Migration and investments in the health of children left behind: the role of remittances in children’s healthcare utilization in Cambodia

Abstract

Remittances, financial support from family members who have migrated for work, are an increasingly important source of income for households left behind in many lower- and middle-income countries. While remittances have been shown to affect the health status of children left behind, evidence is very limited as to whether and how they affect children’s healthcare utilization. Yet, this is an important consideration for policymakers seeking to improve equitable access to quality care in settings where migration is common. I examine whether children under age five whose household receives remittances are more likely to utilize higher quality healthcare providers than those without remittances in Cambodia, a country with high rates of migration and a pluralistic health system. The analysis includes 2230 children reporting recent illness in three waves of the Cambodia Socio-Economic Survey with data on migration, remittances and children’s health expenditures. I use mixed-effects and fixed-effects regression analysis to estimate the effect of remittances on children’s likelihood of entering care with a formally trained provider, and among those attending a formally trained provider, likelihood of using a public-sector facility. Treatment expenditures are lower among households with remittances, while transportation expenditures do not vary significantly by remittance status. In mixed-effects and fixed-effect regression models, children who receive remittances have a lower likelihood of utilizing qualified providers (adjusted OR = 0.66, 95% confidence interval 0.44–0.98), though this effect is attenuated in fixed-effects models, and there is no association between remittances and attending a public-sector facility. These findings underscore that remittances alone are not sufficient to increase children’s utilization of qualified providers in migrant-sending areas, and suggest that policymakers should to address barriers to care beyond cost to promote utilization and equity of access to higher quality care where remittances are a common source of income.

Key Messages:

• Migration is an increasingly common livelihood strategy in many lower- and middle-income countries, yet as more households rely on remittances from migrant family members, it is unknown how this income affects children’s utilization of curative healthcare.

• In Cambodia, households that receive remittances spent less than households without remittances on children’s treatment, and had no difference in expenditures for transportation to care.

• Children who receive remittances are not more likely to utilize formally trained providers when ill.

• Remittances alone are not sufficient to increase utilization of qualified providers; policymakers should address other barriers to care in high-migration settings to improve child health equity.

Introduction

Targeted interventions, health system improvements and socio-economic development have contributed to significant improvements in child health since 2000 in a majority of low- and middle-income countries (LMICs) (Black et al., 2010). Yet, social and structural barriers persist that prevent many children from accessing necessary medical care. Migration is both a response to this disadvantage, and a potential strategy for parents and families to improve children’s health and opportunities in many LMICs, such as Cambodia.

With greater labour migration across the Global South, an increasing proportion of children are left behind as their parents or other family members migrate (Parreñas, 2005; Nobles, 2013). In Cambodia, recent economic growth driven by the garment and construction sectors has contributed to high-migration rates among young adults, many of whom provide remittances, or financial support, to their children and family members left behind (Ministry of Planning, 2012). Globally, migrants remitted $441 billion to LMICs in 2016 (Global Knowledge Partnership on Migration and Development, 2016). Remittances reduce the proportion of households living in extreme poverty and are associated with increased health expenditures (Adams and Page, 2005). This additional income may favourably impact children’s health, especially if remittances facilitate access to higher quality healthcare by allowing families to spend more on treatment and/or transportation to care when children fall ill. Thus, the rise of remittances as a source of income in LMICs could shift children’s healthcare utilization towards higher quality, formally trained providers in the public and private sectors by improving access to more distant facilities and reducing reliance on informal providers. These potential shifts in utilization in settings with high rates of migration could affect health systems’ ability to deliver quality care, also affecting equity of access to care for young children.

In the pluralistic health systems found in many LMICs, including Cambodia, parents and caregivers may access a range of sources of care for a sick child, including public-sector primary health centres, district and tertiary hospitals; private clinics and hospitals; pharmacies, informal and ambulatory drug sellers who lack formal health training; and traditional or religious healers, among others. In Cambodia, public-sector facilities and private clinics and hospitals are generally staffed by formally trained providers (most often doctors or nurses), although there is less regulation of private-sector facilities (Ovesen and Trankell, 2010; Meessen et al., 2011). On the other hand, pharmacies in low- and middle-income Asian countries suffer from a range of issues that engender a persistently low quality of care (Miller and Goodman, 2016). These include poor referral linkages; provision of clinically inappropriate drugs or dosages, including incomplete courses of antibiotics; and limited counselling. Even where pharmacies are owned or staffed by a trained pharmacist, they are often not present during open hours. Quality of care is even lower among informal drug sellers and other providers who lack formal health education (Ahmed et al., 2009; Beyeler et al., 2015).

Globally, children are less likely to experience positive treatment outcomes if they utilize untrained, unqualified providers or are otherwise delayed in accessing quality care (Awor et al., 2012; Mohanan et al., 2015). When a child falls ill, parents and caregivers must make multiple decisions throughout the illness episode, from whether to seek care to the choice of provider and amount of health expenditure (Pokhrel and Sauerborn, 2004). Colvin et al.’s child healthcare-seeking framework describe several steps in the cascade of care: first, caregivers must recognize and respond to illness, then seek advice and negotiate access to care outside the home, eventually attending ‘middle layer’, community-based informal providers before accessing formal biomedical services (Colvin et al., 2013). Families consider factors such as cost, distance, availability, prior experience and provider reputation and perceived quality when making these decisions, weighing this information against available resources such as time and ability to pay (Jacobsen et al., 2012; Leonard, 2014; Scott et al., 2014). In many cases, informal or unqualified providers remain attractive to families due to their proximity, convenient open hours and greater propensity to accept flexible payment mechanisms (Sudhinaraset et al., 2013). Among those who use formal biomedical providers, the more personal nature of relationships with private-sector providers leads to increased trust in these providers; more flexible payment mechanisms compared to the public may drive use their use despite the generally increased cost over public-sector facilities (Ozawa and Walker, 2011). Despite increased costs, private sector use remains high, even in areas where subsidy programs are in place (Jacobs et al., 2018).

In Cambodia, high user fees may discourage the poor from utilizing public-sector services, though this has been mitigated to some extent by the introduction of Health Equity Funds (Bigdeli and Annear, 2009; Fernandes Antunes et al., 2018). However, even moderate health expenditures for a sudden, emergent illness can lead to catastrophic debts, particularly among the poorest (Van Damme et al., 2004; Fernandes Antunes et al., 2018). Thus, cost or inability to access credit for services or transportation can significantly influence where children seek care, in turn affecting the quality of care received and subsequent health outcomes (Jacobsen et al., 2012; Colvin et al., 2013). Where migrants provide financial support to their household of origin, this additional remittance income may mitigate cost-related barriers to care for children left behind, allowing families to access providers formerly out-of-reach. Returning to Colvin et al.’s framework (Colvin et al., 2013), remittance income may translate to a greater access to more expensive and/or more distant health services, allowing children in these households to progress to formal biomedical care more quickly and to access more.

Previous studies suggest remittances differentially impact various aspects of children’s health depending on the ways in which they are invested. For example, in Mexico, households that receive remittances have lower rates of infant mortality (Hildebrandt et al., 2005). However, remittances do not result in lower prevalence of pneumonia or diarrhoea among children in Ecuador (Ponce et al., 2011), and findings regarding the effects of remittances on children’s nutritional status is conflicting (Antón, 2010; Carletto et al., 2011; Davis and Brazil, 2016; Nguyen, 2016). A study in Ecuador indicates remittances may increase children’s access to preventive services such as vaccination and deworming (Ponce et al., 2011). In Vietnam, remittances are associated with a greater number of outpatient visits per year among children in migrant-sending households (Nguyen and Nguyen, 2015). However, these studies use dichotomous measures of access to care, and do not examine where children utilize health services, which has important implications for quality and outcomes.

Using three waves of the Cambodia Socio-Economic Survey (CSES), I examine whether households receiving remittances invest this income in curative healthcare for acutely ill children, and if so, how. Cambodia has high rates of out-migration, yet many of its public policies fail to consider or address the impacts of migration or remittances (OECD, Cambodia Development Resource Institute, 2017). To assess how remittances affect children’s healthcare utilization during acute illnesses, I examine whether remittances are associated with a higher likelihood of entering care with a formally trained provider or at a public-sector facility among recently ill children, and how care-seeking behaviours and health expenditures for young children vary across households that do and do not receive remittances to identify possible mechanisms of investment. Because remittances generally represent additional household income, I hypothesize that remittances result in (1) a higher likelihood of first attending care with a formally trained provider vs community-based informal providers as cost-related barriers to care are diminished; (2) a higher likelihood of accessing more distant public-sector facilities among those who access formal biomedical as cost-related barriers to transportation are diminished. Understanding the role of remittances in children’s health utilization sheds light on familial decision-making strategies for child health, which hold important implications for the design of effective policies and interventions to improve child health outcomes and equity of access to quality care in migrant-sending areas.

Methods

Data and sample

This study uses data from the Cambodian Socio-Economic Survey (CSES), a nationally representative, repeated cross-sectional survey. CSES collects information on individuals’ and households’ socio-demographic characteristics, labour, income, consumption and health. In each sampled village, a village leader provides information on village characteristics, such as health and education infrastructure, employment, agricultural production, and wages and retail prices. I use CSES data from 2009, 2010 and 2011, which uniquely include information about migration, remittances and health expenditures. Data from multiple survey waves are pooled to increase power.

CSES uses a multistage random sampling technique. In the first stage, villages, the primary sampling unit (PSU), are randomly selected within each province. In the second stage, an enumeration area (EA) within each village is randomly selected, and in a third stage, households within each EA are randomly sampled. In each wave, the response rate was over 99% (Ministry of Planning, 2017a,b,c). Sampling weights are calculated for each wave.

This analysis is restricted to children under age five who report recent illness in any wave (N = 2280, 29.2% of all children under five). There is no significant difference in illness prevalence among children whose households do and do not receive remittances (N = 8733, P = 0.69), nor in underlying nutrition status measured by stunting and wasting among the subset of children with available anthropometric data (N = 5455, P = 0.62, P = 0.82, respectively). Among children included in the analytical sample with anthropometric data, the lack of association between remittance income and stunting and wasting remains (N = 1339, P = 0.27, P = 0.86, respectively). I exclude 50 children who reside in households that sent a migrant but did not receive remittances in the preceding year because these children likely systematically differ from children in non-migrant households who do not receive remittances, due to their migration exposure. Exclusion of these children permits examination of the effect of receiving remittances from a migrant family member vs residing in a non-migrant household without remittances. The final analytical sample includes 2230 children. However, regression analyses are restricted to 2072 children who sought care with any provider, excluding 117 children who did not seek any care and 41 children for whom the location of their village is missing.

Measures

CSES includes 16 distinct categories for facility or provider type, allowing for specificity in identifying where a child received care. These include public-sector hospitals and primary health centres, private clinics and hospitals, home visits by trained private health providers, pharmacies, drug shops and informal providers such as traditional and religious healers. I use two dichotomous dependent variables in this analysis to assess type of care sought: the first is whether the child was first brought for care with a formally trained provider. This includes children who attended any public- or private-sector facility, which are generally staffed by at least one formally trained health worker (World Health Organization and Ministry of Health, 2012), or those who received a home visit from a trained provider. Alternately, children who first sought care at pharmacies, drug shops, with traditional or religious healers, or elsewhere are considered to have attended informal providers. The second is whether the child was first brought for care at a public-sector facility, which include national, provincial and district hospitals, as well as primary health centres and posts.

Heads of household report treatment and transportation expenditures for recent illnesses in Cambodian Riel, converted to US dollars (4000 Riel = $1). Treatment and transportation expenditures for single illness episodes are reported as the total amount spent on treatment and transportation to all providers attended during the illness episode. For tests of significance, these values are log-transformed.

The independent variable of interest is whether a child resides in a household that received remittances in the preceding year. If the head of household reports any remittance income from at least one migrant family member in the year preceding the survey, the child is coded as receiving remittances, vs children whose household reports no remittances.

CSES includes a continuous measure of household wealth for all households in each survey wave using an aggregate score of household consumption, standardized to Phnom Penh prices using a 2009 Riel benchmark. I determine separate wealth quintile rankings for urban vs rural areas using this aggregate consumption score. Urban and rural living standards in Cambodia are dissimilar in terms of wages and spending; therefore, separate wealth indices for urban and rural areas more accurately reflect households’ relative wealth.

Regression models control for the following covariates: child’s age (years), child’s sex (male/female), number of household members, type of place of residence (Phnom Penh/other urban/rural), whether the family has access to subsidized healthcare, education of the household member with the highest educational attainment (none; primary; junior secondary; secondary, technical, vocational or post-secondary; missing), household wealth (quintiles), the child’s living arrangements (nuclear; multigenerational; single parent; apart from parents), village distance to the district centre (kilometres) and whether there is a public health centre, private clinic, pharmacy or informal drug shop in the village.

Analytical approach

Among all children who sought care with any provider, I estimate logistic regressions to assess how the relationship between remittances and utilization of qualified and public-sector providers changes when controlling for relevant biological and socio-demographic characteristics, accounting for the hierarchical structure of the data (children nested within households within villages). In models assessing use of a formally trained provider, the sample includes all children with recent illness; in models assessing use of a public-sector facility, the sample is limited to children with recent illness who sought care with a formally trained provider in order to test determinants of public vs private-sector providers while accounting for the differences that might drive use of a higher-level, more expensive provider vs an informal provider, such as perceived quality.

To account for between-village heterogeneity, I employ multilevel mixed-effects models with random effects (Raudenbush and Bryk, 2002). Covariates are included as fixed effects, while village is entered as a random effect. Given that few households have more than one child with recent illness, I omit a household-level random effect. In a final set of models, I use village-level fixed-effects models to account for unobserved village-level characteristics, which may include relevant characteristics related to migration networks and the healthcare landscape. Analysis was conducted in Stata 14.1.

In the mixed-effects models, a first model assesses the outcome by remittance status. A second model enters survey year and child- and household-level physical and socio-demographic characteristics (child age and sex, household size, access to subsidized care, type of place of residence, village distance to district centre, access to public and private health facilities and in models related to attending a formally trained provider, pharmacies and drug shops). A third model adds household wealth and educational attainment, and children’s living arrangements, which may mediate the relationship between migration, remittances and child health. I then test unadjusted and adjusted fixed-effects models, where adjusted models include controls for child and household characteristics (child age and sex, household size, access to subsidized care, household wealth, household educational attainment and children’s living arrangements).

Sensitivity analyses include using an alternate categorization of household wealth, a log-transformation of the annual remittance amount in USD and threshold effects of minimum total remittance amounts. I test a model that includes controls for type of illness, which is available in the 2010 and 2011 waves (diarrhoea, fever, cough, other), with a control for missing type of illness (2009 wave). I test models of public-sector facility use among the sample of all children reporting recent illness. Finally, I test a model that include whether the child received vitamin A supplementation and a model that excludes 74 children reporting chronic illness to assess how results change when measures of longer-term health available for the full sample are included.

Results

The sample includes 2230 children under five reporting illness in the month preceding the survey (Table 1). The mean age of children in the sample is 1.8 years (SD = 1.4). The majority live in rural areas (N = 1733, 84.9%). Compared to the overall sample of households in CSES, many young children reside in poorer or middle-income households. Household educational attainment is low. About 10% of children (N = 263) reside in households that received subsidized healthcare in the preceding year. Availability of healthcare within villages is generally poor. For example, 229 children (8.4%) have a public primary health centre in their village, while 262 (9.6%) have a private clinic in their village. Access to shops selling drugs is greater: 388 children (15.5%) reside in a village with a pharmacy, and over a quarter have an informal drug seller in their village (N = 655, 29.5%).

Table 1.

Characteristics of children under five reporting recent illness (N = 2230)

	Received remittances		Did not receive remittances		Total		P-value
	N	%	N	%	N	%
Total	284	14.2	1946	85.8	2230	100%
Mean age in years (SD)	1.9 (1.4)		1.8 (1.4)		1.8 (1.4)		0.43
Child is female	129	43.4	941	47.9	1070	47.3	0.32
Type of place of residence							0.01
Phnom Penh	17	3.8	200	7.8	217	7.3
Other urban area	28	6.6	270	9.4	298	9.0
Rural area	239	89.6	1476	82.8	1715	83.7
Mean distance to district centre in kilometres (SD)	10.0 (9.6)		11.4 (13.1)		11.2 (12.7)		0.20
Household wealth quintile							0.62
Poorest	69	22.6	477	21.9	546	22.0
Poor	65	24.2	410	22.9	475	23.0
Middle	54	22.5	386	21.1	440	21.3
Rich	59	19.8	352	18.2	411	18.5
Richest	37	10.9	321	15.9	358	15.2
Mean size of household (SD)	6.0 (1.8)		5.0 (1.9)		5.2 (1.9)		<0.001
Child’s living arrangements							<0.001
Resides with both parents	39	15.3	1469	78.4	1508	69.4
Resides with single parent	12	2.7	53	18.3	65	2.8
Resides with grandparent(s) and one/both parent(s)	196	67.9	410	2.8	606	25.3
Does not reside with parents	37	14.1	14	0.6	51	2.5
Highest educational attainment in household							0.01
None	41	12.9	271	14.0	312	13.8
Primary	80	31.3	707	37.3	787	36.4
Junior secondary	104	37.2	538	27.9	642	29.2
Secondary, post-secondary, technical or vocational	52	17.7	340	16.5	392	16.7
Missing	7	1.0	90	4.3	97	3.9
Household received subsidized healthcare in preceding year	38	15.1	225	9.8	263	10.4	0.44
Healthcare availability in village
Primary health centre or post (public)	22	6.0	207	8.8	229	8.4	0.19
Private clinic	22	5.6	240	10.3	262	9.6	0.06
Pharmacy	39	10.3	349	16.4	388	15.5	0.16
Informal drug shop	61	19.3	594	31.22	655	29.5	0.007

Among children under five reporting recent illness, 284 (14.2%) reside in households that receive remittances. Children whose households receive remittances are significantly more likely to live in rural areas (P = 0.01) and live in households with higher educational attainment (P = 0.01). Children whose households receive remittances are significantly less likely to reside in villages with a private clinic (P = 0.06) or informal drug shop (P = 0.007). Children’s living arrangements are closely related to the receipt of remittances; those with remittances are significantly more likely to live with grandparents or apart from their parents (P < 0.001).

Less than half of children first sought care at a place staffed by a formally trained, or qualified, provider (Table 2). Overall, 447 children (20.5%) first attended care in the public sector, 647 (27.7%) in the private sector and 1019 (48.2%) with an informal provider, while 117 (3.6%) did not seek care with any provider. Children most frequently first sought care at drug shops and pharmacies, followed by private clinics and primary-level public-sector facilities. The distribution of sites of care attended by children with and without remittances varies significantly (P = 0.006). Children with remittances tend to visit shops selling drugs more frequently, and less frequently attend pharmacies or qualified providers in the public or private sectors. Whether households report any expenditure for treatment or transportation does not vary significantly by remittance status or sector of care, although households with remittances have significantly lower treatment expenditures (P = 0.013).

Table 2.

Description of care seeking and health expenditures among recently ill children under five (N = 2230)

		Received remittances (N = 284)		Did not receive remittances (N = 1946)		Total (N = 2230)		P-valuea
		N	%	N	%	N	%
Type of facility first attended								0.006
Qualified
Primary health centre or other public		35	12.0	275	14.8	310	14.4
Public hospital		14	5.9	123	6.1	137	6.1
Private clinic or other private		41	14.4	298	15.3	339	15.2
Private hospital		2	0.4	36	1.5	38	1.3
Private trained health worker (home visit)		31	8.5	239	11.6	270	11.2
Unqualified
Pharmacy		44	15.6	408	20.6	452	19.9
Shop selling drugs		103	39.2	436	25.7	539	27.6
Other informal providersb		5	1.1	23	0.7	28	0.7
No care outside home		9	3.0	108	3.7	117	3.6
Any expenditure for treatment		165	94.0	1754	91.1	2019	91.6	0.10
Mean expenditure for treatment, USD (SD)c (N = 2018)		5.62 (11.0)		5.87 (12.7)		5.83 (12.4)		0.01
Any expenditure for transport to care		85	32.5	643	34.6	728	34.3	0.37
Public (N = 447)		25	45.4	223	56.3	248	54.9	0.57
Private (N = 647)		26	41.0	223	42.5	249	42.3	0.52
Informal (N = 1019)		34	26.6	197	22.9	231	23.5	0.94
Mean expenditure for transportation, USD (SD)c (N = 726)		2.33 (5.80)		3.59 (14.8)		3.4 (13.9)		0.83

^*** P < 0.001,

^** P < 0.01,

^* P < 0.05,

^**** P < 0.10.

^aComparing remittance-receiving households to households without remittance income using unadjusted logistic or multinomial regression with standard errors clustered by PSU.

^bIncludes untrained providers such as traditional or religious healers.

^cIncluding all non-zero values of expenditures.

Remittances are negatively associated with a lower likelihood of first seeking care with a qualified provider in all unadjusted and adjusted mixed-effects models, restricted to 2072 children seeking care with any provider (Table 3). Adjusting for child, household and village-level covariates (Model 2), older age is negatively associated with attending a qualified provider [odds ratio (OR) = 0.89, 95% confidence interval (CI) 0.82–0.96]. Model 3 adds controls for household wealth, education and children’s living arrangements. Wealth is significantly and negatively associated with attending a qualified provider, controlling for remittances; the poorest children have 0.61 times the odds of attending a qualified provider as the richest (95% CI 0.43–0.86). Controls for year of survey 2011 are negatively associated with the outcome in all adjusted models.

Table 3.

Mixed-effects regression models of determinants of entering care with a qualified provider among children under five (N = 2072)

	Model 1	Model 2	Model 3
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Household received remittance	0.62* (0.48–0.94)	0.67* (0.47–0.95)	0.66* (0.44–0.98)
Child’s age (years)		0.89** (0.82–0.96)	0.89** (0.82–0.97)
Child is female		1.05 (0.84–1.31)	1.05 (0.84–1.32)
Household size		0.99 (0.93–1.05)	1.03 (0.96–1.11)
Resides in Phnom Penh (vs rural)		0.53**** (0.26–1.06)	0.49**** (0.24–1.02)
Resides in urban area (vs rural)		0.65 (0.36–1.16)	0.75 (0.41–1.38)
Household receives subsidized healthcare		1.15 (0.79–1.68)	1.26 (0.85–1.85)
Kilometres to district centre		1.01 (0.99–1.03)	1.01 (0.99–1.03)
Village has a primary health centre		1.27 (0.72–2.22)	1.23 (0.69–2.19)
Village has a private health clinic		0.92 (0.52–1.63)	0.85 (0.47–1.53)
Village has a pharmacy		0.88 (0.54–1.44)	0.83 (0.50–1.37)
Village has an informal drug shop		1.04 (0.75–1.45)	1.09 (0.77–1.53)
Survey year 2010 (vs 2009)		0.90 (0.66–1.23)	0.93 (0.68–1.28)
Survey year 2011 (vs 2009)		0.59** (0.42–0.82)	0.61** (0.43–0.86)
Poorest wealth quintile (vs richest)			0.42*** (0.27–0.66)
Poor wealth quintile (vs richest)			0.50** (0.34–0.76)
Middle wealth quintile (vs richest)			0.57** (0.38–0.85)
Rich wealth quintile (vs richest)			0.69**** (0.46–1.02)
Household educational attainment: primary (vs none)			1.15 (0.79–1.68)
Household educational attainment: junior secondary (vs none)			0.86 (0.59–1.27)
Household educational attainment: secondary or higher (vs none)			1.30 (0.84–2.02)
Household educational attainment: missing (vs none)			1.51 (0.77–2.96)
Child co-resides with one parent (vs two parents)			1.03 (0.74–1.45)
Child co-resides in multigenerational household (vs two parents)			1.09 (0.54–2.18)
Child resides apart from parents (vs two parents)			0.90 (0.39–2.08)
Village-level variance	0.89 (0.58–1.21)	0.86 (0.54–1.18)	0.94 (0.62–1.26)
Intraclass correlation (ICC)	0.42 (0.35–0.50)	0.42 (0.34–0.50)	0.44 (0.36–0.52)
Constant	1.23	1.83	2.25
Log likelihood	−1314.47	−1297.67	−1285.06

^*** P < 0.001,

^** P < 0.01,

^* P < 0.05,

^**** P < 0.10.

Among recently ill children who sought care with a formally trained provider, remittances are not significantly associated with odds of attending a public-sector facility in unadjusted or adjusted models (Table 4). In a model adjusted for child, household and village characteristics (Model 2), older children are significantly less likely to attend a public facility (OR = 0.83, 95% CI 0.74–0.94). Children whose households have subsidized healthcare have 1.69 times the odds of attending a public facility compared to children without access to subsidies (95% CI 1.01–2.83). Availability of healthcare within the village shapes children’s propensity to attend a public facility: those with a primary health centre in their village are significantly more likely to seek care in the public sector (OR = 3.00, 95% CI 1.47–6.12), while those with a private clinic in their village are marginally less likely to attend public-sector facilities (OR = 0.46, 95% CI 0.21–1.03). Adding controls for household wealth, education and children’s living arrangements in Model 3, these relationships hold with the exception of subsidized care, which is no longer significantly associated with attending a public-sector facility.

Table 4.

Mixed-effects regression models of determinants of using a public-sector facility among children under five who sought care with a formally trained provider (N = 1076)

	Model 1	Model 2	Model 3
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Household received remittance	1.07 (0.63–1.81)	0.97 (0.57–1.64)	1.00 (0.55–1.84)
Child’s age (years)		0.83** (0.74–0.94)	0.85* (0.75–0.96)
Child is female		0.75**** (0.54–1.04)	0.73**** (0.52–1.01)
Household size		1.07 (0.98–1.17)	1.06 (0.94–1.17)
Resides in Phnom Penh (vs rural)		1.11 (0.44–2.81)	1.09 (0.42–2.83)
Resides in urban area (vs rural)		1.26 (0.62–2.59)	1.06 (0.51–2.21)
Household receives subsidized healthcare		1.69* (1.01–2.83)	1.47 (0.87–2.49)
Kilometres to district centre		1.01 (0.99–1.03)	1.01 (0.99–1.03)
Village has a primary health centre		3.00** (1.47–6.12)	3.14** (1.53–6.47)
Village has a private health clinic		0.46**** (0.21–1.03)	0.46**** (0.20–1.04)
Survey year 2010 (vs 2009)		0.79 (0.51–1.23)	0.79 (0.50–1.24)
Survey year 2011 (vs 2009)		1.33 (0.80–2.21)	1.37 (0.82–2.29)
Poorest wealth quintile (vs richest)			1.53 (0.84–2.78)
Poor wealth quintile (vs richest)			1.01 (0.57–1.79)
Middle wealth quintile (vs richest)			0.70 (0.39–1.24)
Rich wealth quintile (vs richest)			0.60**** (0.34–1.06)
Household educational attainment: primary (vs none)			2.11* (1.18–3.77)
Household educational attainment: junior secondary (vs none)			1.71**** (0.93–3.12)
Household educational attainment: secondary or higher (vs none)			2.20* (1.12–4.32)
Household educational attainment: missing (vs none)			3.99** (1.55–10.30)
Child co-resides with one parent (vs two parents)			1.14 (0.70–1.86)
Child co-resides in multigenerational household (vs two parents)			0.76 (0.27–2.17)
Child resides apart from parents (vs two parents)			0.94 (0.26–3.39)
Village-level variance	0.89 (0.45–1.32)	0.74 (0.28–1.21)	0.72 (0.25–1.20)
ICC	0.42 (0.32–0.53)	0.39 (0.29–0.50)	0.39 (0.28–0.50)
Constant	0.62	0.53	0.33
Log likelihood	−673.70	−655.94	−642.75

^*** P < 0.001,

^** P < 0.01,

^* P < 0.05,

^**** P < 0.10.

In the fixed-effects models shown in Table 5, the effect of remittance income on likelihood of attending a formally trained provider is attenuated in unadjusted and adjusted models, and remittances remain unassociated with attending a public-sector facility among those attending a formally trained provider. In an adjusted model, older children have significantly lower odds of attending a formally trained provider (OR = 0.88, 95% CI 0.80–0.97), as are children in poorer households. Children in the poorest households have 0.34 times the odds of attending a formally trained provider as those in the richest households (95% CI 0.20–0.55). Turning to the adjusted model for seeking care in the public sector, educational attainment of the household head is marginally associated with public-sector care. Children in households where the head has attended primary or secondary school have more than twice the odds of seeking care in the public sector than children whose household head has no formal schooling (OR = 2.06, 95% CI 0.94–4.51; OR = 2.18, 95% CI 0.90–5.31, respectively). Missing educational status is also significantly associated with higher odds of attending public-sector care (OR = 3.20, 95% CI 0.97–10.54).

Table 5.

Fixed-effects models of determinants of care-seeking behaviours among children under five

	Formally trained provider		Public sector
	Unadjusted	Adjusted	Unadjusted	Adjusted
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Household received remittance	0.71**** (0.49–1.03)	0.70 (0.45–1.08)	0.94 (0.60–1.48)	1.01 (0.48–2.10)
Child’s age (years)		0.88** (0.80–0.97)		0.89 (0.77–1.03)
Child is female		1.05 (0.82–1.34)		0.76 (0.50–1.15)
Household size		1.05 (0.97–1.14)		1.12**** (0.98–1.27)
Household receives subsidized healthcare		1.17 (0.77–1.79)		1.21 (0.65–2.27)
Survey year 2010 (vs 2009)		1.17 (0.81–1.68)		0.55* (0.31–0.97)
Survey year 2011 (vs 2009)		0.70**** (0.47–1.05)		0.69 (0.33–1.42)
Poorest wealth quintile (vs richest)		0.34*** (0.20–0.55)		0.95 (0.44–2.03)
Poor wealth quintile (vs richest)		0.51** (0.32–0.81)		0.72 (0.36–1.45)
Middle wealth quintile (vs richest)		0.54** (0.34–0.84)		0.56 (0.27–1.16)
Rich wealth quintile (vs richest)		0.76 (0.50–1.17)		0.54 (0.27–1.09)
Household educational attainment: primary (vs none)		1.15 (0.76–1.74)		2.06**** (0.94–4.51)
Household educational attainment: junior secondary (vs none)		0.83 (0.54–1.25)		1.69 (0.75–3.79)
Household educational attainment: secondary or higher (vs none)		1.22 (0.76–1.96)		2.18**** (0.90–5.31)
Household educational attainment: missing (vs none)		1.49 (0.70–3.15)		3.20**** (0.97–10.54)
Child resides with one parent (vs two)		0.96 (0.67–1.38)		1.12 (0.61–2.04)
Child resides in multigenerational household (vs two parents)		0.91 (0.41–2.02)		1.26 (0.36–4.41)
Child resides apart from parents (vs two parents)		0.96 (0.38–2.45)		1.21 (0.26–5.70)
N	1338	1338	542	542
Number of groups (villages)	234	234	135	135
Log likelihood	−557.90	−535.05	−410.21	−193.41

^*** P < 0.001,

^** P < 0.01,

^* P < 0.05,

^**** P < 0.10.

Discussion

These findings indicate households with remittance income are less likely to utilize formal biomedical providers for childhood illness. However, they do not differ in their likelihood of accessing public-sector facilities vs private, and are no more likely to make additional investments in treatment for childhood illnesses compared to households not receiving remittances. The findings are contrary to both hypotheses. Identifying the impact of remittances on behaviours related to children’s health and illness provides insight into how remittances are—and are not—used in households supported by this increasingly common source of income, and highlights considerations for child health policies and programs in high-migration contexts. Given increasing reliance on remittance income across the Global South, whether and how families utilize such income to access particular providers is an important determinant of child health outcomes and equity in pluralistic health systems, where quality and cost can vary widely.

While previous studies show remittances improve children’s access to preventive care (Ponce et al., 2011) and increase children’s number of total annual health visits (Nguyen and Nguyen, 2015), in contrast, I find remittances are not associated with utilization of higher quality formal care when children are ill. Because the mechanisms of access and costs of preventive and curative care vary, it is plausible that remittances shape decisions about children’s preventive and curative care differently, and/or that migrant-sending families prioritize remittances for other domains of child well-being, such as nutrition or education (Adams and Cuecuecha, 2010; Antón, 2010). In Cambodia, vaccinations are offered freely in the public sector, yet curative care generally requires user fees except when subsidized by Health Equity Funds (Flores et al., 2013). When families face competing financial priorities, remittances may not provide sufficient insulation against the high costs of care for children’s illness if some portion of remittances are not set aside in advance. In Mexico, households that receive remittances tend to consume this income rather than save it (Massey and Parrado, 1994). Thus, across migrant-sending settings, households with remittances may be no more prepared than other households to handle the financial shock of a child’s illness, contributing towards a lack of difference in children’s utilization of curative healthcare among these households.

Costs associated with care are likely a particularly important driver of provider/facility choice in this setting, consistent with other studies of child healthcare utilization globally (Jacobsen et al., 2012; Colvin et al., 2013; Geldsetzer et al., 2014). Households with remittances report significantly lower expenditures on children’s treatment, which is likely related to their lower likelihood of using more expensive formal medical care. I find that the poorest households are significantly less likely to utilize trained providers for care, while subsidies are not consistently associated with care-seeking patterns. Even controlling for access to subsidized care, households with remittances are no more likely to utilize public-sector facilities, which are often more distant. Therefore, factors beyond cost may also explain the lack of association between remittances and children’s care-seeking outcomes. For example, households using remittances to finance children’s health expenditures may face difficulties accessing these funds to pay at the time of service, which may drive them to use informal or formal private-sector providers that accept deferred or in-kind payments (Ozawa and Walker, 2011; Sudhinaraset et al., 2013). Lack of funds for transportation and/or perceptions around provider quality may also drive this lack of difference (Ozawa and Walker, 2011; Flores et al., 2013). Additionally, if migration or remittances fail to affect parents’ and/or caregivers’ perceptions of quality or care-seeking preferences, children may continue to utilize care at the same sites despite an influx of income.

This analysis holds several implications for the delivery of equitable children’s healthcare in the context of out-migration. Subsidized care has been shown to increase children’s access to formal providers in other LMICs (Qian et al., 2009; Li et al., 2017). Subsidies like the Health Equity Funds may further mitigate cost-related barriers to public-sector care if extended to slightly more well-off households, who still face challenges in paying for healthcare (Van Damme et al., 2004; Khun and Manderson, 2008; Jacobs et al., 2018). Although Health Equity Funds reimburse patients for transportation costs (Flores et al., 2013), patients may not be able to afford up-front costs of transportation, or may be unaware of this benefit. Finding alternate mechanisms to address transportation-related barriers may increase children’s utilization of higher quality, distant providers, particularly in the public sector. Such strategies may be especially important for children left behind whose households are reliant upon remittances to meet daily needs, limiting their ability to use remittances to access higher quality facilities. Finally, flexible payment mechanisms may be particularly attractive to migrant-sending households. Within the public sector, where fees are required at the time of care, the implementation of flexible payment schemes may increase access for families dependent upon remittances (Khun and Manderson, 2008) In order to encourage use of public-sector facilities that offer higher quality care than community-based informal providers and result in lower out-of-pocket expenditures than private formally trained providers, the government should consistently include primary health centres in equity schemes, expand access to these schemes, and engage in community participation and promotion activities to promote trust in these facilities (Jacobs et al., 2012, 2018). Lower rates of access to subsidized care in Cambodia, especially at the time of data collection, may partially explain the contrast of the present findings to research in Mexico, where remittances are associated with higher health expenditures and where much of the population has access to health insurance (Amuedo-Dorantes and Pozo, 2011).

The findings regarding socio-demographic characteristics and care seeking are consistent with the child health literature globally. For example, similar to studies elsewhere in Asia and Africa, I find older children are less likely to be taken for care with a formally trained provider, as are children of higher socio-economic status (Nasrin et al., 2013; Geldsetzer et al., 2014). Although household wealth and educational attainment are often closely intertwined, the mixed-effects and fixed-effects models suggest education and wealth lead to different child healthcare-seeking preferences: while wealth is associated with utilization of formally trained providers over informal community-based providers, among those utilizing formal providers, education is associated with public-sector care. These results are consistent in a sensitivity analysis of public-sector facility use that includes all recently ill children. Proximity to private providers is a factor in their utilization in rural Cambodia (Van Damme et al., 2004), while public-sector facilities may require greater investments in transportation and time. Better-educated parents and caregivers may recognize and prefer additional dimensions of quality that lead them to bypass closer formally trained private providers in favour of the public sector (Akin and Hutchinson, 1999; Leonard, 2014). Children in major urban areas such as Phnom Penh enjoy access to a greater array of health services, particularly pharmacies and drug shops, which may explain their lower propensity to use formally trained providers.

This analysis has several limitations. As with all studies of migration, remittances and health, there may be unobserved factors that drive both household decisions to migrate and remit, and household responses to children’s illness; this analysis is unable to control for such factors. Because CSES is a repeated cross-sectional survey, I am unable to compare within-household healthcare utilization before and after receipt of remittances. Remittances are reported as a sum over the 12-month period preceding the survey; however, timing of remittances relative to recent child illness is not included in the survey. Given the long recall periods for recent illness and remittance income, these measures may be subject to recall bias. It is also unknown with what frequency households receive remittances, though many Cambodian households report remittances are received irregularly (Ministry of Planning, 2012; Bylander, 2014). Thus, estimates are potentially biased by measurement error regarding the amount and timing of remittance relative to health expenditures. This may result in non-differential misclassification bias, which biases estimates towards the null. Education and other socio-demographic characteristics are not reported for migrants. Thus, there is a significantly higher degree of missingness of parental and household head education data in migrant-sending households. Parental education may have a different effect on care-seeking behaviours than the educational attainment of other household members, even where parents have migrated. CSES excludes information about severity of illness, which is a determinant of care-seeking behaviours. Despite these limitations, the findings of this study contribute novel insights on the relationship between remittances and children’s healthcare utilization in an analytical approach that addresses potential sources of endogeneity.

Conclusions

While remittances may provide some important benefits for children’s nutrition and underlying health status, this analysis suggests they are neither associated with increased utilization of higher quality providers for child illness, nor increased utilization of lower-cost public-sector facilities among those accessing formal biomedical care. Thus, without further policy or programmatic intervention, remittances are unlikely to significantly reshape care-seeking patterns for childhood illness in migrant-sending areas. Given increasing rates of migration in Cambodia and other LMICs, and the concomitant increasing prevalence of children left behind, the lack of an association of remittances with utilization of higher quality and/or lower-cost healthcare is an important consideration for policymakers and other stakeholders seeking to improve child health and health equity, particularly in high-migration settings. As migration shifts children’s and families’ opportunities and livelihoods, policymakers must identify and address the specific needs of these children and families to promote their access to quality, timely, affordable healthcare.