Infant and young child feeding practices in urban Philippines and their associations with stunting, anemia, and deficiencies of iron and vitamin A

Transcription

1 Infant and young child feeding practices in urban Philippines and their associations with stunting, anemia, and deficiencies of iron and vitamin A Fabian Rohner, Bradley A. Woodruff, Grant J. Aaron, Elizabeth A. Yakes, May Antonnette O. Lebanan, Pura Rayco-Solon, and Ofelia P. Saniel Abstract Background. The prevalence of stunting, underweight, and micronutrient deficiencies are persistently high in young children in the Philippines, and among other factors, suboptimal infant and young child feeding behavior may contribute to these forms of malnutrition. Objective. To improve the understanding of contributors associated with the nutritional status of children 6 to 23 months of age living in urban areas of the Philippines. Methods. A cross-sectional survey was conducted covering five urban centers in the Philippines. Data on infant and young child feeding and nutritional status (including wasting, stunting, underweight, anemia, iron deficiency, and vitamin A deficiency) were collected for 1,784 children. Results. Among children from urban and predominantly poor and very poor households, 26% were stunted, 18% were underweight, and 5% were wasted. Forty-two percent were anemic, 28% were iron deficient, and 3% were vitamin A deficient. About half of the children were breastfed within an hour after birth, were breastfed at the time of the survey, and had been continuously breastfed up to 1 year of age. Of the factors investigated, low socioeconomic status, use of cheaper cooking fuel, and nonuse of multivitamins were all independently associated with stunting. The prevalence of anemia, iron deficiency, and vitamin A deficiency were independently associated Fabian Rohner is affiliated with GroundWork LLC, Crans-près-Céligny, Switzerland, and the Global Alliance for Improved Nutrition (GAIN), Geneva; Switzerland. Bradley A. Woodruff is affiliated with GroundWork LLC, Crans-près- Céligny, Switzerland; Grant J. Aaron is affiliated with GAIN, Geneva; Elizabeth A. Yakes is affiliated with the University of New Mexico, Albuquerque, New Mexico, USA; May Antonette O. Lebanan and Ofelia P. Saniel are affiliated with the University of the Philippines, College of Public Health, Manila; Pura Raycon-Solon is affiliated with UNICEF, Manila. Please direct queries to the corresponding author: Ofelia P. Saniel, Department of Epidemiology, College of Public Health University of the Philippines Manila, 625 Pedro Gil Street, Ermita, Manila, Philippines 1000; opsaniel@gmail.com. with the same factors and poorer sanitation facilities, lower maternal education, current unemployment, and inflammation. Conclusions. These factors merit attention in future programming, and interventions may include promotion of the timely introduction of appropriate fortified complementary foods, the use of affordable multiple micronutrient preparations, and measures to reduce infections. Key words: Micronutrient deficiencies, Philippines, stunting, young child feeding practices Introduction Elimination of extreme poverty and hunger and reduction of child mortality are among the Millennium Development Goals (MDGs) set by the United Nations to be achieved by With only 2 years left, the Philippines and many other countries are trying to further reduce mortality rates and the high prevalence of undernutrition among infants and young children. Yet, although progress was made in reducing underfive and infant mortality rates over the past decade [1] in the Philippines, the prevalence of stunting and underweight among children under 5 years of age has remained constant, with about a quarter of this age group being affected in 2008 [2]. Stunting, underweight, and wasting have repeatedly been identified as contributors to mortality, and therefore, achievement of MDG 5 will also depend on reductions in stunting, underweight, and wasting [3]. The prevalence of anemia among children under 2 years of age has remained persistently high in the face of marked reductions in anemia among older children and pregnant and lactating women in the Philippines. Anemia data from the 2008 National Nutrition Survey [2] showed that about 20% of the total population was anemic. However, among children 6 to 11 and 12 to 23 months of age, the prevalence of anemia was 56% and 41%, respectively, which represents only a very Food and Nutrition Bulletin, vol. 34, no. 2 (supplement) 2013, The United Nations University. S17

2 S18 F. Rohner et al. moderate improvement since 2003 [2]. Infants and young children also have a markedly higher prevalence of vitamin A deficiency than the general population. Vitamin A data from the 2008 National Nutrition Survey demonstrate that the national prevalence of vitamin A deficiency was 6%, whereas in children aged 6 to 59 months the prevalence was 15%; however, this prevalence is considerably lower than that in the period before 2003, when the prevalence of vitamin A deficiency in young children fluctuated around 40%. The decrease in vitamin A deficiency may be due to the enactment of mandatory fortification of edible oil [2]. Suboptimal feeding patterns, low household socioeconomic status and parental education levels, poor child health status, and inadequate sanitation have all been previously linked to child mortality and malnutrition in the Philippines [1, 4]. In order to ameliorate this, a multiple stakeholder program was instituted to build on government initiatives to improve rates of exclusive breastfeeding during the first 6 months of life and ageappropriate complementary feeding practices among children 6 to 23 months old. The premise behind this is that improved infant and young child feeding (IYCF) should lead to improved child nutrition and micronutrient status, and consequently to reductions in child mortality. The data presented in this study were collected as part of a larger study examining IYCF patterns and their influence on malnutrition, with a focus on the assessment of micronutrient status. The goal of this study is to improve the understanding of how determinants such as demographic variables, household characteristics, and infant and young child feeding practices relate to anthropometric and micronutrient status among children aged 6 to 23 months living in the Philippines. Design and methods Survey design This survey collected baseline data in five urban areas (Manila, Cebu, Zamboanga, Naga, and Iloilo cities) where strengthened efforts to improve IYCF practices were to be implemented. The survey participants in each urban area were selected by stratified two-stage cluster sampling. Within each urban area, the barangays (the smallest administrative divisions in the Philippines) served as the basis for forming primary sampling units (PSUs). Most barangays were by themselves one PSU; however, small barangays with fewer than 500 households were merged with adjacent ones to form a single PSU. PSUs were then selected systematically with probability proportional to the number of households based on 2007 census data [5]. Some large PSUs were selected more than once due to their size. In Manila City, 29 PSUs were selected, in Cebu City 27, in Naga City 10, in Iloilo City 27, and in Zamboanga City 33. The secondary sampling units were children under 2 years of age, randomly selected from updated census lists for the selected barangays. For blood sampling, only children aged 6 to 23 months were included, and in Zamboanga City, a random subsample was selected for collection of biologic specimens due to the large number of children in the PSUs. Sample size The calculation of the minimum sample size was based on the national estimates of prevalence of anemia of 55% and 41% among children 6 to 11 and 12 to 23 months of age, respectively [2]. Since these data will be used as the baseline for several IYCF interventions, a reduction of anemia prevalence by one-third in each age group category per site within 2 years was set as the target; a separate sample size was calculated for anthropometric outcomes as well (not presented here). The computations were based on an alpha error of 5% (one-tailed), a power of 80%, and a design effect of 1.5. After adjustment for a nonresponse rate of 20%, 170 children 6 to 11 months of age and 278 children 12 to 23 months of age were calculated to be required for each site, or a total of 448 children 6 to 23 months of age. Because Naga and Iloilo were treated as one stratum, this sample size was multiplied by four for a total desired sample size of 1,792 in all four strata. Ethical review and consent The Ethics Review Committee of the College of Public Health, University of the Philippines Manila, approved the survey protocol (permit number ). Inclusion in the survey was dependent on the legal caretaker of the child giving written informed consent for participation. To compensate for time lost due to study participation and travel to the phlebotomy sites, caretakers of participating children were given a small financial and food contribution. Severely anemic children were referred to local health centers for appropriate diagnosis and treatment; moderately or mildly anemic children were provided with 30 micronutrient powder sachets. Data collection After the children were selected and consent was obtained, all eligible subjects were registered according to household, sex, age, date, and identification number. The caregivers of the participants answered questions on demographic factors, health (recent morbidity and

3 Infant and young child feeding practices in urban Philippines treatment), socioeconomic status (household income, housing quality, access to water, electricity, and transport), and child feeding practices and behaviors. In addition, anthropometric data and biochemical samples were collected. Infant and young child feeding (IYCF) World Health Organization (WHO)-recommended infant and young child feeding (IYCF) indicators were used to assess child feeding practices [6]. These indicators cover a range of practices, and provide specific criteria for assessing the adequacy of those practices at the population level, including age-appropriate breastfeeding practices (timing, duration, and exclusivity), and timely and adequate introduction of high-quality complementary foods. Several of the eight core IYCF indicators as well as optional indicators were calculated based on these WHO guidelines:» Initiation of breastfeeding within 1 hour after birth indicates whether or not the newborn was breastfed within 1 hour of delivery.» Currently breastfeeding was measured by asking the mother or caretaker if the child had consumed breastmilk (by suckling, cup, or bottle feeding) during the day and night prior to the interview.» Continued breastfeeding at 1 year applies this same definition of current breastfeeding to the subgroup of children 12 to 14.9 months of age.» Introduction of complementary foods was measured by asking respondents how many times during the prior day and night a child 6 to 8.9 months of age had consumed solid, semisolid, or soft foods.» Minimum dietary diversity is met if the child had consumed foods from at least four major food groups, which included grains and roots, legumes and nuts, dairy products, meat, eggs, vitamin A rich fruits and vegetables, and other fruits and vegetables, during the prior day and night.» The definition of minimum meal frequency varies by age and breastfeeding status; for breastfed children, the frequency of eating solid, semisolid, and soft foods is counted, and for such children 6 to 8.9 months of age, the minimum frequency is two times in a 24-hour period. For breastfed children 9 to 23 months of age, the minimum frequency is three times. For nonbreastfed children, the frequency of milk feeds plus eating solid, semisolid, and soft foods is counted, and the minimum frequency for all such children 9 to 23 months of age is four times in a 24-hour period.» The minimum acceptable diet meets standards for both minimum dietary diversity and minimum meal frequency. For breastfed children, both standards must be met for their diet to be considered minimally acceptable. For nonbreastfed children, the minimum meal frequency used to define minimum dietary acceptability no longer includes milk feeds. S19 Such children must have at least two milk feeds (nonbreastfed) per day along with minimum dietary diversity and minimum meal frequency of foods other than milk. Animal-source foods include meat, milk, and eggs. It must be noted that the WHO indicator for consumption of iron-rich or iron-fortified foods was not assessed in this study. Water and sanitation quality Several household-related characteristics were clustered together to create three composite indicators of household water and sanitation quality [1]. Improved drinking water source is defined as piped water or water from a borehole, protected or semiprotected dug well, or protected spring or rainwater collection. Safe drinking water has been treated by methods to make it safer to drink: boiling, bleaching or chlorine treatment, water filter, solar disinfection, improvised filter, or letting water stand and settle. Adequate sanitation facilities refers to a toilet that flushes or empties into a piped sewer or septic tank, a closed pit latrine, an improved pit latrine, or a pit latrine with slab. Anthropometry Anthropometric measurements were taken at the household by experienced and trained anthropometrists using standard techniques [7] and equipment (weighing scale, Seca 334; length measuring board, Seca 417, Hamburg, Germany). The length-for-age (LAZ), weight-for-age (WAZ), and weight-for-length (WLZ) z-scores for each child were determined using the WHO 2006 Child Growth Standards [8]. A child with LAZ < 2.0 is classified as stunted, one with WAZ < 2.0 as underweight, and one with WLZ < 2.0 as wasted. Severe stunting, wasting, and underweight are defined as z-scores < 3.0 for LAZ, WLZ, and WAZ, respectively. Blood sampling and analysis The mothers or caregivers were asked to bring their children to a designated health center for blood collection. Trained phlebotomists collected capillary blood at designated phlebotomy sites by pricking the heel of children 6 to 11 months of age and the index or middle finger of children 12 to 23 months of age. For hemoglobin, one drop of capillary blood was collected and measured with the Hemocue device (Hb 201+, Ängelholm, Sweden). An additional 300 µl of blood was collected into a lithium-heparin-coated microtainer (Microvette CB300, Sarstedt, Switzerland), always maintaining aseptic precautions to protect the infant from any potential infection due to the prick. Whole blood was stored on ice for later blood processing in the field, where whole blood was centrifuged, and µl of the supernatant plasma was aliquoted for

4 S20 F. Rohner et al. later analysis and as a backup. Once aliquoted, plasma samples were stored at 20 C until further shipment on dry ice for analysis. Plasma samples for retinol-binding protein (RBP), plasma ferritin (PF), C-reactive protein (CRP), and alpha 1 -acid-glucoprotein (AGP) were analyzed in one run from < 100 μl plasma using an in-house sandwich ELISA (VitA-Iron-Laboratories, Willstaett, Germany) [9]. The laboratory is a participant in the US Centers for Disease Control and Prevention (CDC) VITAL- EQA interlaboratory comparison rounds and has a rigorous internal quality control system. In children 6 to 23 months of age, anemia was defined as a hemoglobin concentration < 110 g/l; mild anemia was defined as hemoglobin concentration of g/l), moderate anemia as a hemoglobin concentration of 70 to 99 g/l, and severe anemia as a hemoglobin concentration < 70 g/l. The following anemia prevalence cutoffs were used to indicate population levels of severity: 40% indicates a severe public health problem, 20% to 39.9% a moderate public health problem, 5% to 19.9% a mild public health problem, and < 4.9% is considered normal [10]. The acute phase proteins CRP and AGP were used to classify inflammation into four categories: normal or no inflammation, defined as no elevated acute phase proteins; incubation period, defined as a CRP concentration > 5 mg/l; early convalescence, defined as a CRP concentration > 5 mg/l and an AGP concentration > 1 g/l; and late convalescence, defined as an AGP concentration > 1 g/l. These four stages of inflammation were then used to adjust ferritin [11] and RBP (assumed equivalent to retinol) according to adjustment factors proposed in the literature [12]. Iron deficiency was defined as a ferritin concentration < 12 µg/l, after adjustment for the effect of subclinical inflammation using both CRP and AGP [11]. Iron-deficiency anemia was defined as the presence of both anemia and iron deficiency. Similarly, for vitamin A deficiency, the threshold for RBP was 0.7 µmol/l after adjustment for inflammation using both CRP and AGP [12]. Socioeconomic classification To calculate socioeconomic status, an approach developed within the Philippines was used [13] whereby households are grouped into six classes: AB (upper class), C1 (upper middle class), broad C (broad middle class), C2 (lower middle class), D (lower class), and E (extremely low class). A composite score derived by summing individual scores for characteristics of the neighborhood, home durability, outdoor and indoor home quality, household assets, and monthly household income and electricity bill was calculated to classify households. Data management and analysis Data entry for the quantitative survey data was completed using EpiInfo software, version 6. SPSS software, version 20, was used for data analysis. Design-based analysis, appropriate for the complex sampling design, was employed where sampling weights were used in the estimation of various indicators. For weightings, absolute population figures for each of the five survey sites were used to assign a statistical weight for each household, mother, and child. For continuous data, means with associated 95% confidence intervals and medians were computed. For dichotomous data, frequencies and proportions were generated with their 95% confidence intervals. Associations between variables were measured by creating tables correlating a potential risk factor with one of the nutrition outcomes; continuous variables were categorized as necessary. The statistical significance of differences in proportions among subgroups was determined using adjusted chi-squares. These analyses did not take into account potential confounding or effect modification. In order to take into account these effects, multiple logistic regression models were constructed separately for each outcome, which included all potential contributory or confounding variables for which the p value was less than.20 in bivariate analysis. Second-order interaction terms were also entered for each of the factors. Backward elimination was then carried out to remove those interaction terms and variables not significantly contributing to the model. Results Demographic information A total of 1,777 households were surveyed, and all eligible children in the household were enrolled, resulting in interview data from 1,784 children 6 to 23 months of age and 1,711 mothers. For 66 children, the data were provided by someone other than the child s mother. For biological specimen data, the sample size varied slightly, depending on the indicator. The characteristics of the participating households are described in table 1. The vast majority of the participating households were from the low and very low socioeconomic classes, with a monthly income of approximately 14,433 PHP (US$340). Households are generally large (6.5 members on average), with all family members sleeping in one room. Of the households interviewed, almost half used water from an improved water source, and only few households (about 7%) with unimproved water sources treated water to render it safer to drink, resulting in about half of households consuming safe drinking water. Most households had adequate sanitation facilities.

5 Infant and young child feeding practices in urban Philippines S21 TABLE 1. Characteristics of households in which selected children lived Characteristic No. of households Weighted % or mean 95% CI No. of household members (mean) 1, , 6.7 Monthly household income (Philippine pesos [PHP]) (mean) 1,777 14,433 11,005, 17,862 Household income score (%) 1 ( 20,000 PHP) 1, % 80.9%, 87.1% 2 (20,001 40,000 PHP) % 9.4%, 14.3% 3 (40,001 60,000 PHP) % 2.3%, 4.8% 4 (> 60,000 PHP) 8 0.8% 0.3%, 2.4% Socioeconomic classification (%) a C % 0.0%, 0.5% Broad C % 0.5%, 1.7% C % 11.4%, 17.5% D 1, % 61.7%, 72.6% E % 12.1%, 24.4% Cooking fuel (%) Electricity or natural gas % 46.7%, 57.4% Kerosene, coal, wood, etc. 1, % 42.6%, 53.3% No. of persons per sleeping room (mean) 1, , 4.8 Improved source of drinking water (%) b No % 50.8%, 59.8% Yes 1, % 40.2%, 49.2% Safe drinking water (%) c No % 43.7%, 52.7% Yes 1, % 47.3%, 56.3% Adequate sanitation facilities (%) b No % 8.3%, 18.1% Yes 1, % 81.9%, 91.7% a. Socioeconomic classification categories collapsed into three groups for subsequent analyses: groups C1, broad C, and C2 combined into group C. b. See text for definitions of improved and adequate. c. Either water was obtained from a safe source or adequate water treatment was done in the home. Table 2 provides an overview of the characteristics of the responding mothers. The mean age of the mothers was 27.6 years. More than half of responding mothers had partially or totally completed high school, with an additional third having attended a university. Despite the high level of education, 68% of the interviewed mothers reported being unemployed. Child characteristics The mean age of the selected children was 15 months, and the sex ratio was close to 1:1 (table 3). More than 90% of children were born by vaginal delivery, and 23.9% had three or more siblings. About a third of the children in the study were reported to have suffered from a fever in the 2 weeks preceding the interview (self-report from caregiver interview). A smaller proportion was reported to have suffered from cough and fast breathing or diarrhea in the 2 weeks prior to the survey (table 3). Table 3 presents the detailed results for breastfeeding and complementary feeding patterns. In brief, early initiation of breastfeeding and current breastfeeding were reported for about half of children 6 to 23 months of age. About half of children 12 to 14 months of age were still being breastfed. Over 90% of children 6 to 8 months of age had been introduced to complementary foods. Among children 6 to 23 months of age, the vast majority met the criterion for adequacy of meal frequency; however, only slightly more than half met the criteria for minimum dietary diversity and diet acceptability. Interestingly, almost three-quarters of mothers reported giving multivitamins to their children. Among children 12 months of age or older, about two-thirds had received a vitamin A supplement in the 6 months prior to the survey. Analysis of the relationship between socioeconomic status and young child feeding behaviors showed that

6 S22 F. Rohner et al. TABLE 2. Characteristics of mothers of selected children Characteristic No. of women Weighted % or mean 95% CI Age (yr) (mean) 1, , 28.1 Age group (yr) (%) < % 7.1%, 10.8% % 52.9%, 59.6% % 26.5%, 33.9% % 3.7%, 6.5% Marital status (%) Never married % 7.3%, 11.5% Married % 38.4%, 46.2% Living together % 42.3%, 49.2% Separated or divorced % 1.4%, 4.0% Widowed 5 0.5% 0.2%, 1.4% Educational level (%) None or preschool only 8 0.4% 0.2%, 1.1% Elementary school a % 7.4%, 13.3% High school or postsecondary a % 55.5%, 62.5% University or postgraduate % 26.8%, 34.6% Occupation (%) Unemployed 1, % 64.5%, 72.1% Professional, clerical, or skilled manual % 5.9%, 10.6% Sales or service % 12.7%, 17.4% Unskilled, domestic, or agriculture % 6.8%, 11.1% Partner s occupation (%) Unemployed % 10.7%, 16.7% Professional, clerical, or skilled manual % 37.5%, 46.0% Sales or service % 10.7%, 15.2% Unskilled, domestic, or agriculture % 28.6%, 35.8% a. Either some schooling or completed schooling. children from poorer households were more likely to be breastfed at the time of the survey, were more likely to have been introduced to complementary foods at 6 to 8 months of age, were less likely to be eating a minimally acceptable diet, were more likely to be breastfed at 2 years of age, and were substantially less likely to be taking multivitamin supplements than their peers from wealthier households (all p <.05; data not shown). The prevalence of acute malnutrition (wasting, WLZ < 2) was low (approximately 5%) in study children 6 to 23 months of age. Just over 26% of children were stunted (LAZ < 2), and 18% were underweight (WAZ < 2). Anemia affected 41.8% of children aged 6 to 23 months, but more than a quarter of the children were only mildly anemic; moderate anemia was somewhat less common (16% of children), and severe anemia was rare. Iron deficiency was present in slightly more than a quarter of children, and about threequarters of these cases manifested as iron-deficiency anemia, meaning that almost half of the anemia in this population of young children was explained by iron deficiency. Vitamin A deficiency was present in 3.3% of the study children. Risk factors for stunting and wasting In a first step, an extensive list of socioeconomic and IYCF variables was assessed for crude associations with stunting or wasting without consideration of possible confounding. Because wasting was relatively uncommon, none of the associations between wasting and any potential risk factors were statistically significant, and these analyses are not presented. Detailed results of the analyses for stunting are presented in table 4. It is noteworthy that many of the household economic factors and mother s education were significantly associated with stunting, although the bivariate associations between stunting and morbidity indicators, and stunting and the WHO core IYCF indicators were not statistically significant. The only dietary factor with a

7 Infant and young child feeding practices in urban Philippines S23 TABLE 3. Characteristics of selected children Characteristic No. of children Weighted % or mean 95% CI Demographic variables Age (mo) (mean) 1, , 15.4 Age group (mo) (%) % 27.9%, 36.6% % 18.8%, 24.3% % 42.2%, 50.8% Sex (%) Male % 47.0%, 52.9% Female % 47.1%, 53.0% Type of delivery (%) Vaginal 1, % 87.6%, 92.3% Cesarean % 7.7%, 12.4% No. of siblings (%) % 28.2%, 35.3% % 24.8%, 31.0% % 13.9%, 19.8% % 7.7%, 12.3% % 6.1%, 9.8% % 4.9%, 8.3% Principal caretaker (%) Mother 1, % 79.1%, 86.0% Father % 2.5%, 4.7% Grandmother or grandfather % 6.2%, 11.4% Other family member % 3.0%, 7.0% Unrelated caretaker % 0.3%, 2.0% Recent morbidity and current inflammation Fever in past 2 wk (%) Yes % 30.2%, 36.7% No 1, % 63.3%, 69.8% Cough and fast breathing in past 2 wk (%) Yes % 10.2%, 14.9% No 1, % 85.1%, 89.8% Diarrhea in past 2 wk (%) Yes % 14.2%, 19.6% No 1, % 80.4%, 85.8% Elevated markers of inflammation (%) None 1, % 70.3%, 77.4% CRP only % 3.4%, 6.8% Both CRP and AGP % 7.3%, 11.3% AGP only % 9.6%, 15.0% Breastfeeding and complementary feeding indicators a Breastfeeding initiated < 1 h after birth (all children) (%) Yes % 44.9%, 52.2% No % 47.8%, 55.1% Currently breastfeeding (all children) (%) Yes % 48.4%, 55.5% No % 44.5%, 51.6% continued

8 S24 F. Rohner et al. TABLE 3. Characteristics of selected children (continued) Characteristic No. of children Weighted % or mean 95% CI Continued breastfeeding at 1 yr (children mo) (%) Yes % 43.1%, 60.9% No % 39.1%, 56.9% Introduction of complementary foods (children mo) (%) Yes % 85.5%, 94.9% No % 5.1%, 14.5% Minimum dietary diversity met (all children) (%) Yes 1, % 54.1%, 61.2% No % 38.8%, 45.9% Minimum meal frequency met (all children) (%) Yes 1, % 87.1%, 91.6% No % 8.4%, 12.9% Diet meets minimum criteria for acceptability (all children) (%) Yes % 53.6%, 62.6% No % 37.4%, 46.4% Ate animal-source food in past 24 h (%) Yes 1, % 81.6%, 92.3% No % 7.7%, 18.4% Micronutrient supplementation Current use of multivitamins (%) Yes 1, % 67.1%, 73.7% No % 26.3%, 32.9% Vitamin A supplementation in past 6 mo (children 12 mo) (%) Yes % 57.8%, 67.7% No % 32.3%, 42.2% Malnutrition Weight-for-length z-score (mean) 1, , 0.29 Wasting (%) Severe % 0.2%, 1.2% Moderate % 3.4%, 6.5% None 1, % 91.3%, 94.9% Overweight % 0.6%, 2.3% Obese 2 0.3% 0.1%, 1.2% Length-for-age z-score (mean) 1, , 1.06 Stunting (%) Severe % 5.8%, 9.3% Moderate % 16.3%, 22.3% None 1, % 69.8%, 77.0% Weight-for-age z-score (mean) 1, , 0.78 Underweight (%) Severe % 2.1%, 4.9% Moderate % 11.7%, 17.4% None 1, % 79.0%, 85.5% continued

9 Infant and young child feeding practices in urban Philippines S25 TABLE 3. Characteristics of selected children (continued) Characteristic No. of children Micronutrient status Weighted % or mean 95% CI Hb concentration (g/l) (mean) 1, , 112 Anemia (%) None 1, % 54.0%, 62.2% Mild (Hb g/l) % 22.0%, 28.8 Moderate (Hb g/l) % 13.5%, 19.4% Severe (Hb < 70 g/l) 3 0.3% 0.1%, 2.0% Iron deficiency (after adjustment for inflammation) (%) Yes % 24.6%, 31.3% No 1, % 68.7%, 75.4% Anemia (%) Anemia with iron deficiency % 16.0%, 21.9% Anemia without iron deficiency % 20.1%, 26.3% No anemia 1, % 54.1%, 62.2% Vitamin A deficiency (after adjustment for inflammation) (%) Yes % 2.3%, 4.8% No 1, % 95.2%, 97.7% AGP, alpha 1 -acid-glucoprotein; CRP, C-reactive protein; Hb, hemoglobin a. See text in Methods section for summary definitions or Daelmans et al. [6] for detailed definitions. TABLE 4. Bivariate relationships between stunting and various potential risk factors Factor Weighted % stunting Adjusted chisquare p value Household socioeconomic status <.001 Poorest (class E) 44.4 Poor (class D) 25.5 Middle (class C) 10.7 Cooking fuel <.001 Kerosene, coal, wood, etc Electricity of natural gas 19.6 Safe drinking water <.001 Yes 32.0 No 20.6 Adequate sanitation facilities <.01 Yes 25.1 No 36.5 Mother s age (yr).43 < Mother s educational level <.001 None or preschool only 59.6 Elementary 35.9 High school or postsecondary 29.8 University or postgraduate 16.7 continued

10 S26 F. Rohner et al. TABLE 4. Bivariate relationships between stunting and various potential risk factors (continued) Factor Weighted % stunting Adjusted chisquare p value Mother s occupation.32 Unemployed 28.0 Unskilled, domestic, or agriculture 27.9 Sales or service 20.2 Professional, clerical, or skilled manual 21.3 Child s age (mo) < Child s sex.78 Male 26.9 Female 26.1 No. of siblings Child had fever in past 2 wk.07 Yes 31.0 No 24.2 Child had cough and fast breathing in past 2 wk Yes 28.0 No 26.2 Child had diarrhea in past 2 wk.62 Yes 28.2 No 26.1 Child has inflammation (CRP or AGP elevated) Yes 28.5 No 25.7 Breastfeeding initiated < 1 h after birth.11 Yes 29.0 No 23.9 Child currently breastfeeding.09 Yes 29.4 No 24.0 Child meets minimum dietary diversity.35 Yes 27.8 No 24.6 Child meets minimum meal frequency.40 Yes 27.2 No 23.6 Child s diet is acceptable.14 Yes 30.1 No continued

11 Infant and young child feeding practices in urban Philippines S27 TABLE 4. Bivariate relationships between stunting and various potential risk factors (continued) Factor Weighted % stunting Adjusted chisquare p value Child ate animal-source food in past 24 h.71 Yes 26.6 No 24.8 Child consumes multivitamins at least weekly.001 Yes 22.2 No 36.7 Child received vitamin A in past 6 mo.48 Yes 32.5 No 29.3 AGP, alpha 1 -acid-glucoprotein; CRP, C-reactive protein statistically significant negative association with stunting was recent daily or weekly intake of multivitamin supplements. Logistic regression corroborated some of these findings (table 5), confirming that a lower socioeconomic status, the use of cheaper household cooking fuel, older child age, and nonuse of multivitamins remained significantly negatively associated with stunting after accounting for the effects of the other variables in the model. TABLE 5. Logistic regression for stunting, anemia, iron deficiency, and vitamin A deficiency as outcomes using factors with statistically significant associations Variable Odds ratio 95% CI P value Outcome: stunting a Socioeconomic group <.001 Poorest (class E) , 9.4 Poor (class D) , 4.4 Middle (class C) Referent Cooking fuel <.05 Kerosene, coal, wood, etc , 2.2 Gas or electricity Referent Child s age (mo) <.001 Change in odds with each month , 1.12 Multivitamin use daily or weekly <.05 No , 2.4 Yes Referent Outcome: anemia a Socioeconomic group <.05 Poorest (class E) , 3.8 Poor (class D) , 1.3 Middle (class C) Referent Cooking fuel <.01 Kerosene, coal, wood, etc , 2.3 Gas or electricity Referent Fever in past 2 wk <.05 Yes , 2.4 No Referent continued

12 S28 F. Rohner et al. TABLE 5. Logistic regression for stunting, anemia, iron deficiency, and vitamin A deficiency as outcomes using factors with statistically significant associations (continued) Variable Odds ratio 95% CI P value Inflammation (elevated CRP or AGP).07 Yes , 2.1 No Referent Child s age (mo) <.01 Change in odds with each month , 0.98 Currently breastfeeding <.001 Yes , 4.6 No Referent Outcome: iron deficiency a Mother s educational level <.001 None, preschool, or elementary school , 4.0 High school or postsecondary , 4.5 University or postgraduate Referent Fever in past 2 wk <.05 Yes , 0.94 No Referent Inflammation (elevated CRP or AGP) <.001 Yes , 0.70 No Referent Currently breastfeeding <.001 Yes , 2.8 No Referent Multivitamin use daily or weekly <.01 No , 2.0 Yes Referent Outcome: vitamin A deficiency a Cooking fuel <.001 Kerosene, coal, wood, etc , 22.2 Gas or electricity Referent Inflammation (elevated CRP or AGP) <.001 Yes , 7.6 No Referent Breastfeeding initiated < 1 h after birth <.05 Yes , 5.6 No Referent AGP, alpha 1 -acid-glucoprotein; CRP, C-reactive protein a. Initial models included the following additional variables, which were excluded during backwards elimination: Stunting: use of safe drinking water, adequate sanitation facilities, mother s educational level, number of siblings, presence of fever in past 2 weeks, currently breastfeeding, breastfeeding initiated < 1 hour after birth, and dietary acceptability. Anemia: use of safe drinking water, adequate sanitation facilities, mother s educational level, mother s occupation, number of siblings, minimum dietary diversity, minimum meal frequency, minimum dietary acceptability, ate animal-source food, and daily or weekly multivitamin consumption. Iron deficiency: socioeconomic group, predominant type of cooking fuel used, adequate sanitation facilities, mother s occupation, number of siblings, child s age, presence of fever in past 2 weeks, ate animal-source food. Vitamin A deficiency: socioeconomic group, use of safe drinking water, number of siblings, currently breastfeeding, minimum dietary diversity, daily or weekly multivitamin consumption. None of the interaction terms were statistically significant in the final models.

13 Infant and young child feeding practices in urban Philippines S29 Risk factors for anemia, iron deficiency, and vitamin A deficiency Household socioeconomic status, type of fuel used in the household, and sanitation facilities were all significantly associated with anemia, iron deficiency, and vitamin A deficiency, with lower prevalence rates among children from wealthier households, households using more expensive fuel sources (gas or electricity), and households with adequate sanitation facilities (table 6). Households with safe drinking water had a higher prevalence of anemia, iron deficiency, and vitamin A deficiency, although the association was significant only for anemia. Mother s age was not significantly related to the prevalence of anemia or iron deficiency, but the association with vitamin A deficiency TABLE 6. Bivariate relationships between anemia, iron and vitamin A deficiency and various risk factors Factor Anemia Iron deficiency Vitamin A deficiency Weighted Weighted Weighted % p a % p a % p a Household socioeconomic status <.001 <.001 <.05 Poorest (class E) Poor (class D) Middle (class C) Predominant household cooking fuel <.001 <.001 <.001 Kerosene, coal, wood, etc Natural gas or electricity Household has safe drinking water < Yes No Household has adequate sanitation <.01 < Yes No Mother s age (yr) < Mother s educational level <.001 < None or preschool only Elementary High school or postsecondary University or postgraduate Mother s occupation <.01 < Unemployed Unskilled, domestic, or agriculture Sales or service Professional, clerical, or skilled Child s age (mo) < Child s sex Male Female continued

14 S30 F. Rohner et al. TABLE 6. Bivariate relationships between anemia, iron and vitamin A deficiency and various risk factors (continued) Factor Anemia Iron deficiency Vitamin A deficiency Weighted Weighted Weighted % p a % p a % p a No. of child s siblings < < Child had fever in past 2 wk <.01 < Yes No Child had cough and fast breathing in past 2 wk Yes No Child had diarrhea in past 2 wk Yes No Child has inflammation (CRP or AGP elevated) < <.001 Yes No Breastfeeding initiated within 1 h of birth <.01 Yes No Child currently breastfeeding <.001 < Yes No Child meets minimum dietary diversity Yes No Child meets minimum meal frequency < Yes No Child s diet is acceptable < Yes No Child ate animal-source food in past 24 h < Yes No Child consumes multivitamins at least weekly <.01 <.01 <.01 Yes No Child received vitamin A in past 6 mo Yes No a. Adjusted chi-square p value.

15 Infant and young child feeding practices in urban Philippines was marginally significant. Maternal educational level and current occupation were significantly associated with anemia, iron deficiency, and vitamin A deficiency. Children of mothers with a college-level education or a professional, clerical, or skilled occupation had a significantly lower prevalence of these deficiencies. The low prevalence seen in children of mothers with no education is probably due to the very small number of children in this category. Inflammation was significantly associated with anemia, iron deficiency, and vitamin A deficiency, with positive associations for anemia and vitamin A deficiency and a negative association for iron deficiency. Other child-related factors, such as age, number of siblings, and recent fever were significantly associated with anemia, but not with iron or vitamin A deficiency. Symptoms of morbidity were not associated with micronutrient deficiencies. The observed bivariate associations between feeding indicators and micronutrient status were relatively weak and sporadic. Children who were breastfed at the time of the survey had a substantially higher prevalence of both anemia and iron deficiency. On the other hand, children with early initiation of breastfeeding had a statistically significantly higher prevalence of vitamin A deficiency than children without early initiation of breastfeeding. Meeting the minimum criteria for meal frequency and acceptability of child s diet was associated with a lower prevalence of anemia. Not surprisingly, daily or weekly consumption of multivitamin supplements was significantly associated with a higher prevalence of anemia, iron deficiency, and vitamin A deficiency. Finally, there were no associations between receiving vitamin A supplementation in the prior 6 months and any micronutrient deficiency investigated in this study. As with stunting, logistic regression demonstrated the strongest independent association between household socioeconomic status and anemia and iron deficiency (table 5). Children presenting with inflammation (as measured by elevated AGP or CRP) had a significantly higher risk of concurrent vitamin A deficiency, whereas inflammation-positive children had a lower risk of concurrent iron deficiency; for anemia, the significance level was borderline (p =.07), with a higher anemia prevalence for children with inflammation. The mother s educational level influences the risks of both anemia and iron deficiency; however, there was no dose response relationship between level of education and level of deficiency. Breastfed children were substantially more likely to be anemic or iron deficient, even after controlling for socioeconomic status, maternal education, and child s age (anemia) and multivitamin use (iron deficiency). Children who lived in households using cheaper liquid or solid cooking fuels and those with early initiation of breastfeeding had a significantly higher prevalence of vitamin A deficiency than those with later onset of breastfeeding. Discussion S31 A Demographic and Health Survey conducted in 2008 revealed that many young children in the Philippines were not being fed according to internationally recommended guidelines [1]. Based on these and earlier findings, a consortium of agencies was established to improve breastfeeding practices among young children and to increase timely introduction of appropriate complementary foods. The survey described here was undertaken to establish a baseline for these interventions, and at the time of writing, implementation of these interventions has begun. This is a cross-sectional study, and the findings should therefore be interpreted cautiously when formulating conclusions about cause effect relationships. The UNICEF framework, however, provides a useful logical model for interpretation of our data. This framework consists of basic, underlying, and direct causes of malnutrition. Household socioeconomic status is related to many of the underlying causes, which include access to food, caring practices, hygiene environment, and healthcare utilization. In our data, socioeconomic status had an important association with the risks of children s being stunted, anemic, or iron or vitamin A deficient; children from poorer households were more prone to suffer from these conditions. This has been previously shown in the analysis of large datasets [14], in which socioeconomic status had a particularly strong relationship with child nutritional status in urban (as opposed to rural) households [15]. The direct causes of stunting, as with other causes of malnutrition shown in the UNICEF framework, are poor dietary intake and high incidence rates of diseases. Some of the intermediate factors, both underlying and immediate causes in the UNICEF framework, through which socioeconomic status may affect nutritional status were measured in our survey using variables such as access to food, access to safe drinking water and sanitation facilities, and caring and feeding practices, which may also be influenced by maternal education and occupation. For example, poor households may have limited access to safe water and sanitation, increasing the risk of diarrheal disease, or may lack the purchasing power to provide a sufficiently varied diet. Although mother s education has long been known to have a major effect on child nutritional status [16], it was associated only with iron deficiency in our study. Maternal education may operate independently from overall household socioeconomic status by improving mother s caring practices for young children [16], or it may merely raise the socioeconomic status of the household without having an independent effect through other mechanisms [17 19]. Our study seems to support the latter hypothesis, given the lack of association between indicators of caring practices and stunting. Yet, the ability of this study to detect

16 S32 F. Rohner et al. an association between maternal education and child nutritional status or to distinguish among the various mechanisms through which maternal education influences child nutritional status is handicapped by the lack of variability in maternal education found in this survey sample; more than 90% of mothers had at least a high-school education. Stunting is a more complex nutritional phenomenon than vitamin A or iron deficiency. Although access to fortified food and/or micronutrient supplements can prevent specific micronutrient deficiencies, these measures may only partially address the range of nutrient deficiencies that underlie stunting. The fact that socioeconomic status remained associated with stunting but not with vitamin A and iron deficiencies in multiple regression analyses indicates that the direct and underlying factors as measured in this survey were not completely able to explain the variation observed in stunting. Since stunting develops over a long period of time, starting at conception, current dietary intake may not adequately measure the adequacy of nutrient supply over the entire period. If this is true, socioeconomic status may be a better proxy of access to nutritious foods and hence nutrient intake. Inflammation was positively associated with anemia and vitamin A deficiency, whereas the association was negative for iron deficiency. In the case of anemia, this could be ascribed to anemia of chronic disease [20], which develops from long-standing, low-level inflammation. The negative association of inflammation with iron deficiency is somewhat more surprising, as one would expect iron metabolism to be disturbed during inflammation [21]. In the data we present, we have adjusted ferritin concentrations for inflammatory responses, but it may be that these adjustments are not sufficient to counterbalance ferritin s function as an acute-response protein. Most of the IYCF indicators were not significantly associated with stunting, anemia, iron deficiency, or vitamin A deficiency. These indicators are meant to provide cross-sectional assessments of behaviors in the population; they have major drawbacks when used as measures of individual-level exposure to risk. For example, they do not include information about the quantity or frequency of food consumption over a longer period of time. As a result, our findings could be influenced by misclassification bias, which would weaken the observed association between IYCF behaviors and malnutrition. In addition, because our survey had a relatively small sample size, we may have failed to achieve the power needed to detect a statistically significant association between dietary factors and nutritional outcomes, such as dietary diversity, which have been shown to be significant in much larger studies examining these factors [22, 23]. Regardless, the confidence intervals for the odds ratios for dietary diversity and stunting from our survey (analysis not shown) overlap with the estimates from these other studies, indicating that our survey s findings are not incompatible with those of other studies. Children who were currently breastfed were more likely to be anemic or iron deficient. This has been repeatedly shown in previous studies, with the hypothesis being that after 6 months of the child s life, iron stores from birth are exhausted and the low iron content of breastmilk does not provide sufficient iron to the child, although breastfeeding might partially displace more iron-rich foods [24]. The mean age of the children in the sample was 15 months, and at that age, breastmilk (which has a very low iron content), if it is replacing other iron-rich foods, may not provide sufficient amounts of dietary iron. In addition, mothers of children who are not healthy or who are growing poorly (and therefore are more likely to be anemic) may breastfeed their children longer a form of reverse causation, as demonstrated in other studies [25]. Current consumption of multivitamin preparations, including syrups, tablets, or drops, was associated with a reduced prevalence of stunting and current iron deficiency. This finding should be further investigated in future studies, and detailed questionnaire modules should be incorporated into nutrition surveys to assess the frequency, dosing, and type of multivitamin preparations consumed by young children. In addition, analyses should characterize which households use such preparations, as multivitamin use not only is an indicator of micronutrient intake but may also be an indicator of socioeconomic status. Furthermore, a more in-depth module on the use of other fortified foods, such as micronutrient powders, should be included; this could possibly expand to consumption of iron-fortified foods or supplements during pregnancy. In our study, the number of caregivers who reported using micronutrient powders was too small for analysis. Both use of cheaper cooking fuel and nonuse of multivitamin preparations were associated with several outcomes in this study. Although neither of these factors was directly used to assess socioeconomic status, they are linked to socioeconomic status (results not shown). In spite of this association, our analysis showed that both factors remained significantly associated with several nutrition outcomes after accounting for the effect of socioeconomic status in regression models. The reason for an association between multivitamin use and lower prevalence of micronutrient deficiencies is somewhat self-evident. However, the association between the type of cooking fuel and nutrition outcomes is less clear. Previous studies have shown higher stunting and anemia prevalence in households using lower-quality fuels or open fires [26, 27]. As was done in our study, the data included in these studies were derived from questions asked about the usual type of fuel used for cooking during interviews carried out in cross-sectional surveys. The authors of the studies did