Communication, Ideation and Contraceptive Use in Burkina Faso: An Application of the Propensity Score Matching Method

"Communication programmes are more effective in changing contraceptive behaviours when people are exposed to consistent messages from multiple sources."

Against a background of poor reproductive health outcomes, the Burkinabe Government worked with a number of national and international organisations to design and implement appropriate reproductive health interventions. One such intervention was the Gold Circle (GO) initiative, implemented under the United States Agency for International Development (USAID)-funded reproductive health project Santé Familiale et Prévention du SIDA (SFPS) with the aim of increasing contraceptive use while fostering quality services. The propensity score matching (PSM) technique was the analytical method used to assess the effects of the multi-component GO communication campaign.

In brief (see also Related Summaries, below), both the design and the implementation of the GO initiative are based on a mixture of theory and empirical data. Based on the results of the formative research, GO targeted both the demand and the supply sides. On the demand side, the GO strategy involved the use of mass media and community activities to empower clients and the community to expect and demand certain standards of quality from providers, to promote GO sites, and to increase contraceptive use. In addition to the activities that directly promoted the GO initiative, the GO strategy included the development and distribution of 4 short radio and television programmes about family planning. These infomercials were 3-4 minutes long and were developed to provide information on methods, debunk myths and rumours about the methods, and provide information on management of side effects.

The theoretical basis underlying the ideation model used in this paper is that a person's ways of thinking (which can be measured by a set of psychosocial variables) form a part of the proximate determinants of behaviour. Furthermore, the model assumes that the ideational determinants mediate the behavioural impact of sociodemographic and contextual variables, including communication. In analysing the effects of the GO campaign, the researchers posit that the campaign has both direct and indirect (through the ideational variables) effects on contraceptive use. They first examine the effects of campaign exposure on pertinent ideational variables. Subsequently, they look at the impact of the campaign on contraceptive use while controlling for the ideational variables. The ideational variables they examine are borrowed from psychosocial theories and include knowledge about contraceptive methods, personal approval, discussion with others, and personal advocacy.

This was a cross-sectional survey conducted in 2001 involving 1,421 women of reproductive age from Ouagadougou and Bobo-Dioulasso who were randomly selected from 54 enumeration areas covered during the 1999 Demographic and Health Survey. The predominantly Muslim sample was on average younger than 35 years, and more than one-third of them were never married. Very few read magazines or newspapers; the majority listens to the radio or watches the television regularly instead.

Almost four-fifths (78.7%) of the respondents reportedly saw or heard at least one of the campaign materials and could recall the messages. Overall, 8.8% could recall the GO promotional campaign messages alone, 17.0% recalled the infomercials alone, and more than half (52.9%) recalled messages from both components of the campaign. Subsequently, the term 'high exposure' is used to describe respondents who were exposed to both components of the campaign. and 'low exposure' those who were exposed to only one component. The major sources of exposure to the campaign were television (50.1%), radio (22.4%), print media (4.4%), and health facility (3.4%).

To assess the effects of exposure on ideational variables, the researchers first compared the outcome between the respondents who were exposed (treated) and those that were not exposed (controls) (see Table 2, where data are provided, for example, to suggest that the campaign is associated with increased knowledge about contraceptives and more favourable attitudes towards family planning). Subsequently, PSM was used to adjust for observed pre-existing sociodemographic characteristics (see Table 3).

Overall, about one-quarter (25.6%) of the respondents were currently using a modern contraceptive method. Contraceptive prevalence increased monotonically with campaign exposure: 9.3% of the respondents with zero exposure, 19.6% of those with low exposure, and 35.2% of those with high exposure reported current use of modern methods. The data showed that contraceptive use among the respondents with low campaign exposure would have increased by 13.2 percentage points had they experienced high exposure. Similarly, high exposure increased contraceptive use by more than 21.8 percentage points compared to zero exposure.

"[T]hree separate models were estimated to account for the multiple-treatment nature of the campaign, and the results are presented in Table 5. In the results presented in this table, the outcome variable is current use of modern methods while the predictor variables are propensity score quintile, campaign exposure and ideational variables. The analyses show that personal approval of family planning does not have significant independent effects on contraceptive use. In contrast, personal advocacy in favour of family planning, knowledge about contraceptive use and discussion of family planning show the expected positive relationship with contraceptive use. The analyses further reveal that in the presence of ideational variables and the propensity for campaign exposure, low campaign exposure is not associated with a significant increase in contraceptive use compared with zero exposure. In contrast, a high level of exposure makes a significant difference compared with zero exposure or low exposure, controlling for ideational variables. In other words, it appears that what makes a significant difference for increased demand for contraception is considerable exposure to the campaign. At a high level, the campaign has a significant direct impact on contraceptive use, which is not absorbed by the ideational variables. In contrast, at a low level of exposure, much of the impact of the campaign is indirect and filtered through the ideational variables."

The researchers suggest that PSM is a useful technique for documenting the effects of communication interventions. They explain that "A serious issue in ascribing causal effects using observational data (i.e. data not derived from randomised experiments) is the counterfactual problem. The PSM technique addresses this problem by creating groups of treated and untreated individuals that are statistically equivalent on a set of background characteristics." PSM advantages include:

1. "[I]t reduces a set of background variables to a single 'composite' score that appropriately summarises the variables. This reduction in dimensionality makes it easier to see the extent to which treated and untreated groups overlap on background variables. Given such overlap, PSM allows straightforward estimation of treatment effects."
2. "[U]nlike regression models, in estimating treatment effects PSM does not rely on any particular functional form for the relationship between the outcome and the predictors within each treatment group."

In conclusion, this study reinforces the evidence that appropriately designed communication programmes using multiple channels can convey to large audiences consistent messages within a short time.