1. Exploring an aggregate longitudinal model to evaluate a mass media campaign to promote exclusive breastfeeding in Viet Nam Danielle Naugle Robert Hornik

2. Intervention National mass media campaign Development of a franchise network of health centers providing quality counseling and care for infant and young child feeding (IYCF) Trained service providers at 800 health facilities in 15 provinces Advocacy and policy change Extend maternity leave to six months Instate a ban on marketing of breastmilk substitutes for children under 1 year

3. Research Questions Were there population-level increases in EBF over the course of the campaign? If so, can we link increases in EBF to the mass media campaign? Can mass media alone impact EBF rates?

4. National Mass Media Campaign “Nurse More”Nurse More “No Water”No Water “Iron-rich Foods”Iron-rich Foods“Little Sun Franchise Promotion”Little Sun Franchise Promotion

6. Timing of Media Bursts and Data Collection

7. Data 4 provinces: Hai Phong, Quang Nam, Dak Lak, Tien Giang Three-stage cluster sampling Within each province, 4 districts purposefully selected (2 franchise districts; 2 mass media only districts) Primary sampling unit, villages, selected based on population proportionate to size Mothers were selected via systematic random sampling 5 independent cross-sectional samples between August 2011 and April 2014 At each wave, face-to-face surveys with approximately 2,000 mothers of children under 6 months

8. Measures Exclusive Breastfeeding (primary dependent variable) WHO definition: Giving only breastmilk, and no other food, water, or infant formula to a baby under the age of 6 months “Thinking about the time period from when (name of infant) woke up yesterday morning until the time she woke up this morning, was she given any plain water [infant formula, other liquids, or semi-solid or solid foods]?” Dichotomous: 0 = not exclusively breastfeeding, 1 = exclusively breastfeeding Exposure (primary independent variable) Aided recall: “Have you ever seen a video clip with these snapshots below?” Confirmed recall: “What are the key messages you could recall after watching the video clips?” Ordinal: 0 = not exposed, 1 = exposed, but recalled no messages, 2 = exposed and recalled 1-2 message, 3 = exposed and recalled 3 or more messages

9. Analyses Four types of analyses Secular change Individual-level association between self-reported exposure and EBF Commune-level over-time analyses Differences between franchise communes and mass media only communes Multivariate logistic and linear regressions using a robust variance estimator to adjust for commune-level clustering Covariates: mother’s ethnicity, age, education, occupation, whether she is a first- time mother, whether she had a cesarean section, whether or not the mother went back to work, the infant’s age, and the presence of the franchise.

10. Results: Descriptive Statistics 11,277 participants from 118 communes Ethnicity: Kinh (90%) Mother’s age: 27.8 Education: 70% had some secondary schooling Occupation: 1/3 farmers, 1/3 salaried non-government employees, 1/3 housewives, self-employed or salaried government employees First-time mothers: 68% Gave birth in a medical facility: 98% Cesarean section: 25% (increased from 21% at wave 1 to 30% at wave 5) Back to work: 10%

11. Results: Secular Change EBF

12. Results: Secular Change EBF

13. Results: Individual-Level Analyses Mass Media Only Communes Franchise Communes Not ExposedExposed, but no message recall Exposed, and recalled 1-2 messages Exposed, and recalled 3+ messages Levels of Exposure

14. Results: Individual-Level Analyses EBF Association between Individual Exposure and EBF

15. The secular change over time suggests that EBF rates only increased significantly in franchise communes However, the individual-level cross-sectional association between self-reported exposure and EBF suggests that higher levels of exposure were associated with higher levels of EBF in both mass media only and franchise communes Threated by causal order and self-selection Conclusions: Secular and Individual-Level Analyses

16. Commune Level Analyses High Exposure Communes Low Exposure Communes EBF

17. Results: Commune Level Analyses - Time EBF

18. Results: Commune Level Analyses - Exposure

19. Conclusions: Commune Level Analyses The mass media campaign only had an effect on commune level EBF rates in franchise communes, not in mass media only communes How did the mass media campaign have an effect in franchise communes?

20. Did exposure drive franchise attendance? Attendance at Franchise

21. Did franchise attendance drive increases in EBF? EBF

22. Did exposure affect EBF above and beyond the effect through franchise attendance? Drop all individuals who attended “Little Sun” franchise from construction of commune level EBF variable Interaction between time and exposure on EBF “Little Sun” clients included: 0.191; p = 0.048) “Little Sun” clients excluded: 0.180; p = 0.052

23. Conclusions Mass media alone did not improve population EBF rates Mass media -> franchise attendance -> EBF Mass media -> EBF (in franchise communes)

24. Significance To change a complex behavior like EBF, multiple intervention strategies may be valuable Mass media can play a role in scaling up interventions In intervention-settings where there is an interpersonal support structure in place, mass media can contribute to population-level changes in EBF

25. Thank you! Danielle Naugle, PhD Candidate Annenberg School for Communication University of Pennsylvania dnaugle@asc.upenn.edu

26. Individual-Level Analyses Model 1Model 2Model 3 EBF exposure 1 (vs. no exposure)-0.131-0.0790.034 exposure 2 (vs. no exposure)0.218 ** 0.246 *** 0.249 * exposure 3 (vs. no exposure)0.659 *** 0.745 *** 0.699 *** ethnicity (kinh vs. other) -0.289 ** -0.288 ** mother’s age (years) 0.032 *** no school (vs. > 12 years) -0.286 1-5 years (vs. 6-9 years) -0.394 ** -0.395 ** 6-9 years (vs. > 12 years) -0.024-0.026 10-12 years (vs. > 12 years) 0.017 farmer (vs. housewife) 0.215 * 0.217 * government (vs. housewife) -0.005-0.001 salaried (vs. housewife) 0.1550.156 self-employed (vs. housewife) -0.057-0.055 primipara -0.106 * -0.105 * cesarean -0.355 *** -0.354 *** month 0 (vs. month 5) 1.355 *** 1.353 *** month 1 (vs. month 5) 1.290 *** 1.293 *** month 2 (vs. month 5) 1.246 *** 1.244 *** month 3 (vs. month 5) 0.876 *** 0.879 *** month 4 (vs. month 5) 0.462 *** 0.464 *** back to work -0.380 *** -0.384 *** franchise (vs. mass media only)0.827 *** 0.835 *** exposure 1#franchise (vs. no exposure#franchise) -0.247 exposure 2#franchise (vs. no exposure#franchise) -0.005 exposure 3#franchise (vs. no exposure#franchise) 0.080 _cons-0.432 *** -2.194 *** -2.201 *** N90409009 * p < 0.05, ** p < 0.01, *** p < 0.001

27. Results: Commune Level Analyses Model 1 EBF Model 2 EBF Model 3 EBF Model 4 EBF Model 5 EBF after (vs. before)0.188 *** 0.03630.303 ** 0.172 franchise (vs. mass media only)0.150 *** -0.101 ** -0.137 *** 0.0607 after#franchise 0.314 *** -0.307 *** 0.172 exposure 0.129 * 0.217 * after#exposure 0.0257-0.0875 franchise#exposure -0.215 after#franchise#exposure 0.278 * _cons0.263 *** 0.190 *** 0.311 *** 0.111-0.0252 N590

28. Is there a significant before-after effect of being in a higher exposure franchise commune compared to being in a lower exposure franchise commune? How about in mass media only communes? Additional Commune-Level Analyses * p < 0.05, ** p < 0.01, *** p < 0.001 Franchise CommunesMass Media Only Communes Model 1AModel 1BModel 2AModel 2B EBF after (vs. before)0.350 *** 0.0004440.03630.172 exposure 0.00185 0.217 * after#exposure0.191 * -0.088 _cons0.210 *** 0.2070.311 *** -0.0252 N285 305

29. * p < 0.05, ** p < 0.01, *** p < 0.001 Mediation Pathway: Exposure-Franchise Attendance-EBF Model 1AModel 1BModel 2AModel 2B Franchise Attendance [95% CI] Franchise Attendance [95% CI] EBF [95% CI] EBF [95% CI] after0.249 *** -0.1650.350 *** 0.242 *** [0.206,0.293][-0.418,0.0880][0.300,0.400][0.158,0.325] exposure0.198 *** 0.0177 [0.0870,0.310][-0.0149,0.0503] after#exposure0.226 ** [0.0846,0.367] franchise attendance 0.656 *** 0.326 ** [0.482,0.829][0.105,0.546] after#franchise attendance 0.412 ** [0.171,0.653] _cons-0.351 ** -0.01950.03860.125 *** [-0.554,-0.147][-0.0735,0.0345][-0.0246,0.102][0.0590,0.191] N285

30. Why did EBF rates decrease so sharply at wave 5 in franchise communes?

31. Why did EBF rates decrease so sharply at wave 5? Wave 1Wave 2Wave 3Wave 4Wave 5 Franchise Attendance1.30%25.72%28.57%30.13%20.45% N57

32. Why did EBF rates decrease so sharply at wave 5? Number of spots aired per week

33. Mechanisms of Effect and Failure Reasoned Action Model Attitudes Perceived social norms Perceived behavioral control (self-efficacy) [Knowledge]

34. Knowledge (α =.65) Which is better for an infant under 6 months, breast milk alone or a combination of breast milk and infant formula? Until what month should a mother give her infant only breast milk and no other foods, water or infant formula? In what month do you think an infant should start receiving plain water in addition to breast milk? In what month do you think an infant should first start to receive liquids other than water in addition to breast milk? After completing what month should an infant first start to receive semi-solid foods?

35. Attitudes Scale (α =.84) *If I am breastfeeding, but do not give my infant water until s/he completes 6 months, my infant will be thirsty. If I feed my infant only breast milk and no other food, water or infant formula, until s/he completes 6 months, I am giving my infant all the nutrients s/he needs to be healthy. *If I feed my infant a combination of breast milk and infant formula until s/he completes 6 months, I am giving him/her the best possible nutrition. * If do not clean my infant’s mouth out with water after breastfeeding, my infant will get thrush. *If I am breastfeeding my 5 month old infant, but do not give my infant water, s/he will be too hot. *If I feed my infant a combination of breast milk and other foods when s/he is between 4 and 6 months of age, I am giving my infant the best possible nutrition. If I feed my infant only breast milk and no other food, water, or infant formula until he completes 6 months, I am giving my infant all the nutrients s/he needs for optimal brain development.

36. Perceived social norms (α =.72) Most people who are important to me (e.g. family members, friends…) think that I should feed my infant only breast milk, and no other food, water, or infant formula for the first 6 months (injunctive norms) Most women who have infants like me feed their infant only breast milk, and no other food, water or infant formula for the first 6 months (descriptive norms)

37. Self-efficacy (α=.69) My breast milk is of good enough quality to nourish my infant so that the infant does not need any other food, water, or infant formula until s/he has completed 6 months. The more I breastfeed my infant, the more breast milk my body will produce. My body can produce enough colostrum to feed my newborn within one hour an infant after birth. My body can produce enough breast milk to feed my newborn only breast milk and no water or infant formula in the first 24 hours. The “first milk” produced by my body is all my newborn needs in the 24 hours after birth.

38. Franchise communes: Before-after changes in commune level cognitions KnowledgeAttitudesNormsSelf-efficacy M=.72, SD=.15M=4.30, SD=.69M=4.18, SD=.87M=4.71, SD=.46 [95% CI] after (vs. before)0.250 *** 1.232 *** 1.514 *** 0.693 *** [0.227,0.273][1.126,1.337][1.364,1.664][0.606,0.780] _cons0.515 *** 3.318 *** 2.974 *** 4.156 *** [0.487,0.543][3.179,3.456][2.814,3.133][4.061,4.250] N285 * p < 0.05, ** p < 0.01, *** p < 0.001

39. Franchise communes: Cross-sectional association between commune level exposure and cognitions KnowledgeAttitudesNormsSelf-efficacy [95% CI] exposure0.140 *** 0.666 *** 0.857 *** 0.459 *** [0.0974,0.183][0.386,0.946][0.526,1.189][0.215,0.704] _cons0.457 *** 3.081 *** 2.612 *** 3.867 *** [0.378,0.536][2.558,3.604][1.990,3.234][3.406,4.328] N285 * p < 0.05, ** p < 0.01, *** p < 0.001

40. Franchise communes: Before-after cognitions by commune level exposure KnowledgeAttitudesNormsSelf-efficacy [95% CI] after (vs. before)0.1290.7980.5370.754 * [-0.015,0.272][-0.028,1.625][-0.749,1.824][0.0572,1.451] exposure0.087 * 0.477 * 0.4320.486 ** [0.013,0.161][0.0184,0.936][-0.185,1.048][0.178,0.794] after#exposure0.0660.2360.532-0.0334 [-0.008,0.141][-0.211,0.683][-0.140,1.205][-0.398,0.331] _cons0.354 *** 2.442 *** 2.182 *** 3.264 *** [0.216,0.493][1.622,3.263][1.033,3.331][2.712,3.816] N285 * p < 0.05, ** p < 0.01, *** p < 0.001

41. Mass media communes: Before-after changes in commune level cognitions KnowledgeAttitudesNormsSelf-efficacy M=.64; SD=.12M=3.89; SD=.54M=3.71; SD=.68M=4.43; SD=.35 [95% CI] after (vs. before)0.110 *** 0.459 *** 0.507 *** 0.264 *** [0.0812,0.138][0.348,0.570][0.365,0.649][0.182,0.346] _cons0.547 *** 3.525 *** 3.309 *** 4.219 *** [0.515,0.580][3.379,3.672][3.147,3.471][4.117,4.321] N305 * p < 0.05, ** p < 0.01, *** p < 0.001

42. Mass media communes: Cross-sectional association between commune level exposure and cognitions KnowledgeAttitudesNormsSelf-efficacy [95% CI] exposure0.138 ** 0.628 ** 0.724 ** 0.315 * [0.047,0.228][0.197,1.058][0.244,1.204][0.0778,0.552] _cons0.421 *** 2.918 *** 2.590 *** 3.941 *** [0.275,0.568][2.222,3.614][1.822,3.359][3.552,4.330] N305 * p < 0.05, ** p < 0.01, *** p < 0.001

43. Mass media communes: Before-after cognitions by commune level exposure KnowledgeAttitudesNormsSelf-efficacy [95% CI] after (vs. before)0.07700.759 ** 0.3080.604 *** [-0.028,0.182][0.287,1.232][-0.310,0.927][0.291,0.918] exposure0.121 * 0.782 ** 0.622 * 0.490 ** [0.007,0.235][0.209,1.355][0.020,1.223][0.198,0.782] after#exposure0.021-0.1930.128-0.219 * [-0.048,0.090][-0.524,0.137][-0.287,0.543][-0.423,-0.016] _cons0.360 *** 2.311 *** 2.344 *** 3.457 *** [0.177,0.543][1.409,3.212][1.382,3.305][2.988,3.926] N305 * p < 0.05, ** p < 0.01, *** p < 0.001