frequent listen-
ers and nonlisteners to have stronger intentions to have HIV testing. Likewise, com-
pared with those who listened for less than 1 year or not at all, longer term listeners
(≥1 year) were twice as likely to have stronger intentions to be tested in the next 3
months (AOR = 2.1).
Talked With Partner About HIV Testing. Among the measures of exposure tested,
only duration of listening had a statistically significant association with talking about
HIV testing (Table 4). Respondents who listened for 1 year or longer were more than
twice as likely as those who listened less than 1 year or not at all to report talking to
their partner about HIV testing (aOR = 2.45).
TABLE 2. Ordinary Least Squares Regression Analysis of Adjusted Models for Stigma Related to HIV
and Three Measures of Exposure to Makgabaneng Radio Serial Drama Among 555 Sexually Active
Respondentsa Reporting Outcomes Related to HIV Testing
Model variablesfrequency beta (t)duration beta (t)identification beta (t)
Frequency of listening (n = 526)
≥ Weekly vs. < weekly
-0.15* (-2.15) NA NA
Duration of listening (n = 501)
≥1 year vs. <1 year
NA-0.26** (-2.71) NA
Identification with program (n = 527)
Attentive and identified vs. notNA NA -0.33** (-2.71)
Attentive only vs. not -0.22** (-2.91)
Male vs. female 0.30** (3.59) 0.29** (3.53) 0.26** (3.02)
15–24 vs. 25–49 0.09 (0.94) 0.06 (0.62) 0.10 (1.15)
None/primary only vs. higher0.32** (3.42) 0.30** (3.30) 0.32** (3.64)
Married/cohabiting vs. not -0.06 (-0.72) -0.04 (-0.42) -0.05 (-0.71)
Urban vs. rural -0.29* (-2.09) -0.27 (-1.90) -0.28* (-2.10)
Age at first sex -0.01 (-1.01) -0.01 (-0.70) -0.01 (-0.83)
0.14 0.14 0.16
Note. aSexually active defined as having sexual intercourse in the 12 months before the survey. bNot applicable, vari-
able not included in the model. *p < .05, **p < .01.
ENTERTAINMENT-EDUCATION AND HIV TESTING 497
In an era of competing demands on audience attention, limited public health re-
sources, and frequent media messages related to HIV prevention, it is noteworthy
that a program such as Makgabaneng can capture and maintain the attention, on
a weekly basis, of a large audience at risk for HIV. Almost one half of all surveyed
respondents reported frequent listening to Makgabaneng. This finding suggests that
the program has broad reach and appeal for the target audience. Having an audience
who consistently tunes in to a program provides public health practitioners with a
regular opportunity to communicate with the audience, to introduce relevant infor-
mation, to promote available services, to address emerging problems, and to model
new behaviors and norms.
We observed positive relationships between exposure to Makgabaneng and in-
termediate outcomes related to HIV testing. These outcomes included lower levels of
stigmatizing attitudes related to HIV, stronger intention to test, and increased talking
with a partner about testing. Intermediate effects are important for understanding
the impact of a communication program on behavior change in that they suggest
progress toward the intended goal (Piotrow et al., 1997). Although no association
between exposure and reported testing behavior was observed, the positive associa-
tions between intermediate outcomes and exposure to Makgabaneng suggest prog-
ress toward the behavioral goal of HIV testing.
TABLE 3. Multinomial Logistic Regression Analysis of Adjusted Models for Intention to Test for HIV
in Next 3 Months and Three Measures of Exposure to Makgabaneng Radio Serial Drama Among
555 Sexually Active respondentsa Reporting Outcomes Related to HIV Testing
sion aor (95% ci)
aor (95% ci)
aor (95% ci)
Frequency of Listening
(n = 337)
≥ Weekly vs. < Weekly
1.81** (1.18–2.86) NA NA
Duration of Listening
(n = 320)
≥ 1 Year vs. < 1 Year
NA 2.12* (1.01–4.55)NA
Identification to program
Attended and identified vs. Not NA NA1.56 (0.79–3.03)
Attended only vs. Not 1.37 (0.90–2.08)
Male vs. Female1.40 (0.92–2.17)1.29 (0.81–2.08) 1.39 (0.88–2.17)
15-24 years vs. 25-49 Years 1.67 (0.92–3.03)1.92* (1.01–3.57) 1.69 (0.91–3.13)
None/Primary Only vs. Higher 1.39 (0.80–2.44)1.56 (0.88–2.78) 1.45 (0.81–2.56)
Married/cohabiting vs. Not1.00 (0.58–1.72)1.00 (0.55–1.81)0.95 (0.54–1.69)
Urban vs. Rural1.01 (0.63–1.61)1.05 (0.63–1.75) 1.04 (0.63–1.69)
Age at First Sex 1.03 (0.96–1.10) 1.03 (0.95–1.09)1.02 (0.95–1.10)
Note. aSexually active defined as having sexual intercourse in the 12 months before the survey. AOR = adjusted odds
ratio; CI = confidence interval; NA = Not applicable, variable not included in the model. *p < .05; **p < .01.
498 PAPPAS-DELUCA ET AL.
Stigma related to HIV has been cited as the greatest challenge to arresting the
HIV epidemic in Botswana (Letamo, 2003). Such stigma has been found to influence
willingness to seek and provide health care (Fombad, 2001; Livingston, 2004); to
adopt preventive behaviors (Godfrey-Faussett et al., 2002; Omari, Luo, Kankasa,
Bhat, & Bunn, 2003; Shapiro et al., 2003); and to seek HIV testing (CDC, 2004;
Fombad, 2001; Hutchinson & Mahlalela, 2006; Lauby, Bond, Eroglu, & Batson,
2006; MacPhail et al., 2006; Rakgoasi, 2005). Although stigmatizing attitudes were
low in our sample overall, positive associations between Makgabaneng exposure
and lower stigmatizing attitudes suggest a positive influence on attitudes toward
people living with HIV/AIDS. Reducing HIV-related stigma may, over time, reduce
an important barrier to residents of Botswana learning their HIV status and adopt-
ing strategies to prevent transmission of HIV.
Positive associations between exposure to Makgabaneng and “intention to test”
among respondents who had never been tested were also encouraging. Although not
a direct measure of behavior, intention to perform a behavior has been identified as
a strong predictor of behavior in behavioral science research (Ajzen, 1991; Ajzen &
Fishbein, 1980; Prochaska et al., 1997). This finding is particularly evident when
intention is measured within a specific time frame, as our measure was (i.e., in the
next 3 months). Furthermore, the association between listening duration and talking
to the most recent sexual partner about HIV testing suggests that listeners may be
taking preliminary steps toward behavior change.
TABLE 4. Logistic Regression Analysis of Adjusted Models for Talking with Partner About HIV
Testing and Three Measures of Exposure to Makgabaneng Radio Serial Drama Among 555 Sexually
Active Respondentsa Reporting Outcomes Related to HIV Testing
AOR (95% CI)
AOR (95% CI)
AOR (95% CI)
Frequency of Listening (n = 531)
≥ Weekly vs. < Weekly
1.45 (0.99–2.12) NA NA
Duration of Listening (n = 506)
≥ 1 Year vs. < 1 Year
NA*2.45 (1.17–5.16) NA
Identification with program (n = 532)
Attended and Identified vs. Not NA NA1.58 (0.71–3.54)
Attended only vs. Not 1.51 (0.97–2.33)
Male vs. Female1.04 (0.66–1.64)1.07 (0.67–1.69) 1.10 (0.70–1.75)
15-24 years vs. 25-49 years*0.61 (0.38–0.99) 0.66 (0.39–1.10)*0.61(0.37–0.99)
None/Primary Only vs. Higher*0.59 (0.37–0.93) *0.62 (0.39–1.00)*0.60 (0.38–0.96)
Married/cohabiting vs. Not1.14 (0.70–1.88)1.15 (0.67–1.98)1.14 (0.69–1.87)
Urban vs. Urban 1.48 (0.95–2.31)1.48 (0.95–2.31) 1.51 (0.97–2.34)
Age at First Sex 1.01 (0.93–1.09)1.00 (0.92–1.08)1.00 (0.93–1.09)
Note. AOR = adjusted odds ratio, CI = confidence interval. CI = confidence interval. NA = Not applicable, variable
not included in the model. aSexually active defined as having sexual intercourse in the 12 months before the survey.
*p < .05; **p < .01.
ENTERTAINMENT-EDUCATION AND HIV TESTING 499
Although we assessed multiple dimensions of exposure to Makgabaneng to cap-
ture audience involvement and identification with the program, only the duration
of listening was positively associated with all three intermediate outcomes. The re-
lationship between duration of exposure to the program and intermediate effects is
consistent with published theory (Bandura, 2004; Piotrow & de Fossard, 2004) and
suggests that longer term listening provides a greater opportunity to develop rela-
tionships with the characters and to observe the steps toward making a change and
the consequences of the specific choices and behaviors. Associations between longer
term listening and intermediate outcomes may indicate that longer-term listeners are
beginning the gradual process of change by adopting the preliminary steps that may,
over time, lead to testing for HIV.
Our findings are subject to a number of limitations. We used a cross-sectional
design to evaluate the effect of program exposure among a large sample of poten-
tial listeners. This approach limits the ability to control for self-selection bias or to
draw conclusions about the time relationships between listenership and HIV-related
outcomes. Two important structural factors influenced the study design of our evalu-
ation. First, The program was broadcast on radio stations transmitting throughout
the country, so random assignment to exposure was not an option. Second, collec-
tion of baseline data was prohibited when the program started, because a national
census of Botswana was under way.
In an effort to address our inability to determine whether respondents were
differentially predisposed to listen to the drama, we compared demographic charac-
teristics of frequent listeners and all other respondents. Frequency of listening was
the exposure measure with the largest number of respondents reporting exposure.
Age was the only demographic factor that showed a statistically significant differ-
ence in the comparison of data on these two groups. This finding suggests that few
measured demographic differences predispose some respondents toward listening to
Makgabaneng (data not shown).
The timing of the assessment also compromised our ability to detect certain
associations. First, during the first 18 months of the Makgabaneng program, the na-
tional voluntary testing and counseling program was being established throughout
the country, increasing the availability of HIV testing services overall. Unfortunately,
we have no way to evaluate the role of increased access to HIV testing centers in
respondent’s decisions to be tested.
Second, although the assessment was designed to provide us with information
on preliminary effects, including intermediate outcomes, the assessment was con-
ducted before resolution of the behavior-change modeling process for one of the
selected outcomes (reported HIV testing). For example, characters in story lines re-
lated to HIV testing (Duke and Ausi Babinang) had not yet been tested for HIV at
the time of data collection. In addition, Mary, a negative character, had not yet ex-
perienced negative health consequences related to her behavior. Negative characters
in Makgabaneng did experience negative consequences. For example, Mary becomes
pregnant, delivers a baby with HIV who dies in infancy, becomes ill herself, and
eventually dies. However, the timing of the assessment preceded much of these story
A lack of negative health consequences for the negative character may be par-
ticularly problematic for several reasons. Some researchers suggested that the influ-
ence of negative outcomes experienced by negative characters may actually be a
more important influence on behavior change than positive outcomes experienced
by transitional and positive characters (Lockwood, Marshall, & Sadler, 2005). The
500 PAPPAS-DELUCA ET AL.
lesson seems to be that illustrating the negative behavior and outcomes to avoid,
rather than the good and positive aims to pursue, may be more motivating to some
persons. Similarly, Green and Witte (2006) suggest that elements of fear (negative
consequences) in HIV prevention campaigns may be more motivating for behavior
change in Africa than in the United States. On the other hand, other researchers de-
scribe how audiences may arrive at positive views of negative characters when these
characters have insufficient consequences for bad behavior (Chandran, Hirata, &
Rogers, 1993), thereby subverting their intended role in the story line.
Had the assessment been conducted after transitional characters Duke and Ausi
Babinang tested for HIV and after Mary began to experience the negative conse-
quences of her behavior, the influence of behavioral modeling and contrast model-
ing may have been more dramatic and the effect of the radio program on outcomes
related to HIV testing may have been greater. For these reasons, we believe the full
impact of Makgabaneng on HIV testing may be underestimated by this assessment.
concLusions and rEcoMMEndations
A recent Cochrane review reported that mass media can be an effective intervention
medium to promote HIV testing (Vidanapathirana, Abramson, Forbes, & Fairley,
2006). The popularity of Makgabaneng combined with preliminary evidence of its
effects on intermediate outcomes related to HIV testing suggest that the radio serial
drama is a promising medium to facilitate behavior change in Botswana. Because
our results are preliminary, additional evaluation of the strategy and program is
warranted. Because of Botswana’s 2004 adoption of an “opt out” approach to HIV
testing, where HIV testing became a part of routine medical visits for nearly all pa-
tients unless they explicitly refuse (Weiser et al., 2006), additional evaluation of the
impact of Makgabaneng on HIV testing behavior is not possible. However, story
lines that emphasize other important outcomes, such as delaying sexual initiation
and being faithful to one partner, offer additional opportunities to evaluate the role
of Makgabaneng in changing behavior.
MARCH programs in Botswana and elsewhere could also benefit from addi-
tional quantitative and qualitative research on audience identification with charac-
ters. Such research could further explore the type and level of connection between
the audience and the program. The strength of the longer term exposure with inter-
mediate outcomes suggests importance of the entertainment factor for the program,
so audiences can be captured and retained over time. Additional quantitative evalu-
ation activities would also help to further explore the role of duration of exposure
in the effects of a program.
To more fully evaluate the role of the MARCH project in changing behaviors,
an evaluation that combines the influence of reinforcement and modeling would
also be desirable. Near simultaneous rollout of reinforcement activities and model-
ing programs would be ideal for future MARCH programming. In addition, future
programming should seek to resolve behavioral modeling of positive and negative
characters in a more proximal time frame, so that negative consequences of negative
character’s behavior can be more effectively contrasted with positive consequences
of positive behavior. MARCH programs in Zambia and Guyana that have been
started since this evaluation attempt to address some of these recommendations in
ENTERTAINMENT-EDUCATION AND HIV TESTING 501
Ajzen, I. (1991). The theory of planned behavior.
Organizational Behavior and Human Deci-
sion Processes, 50, 179–211.
Ajzen, I., & Fishbein, M. (1980). Understanding
attitudes and predicting social behavior. En-
glewood Cliffs, NJ: Prentice Hall.
Allen, S., Meinzen-Derr, J., Katzman, M., Zulu,
I., Trask, S., Fideli, U., et al. (2003). Sexual
behavior of HIV discordant couples after
HIV counseling and testing. AIDS, 17(5),
Allen, S., Tice, J., Van de Perre, P., Serufilira, A.,
Hudes, E., Nsengumuremyi, F., et al. (1992).
Effect of serotesting with counselling on
condom use and seroconversion among
HIV discordant couples in Africa. British
Medical Journal, 304(9), 1605–1609.
Bandura, A. (1997). Self-efficacy: the exercise of
control. New York: Freeman.
Bandura, A. (2004). Social cognitive theory for per-
sonal and social change by enabling media.
In A. Singhal, M. J. Cody, E. M. Rogers, &
M. Sabido (Eds.), Entertainment-education
and social change: History, research, and
practice (pp. 75–96). Mahwah, NJ: Erl-
Bertrand, J. T. (2004). Diffusion of innovations
and HIV/AIDS. Journal of Health Commu-
nication, 9, 113–121.
Bond, L., Lauby, J., & Batson, H. (2005). HIV
testing and the role of individual- and struc-
tural-level barriers and facilitators. AIDS
Care, 17(2), 125–140.
Centers for Disease Control and Prevention.
(2004). Introduction of routine HIV testing
in prenatal care—Botswana, 2004. Morbid-
ity and Mortality Weekly Report, 53(46),
Chandran, A. S., Hirata, T. M., & Rogers, E. M.
(1993, May). Using entertainment for de-
velopment: Viewer identification with a pro-
social Indian soap opera. Paper presented at
the meeting of the International Communi-
cation Association, Washington, DC.
Cohen, J. (2001). Defining identification: A theo-
retical look at the identification of audienc-
es with media characters. Mass Communi-
cation and Society, 4(3), 245–264.
DeCock, K. M., Marum, E., & Mbori-Ngacha,
D. (2003). A serostatus-based approach to
HIV/AIDS prevention and care in Africa.
The Lancet, 362, 1847–1849.
Fombad, C. M. (2001). The crisis of confidentiality
in the control of the HIV/AIDS pandemic
in Botswana. International Social Science
Journal, 53(4), 643–656.
Galavotti, C., Pappas-DeLuca, K. A., & Lansky,
A. (2001). Modeling and Reinforcement to
Combat HIV: The MARCH approach to
behavior change. American Journal of Pub-
lic Health, 91(10), 1602–1607.
Godfrey-Faussett, P., Maher, D., Mukadi, Y. D.,
Nunn, P., Perriens, J., & Raviglione, M.
(2002). How human immunodeficiency
virus voluntary testing can contribute to
tuberculosis control. Bulletin of the World
Health Organization, 80(12), 939–945.
Green, E. C., & Witte, K. (2006). Can fear arousal
in public health campaigns contribute to
the decline of HIV prevalence? Journal of
Health Communication, 11, 245–259.
Houston, S., Archibald, C. P., Strike, C., & Suther-
land, D. (1998). Factors associated with
HIV testing among Canadians: Results of
a population-based survey. International
Journal of STD and AIDS, 9, 341–346.
appEndix a. itEMs assEssing Hiv-rELatEd stigMa
Please tell me whether you agree or disagree with the statement and how strongly you
agree or disagree. If you have no opinion, let me know.
1. You would let children from this family play with children from another family even
if you suspected that the other children had the AIDS virus.
2. You would be afraid -to be around a person who has the AIDS virus.
3. You would be willing to buy fresh fruits and vegetables or prepared food from a
shopkeeper even if you thought the shopkeeper had the AIDS virus.
4. You would let children from this family go to school if their teacher had the AIDS
virus but was not sick.
5. You would be willing to live with somebody from your extended family member even
if they had the AIDS virus.
6. You would be willing to care for an extended family member who had the AIDS
502 PAPPAS-DELUCA ET AL.
Hutchinson, P. L., & Mahlalela, X. (2006). Utili-
zation of voluntary counseling and testing
services in the Eastern Cape, South Africa.
AIDS Care, 18(5), 446–455.
Joint United Nations Programme on HIV/AIDS.
(UNAIDS). (2000). National AIDS pro-
grammes: A guide to monitoring and evalu-
ation. Geneva, Switzerland: Author.
Kamenga, M., Ryder, R. W., Jingu, M., Mbuyi, N.,
Mbu, L., Behets, F., et al. (1991). Evidence
of marked sexual behavior change associ-
ated with low HIV-1 seroconversion in 149
married couples with discordant HIV-1 se-
rostatus. Experience at an HIV counseling
center in Zaire. AIDS, 5, 61–67.
Lauby, J. L., Bond, L., Eroglu, D., & Batson, H.
(2006). Decisional balance, perceived risk
and HIV testing practices. AIDS and Be-
havior, 10(1), 83–92.
Letamo, G. (2003). Prevalence of, and factors as-
sociated with, HIV/AIDS-related stigma
and discriminatory attitudes in Botswana.
Journal of Health Population and Nutri-
tion, 21(4), 347–357.
Liebes, T., & Katz, E. (1990). The export of mean-
ing: Cross-cultural readings of “Dallas.”
New York: Oxford University Press.
Livingston, J. (2004). AIDS as chronic illness: Epi-
demiological transition and health care in
south-eastern Botswana. African Journal of
AIDS Research, 3(1), 15–22.
Lockwood, P., Marshall, T. C., & Sadler, P. (2005).
Promoting success or preventing failure:
Cultural differences in motivation by posi-
tive and negative role models. Personal-
ity and Social Psychology Bulletin, 31(3),
Maccoby, E. E., & Wilson, W. C. (1957). Identi-
fication and observational learning from
films. Journal of Abnormal Social Psychol-
ogy, 55, 76–87.
MacPhail, C. L., Pettifor, A., Coates, T., & Rees, H.
(2008). “You must do the test to know your
status”: Attitudes to HIV voluntary coun-
seling and testing for adolescents among
South African youth and parents. Health
Education & Behavior, 35(1), 87-104.
National AIDS Coordinating Agency. (2005). Bo-
tswana AIDS Impact Survey II: Popular re-
port. Gaborone, Republic of Botswana.
Noar, S. (2006). A 10-year retrospective of research
in health mass media campaigns: Where do
we go from here? Journal of Health Com-
munication, 11, 21–42.
Omari, A., Luo, C., Kankasa, C., Bhat, G. J., &
Bunn, J. (2003). Infant-feeding practices of
mothers of known HIV status in Lusaka,
Zambia. Health Policy and Planning, 18(2),
Papa, M. J., Singhal, A., Law, S., Pant, S., Sood, S.,
Rogers, E. M., et al. (2000). Entertainment-
education and social change: an analysis of
parasocial interaction, social learning, col-
lective efficacy, and paradoxical communi-
cation. Journal of Communication, 50(4),
Piotrow, P., & de Fossard, E. (2004). Entertain-
ment-education as a public health inter-
vention. In A. Singhal, M. J. Cody, E. M.
Rogers, & M. Sabido. (Eds.), Entertain-
ment-education and social change: History,
research, and practice (pp. 39-60). Mah-
wah, NJ: Erlbaum.
Piotrow, P. T., Kincaid, D. L., Rimon II, J. G., &
Rinehart, W. (1997). Health communica-
tion: lessons from family planning and re-
productive health. Westport, CT: Praeger.
Prochaska, J. O., & DiClemente, C. C. (1983).
Stages and processes of self-change of smok-
ing: Toward an integrative model of change.
American Psychological Association, 51(3),
Prochaska, J. O., Redding, C. A., & Evers, K. E.
(1997). The transtheoretical model and
stages of change. In K. Glanz, F. Marcus
Lewis, & B. K. Rimier (Eds.), Health behav-
ior and health education: Theory, research,
and practice (2nd ed., pp. 60–84). San Fran-
Pronyk, P. M., Kim, J. C., Makhubele, M. B., Har-
greaves, J. R., Mohlala, R., & Hausler, H. P.
(2002). Introduction of voluntary counsel-
ing and rapid testing for HIV in rural South
Africa: From theory to practice. AIDS Care,
Rakgoasi, S. D. (2005). HIV counselling and test-
ing of pregnant women attending antenatal
clinics in Botswana, 2001. Journal of Health
Population and Nutrition, 23(1), 58–65.
Rogers, E. M. (2003). Diffusion of innovations
(5th ed.). New York: Free Press.
Rubin, R. B., & McHugh, M. P. (1987). Develop-
ment of parasocial interaction relationships.
Journal of Broadcasting and Electronic Me-
dia, 31(3), 279–292.
Sebert Kuhlmann, A. K., Kraft, J. M., Galavotti,
C., Creek, T., Mooki, M., & Ntumy, R.
(2008). Radio role models for the preven-
tion of mother to child transmission of HIV
(PMTCT) and HIV testing among pregnant
women in Botswana. Health Promotion In-
ternational, 23(3), 260-268.
Shapiro, R. L., Lockman, S., Thior, I., Stock-
ing, L., Kebaabetwse, P., Wester, C., et al.
(2003). Low adherence to recommended in-
fant feeding strategies among HIV-infected
women: Results from the pilot phase of a
randomized trial to prevent mother-to-child
transmission in Botswana. AIDS Education
and Prevention, 15(3), 221–230.
Singhal, A., & Rogers, E. M. (1999). Entertain-
ment-education: A communication strategy
for social change. Mahwah, NJ: Erlbaum.
Singhal, A., & Rogers, E. M. (2004). The status
ENTERTAINMENT-EDUCATION AND HIV TESTING 503 Download full-text
of entertainment-education worldwide. In
A. Singhal, M. J. Cody, E. M. Rogers, &
M. Sabido. (Eds.). Entertainment-education
and social change: History, research, and
practice (pp. 3–20). Mahwah: Erlbaum.
Thapisa, A. P.N., & Megwa, E. R. (2002). Report
on the National Survey on Print and Elec-
tronic Media in Botswana, 2001–2002. Ga-
borone, Botswana: Department of Informa-
tion and Broadcasting.
Valente, T. W., & Fosados, R. (2006). Diffusion of
innovations and network segmentation: The
part played by people in promoting health.
Sexually Transmitted Diseases, 33(Suppl.
Vidanapathirana, J., Abramson, M. J., Forbes, A.,
& Fairley, C. (2006). Mass media interven-
tions for promoting HIV testing: Cochrane
systematic review. International Journal of
Epidemiology, 35(2), 233-236.
Weiser, S. D., Heisler, M., Leiter, K., Percy-de Ko-
rte, F., Tlou, S., DeMonner, S., et al. (2006).
Routine HIV testing in Botswana: a popula-
tion-based study on attitudes, practices, and
human rights concerns. PLOS Medicine 3
(7). Retrieved from http://medicine.plos-
journals.org. e261 doi: 10.1371/journal.