Women's Contraceptive Profiles in Burundi: Knowledge, Attitudes, and Interactions with Media and Health Services

"...make data on contraception more useful to reproductive health programs by applying market segmentation approaches, with the aim that such programs can better tailor their messages and services to target potential and existing clients."

As human-centred design (HCD) approaches have burgeoned in global health, renewed attention has been given to segmentation, which involves identifying distinct subpopulations who have different needs, attitudes, and behaviours. An emerging body of research has used market segmentation approaches to identify targeted groups for family planning (FP) interventions. Using sequence and cluster analysis of contraceptive calendar data from the 2016-17 Burundi Demographic and Health Survey (DHS), this study identified discrete clusters characterising patterns in women's contraceptive and pregnancy behaviours over the previous 5 years. The study pairs these clusters with data on factors typically targeted in social and behaviour change (SBC) interventions - knowledge, attitudes, and women's interactions with media and health services - to create composite profiles of women in these clusters.

The analytic sample consists of 13,293 women age 15-44 at the start of their calendar sequence (age 20-49 at the time of the survey). The analysis identified 6 distinct clusters from these women's DHS contraceptive calendar sequences. They comprise 3 clusters with no discernible contraceptive use: (i) Quiet Calendar (42% of women), characterised by women who did not experience pregnancy or use any methods of contraception; (ii) Family Builder 1 (25%) and (3) Family Builder 2 (18%), which are similar in that they are both characterised by women who did not use any method and experienced two pregnancies, but varied in terms of timing during the calendar sequence. The next 3 clusters, all of which are marked by contraceptive use, are: (iv) Modern Mother (8%), characterised by women who adopted short-term modern methods toward the end of year 2 after a period of nonuse and a pregnancy; (v) Consistently Covered Mother (6%), characterised by women who adopted long-acting or permanent methods (LAPMs) after a period of nonuse and a pregnancy; and (vi) Traditional Mother (2%), characterised by those who adopted traditional methods at the end of year 2 after nonuse and a pregnancy.

For the study's regression models, the research select 7 covariates that describe knowledge and attitudes and 6 covariates that describe media exposure and interactions with the health system. These variables align with conceptual frameworks describing how SBC interventions may influence contraceptive behaviour. In particular, the frameworks predict that SBC interventions that engage people through mass media initiatives and interpersonal communication (typically with providers) influence contraceptive behaviour, both directly and indirectly, by increasing knowledge and shifting attitudes and norms.

The analysis found that media exposure and attitudes regarding sex preference, wife beating, and self-efficacy largely do not explain cluster membership. Contraceptive knowledge is positively associated with two clusters (Family Builder 1 and Traditional Mother) and negatively associated with a third (Quiet Calendar). Clusters also differ in their members' fertility desires, contraceptive intentions, and interactions with health services. For example, visiting with a health worker and discussing FP with that health worker is positively associated with membership in the Modern Mother cluster and both Family Builder clusters. The latter two clusters (Family Builders) are both characterised by the presence (but not timing) of multiple pregnancies in their calendar histories, but they differ in that women with high contraceptive knowledge, intentions to use contraception, and well-articulated family size ideals are characteristic of one cluster (Family Builder 1), and low contraceptive knowledge, no use of contraception, and vague family size preferences are characteristic of the other (Family Builder 2).

The detailed profiles of each cluster presented in the paper may guide SBC programming. For example, women in the Quiet Calendar cluster tend to be more educated but also report low contraceptive knowledge and limited access to health services - described here as a somewhat unusual combination of factors and a contrast from women in other clusters. It may be that these women distrust health services, are of a stage of the life course in which they do not perceive health services to be relevant or designed for them, or face other barriers to health services that are unique to them. Therefore, innovative approaches may be required to reach such women with information and connect them to services. For example, ownership of a mobile phone is common in this cluster. However, mobile health apps or outreach by SMS (text messaging) may be effective only for younger, unmarried Quiet Calendar women, but different modalities may be needed to reach older women.

In conclusion, by making use of behaviourally defined clusters of women, this paper represents "a novel application of segmentation analysis to DHS calendar data....The results can inform and guide the design of reproductive health programs as they target SBC and other interventions to the unique subpopulations they seek to serve."