Upload
corby
View
27
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Program-stimulated change in network composition and behavior related to family planning in Ghana. Nepal. Marc Boulay Dynamics of Networks and Behavior Symposium XXIV International Social Network Conference Portoroz, Slovenia May 11, 2004. Social Networks and FP. - PowerPoint PPT Presentation
Citation preview
Program-stimulated change in Program-stimulated change in network composition and network composition and behavior related to family behavior related to family planning in Ghanaplanning in Ghana
Marc BoulayMarc Boulay
Dynamics of Networks and Behavior Symposium Dynamics of Networks and Behavior Symposium XXIV International Social Network ConferenceXXIV International Social Network ConferencePortoroz, SloveniaPortoroz, SloveniaMay 11, 2004May 11, 2004
Nepal
Social Networks and FPSocial Networks and FP
• Widely believed that social networks exert Widely believed that social networks exert a strong influence on contraceptive a strong influence on contraceptive behaviorbehavior
• This influence can be positive or negativeThis influence can be positive or negative• Favorable networks consistently Favorable networks consistently
associated with contraceptive adoptionassociated with contraceptive adoption• Survey and anecdotal evidence has linked Survey and anecdotal evidence has linked
non-use to negative information women non-use to negative information women learn from peerslearn from peers
FP communication programsFP communication programs
• Often promote discussions about FP as Often promote discussions about FP as one mechanism for behavior changeone mechanism for behavior change
• Evaluations consistently show an Evaluations consistently show an association between exposure and FP association between exposure and FP discussiondiscussion
• Unclear whether these program-stimulated Unclear whether these program-stimulated discussions result in a network favorable discussions result in a network favorable to contraceptive useto contraceptive use
Other factors that may also influence Other factors that may also influence composition of FP discussion networkcomposition of FP discussion network
• ProximityProximity• Likelihood of interactionLikelihood of interaction
• Ethnic HomophilyEthnic Homophily• Similarity in languageSimilarity in language
• Contraceptive use statusContraceptive use status• Information from expert sourcesInformation from expert sources• Confirmation/Dissonance reductionConfirmation/Dissonance reduction
Research QuestionsResearch Questions
• What factors influence the selection of FP What factors influence the selection of FP discussion partners within the study discussion partners within the study villages?villages?• Ethnic homophily, contraceptive use Ethnic homophily, contraceptive use • Does program exposure influence selection, Does program exposure influence selection,
particularly of contraceptive users?particularly of contraceptive users?
• Is a positive change in network Is a positive change in network composition associated with composition associated with contraceptive adoption?contraceptive adoption?
Nepal Radio Communication Nepal Radio Communication Project (RCP)Project (RCP)
• USAID-funded project to promote FP USAID-funded project to promote FP among couples in Nepalamong couples in Nepal
• Weekly radio dramaWeekly radio drama• Residents of fictional village modeled Residents of fictional village modeled
interpersonal communication regarding FPinterpersonal communication regarding FP
• Broadcast between 1995-2000Broadcast between 1995-2000
Study DesignStudy Design
• Site: 3 villages in Dang District, NepalSite: 3 villages in Dang District, Nepal• Sample: Nearly all CMWRA (15-49 years) Sample: Nearly all CMWRA (15-49 years)
in November 1997 & March 1999in November 1997 & March 1999• Wave 1: n = 350 (response rate = 98.6%)Wave 1: n = 350 (response rate = 98.6%)• Wave 2: n = 337 (response rate = 83.0%)Wave 2: n = 337 (response rate = 83.0%)• Present in both surveys: n = 281 Present in both surveys: n = 281
• Collected sociometric dataCollected sociometric data• Women living in your village with whom you Women living in your village with whom you
have discussed FP during the past six monthshave discussed FP during the past six months
Description of the study Description of the study villagesvillages
Village2 3 6
% Tharu ethnic group 91.5 44.0 44.9% exposed to RCP – 1997 44.6 52.0 61.4% using a modern FPmethod – 1997
28.5 34.0 45.5
% adopting a modern FPmethod between 1997 and1999
14.9 29.2 21.8
N 130 100 177
Analysis plan for change in Analysis plan for change in network compositionnetwork composition
• Description of change in networksDescription of change in networks
• Bivariate examination of change using Bivariate examination of change using binomial testbinomial test
• Multivariate analysis using SIENAMultivariate analysis using SIENA
Change in FP discussion Change in FP discussion networksnetworks
Village2 3 6
% discussed FP19971999
37.579.2
57.352.0
62.754.5
# of ties19971999
45168
7388
140105
% of 1997 ties that weredropped
71.1 74.0 85.0
% of 1999 ties that wereadded
92.3 78.4 80.0
Odds ratios for addition of a FP Odds ratios for addition of a FP discussion partner to networkdiscussion partner to network
1.55
1.92
1.49
0
1
2
3
Tharu vs Non-Tharu FP user vs Non-user* Exposed vs Unexposed
Source: 1997 and 1999 Nepal Social Network Surveys*p<0.05Adjusted for age, education, parity, village of residence, and ethnic groupH-W standard errors accounting for within-village correlation
Observed and expected addition of Observed and expected addition of Tharus to network, by ethnicity of egoTharus to network, by ethnicity of ego
Addition of discussion partnersfrom Tharu ethnic group
Ego’s Ethnicity Observed Expectedn p-value
VILLAGE 2Tharu 131 121.7 133 0.001Other 12 20.1 22 <0.001
VILLAGE 3Tharu 32 14.1 32 <0.001Other 6 16.3 37 <0.001
VILLAGE 6Tharu 37 16.6 37 <0.001Other 2 21.1 47 <0.001
Observed and expected addition of Observed and expected addition of contraceptive users to network, by contraceptive users to network, by contraceptive use of egocontraceptive use of ego
Addition of discussionpartners that are FP Users
Ego’s FP UseStatus in 1997
Observed Expectedn p-value
VILLAGE 2User 38 18.2 64 <0.001Non-user 37 25.7 90 0.010
VILLAGE 3User 20 9.5 28 <0.001Non-user 23 14.3 42 0.008
VILLAGE 6User 37 23.7 52 <0.001Non-user 29 14.6 32 <0.001
Observed and expected addition of Observed and expected addition of contraceptive users to network, by contraceptive users to network, by program exposure of egoprogram exposure of ego
Addition of discussion partnersthat are FP Users
Ego’s Exposureto RCP in 1997
Observed Expectedn p-value
VILLAGE 2Exposed 35 22.5 79 0.003Unexposed 41 21.7 76 <0.001
VILLAGE 3Exposed 22 13.6 40 0.007Unexposed 20 9.9 29 <0.001
VILLAGE 6Exposed 49 27.3 60 <0.001Unexposed 17 10.9 24 0.014
SIENA AnalysisSIENA Analysis
• Allowed for changing composition of Allowed for changing composition of networksnetworks
• Assumed a constant rate functionAssumed a constant rate function• Stepwise addition of variables into Stepwise addition of variables into
objective functionobjective function• Final models confirmed using an Final models confirmed using an
independent run (initial parameter independent run (initial parameter estimates = 0)estimates = 0)
Variables in SIENA AnalysisVariables in SIENA Analysis
• Network effectsNetwork effects• Ethnic similarityEthnic similarity
• Dichotomous variable (Tharu, non-Tharu) Dichotomous variable (Tharu, non-Tharu)
• Contraceptive Use in 1997Contraceptive Use in 1997• Ego use, alter use, similarityEgo use, alter use, similarity
• Program Exposure in 1997Program Exposure in 1997• Ego exposure, alter exposure, Ego exposure, alter exposure,
• Ego Exposure X Alter FP Use InteractionEgo Exposure X Alter FP Use Interaction• Dyadic covariate (= 1 if Dyadic covariate (= 1 if ii exposed and exposed and jj using FP, using FP,
otherwise = 0)otherwise = 0)
Final SIENA ModelsFinal SIENA Models
Village 2 Village 3 Village 6Est. S.E. Est. S.E. Est. S.E.
Basic RateParameter
3.76 0.80 3.46 0.76 2.45 0.49
Density -2.98 0.12 -3.49 0.17 -3.72 0.42Reciprocity 2.99 0.54 2.19 0.27 2.06 0.29Transitivity 0.95 0.15 1.36 0.23
Cuse97 Similarity 0.33 0.14Cuse97 Alter 0.22 0.12 -0.44 0.21 0.47 0.28Cuse97 Ego 0.50 0.20 0.71 0.32
Exp97 Alter 0.34 0.14Exp97 Ego 0.40 0.20
Analysis plan for adoption of Analysis plan for adoption of FP between 1997 and 1999FP between 1997 and 1999
• Analysis restricted to 170 non-users in Analysis restricted to 170 non-users in 1997 interviewed in both survey waves1997 interviewed in both survey waves
• Huber-White standard errors used to Huber-White standard errors used to account for interdependence of account for interdependence of observations observations • Interdependence based on shared membership Interdependence based on shared membership
in a weakly-connected componentin a weakly-connected component• 216 connected components identified216 connected components identified
Percent of adopters by Percent of adopters by presence of a contraceptive presence of a contraceptive user in discussion networkuser in discussion network
22.828.6
20.7 17.5
0
20
40
60
80
100
Contraceptive user in1997 network
Added a user to networkbetween 1997 and 1999
% o
f n
on
-us
ers
in
19
97
ad
op
tin
g b
y 1
99
9
Yes
No
Source: 1997 and 1999 Nepal Social Network Surveys
Adjusted coefficients from a Adjusted coefficients from a logistic regression model logistic regression model predicting adoption of a FP methodpredicting adoption of a FP method
Source: 1997 and 1999 Nepal Social Network SurveysN=170*p<0.05Adjusted for age, education, parity, village of residence, and ethnic groupH-W standard errors accounting for within-component correlation
ConclusionsConclusions
• Effect of contraceptive use and program Effect of contraceptive use and program exposure on network change variesexposure on network change varies
• No evidence that program exposure No evidence that program exposure directly promotes network composition directly promotes network composition favorable to FPfavorable to FP
• Addition of a FP user to network appears Addition of a FP user to network appears to facilitate contraceptive useto facilitate contraceptive use
LimitationsLimitations
• Small number of networks limits ability to Small number of networks limits ability to assess differences across networksassess differences across networks
• Underreporting of FP discussion partnersUnderreporting of FP discussion partners• Time between measures was fairly longTime between measures was fairly long• With only 2 waves, it is still difficult to With only 2 waves, it is still difficult to
establish time order between addition of establish time order between addition of FP user to network and FP adoptionFP user to network and FP adoption
Next stepsNext steps
• Rerun analyses with Ghana data when Rerun analyses with Ghana data when availableavailable• lower prevalence of FP Uselower prevalence of FP Use• greater number of networks (9)greater number of networks (9)
• Explore role of community group Explore role of community group participation on network changeparticipation on network change