Mapping of research on maternal health interventions in low- and middle-income countries: A review of 2292 publications between 2000 and 2012

Background: Progress in achieving maternal health goals and the rates of reductions in deaths from individual conditions have varied over time and across countries. Assessing whether research priorities in maternal health align with the main causes of mortality, and those factors responsible for inequitable health outcomes, such as health system performance, may help direct future research. The study thus investigated whether the research done in low- and middle-income countries (LMICs)matched the principal causes of maternal deaths in these settings. Methods: Systematic mapping was done of maternal health interventional research in LMICs from 2000 to 2012. Articles were included on health systems strengthening, health promotion; and on five tracer conditions (haemorrhage, hypertension, malaria, HIV and other sexually transmitted infections (STIs)). Following review of 35,078 titles and abstracts in duplicate, data were extracted from 2292 full-text publications. Results: Over time, the number of publications rose several-fold, especially in 2004-2007, and the range of methods used broadened considerably. More than half the studies were done in sub-Saharan Africa (55.4%), mostly addressing HIV and malaria. This region had low numbers of publications per hypertension and haemorrhage deaths, though South Asia had even fewer. The proportion of studies set in East Asia Pacific dropped steadily over the period, and in Latin America from 2008 to 2012. By 2008-2012, 39.1% of articles included health systems components and 30.2% health promotion. Only 5.4% of studies assessed maternal STI interventions, diminishing with time. More than a third of haemorrhage research included health systems or health promotion components, double that of HIV research. Conclusion: Several mismatches were noted between research publications, and the burden and causes of maternal deaths. This is especially true for South Asia; haemorrhage and hypertension in sub-Saharan Africa; and for STIs worldwide. The large rise in research outputs and range of methods employed indicates a major expansion in the number of researchers and their skills. This bodes well for maternal health if variations in research priorities across settings and topics are corrected. The paper summarises the findings of a large-scale systematic mapping of all maternal health interventional research in LMICs published between 01/01/2000 and 31/08/2012 [11]. The systematic mapping, using full text publications, covers a broad body of literature on maternal health, and differs from a classic systematic review that addresses a single, narrowly-defined research question [12]. It uses established methods for producing systematic maps, including those with an emphasis on health equity [13, 14]. In the mapping we identify and describe all papers published on this broad topic, but do not assess the quality of the included research. A sensitive search strategy was developed for the mapping, using both controlled vocabulary and free-text terms to identify studies in Medline (PubMED). The search strategy was then adapted for searching other electronic sources, namely CINAHL, Embase, LILACS, PopLINE, PsycINFO and Web of Knowledge (Additional file 1). Methods used in the search strategy are described elsewhere in more detail [15]. We included studies in LMICs that targeted women in pregnancy, during childbirth or within 2 years postpartum, or men within maternal health services. Studies had to include health systems, health promotion or community-based interventions; or interventions on one of five clinical tracer conditions: haemorrhage, hypertension, HIV, sexually transmitted infections other than HIV (STIs), or malaria. The first two of these tracers were selected as they constitute the leading causes of direct maternal deaths [6, 16, 17], while HIV and malaria are the principal causes of indirect maternal deaths in some LMICs [1, 6]. STIs other than HIV remain a key, but neglected, cause of maternal and newborn morbidity and mortality [18, 19]. General health system interventions were included only if they reported outcomes in a maternal health population. We excluded articles related to infertility. Descriptions of population coverage of routine services were also excluded, given difficulties in standardising data extraction from these very diverse studies over a large study team (15 reviewers across 8 countries). All study designs were included, aside from narrative reviews and policy discussion papers. Studies could be in Arabic, English, French, Portuguese or Spanish. Management of the database, screening for eligibility and data extraction were done using online systematic review software (EPPI-Reviewer 4; http://eppi.ioe.ac.uk/cms/). Data extraction codes were piloted and then refined. Reviewers received training on screening articles for eligibility and data extraction. Of the 45,959 articles initially uploaded, 10,881 were duplicate items (23.7 %; Flow Chart). The titles and abstracts of the remaining records (35,078) were screened independently by two reviewers. Differences between reviewers were resolved by a third, more senior reviewer. A total of 18,386 articles were identified on maternal health, of which 16,094 were excluded. The most common reason for exclusion was the absence of an intervention or study outcome (10,536). Almost 4500 studies of clinical interventions other than the tracer conditions were identified and excluded (4450). Of 4175 full text papers reviewed, 2292 were included in the final mapping (54.9 %); data were then extracted from these papers. The list of LMICs and their respective economic category (low-income, low-middle income and upper-middle income) was based on the World Bank classification [20]. Each study was categorised as being a systematic review, effectiveness study (non-experimental quantitative assessment of effectiveness, such as time series and cohort studies), randomised controlled trial (RCT), qualitative study, modelling study or mixed methods research. Systematic reviews and modelling studies were not classified as pertaining to research activities in a particular country, unless it specifically focused on a country. Interventions were classified as targeting one or more of pregnancy, intrapartum or the postpartum, and a specific population, such as women, men, traditional birth attendants (TBAs), and programme managers. Attention given to health inequities was assessed by identifying the proportion of articles that determined intervention outcomes across different categories of social differentiation, as defined by the mnemonic PROGRESS-Plus: Place of Residence, Race/Ethnicity, Occupation, Gender, Religion, Education, Socioeconomic status and Social Capital, Age and Disability [21]. As per the WHO framework for health systems [22], health systems interventions were defined as actions undertaken to improve the functioning of one or more of the five WHO Health Systems Building Blocks, or to enhance access, coverage, efficiency, or quality of maternal health services. Health promotion interventions-implemented either within communities or at health facilities-encompassed activities targeting, for example, TBAs, men, transport and demand-side financing [23]. Data on the journal’s Impact Factor were downloaded from Thomson Reuters [24] and the total number of health publications in 2000–2011 was extracted from Rottingen JA et al. [25]. Data checks were performed in the EPPI-Reviewer software and in Stata 13 (StataCorp LP, College Station, TX, USA); the latter was also used for analysis. Characteristics of research on health systems and health promotion, and the five clinical tracers were compared across settings, time, study design and populations targeted. We also examined the proportion of studies on each of the clinical tracers that included a health systems or health promotion component. The distribution of Impact Factor of the journal and proportion of studies that were RCTs were used as a proxy for the quality of the research done. Only countries with five or more publications were included in cross-country comparisons; those with fewer papers were grouped together. To identify changes over time, publication rates were calculated for three time periods (2000–2003, 2004–2007 and 2008–2012). For comparison, the total number of papers in the last time period (01/2008-08/2012) was multiplied by a factor of 0 · 86, as this period was 4 · 67 years, while the other periods were 4 years each. To assess alignment between the research outputs and the burden of disease from different conditions, we compared the number of articles with the estimated number of women dying from the condition in different settings [6]. Total numbers of papers in the review from each country and geographical region was contrasted with the burden of maternal deaths in that country or region [6]. We compared the number of publications on HIV interventions in different countries with the number of HIV-infected pregnant women in 2012 [26]. The numbers of papers per country and per region were also compared with the average GDP (2000–2012) [27]. Finally, to assess the priority given to maternal health research in each country, we determined the proportion of all health publications in 2000–2011 that described a maternal health intervention. Chi square tests were used to detect associations between categorical variables, and the Chi square test for trend to identify changes over time. The Mann-Whitney U test identified associations between Impact Factor and other study variables. Multiple responses were possible for many variables, given that some studies involved more than one country, or addressed several populations and topics. The sum of percentages for these variables may thus exceed 100 %.