Exploring trust in religious leaders and institutions as a mechanism for improving retention in child malnutrition interventions in the Philippines: a retrospective cohort study.

OBJECTIVES: In the context of persistent child malnutrition in the Philippines, the objective of this study was to examine how different dimensions of trust affected programme retention and physiological outcomes when a faith-based organisation (FBO) addressed moderate and severe acute malnutrition among children from households experiencing extreme poverty.
SETTING: We retrospectively analysed survey data collected by International Care Ministries (ICM) in 2012-2013 across 150 communities in eight provinces (Negros Oriental, Negros Occidental, Bohol, Palawan, Sarangani, South Cotabato, Sultan Kudarat and Zamboanga del Norte) of the Philippines. STUDY PARTICIPANTS: Caregivers of 1192 children experiencing moderate acute malnutrition and severe acute malnutrition between the ages of 6 and 60 months. INTERVENTION: A 16-week child malnutrition treatment programme called Malnourished Child Outreach offered by ICM in partnership with local religious leaders and institutions. PRIMARY AND SECONDARY OUTCOME MEASURES: Programme dropout and weight-for-height z-score (WHZ) at the end of the programme for enrolled children were the two outcomes of interest. A logistic mixed-effects model was built to assess factors associated with programme dropout and a linear mixed-effects model for factors associated with WHZ at the end of the programme.
RESULTS: Trust in religious leaders or institutions (-0.87 (95% CI: -1.43,-0.26)) was negatively associated with programme dropout, suggesting that with increasing levels of trust, decreasing proportions of children dropped out of treatment. Retention in the programme led to improved WHZ among participating children (-0.38 (95% CI: -1.43, 0.26)). Various measures of social capital, including trust in religious and public institutions, were not associated with WHZ at the end of the programme.
CONCLUSIONS: Leveraging pre-existing trust in religious leaders and institutions among households experiencing extreme poverty is one way that ICM, and potentially other FBOs, can promote retention in child nutrition interventions among vulnerable populations. © Author(s) (or their employer(s)) 2020. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

The study was conducted in a unique setting for examining the underlying mechanisms associated with the retention of children enrolled in malnutrition interventions delivered by a faith-based organisation to children in households experiencing extreme poverty.

The retrospective cohort design of this study is an opportunistic and practical approach for exploring the role of religious leaders in health interventions.

Linear and logistic modelling is used to explore the pathways through which social capital shapes nutrition outcomes in malnourished children who concurrently experience extreme poverty in the Philippines.

Multi-level modelling allows analyses to account for the hierarchical structure of variables that are present in the communities where the child malnutrition intervention described in this study took place.

While multiple factors were considered to examine treatment outcomes in child malnutrition interventions, the study could not account for all potential confounders within the complex social settings where the study was conducted.

Introduction

Faith-based organisations (FBOs) play a critical role in delivering healthcare in low resource settings. Compared with public health facilities and providers, programmes and interventions offered by FBOs in some settings may have increased geographical and socioeconomic coverage, greater social and physical capital, and more flexible governance and funding structures.1–3 Additionally, many FBOs focus exclusively on serving poor and vulnerable people, addressing limited reach of public health systems.4 5 As a result, partnerships between public health institutions and FBOs are increasingly being viewed as an important strategy for improving healthcare access and health outcomes in low-and-middle-income countries.3 6

Despite the critical role that FBOs play in delivering healthcare in low resource settings, the capabilities and assets of some FBOs have been underused and underexplored.1 7 This underutilisation may be influenced by concerns that the religious underpinnings of FBOs contradict human rights and associated health outcomes, such as in the case of sexual and reproductive health or vaccination campaigns.1 8 Challenges also exist with the alignment of health priorities between FBOs and national health systems, inconsistent funding and governance of local FBOs, and their limited capacity to adapt to changing health systems.2 7 Additionally, there are gaps in evidence with regard to the quantity and distribution of FBOs, the quality of care provided by these organisations, and the factors that contribute to the success of programmes and interventions led by FBOs.1 5 7

Collaboration with local religious leaders and their institutions is one factor contributing to the success of health interventions implemented by FBOs. Many communities view religious leaders and institutions as a trustworthy and credible source of health advice and information, with research finding that religious leaders’ opinions can strongly influence social and behavioural norms.9–12 As embedded members of their communities, local religious leaders frequently have intimate knowledge of existing histories, networks and sociocultural dynamics influencing individual and community health and well-being, which positions them as important resources for health interventions.13 As such, local religious leaders have been identified as change agents key to promoting health awareness, disseminating health education, developing and implementing health interventions, and influencing health-seeking behaviour.9–12 14 15

Turning to the Philippines, recent impressive national economic growth has not translated into a meaningful reduction in chronic and acute child malnutrition. According to the 2015 National Nutrition Survey, between 2013 and 2015, the national prevalence of under-five underweight increased from 20% to 21.5% and under-five stunting increased from 30.3% to 33.4%.16 17 Prevalence of under-five wasting decreased slightly from 8.0% to 7.1% over this time period.18 Concurrently, an estimated 8.1% of the total population lived in extreme poverty in 2015,19 which represents a high national-level prevalence of poverty when compared with neighbouring Asia-Pacific countries.20

Achieving universal healthcare with an emphasis on health equity is a core mandate for the Department of Health in the Philippines. However, gaps remain in service provision for households that simultaneously experience extreme poverty and child malnutrition, which is especially problematic in cases of moderate acute malnutrition (MAM) and severe acute malnutrition (SAM). Where such service gaps exist, civil society organisations including FBOs, in addition to multi-lateral institutions, such as the World Food Programme and UNICEF, may step in to provide complementary care. Notably, there is increasing attention and interest in the longstanding role of FBOs in delivering healthcare to complement existing public healthcare infrastructure.1 4 21

Trust, social relationships, cooperation and reciprocity, or social capital, play a critical role in the well-being of income poor households.22 23 Social capital is theorised to have both an internal (via bonding relationships between members of a group)24 and an external (via bridging connections to external supports beyond a group) function.25 In terms of external functionality, trust in healthcare providers and institutions is an important mechanism underlying healthcare decisions and treatment adherence.26 Additionally, trust in healthcare personnel, a facility or the healthcare system more broadly is often cited as a determinant of health-seeking behaviour and connected with positive health outcomes.27–29 What is less clear is how trust in religious leaders and institutions in partnership with FBOs delivering health interventions interacts with other structural and socioeconomic barriers to influence healthcare access and use for households that experience extreme poverty. Moreover, there is limited research examining how trust in the religious leaders and institutions associated with these organisations influences subsequent retention of participants within programme interventions aimed at addressing childhood malnutrition.

To engage with these questions, we retrospectively analysed data collected by a Philippine FBO (International Care Ministries; ICM) that delivered a programme to address MAM and SAM in children living in ultrapoor households (defined as less than US$0.50 per person per day) in partnership with local religious leaders and institutions. The objective of this study was to examine how different dimensions of trust, in addition to other indicators of social capital, affected programme retention and physiological outcomes among participating children.

Methods

Intervention

ICM implemented three rounds of a treatment programme targeted at acute child malnutrition from 2012 to 2013. The programmes ran from June 2012 to September 2012, October 2012 to January 2013 and February 2013 to May 2013. A total of 1219 children from 1010 households representing 150 unique communities were treated and surveyed across the provinces of Negros Oriental, Negros Occidental, Bohol, Palawan, Sarangani, South Cotabato, Sultan Kudarat and Zamboanga del Norte in the Philippines.

The treatment programme, called Malnourished Child Outreach (MCO), was a 16-week site-based feeding programme for moderately and severely wasted children between the ages of 6 and 60 months. SAM was defined as weight-for-height z-score (WHZ) ≤−3SD from median reference values, and MAM was defined as WHZ ≤−2SD and >−3SD from reference values according to international standards.30 ICM initiated programmes when a local volunteer pastor was able to identify 10–15 malnourished children within the vicinity of his or her church. Pastors, who were associated with various Protestant denominations, consulted a list of malnourished children kept by local health centres. These pastors then conducted house-to-house visits to complete enrolment. All malnourished children were eligible for enrolment, regardless of religious affiliation. Once the enrolment target was met, ICM would provide the food, protocol and staff to complement the pastor and church volunteers for programme delivery. In a fixed location in or near the volunteer pastors’ church, ICM staff would prepare a single meal which was fed to children, assisted by their caregivers, for 5 days per week over a 16-week period. The product used for the feeding programme was a micronutrient fortified rice-based soy blend which required cooking. Other programme components included deworming, a health assessment, weight monitoring, weekly health, nutrition, health education for caregivers and home-based vegetable gardening. Children who remained SAM at the end of the 16 weeks were referred to local government clinics for additional assessment and management.

Study design

This was an opportunistic study, retrospectively designed to use household surveys that ICM administered to the caregivers of the children enrolled in the MCO programme. As a result, all households with complete treatment outcome data were included as study participants within the retrospective cohort. To understand household characteristics, caregivers of enrolled children were interviewed at baseline by trained enumerators prior to the start of the treatment. Written informed consent was obtained at two points in time: first from the guardian when children were enrolled into the programme, and again at the start of the baseline survey. Questions covered household demographics, economic well-being, general health, asset-based poverty measures and hygiene. Indicators of pre-existing social capital were also explored including group membership, trust in local religious leaders and institutions, and trust in local public healthcare facilities (online supplementary appendix 1). These survey data were linked with weekly monitoring and outcomes data. Monitoring data captured weekly weights, number of feeding sessions attended, and outcomes included treatment completion (did not drop out), and discharge weight and height measures (online supplementary appendix 2).

Patient and public involvement

We did not include patient and public involvement in the design, conduct or analysis in this study. The preliminary findings have been discussed with carers and providers, with plans to disseminate implications to the wider nutrition community in the Philippines.

Statistical approach

There were two outcomes of interest: (1) dropout (categorical variable), defined as children who were withdrawn from the programme by their caregiver or missed repeated feeding sessions and lost to follow-up and (2) WHZ at the end of the programme (continuous variable). Independent variables were at both the individual/household level and the community level (see table 1). The geographical type of community was categorised into Urban Slum, Rural Plain/Rural Slum, Rural Mountain, or Coastal/Fishing by ICM staff. We adopted and revised the measure for intensity of poverty, or A, as defined in the Alkire-Foster Method for measuring multi-dimensional poverty.31 Intensity of poverty is defined as the average proportion of indicators in which a household is deprived in, and a household is categorised as ‘experiencing poverty’ if they are deprived in at least one-third of the weighted indicators. One important feature of A is the ability to quantitatively estimate poverty at the household level, and therefore include in model building (see online supplementary appendix 3 for more detail).

Table 1

List of model outcomes and variables

Level	Item	Variable	Description
Logistic model
Outcome	Dropout	πij	Binary result if child dropped out of the programme
1 (household level)	Sex	XIij	Sex of the child
	Trust in religious leader or church	XIIIij	How much do you trust your religious leaders or church? (5-point Likert Scale)
	Trust in barangay*	XIVij	How much do you trust your local barangay? (5-point Likert Scale)
	Trust in neighbours	XVIIij	How much do you trust your neighbours? (5-point Likert Scale)
	Family satisfaction	XVIIIij	How satisfied are you with your family life? (5-point Likert Scale)
	WHZI	XVij	Baseline weight-for-height z-score of the child
	A	XVIij	Intensity of poverty
Linear model
Outcome	WHZII	Yij	Discharge weight-for-height z-score of the child
1 (household level)	Sex	XIij	Sex of the child
	Age	XIIij	Age of the child at baseline
	Trust in religious leader or church	XIIIij	How much do you trust your religious leaders or church? (5-point Likert Scale)
	Trust in barangay	XIVij	How much do you trust your local barangay? (5-point Likert Scale)
	WHZI	XVij	Baseline weight-for-height z-score of the child
	A	XVIij	Intensity of poverty
2 (community level)	Geography	Zj	Geographical type of the community

*A ‘barangay’ is the smallest administrative division in the Philippines and represents the local government.

WHZ, weight-for-height z-score.

The hierarchical structure in which this intervention was set (households in communities) required the utilisation of mixed-effects modelling to explore the potential causal relationship of varied dimensions of social capital at the household and the geographical context at the community level.

For both outcomes, a series of covariates such as sex of caregiver, age of caregiver, household size and others were explored. The most parsimonious models that minimise deviance were chosen. All analyses were conducted using R (V.3.2.3). Detailed statistical methods are described in the supplementary materials (online supplementary appendix 3).

Results

A total of 1219 treated children were included in this study, however 27 cases had incomplete treatment records and an additional 176 cases had incomplete weight data. As a result, final logistic analyses included 1192 children, while the linear analyses included 1016 children. Significant differences were not detected between the logistic and linear cases at baseline (tables 2 and 3). The average age of children was 33.13 months and 32.96 months in the logistic and linear models, respectively. The sex of the children was balanced, while measures of trust were found to be slightly higher for religious leaders or churches than for local government (barangay) and neighbours.

The final mixed-effects model on dropout included sex, three measures of social capital, intensity of poverty (A) and a series of random effects (table 4). In the most parsimonious model, which includes predictors and interaction terms, the caregiver’s trust in religious leaders or church was negatively associated with dropout (−0.87 (95% CI: −1.43, 0.26)), suggesting that each increased level of satisfaction or trust was associated with a decreased proportion of dropouts from the treatment programme. The caregiver’s trust in the local barangay was associated with dropout in the reverse direction (0.81 (95% CI: 0.22, 1.40)), interpreted as those with higher levels of trust in local government dropping out more. These estimates reveal that trust in specific entities can be correlated with likelihood of dropout in opposing directions, depending on whom or where the trust is directed towards. Households reporting a higher intensity of poverty were also significantly linked with a lower rate of dropout (−4.21 (95% CI: −7.76, 0.66)).

Table 4

Logistic mixed-effects model on dropout from the MCO programme

	Model 1 (intercept only)				Model 2 (with predictors)				Model 3 (with predictors and interaction)
	Coefficient	SE	95% CI	Sig*	Coefficient	SE	95% CI	Sig*	Coefficient	SE	95% CI	Sig*
Fixed items
Intercept	−2.04	0.15	−2.33, 1.75	***	−0.65	1.01	−2.63, 1.33		−1.01	1.05	−3.17, 1.05
Sex (male)					−0.28	0.22	−0.71, 0.15		−0.26	0.22	−0.69, 0.17
A†					−1.34	0.88	−3.06, 0.38		−4.21	1.81	−7.76, 0.66	*
Family satisfaction					−0.35	0.22	−0.78, 0.08		−0.38	0.23	−0.83, 0.07
Trust in religious leader or church					−0.85	0.31	−1.46, 0.24	**	−0.87	0.31	−1.48, 0.26	**
Trust in local barangay					0.73	0.29	0.16. 1.30	*	0.81	0.30	0.22, 1.40	*
Geographical type
Urban slum (reference)					–	–	–		–	–	–
Rural plain/rural slum					0.05	0.43	−0.79, 0.89		0.20	0.46	−0.70, 1.10
Rural mountain					0.12	0.48	−0.82, 1.06		0.41	0.51	−0.59, 1.41
Coastal/fishing					0.009	0.56	−1.09, 1.11		0.30	0.57	−0.82, 1.42
Interactions
A × Urban slum (reference)									–	–	–
A × Rural plain/rural slum									2.11	2.11	−2.03, 6.25
A × Rural mountain									4.05	2.25	−0.36, 8.46)
A × Coastal/fishing									6.45	2.45	1.65, 11.25	**
Random items
σ0	1.12	1.06	−0.96 to 3.20
σ034578					27.21	5.22	16.98, 37.44		28.86	5.37	18.33, 39.39
σ3					1.07	1.04	−0.97, 3.11		1.31	1.14	−0.92, 3.54
σ4					19.56	4.42	10.90, 28.22		12.88	3.59	5.84, 19.92
σ5					0.62	0.79	−0.93, 2.17		0.68	0.82	−0.93, 2.29
σ7					1.40	1.18	−0.91, 3.17		1.62	1.27	−0.87, 4.11
σ8					1.07	1.03	−0.95, 3.09		0.98	0.99	−0.96, 2.92
Deviance	960.1				893.1				884.8

*Statistical significance: when *p<0.05, **p<0.01, ***p<0.001.

†Intensity of poverty.

MCO, Malnourished Child Outreach.

The estimates in table 5 describe the results of mixed-effects modelling on WHZ2 as an outcome. The intercept of Model 1 (intercept only) estimated at −0.38 is the unadjusted decrease in average WHZ at discharge for children treated across all communities. The direction of this coefficient suggests that on average, children that completed the programme experienced movement toward normal WHZ. The intraclass correlation coefficient calculated for the model was 0.27, representing that 27% of variance in WHZ2 is attributed to the community level covariate of geographical type. Age was negatively correlated to WHZ2, indicating that older children experienced diminished growth compared with younger children. WHZ1 was positively correlated with WHZ2, which can be interpreted as children who were closer to normal weight at the beginning of the programme achieved a higher WHZ2 by the end of the programme. Intensity of poverty was also found to be significant. The coefficient of −0.47 in Model 3 is not directly interpretable, but the direction shows that greater intensity of household poverty was linked to lower WHZ2. Neither measure of caregiver trust (in religious leaders and church, or local government) was found to be significantly correlated with WHZ2. Additional modelling using centred coefficients to increase parsimony were conducted but are not reported as they were not found to improve the model.

Table 5

Linear mixed-effects model on weight-for-height z-score (WHZ) at discharge

	Model 1 intercept only)				Model 2 (with predictors)				Model 3 (with predictors & interaction)
	Coefficient	SE	95% CI	Sig*	Coefficient	SE	95% CI	Sig*	Coefficient	SE	95% CI	Sig*
Fixed items
Intercept	−0.38	0.06	−0.50, 0.26	***	0.71	0.30	0.12, 1.30	**	1.02	0.45	0.14, 1.90	**
Sex (male)					0.03	0.06	−0.09, 0.15		0.01	0.06	−0.11, 0.13
Age (months)					−0.48	0.08	−0.64, 0.32	***	−0.47	0.08	−0.63, 0.31	***
WHZ1					0.45	0.05	0.35, 0.55	***	0.45	0.05	0.35, 0.55	***
A†					−0.47	0.16	−0.78, 0.16	**	−0.48	0.16	−0.79, 0.17	**
Trust in religious leader or church					0.09	0.06	−0.03, 0.21		0.02	0.09	−0.16, 0.20
Trust in local barangay					−0.006	0.05	−0.10, 0.09		−0.007	0.05	−0.11, 0.09
Geographical type
Urban slum (reference)					–	–			–	–
Rural plain/rural slum					0.32	0.14	0.05, 0.59	*	0.14	0.60	−1.04, 1.32
Rural mountain					0.21	0.16	−0.10, 0.52		−0.04	0.54	−1.10, 1.02
Coastal/fishing					0.82	0.20	0.43, 1.21	***	−1.21	0.71	−2.60, 0.18
Interactions
Trust in religious leader or church × Urban slum (reference)
Trust in religious leader or church × Rural plain/rural slum									0.04	0.13	−0.21, 0.29
Trust in religious leader or church × Rural mountain									0.05	0.12	−0.19, 0.29
Trust in religious leader or church × Coastal/fishing									0.55	0.18	0.20, 0.90	**
Random items
σe	1.05	1.02	−0.95, 3.05		0.71	0.84	−0.94, 2.36		0.71	0.84	−0.94, 2.36
συ0	0.38	0.62	−0.84, 1.60
συ013					0.89	0.94	−0.95, 2.73		0.83	0.91	−0.95, 2.61
συ1					0.11	0.33	−0.54, 0.76		0.12	0.34	−0.55, 0.79
συ3					0.09	0.31	−0.52, 0.70		0.10	0.31	−0.51, 0.71
συ06					0.91	0.95	−0.95, 2.77		0.64	0.80	−0.93, 2.21
συ6					0.02	0.13	−0.23, 0.27		0.009	0.10	−0.19, 0.21
Deviance	3100.2				2808.8				2800.4

*Statistical significance: when *p<0.05, **p<0.01, ***p<0.001.

†Intensity of poverty.

Discussion

While social networks are critical for the poorest households to access and navigate health and social services, the multi-dimensional vulnerabilities and exclusion that these households often experience make them the least able to effectively leverage relationships for household benefit.32 33 This reality is evident in the Philippines, with a previous study demonstrating that poor households in a slum area with few social ties had less access to key municipal services, such as water.34 We suggest that when households experiencing extreme poverty trust local religious leaders and institutions (ie, local pastors and churches), they are more likely to remain enrolled in services provided through these networks. Thus, an opportunity exists for these actors to improve the delivery of health and social services. In addition to trust, we recognise that need also drives programme retention as households with the greatest intensity of poverty were the least likely to drop out of the programme offered by ICM.

There is a lack of consensus regarding the effectiveness of FBOs to deliver primary healthcare across low-resource settings among populations experiencing extreme poverty.35–38 Similarly there is a need for more information to understand if and under what conditions these organisations contribute to positive health outcomes.5 However, the trusting relationships that FBOs as well as religious leaders and institutions often hold in the communities in which they are embedded in and operate have been highlighted as an important feature and possible mechanism to ensure effective and meaningful service delivery.9 13 15 The logistic mixed effects model showed a significant negative association (−0.87 (95% CI: -1.43, 0.26)) between dropout and caregiver trust in religious leaders and institutions, confirming that this type of trust was a determinant of retention among participants attending the MCO programme administered by ICM in partnership with local religious leaders. This model of service delivery provides an example of a potential strategy FBOs in the Philippines, and elsewhere, can use to contribute to addressing acute child malnutrition— namely, leveraging trust in religious leaders and institutions prior to the intervention among households experiencing extreme poverty to promote health programme retention among vulnerable populations.

High levels of structural social capital (ie, group membership and the presence of social support) among caregivers have been hypothesised to contribute positively to the nutritional status of their children. Structural social capital is thought to lead to access to food resources, improved living conditions, access to knowledge networks and access to health services, which in turn, may create conditions of increased food security, reduced childhood illness and an increased ability to care for children.39 In our study, initial weight-for-height, age at baseline and the intensity of household poverty were associated with physiological outcomes among acutely malnourished children following treatment. Additionally, children who completed the treatment programme experienced movement toward a normal WHZ. However, various structural dimensions of social capital among caregivers were not directly associated with improved physiological outcomes in acutely malnourished children. This finding pushes us to more closely examine the relationship between participant retention in malnutrition interventions and structural social capital as the mechanism through which structural social capital influences child nutrition outcomes.

This study faced several limitations. First, the findings represent the outcomes of one programme implemented by a specific Christian FBO in the Philippines and its partnership with Protestant religious leaders and churches, which might not be readily generalisable to other settings. Second, the data collected were limited to enumerators directly asking caregivers of children in income poor communities to respond to questions about their trust in individuals and groups connected to the organisation providing them with services or resources. Third, the models presented were restricted to exploring the covariates included in the baseline survey. Fourth, there was the lack of distinction between interpersonal trust in religious leaders and trust in religious institutions in the survey. As a result, we were unable to distinguish between these types of trust within this study. Finally, the sustainability of these results over time could be questioned as the data were collected up to and including May 2013. To address this limitation, future replication studies are planned.

Given the high burden of child malnutrition in the Philippines, there is a need for effective strategies to deliver care for MAM and SAM. Experiences of social exclusion (eg, limited trust in public institutions) have been found to influence health-seeking behaviour and contribute to gaps in healthcare provision.40 We suggest that when individuals feel socially excluded from public health services or institutions, providing health services through collaborations between FBOs and trusted religious leaders and institutions can act as a critical alternative. Community-based care offered by FBOs in partnership with local religious leaders and institutions presents an opportunity to engage with and support income poor households with weak social networks. Based on this finding, we suggest that the delivery of healthcare through FBOs that build on pre-existing trusting relationships with local religious leaders and institutions, should be further explored and evaluated.

Table 2

Baseline values of continuous independent variables

Variable	Logistic model(n=1192)		Linear model(n=1016)
	Mean	SD	Mean	SD
Age (months)	33.13	15.56	32.96	15.55
WHZ1	−2.42	1.00	−2.42	1.01
A	0.36	0.22	0.37	0.21
Trust in religious leader or church	4.20	0.83	4.21	0.83
Trust in local barangay	3.96	0.93	3.94	0.94
Trust in neighbours	3.91	0.92	–	–
Family satisfaction	3.77	1.02	–	–

WHZ, weight-for-height z-score.

Table 3

Baseline values of categorical independent variables

Variables	n	%
Logistic model
sex (male)	585	49
sex (female)	607	51
Linear model
sex (male)	504	50
sex (female)	512	50
Geographical types
Urban slum	225	19
Rural plain/rural slum	305	26
Rural mountain	487	41
Coastal/fishing	175	15