Strategies for improving health care seeking for maternal and newborn illnesses in low- and middle-income countries: a systematic review and meta-analysis

Background Lack of appropriate health care seeking for ill mothers and neonates contributes to high mortality rates. A major challenge is the appropriate mix of strategies for creating demand as well as provision of services. Design Systematic review and meta-analysis of experimental studies (last search: Jan 2015) to assess the impact of different strategies to improve maternal and neonatal health care seeking in low- and middle-income countries (LMIC). Results Fifty-eight experimental [randomized controlled trials (RCTs), non-RCTs, and before-after studies] with 310,652 participants met the inclusion criteria. Meta-analyses from 29 RCTs with a range of different interventions (e.g. mobilization, home visitation) indicated significant improvement in health care seeking for neonatal illnesses when compared with standard/no care [risk ratio (RR) 1.40; 95 confidence interval (CI): 1.17–1.68, 9 studies, n=30,572], whereas, no impact was seen on health care seeking for maternal illnesses (RR 1.06; 95% CI: 0.92–1.22, 5 studies, n=15,828). These interventions had a significant impact on reducing stillbirths (RR 0.82; 95% CI: 0.73–0.93, 11 studies, n=176,683), perinatal deaths (RR 0.84; 95% CI: 0.77–0.90, 15 studies, n=279,618), and neonatal mortality (RR 0.80; 95% CI: 0.72–0.89, 20 studies, n=248,848). On GRADE approach, evidence was high quality except for the outcome of maternal health care seeking, which was moderate. Conclusions Community-based interventions integrating strategies such as home visiting and counseling can help to reduce fetal and neonatal mortality in LMIC.

With 99% of maternal, newborn, and child deaths occurring in low-and middle-income countries (LMICs), increasing health resources and appropriate intervention in these countries is an urgent priority and global responsibility for reducing the burden of maternal and child mortality (6,7). Antenatal care provides an opportunity to not only detect potential complications but also to prevent them. Birth preparedness Á an easy to deliver and inexpensive intervention Á can avert the brunt of maternal and perinatal mortalities. It includes different interventions such as identifying SBAs, the closest appropriate health facility, and sometimes funds for emergency transportation and consultation, all of which can reduce delays in obtaining care (8). During the last decade a number of systematic reviews have been published which have assessed interventions for improving maternal and newborn health (9Á38). However, none of these have specifically focused on strategies to improve maternal and newborn health care seeking, the aim of this systematic review and meta-analysis.

Methods
All experimental studies from LMICs that assessed the health care seeking behavior or pattern for maternal and newborn health care and illnesses were included. The population for this review included pregnant women at any gestation, postpartum women up to 6 weeks after giving birth, and neonates less than 28 days of life. We included studies that provided information and education for empowerment and change in the form of group meetings or individual one-to-one counseling at home or at primary health care facilities and compared them with standard/no care. The primary outcomes assessed were health care seeking for maternal and newborn illnesses. The secondary outcomes included maternal, neonatal, and perinatal mortality, stillbirths, (Panel 1) and maternal and newborn care outcomes, such as antenatal care, institutional births, and early initiation of breastfeeding.
The protocol for this systematic review and meta-analysis was registered with PROSPERO 2012:CRD42012003 236 (www.metaxis.com/prospero/full_doc.asp?RecordID0 3236). This review was conducted in accordance with methods of the Cochrane Collaboration (39). Ovid platform was used to search PubMed, MEDLINE, and EMBASE; Popline, the Cochrane Library, and Google Scholar were also searched up to 12 January2015. Search terms were a combination of and synonyms of ('care seeking' OR 'care-seeking' OR 'health care' OR 'health care seeking' OR 'community based intervention*' OR 'community-based intervention*') AND (mother* OR maternal OR women OR newborn* OR neonat*) used as medical subject headings and keyword terms in the title/ abstract (Supplementary File 3). No language restrictions were applied. Grey literature (materials and research produced by organizations, [such as community health workers (CHWs) central, High Impact Practices etc.] outside of the traditional commercial or academic publishing and distribution channels) and reference lists of included studies were also searched to identify studies.
ZSL and PM independently reviewed the retrieved articles in two stages; first assessing relevance from the title and abstract, and if relevance was still unclear, reading the full text. Any disagreement was referred to a third reviewer (CC and ZAB). Studies were analyzed according to their study design i.e. randomized (and cluster) controlled trials (RCTs), non-randomized controlled trials (non-RCTs), and before-after studies.
ZSL and PM extracted data independently from each included study. Study design, country of study, participants, intervention, comparison, and duration of intervention were recorded for each study. If information was missing, authors were contacted. The methodological quality of studies was evaluated using standardized forms. The quality of controlled trials was assessed according to Cochrane methods (40). Prospective studies were graded using the methods described by the Effective Practice, Organization and Communication Cochrane review group (EPOC 2009) (41).
We performed statistical analysis of RCTs, non-RCTs, and before-after studies using the Review Manager software (42). For dichotomous data, we presented results as summary risk ratio (RR) and for continuous data we used mean difference (MD) with 95% confidence intervals (CIs). We included cluster-randomized trials in the analyses along with individually randomized trials and therefore their sample sizes were adjusted by the methods described in the Cochrane Handbook (43) using a design effect reported from the trial.
We have set out the mortality outcomes of the review in summary of findings tables prepared using the GRADE approach (44) using GRADE profiler software. For each of these outcomes, we assessed the quality of the evidence, considering within-study risk of bias (methodological quality), directness of evidence, heterogeneity, precision of effect estimates, and risk of publication bias. We have rated the quality of the body of evidence for each key outcome as 'high', 'moderate', 'low', or 'very low'.
The level of attrition was noted for each study. Heterogeneity between trials was assessed using the I-squared statistic, P value of B0.1 (x 2 ), and by visual inspection of forest plots. When high levels of heterogeneity between trials (I-squared exceeding 50%) were identified, further exploration was conducted by subgroup analysis and was tested by interaction tests. We applied random-effects meta-analysis as an overall summary when substantial methodological heterogeneity between and among the studies was found. A priori subgroup analyses were planned to identify the impact on health care seeking with different strategies (community mobilization, home visitation, combination of two, or perinatal health care/ education); and the extent of intervention (birth preparedness, birth preparedness, and recognition and referrals), or (birth preparedness, recognition and referrals and funds for emergency transportation). Potential publication bias was assessed using funnel plots (45).

Results
Our initial search yielded 20,627 articles, 389 of which had relevant titles and abstracts. After reading the full text of these, 72 appeared to meet our inclusion criteria (Panel 2). After finding 14 of these 72 articles did not meet our inclusion criteria, we included and analyzed 58 original studies (90 published papers), of which 29 were RCTs, 15 were non-RCTs, and 14 were before-after studies ( Fig. 1) (characteristics of included studies Á Supplementary File 1).
A variety of different interventions and behaviors were assessed in the studies that met the eligibility criteria for inclusion (Panel 3). These interventions and behaviors included promoting routine antenatal care, institutional births, and early breastfeeding; provision of clean delivery kits; training of CHWs, SBA, and health care staff on birth preparedness; and provision of maternal and newborn health interventions. In several included studies these interventions were provided in the form of packages of different strategies including community mobilization, home visitation, or a combination of two. Heterogeneity was more than 85% for both these primary outcomes ( Fig. 1a and b).
Subgroup analyses, based on intensity of interventions, suggested that birth preparedness alone as an intervention had no impact on improving health care seeking for maternal illnesses ( . 2a and b). Estimates from non-RCTs   mortality; however, it was high quality for the rest of the other mortality outcomes (Fig. 3).

Discussion
Adequately addressing women's and children's health care needs would resolve a considerable proportion of global health problems. Improving health care seeking for the health of mothers and newborns can prevent many avoidable deaths. Although there was a paucity of included studies reporting health care seeking as an outcome, the systematic review found promising results of the several interventions for improving health care seeking for maternal and newborn illnesses. Although the impact was not Only Bashour 2008 (with 2 subgroups Á single visit and 4 visits) did not have birth preparedness component in the intervention. Significant estimates are provided in BOLD.
significant for health care seeking for maternal illnesses, care seeking for neonatal illnesses improved by 40% overall. The impact was enhanced when the intervention was provided by CHWs though home visiting (45% increase) or when combined with community mobilization (62%), however the later evidence came from a single study with a positive impact. Impact was even larger when promotion of birth preparedness was combined with interventions where CHW recognized illnesses and provided referrals (65% increase). While interpreting the results, it is important to consider that studies were not similar across the subgroups for health care seeking for maternal illnesses and neonatal illnesses. The included studies did not find any impact for any of these interventions on improving maternal mortality. Probably these studies were not powered to detect small but important differences. Significant improvements were observed for neonatal mortality (21% reduction) including early (30%) and late neonatal mortality (23%), stillbirths (18%), and perinatal mortality (18%). A similar direction of effect, although not significant, was found from non-RCTs and before/after studies. Although impact on mortality was more convincing when interventions were given in the form of community mobilization in combination with home visiting, the degree of heterogeneity was high. Mortality substantially improved when birth preparedness was combined with recognition of illnesses and provision of referrals; and was even more effective when interventions involved collection of funds for emergency transportation. However, the number of studies with increasing intensity of intervention decreased and there were too few studies in the highest level of intensity to make robust claims.
The review found positive impacts for these interventions from RCTs on improving antenatal care (27%), uptake of tetanus toxoid immunization (8%), iron/folate supplementation (49%), institutional births (16%), and initiation of breastfeeding (85%). Similar direction of effects was observed from other less rigorous study designs.
The subgroup analyses suggest that the impacts on health care seeking, mortality, and morbidity were greater when interventions included recognition of illnesses and provision of referrals. However, the qualitative findings from these trials were scarce and little or no information was provided to relate these findings with the contextual factors of delays in those scenarios. The literature suggests health service demand is not determined by recognition of problems and perceived seriousness alone; there are underlying beliefs which play a vital role in determining health care utilization patterns (46Á48). Ineffective or inequitable health decision making at the household level is a major obstacles in accessing health care (49,50). Timely recognition of danger signs, autonomy of decision making, availability of finances, accessibility of the health facility, and perceived quality of care are necessary considerations when making the decision to seek formal care.
Even though modest improvement in maternal and neonatal health outcomes has been achieved in the last decade, these can be further improved. While the use of advocacy groups and mobilization campaigns can help to optimize the implementation of these strategies; health system investment training the community and facility health staff and equipping them with essential supplies can help them care for a high risk pregnancy, as well as respond to any emergency that may arise. A specific implementation strategy could be the provision of birthing kits to the Traditional Birth Attendant (TBA)s which will ensure access for those residing in remote areas. This is likely to reduce mortality arising from delay in the provision of emergency medical aid during childbirth. In addition, indirect health care costs such as transportation and certain minor charges at the facility should be minimized. Full implementation of these changes will go a long way to improve not only maternal and neonatal health-seeking behavior, but also their health outcomes.

Conclusions
This systematic review identified that strategies such as mobilization and home visitation can improve health care seeking for neonatal illnesses and can reduce perinatal mortality. Further analyses based on strategies which combined birth preparedness counseling with recognition of illnesses and provision of referrals by CHWs showed an improvement in both maternal and neonatal health care seeking. Similarly, strategies which used mobilization with home visitation showed an improvement in both maternal and newborn health care seeking; however the evidence was only derived from a single study. These interventions had a significant impact on reducing stillbirths, perinatal deaths, and neonatal mortality. Most of the included studies were conducted in Asia, with very a limited number of studies from other LMIC countries such as Africa. Thus, there is a clear need for additional high quality research from other LMIC regions. There is also a need to identify the cost-effectiveness of identified strategies to provide interventions in affordable ways to hard-to-reach communities to prevent illnesses and promote health.

Conflict of interest and funding
ZAB was part of the following trials; however, he played no role in the quality assessment of these trials: Bhutta  This review was part of the first author's doctoral thesis which was funded as part of International Postgraduate Research Scholarship by University of Adelaide, Australia. The funders had no role in the study design, study conduct, data analysis, data interpretation, or writing of the report. All authors take responsibility for the integrity and the accuracy of the data. The corresponding author had final responsibility to submit the report for publication.

Paper context
Reducing maternal and newborn mortality will require rigorous efforts to scale up evidence-based interventions, especially community-based preventive, promotive and therapeutic strategies, as well as availability of commodities and health workers in primary care. A lack of appropriate care seeking for ill mothers and neonates is thought to contribute to high mortality rates; therefore, a major challenge is the appropriate mix of strategies for demand creation as well as provision of services. Prior published reviews have systematically assessed the prevalence of care seeking in neonates; however the prevalence of care seeking among mothers has not been systematically assessed. Also there are no reviews which have assessed the strategies or interventions to increase care seeking for maternal and neonatal health. There is a need to systematically review the literature to determine the interventions that can improve the care seeking pattern among mothers and their newborns at primary care setting in developing countries.