Target 3.1 of the Sustainable Development Goals (SDGs) aims to reduce the global maternal mortality ratio (MMR), defined as the number of maternal deaths attributed to pregnancy-related complications per 100,000 live births, to less than 70 deaths per 100,000 live births by the year 2030 [1]. Achieving this ambitious goal will require a concerted effort from respective country health authorities due to the numerous challenges impeding the prevention of maternal mortality [2].
With an institutional MMR of 133.3% per 100,000 live births in 2013/2014 [3] and 105.9 in 2018/2019 [4], South Africa (SA) has made notable progress in reducing its MMR. Although SA failed to achieve the Millennium Development Goals (MDG) 4a and 5a of reducing under-five mortality by two-thirds and maternal mortality by three-quarters between 1990 and 2015, significant progress was observed [5]. However, this figure is still high and far from the SDG target. The high MMR in SA is in part attributed to the lack of adequate maternal health care complemented by underlying social determinants of poor nutrition and poverty [6, 7].
Although SA’s MMR has improved from 2014 to 2019, complications of hypertension in pregnancy and obstetric hemorrhaging are still the main issues causing maternal mortality, which can be prevented through early uptake of antenatal care (ANC) and postnatal care (PNC) services [8]. The first ANC visit allows for easy detection of pregnant women requiring special attention, referral, and more ANC visits. Early ANC visits are also associated with adherence to antiretroviral therapy during pregnancy among HIV positive women, which reduces the chances of mother to child transmission [9]. Data from SA show that the rate of ANC visits within 20 weeks of gestation was 44.0% in 2012/2013 [9], 53.8% in 2014, and 68.1% 2018/2019 with a different point percent of 14.3 since 2014 [4]. PNC visits within the first 6 days of delivery’ were 73.0% in (2013/2014) [10].
In keeping with the SDG 3 target, SA developed a strategic plan to reduce maternal and child mortality by improving the uptake of maternal and child (MCH) services. To achieve this goal, mobile health technology (mHealth) was integrated into the healthcare system as a strategy to overcome barriers to universal health coverage and promote MCH [11]. The National Department of Health (NDoH) in 2014 launched the flagship MomConnect program to strengthen the quality of MCH services and improve mortality outcomes [12]. The MomConnect program was specifically adopted to address the peripartum factors perpetuating the high maternal and infant mortality rates in SA [12].
The MomConnect program is a digital health communication program [13], which was informed by lessons, experiences, and successes of the MAMA SA program [14, 15]. MAMA SA focused on the prevention of mother-to-child transmission (PMTCT) of HIV and consisted of a multi-channel mHealth approach to communicate healthy pregnancy and newborn child support including mobile phone-based chats and voice messages [16].
The MomConnect program was also guided by an emergent body of work around the use of mobile phones to enhance MCH programs [16,17,18,19,20,21]. Evidence suggests that communicating with pregnant women and new mothers via regular SMS and/or voice-based messages supports the adoption of healthy behaviors and increases the uptake of health services [12]. The MomConnect program leveraged elements from the MAMA SA program based on research to determine the content that would be provided in combination with the NDoH’s messaging on MCH and MCH services.
The program aimed at (i) registering all pregnant women using the public health sector; (ii) sending targeted health promotion messages during pregnancy and up until the first year of age to encourage healthy and health-seeking behavior [14]; and (iii) providing pregnant women with a feedback platform to rate the services they received [18]. MomConnect is a two-way communication channel, providing women the opportunity to rate the MCH services received at the clinics and ask questions regarding their health and that of their babies.
Although no study has shown the effectiveness of the MomConnect program in improving the uptake of MCH services, there is evidence that the national ANC visits within 20 weeks of gestation rate have been increasing in the 3 years following the implementation of MomConnect [4, 10]. Even though MomConnect has a framework for explaining how the program is expected to contribute to expected outcomes, the mechanisms underlying and health systems context that influence the outcome of the MomConnect program remain elusive. Despite this limited understanding, MomConnect was rapidly rolled out. An important step in improving the impact of MomConnect is by unearthing its program theory – the explicit and implicit assumptions of how the program is expected to work.
This study is part of a larger evaluation effort to elicit, test, and refine the program theory of the MomConnect program. The larger study has three phases: Phase 1 will focus on eliciting the initial program theory using three different approaches: a scoping review, document review, and stakeholders’ interviews. Phase 2 will be based on testing the initial program theory, while Phase 3 will entail refining the initial program theory [22]. This paper is part of Phase 1 whereby we sought to glean the initial program theory of MomConnect based on analysis of program-related documents using a combined Theory of Change (ToC) and Realist Evaluation (RE) approaches.
Methodological approaches
We conducted a document analysis using design- and implementation-related documents of the MomConnect program. Document analysis is a systematic procedure for reviewing or evaluating both printed and electronic (computer-based and Internet-transmitted) material [23]. Our document analysis was informed by the ToC and RE approaches [24].
ToC and RE are members of the theory-driven evaluation (TDE) or theory-based evaluation (TBE) family. TDE/TBE are regarded as key to untangling the multiple processes between policy intent and outcomes by investigating program implementation, and the causal link processes that trigger outcomes [25, 26]. Traditionally, ToC and RE are applied separately in different studies to obtain corresponding outputs. However, we considered the two approaches together in our study to obtain a more robust initial program theory for the MomConnect program.
According to Taplin et al. [27], ToC makes underlying assumptions explicit by delineating the different components: input, outputs, outcomes, and impact [28, 29]. ToC can be viewed as both a product and process [27]. As a product, its inquiry results in specific outcomes in a narrative and/or visual form and offers a framework for sense-making that needs to be used, revisited, and adapted as the project or program progresses. As a process, the ToC model shows how a goal will be reached [27].
While ToC illustrates ‘how’ change occurs through program implementation, RE goes deeper by establishing ‘why’ through a mechanism-based generative causal framework. Blamey and Mackenzie [30] propose that the ToC can be used as a means of explicating intervention implementation theory for program planning, improvement, and the development of robust monitoring systems at a whole program level; while RE can be used to examine, in detail, how and why the different aspects of the program leads to the observed or intended outcome – program theory (Fig. 1).
RE focuses on providing a ‘causation-based explanation’ – depiction of cause-effect relationships among elements [24] – to provide explanations of program effectiveness [32]. RE is a theory-driven approach to evaluation informed by principles of critical realism – social structures and agents having underlying causal powers that interact to cause an observed behavior [33]. RE aims to unearth theories to explain how and why the intervention is supposed to trigger change [34]. RE theories are formulated by conceptualizing the relationship between the context (C) within which the program is implemented, the generative mechanisms of change (M), and the observed outcomes (O). This conceptualization is achieved by formulating the Context-Mechanisms-Outcome configurations (CMOs) [35]. Of importance, nevertheless, is the notion that the Intervention (I) must be accepted by the users, Actors (A), for it to be successful. Therefore, we considered adding elements of the Actors (A) and intervention (I) modalities to the original CMOs heuristic tool and adopted the intervention-context-actor-mechanism-outcomes (ICAMO) configuration [33].
There are multiple definitions of ‘mechanism’. In RE terms, a mechanism (M) relates to a combination of intended and unintended resources provided by the program and the response to those resources by stakeholders [36]. A mechanism is also defined as the reasoning and responses that participants attribute to the resources, opportunities, and constraints of a program to bring about the observed outcomes [37, 38]. Others have defined mechanisms as the underlying entities, process, or structures, which operate in a particular context to generate outcomes of interest [38]. Context (C) is the physical or perceived conditions, which allow the mechanisms to come into operation or remain inactive. The context can also be defined as circumstances that configure the settings in which the intervention takes place, and the action sets of activities that will trigger the change [24]. An outcome (O), which can be represented in short and intermediate terms, is what is observed resulting from the interaction between the context and mechanism that is triggered by the intervention, and could be measured as the impact of the intervention [39].
By combining the ToC and RE approaches we intended to unearth, in a systematic manner, the assumptions underlying the MomConnect program and speculate the changes likely to take place as a result of the intervention [40]. Our intention to glean the initial program theory of the MomConnect program informed by ToC and RE approaches was based on harnessing the overlapping and complementary features of these two approaches. To this end, we adopted a framework proposed by Dhillon and Vaca [24] (Fig. 2) integrating the features of both ToC and RE approaches and how they complement each other to inform the elicitation of a robust and testable program theory.