Setting
Gujarat is considered a strong Indian state in terms of administrative capacity and ranks high on economic development. Yet its health indicators are lower than other states that are less economically endowed. Significant gaps in provision and quality of maternal health care services exist in Gujarat [
22] (Table
1). Efforts to work with civil society to strengthen community participation and action for health are particularly important given that the state has resources and capacity, but faces challenges in ensuring that services reach those most in need.
Table 1
Key maternal and child health indicators for intervention districts, state and national levels
Anand | 78.5 | 92.6 | 93.8 | 99.1 | 21.7 |
Panchmahal | 51.2 | 79.6 | 79.0 | 63.7 | 36.3 |
Dahod | 39.2 | 84.3 | 82.3 | 65.3 | 24.9 |
Gujarat | 70.6 | 88.7 | 87.3 | 87.9 | 26.4 |
India | 51.2 | 78.9 | 81.4 | 91.9 | 21.0 |
From 2012 to 2015, the NGO Society for Health Alternatives (SAHAJ), led the 'Community Action for Maternal Health Project,' implemented in 45 villages with approximately 108,000 people. The project worked in two different contexts in Gujarat: Dahod and Panchmahal districts (remote, tribal populations) as compared to Anand district (more urban and affluent). This NGO-community partnership simultaneously addressed demand and supply side constraints to poor utilization and low quality of services by a) raising awareness of maternal health entitlements, b) supporting community monitoring of services, and c) facilitating dialogue with health providers and other key stakeholders (Table
2).
Table 2
Community action for maternal health project aims and activities
1. Awareness of entitlements by pregnant women and community | Focus group discussions and participatory methods with women's groups to elicit local understanding and preferences for safe delivery |
Community wide meetings and group-specific meetings. |
Pictorial banner for group discussions (toran) and individual poster for woman about entitlements (mahiti patrika) |
2. Community monitoring of receipt and delivery of services | Home visits with individual women at 8 months pregnancy and postpartum using healthy mother tool (warli madi tool) |
Monitoring of outreach antenatal services at Village Health and Nutrition Day (VHND) (VHND tool) |
Maternal death tracking to triangulate government tracking |
3. Dialogue with stakeholders about gaps identified | Development of report cards of Primary Health Centre (PHC) functioning |
Support of community members during the Maternal Death Review process |
Raising awareness
Community meetings first elicited local perceptions of safe deliveries among women and providers, before developing a common understanding of essential services and entitlements [
10,
23]. Subsequent meetings imparted information about these entitlements; raised awareness about nutrition, antenatal checkups, high-risk symptoms, newborn care, immunization; and served as a space to follow up on other health problems, such as tuberculosis [
24]. The meetings also built community ownership around village-level health issues and collective decision-making on who would follow up on actions decided upon. In addition, monthly meetings were convened by volunteers for additional problem solving related to the Public Distribution System, water, early childhood care centers (
Anganwadi), Below Poverty Line cards, access to other entitlements, based on other findings from community monitoring efforts [
25].
In an effort to support birth preparedness, a poster, called a
Mahiti Patrika, was developed to provide information regarding maternal health services and emergency contact numbers of health providers [
23]. This poster was displayed on the exterior house walls of all pregnant women visited. Volunteers reported that women and families used the
Mahiti Patrika to call the numbers listed. NGO project documentation notes that, in addition to dialing 108 (the emergency transport number) for referral transport, pregnant women and family members also used the information on the
Mahiti Patrika to follow up on immunization for their child and to solicit help from health providers with deliveries, including those that resulted in complications [
25].
Another educational resource utilized was a pictorial banner, called a
Toran, to increase awareness among women of the services provided through outreach antenatal care services. Project staff noted that the cartoons initially used were not understood due to low literacy levels. Instead, photographs were found to be more effective, particularly those that depicted local women and government health providers [
25]. Medical officers found it motivating to be depicted in the
Toran and were reported by NGO members to be more responsive after receiving such positive publicity. NGO respondents also noted that local political parties also used the
Toran in their campaigns (Interview 54).
Community monitoring and dialogue with health authorities
The project also developed the
Warli-Madi (Healthy Mother
) tool, a short pictorial checklist of maternal health entitlements and services for community members to prospectively track pregnant and postpartum receipt of key services [
23]. Trained volunteers visited households twice with the tool, once during a woman's 8th month of pregnancy and again 10–20 days post-delivery [
10]. The tool was finalized in early November 2012 after seven months of development and testing, building on women's own perspectives of safe delivery and professional standards of quality care and subsequently further simplified based on a mid-term review assessment [
23,
24].
SAHAJ took primary responsibility for collating data from the monitoring tool into the report cards, investing significant time in checking forms and liaising with ANANDI and KSSS about the data. A color-coded system was developed to denote whether levels of service receipt were poor (red), average (yellow) or good (green). NGO staff noted that over time seeing some of the indicators change in color was highly motivating to community members and health providers alike (Interviews 39–41).
In Dahod and Panchmahal districts, the local NGO shared report card results at trainings with volunteers, during women's collectives meetings and ward meetings. Report cards were first shared with local health authorities and medical officers in March 2013. Subsequently, medical officers were interested in receiving primary health center-wise report cards and discussed the results with their staff [
23]. Changes documented include the restarting of services (increasing the number of outreach clinics in hard to reach areas, initiating deliveries in a previously defunct facility), repairs that improved the quality of the service environment (fixing leaks and toilets), better relationships between community members and government providers (health trainings by government providers for women's collectives, invitation to NGO partners to attend block level maternal death review meetings), and addressing inappropriate practices (kick-backs between female community level providers and private providers, private hospital not providing services as per the public-private insurance scheme) [
25,
26].
Methods
Qualitative data included project document review (annual reports, presentations, other project material) and 56 interviews undertaken in 2015 with purposively sampled community members (n = 22), health providers (n = 16), health authorities (n = 8), and project personnel from implementing NGOs (n = 10).
For the qualitative component, extensive consultations were undertaken with the NGO to collect data from villages across districts that were categorized by the NGO as having strong vs. weak implementation to ensure that diverse project experiences would be captured. Semi-structured interview guides were reviewed with implementing NGOs and piloted before being finalized for use. Interviews were done by four data collectors. A local research assistant, a foreign postgraduate student with family in the region, and two more senior medical doctors with extensive community medicine, civil society and government experience. Interviews were carried out in June–July 2015. Interviews were done in the language the respondent was most comfortable with given that the data collectors were fluent in Gujarati, Hindi and English. While the NGO facilitated transport and introductions to key project personnel and leaders in the villages, written consent was obtained by the interviewers without involvement of the NGO. Interviews were recorded electronically if permission was given and detailed notes were taken irrespective of whether the interview was electronically recorded or not. After listening to the interviews, only those that were the most rich across key stakeholders were transcribed.
The research team developed and pilot tested a code book. The final set of codes were derived from the research aims, the interview guide and from the transcripts following three overarching domains; project processes, outcomes and context. We undertook thematic analysis of interviews applying the codes to the interview notes and transcripts from varied stakeholders at different levels of the health system. Triangulation across project documentation and interviews helped understand project processes more fully. Peer review debriefing between the interviewers and key NGO staff helped improve the quality of the data analysis. Research was planned closely with local NGOs to ensure consideration of the local context and accurate interpretation of findings.
The design of the project was to empower women through the collection, analysis and dissemination of data at village level. Volunteers collected the data on pregnant women prospectively through household visits in their respective villages. Hence the data are not based on any pre-determined sampling design and represent the efforts of volunteers to gather data from as many women as possible. The data collected by the volunteers was then collated and analyzed by the NGO, with a view to sharing results back to the community. For the quantitative data, we analyzed the available data that had been collected by the volunteers through monitoring of pregnant and postpartum women (Dahod and Panchmahal districts, n = 1145 pregnant and postpartum women, Anand district n = 1250 pregnant and postpartum women). The number of women was determined by the capacity of the village health volunteers to track the women and varied from village to village based on the local volunteers' time availability and motivation. The results presented in the paper use the data collected by the volunteers without any sampling modifications.
In analysing the household level data from the project database, we restricted analysis of indicators to variables in districts with the longest implementation, where definitions were unchanged, and data collection was considered reliable. Frequencies and proportions were used to analyze the joint distribution of predictor variables (like demographic and socioeconomic characteristics) and care seeking variables including awareness of maternal health entitlements, content of antenatal care and care seeking for delivery care across years of implementation.
Multivariable logistic regression was focused on the delivery care seeking variable, whether there were any changes in the pattern of care seeking for delivery care among the most vulnerable groups (Schedule Caste and Schedule Tribes (SC/ST) compared to the less vulnerable groups (General Caste (GC) and Other Backward Castes (OBC)) over the duration of the project. Two indicators for patterns in care seeking for delivery were analyzed: a) proportion of all pregnant women using health facilities for delivery care; b) proportion of women delivering at facilities who use government facilities. Other explanatory variables include mother's education, occupation, ownership of mamta card (maternal and child health card) and type of family. Table
3 describes the different variables used in the regression model. All the independent variables were fit using the multivariable logistic regression model. To model the effect of time (program effect) on different groups of vulnerability, an interaction term of time and dimension of vulnerability (by caste) was added to a logistic regression model. All women were used for the outcome of facility delivery, while the model for type of facility used for delivery was restricted to the women who reported delivering at a facility
$$ probability\kern0.3em of\kern0.34em outcome(y)=b0+b1 vulnerability+b2 time+b3{time}^{\ast } vulnerability+b4\left( Other\mathit{\exp}\mathit{\ln } atory\mathit{\operatorname{var}} iables\right) $$
Table 3
Variables used in multivariable regression
Facility delivery | Delivered at a facility Delivered at home or on the way |
Type of facility | Government/public facility Private facility |
Mother's education | No schooling Primary Secondary or higher |
Mother's occupation | Not employed Employed |
Ownership of mamta card (maternal and child health card) | No Mamta card Have Mamta card |
Type of families | Joint Nuclear |
Time | Year 2013 Year 2014 Year 2015 |
Vulnerability | Least vulnerable - General Caste and Other Backward Castes Most vulnerable - Schedule Caste and Schedule Tribes |
Clustering of responses at the village level was accounted for by the use of robust variance estimators based on a first-order Taylor series linear approximation. The adjusted prevalence of the outcome indicators for Dahod and Panchmahal over the period of the project are presented graphically with 95% confidence intervals. For Anand, the interaction models did not achieve convergence and hence the proportions are presented.