Evaluating Botswana’s Campaign on Concurrent Partnerships using Coarsened Exact Matching

Iris Halldorsdottir, Noah Taruberekera,  Rebecca Firestone, Lung Vu, Virgile Capo-Chichi, Richard Harrison

25 October 2012

Campaign background Campaign structure Impact evaluation Methods Results Lessons learned

Outline

PAGE 2

Goal: zero by 2016 2006 SADC Think Tank

– Key drivers: high CP, low levels of consistent condom use and male circumcision

National Operational Plan (2008-2010) – HIV Prevention scale-up focus on drivers

2007 CP Point Prevalence– 15-35 years, at least 1 partner– Men 17.5%, Women 17.4%

2007 CP Formative (Soul City) 2008 Development of CP

Campaign Plan

Background

page 3

Botswana’s Concurrent Partnerships CampaignCampaign team •National AIDS Coordinating Agency

•Natl Prevention Technical Advisory Cmte

•PSI

Campaign strategy •Teaser campaign – get people talking

•Phase 1 – address risk perceptions, knowledge, awareness of CP

•Phase 2 – assess consequences, personalize risk

Mechanisms•Social mobilisation – reach key influencers and leaders

•Mass media – TV, radio, print, billboards, combis• Interpersonal communication – CBOs, FBOs, door-to-door, bars/shebeens

Theory of change

page 5

Behaviour - harm reduction

within MCP

IPC

Direct exposur

e

Exposure from

all sources

Drivers of

individual behaviour

Opportunit

yAbility

Motivation

Intention to

avoid MCP

Behaviour-

reduce MCP

Reduced HIV inciden

ce

Community

sensitisation

Mass media

Integration

Key influencer

s

Media

Social diffusion

Institution

diffusion

Intention to

reduce harm within MCP

Adapted from Hornik and Yanovitzky, 2003

Did exposure to campaign messages among adults aged 18-35

(1) reduce concurrent partnerships and factors associated with having concurrent partnerships?

• Point prevalence of CP• Cumulative prevalence of CP

(2) increase HIV risk reduction behaviors and associated factors?

• Consistent condom use• HIV testing

Campaign evaluation questions

PAGE 6

National two-stage cluster sampling survey (n=1237)–Stage 1

• PPS sampling of enumeration areas (EAs) from census sampling frame• EAs in high exposure areas prospectively matched to EAs in low exposure areas,

based on geography and availability/coverage of radios and televisions– Stage 2

• Households within selected EAs selected using simple random sampling• One eligible individual in each household randomly interviewed.

Coarsened exact matching to compare exposed/non-exposed– Retrospective matching on radio and/or television access and or ownership, and

place of residence• L1 (Global Imbalance measure) = 0.059

Multiple logistic regression on sub-sample (n=1138)– Two exposure variables

• Exposure to at least one national campaign• Exposure to specific campaigns

– Stratification by gender

Evaluation Methodology

page 7

Descriptive statistics

page 8

Point preva-lence of CP

Cumulative prevalence of

CP

Consistent condom use

Ever tested for HIV

Campaign exposure

Employed Secondary education or

above

0

10

20

30

40

50

60

70

80

90

100

8.7

17.2

48.5

92.0

71.0

34.0

92.0

13.7

24.9

53.5

76.379.2

48.0

92.6

11.2

21.0

51.0

84.2

75.1

40.9

92.3

women

men

overall

Did campaign exposure reduce concurrent partnerships or increase HIV risk reduction behaviors?

Program exposure effects (vs. no exposure) on behaviors and behavioral factors

Adjusted OR (95% CI)

Effects for concurrency

Concurrency point prevalence 1.3 (0.8, 2.1)

Concurrency cumulative prevalence 1.1 (0.8, 1.6)

Peer pressure not to engage in CP 1.7 (1.1, 2.7)

Negative attitude to having variety of partners 1.6 (1.1, 2.3 )

Effects for HIV risk reduction

Consistent condom use 1.3 (1.01, 1.7)

Tested for HIV 1.6 (1.1, 2.4)

Condom use self-efficacy 1.4 (1.02, 1.8)

Perceived HIV risk 1.5 (1.1, 2.0)

• No evidence the campaign influenced concurrent partnerships

• Campaign was associated with HIV risk reduction behaviors

Campaign effects concentrated in men

PAGE 10

Program exposure effects (vs. no exposure) by gender

Men AOR (95% CI) a

Women AOR (95% CI) a

Effects for concurrency

Negative attitude to having variety of partners 2.0 (1.3, 3.1) 0.9 (0.4, 1.7)

Effects for HIV risk reduction

Consistent condom use 1.7 (1.1, 2.6) 1.1 (0.7, 2.8)

Condom use self-efficacy 1.6 (1.02, 2.4) 1.2 (0.8, 1.9)

Perceived HIV risk 3.0 (1.9, 4.6) 0.8 (0.5, 1.2)

Cross-sectional survey Recall bias Not fully nationally representative, but equally

covered low and high implementation areas

Limitations

PAGE 11

Conclusions

page 12

Campaign exposure was associated with risk reduction strategies - more consistent condom use; greater HIV testing; greater confidence in condoms as an HIV risk avoidance strategy

Risk reduction effects primarily concentrated among men, along with some evidence of changes in attitudes to view sexual networking more negatively

Stand-alone campaigns on concurrency likely not sufficient in duration or intensity to change a culturally entrenched, socially tolerated behaviour

Adults in this context need a range of HIV risk reduction strategies

Combination prevention

Clear identification of the counterfactual Only one round of data required Superior to other common matching methods in the

quality of the match it can provide   Easy to implement with the right software

Lessons learned in using CEM

page 13

National AIDS Coordinating Agency, Bostwana Centers for Disease Control and Prevention US Agency for International Development/Pepfar

Contact: rfirestone@psi.org

Acknowledgements

page 14