Child stunting remains an important global health issue, with 157 million children under five years of age estimated to be stunted in 2014. Until recently, stunting was thought to occur in the first 1000 days of life (between conception and 2 years of life), and was thought to be largely irreversible thereafter. However, emerging research suggests that children can transition between stunted and non stunted status up to 15 years of age, with studies also suggesting potential implications in terms of cognitive status. Despite this, there is little research on stunting and its potential determinants among children of older ages, with most current studies confined to those under five. This study aims to assess the prevalence of stunting and examine potential sociodemographic determinants of stunting (including individual, maternal and household level indices) among older children (aged 6-19 years) in a Malaysian population.

This analysis is based on existing data collected by a health and demographic surveillance system operating in Segamat, Malaysia, and data for all individuals meeting the stated inclusion criteria are used in the study.

There is not a specific control treatment in this study. Rather, we calculate the risk of stunting associated with (1) a unit increase in each exposure, or (2) categories of exposure with respect to a referent category. Specifically, for the primary exposures as listed above:

• Age: risk per year increase
• Sex: risk in girls versus boys (referent)
• Ethnicity: risk in (1) Indian, (2) Chinese, (3) Indigenous or (4) Other ethnicity, versus Malay (referent)
• BMI-for-age status: risk in (1) underweight or (2) overweight, versus normal weight (referent)
• Birth order: risk in children of second, third and fourth or higher birth order, versus first born (referent)
• Maternal height: risk in children of mothers with height 155-159cm, 150-154cm, 145-149cm, and <145cm, versus those with height 160cm or higher (referent)
• Maternal current underweight: risk in children of mothers with BMI <18.5 kg/m2, versus those with BMI >=18.5 kg/m2 (referent)
• Rooms (bedrooms, bathrooms, living areas) per household member: risk per unit increase in room to person ratio for each room type
• Type of toilet: risk among children in households with (1) Bore hole toilet, (2) Pour flush toilet, (3) Flush toilet with septic tank, (4) Flush toilet connected with sewerage system, versus those living in households with none, bucket or hanging latrine (referent)
• Toilet shared with other household (yes/no): risk among children in households with a shared toilet, versus those living in households without one
• Main source of drinking water: risk among children in households using water from (1) Public standpipe or other protected source, (2) Piped into yard, (3) Piped into house, versus those living in households using an unprotected source (referent)
• Main method of garbage disposal: risk among children in households having their garbage (1) Collected and thrown for recycling, (2) Collected irregularly by local authority, (3) Collected regularly by local authority, versus in those in households where garbage is buried, burned or thrown (referent)

Methods for crude analysis and gaining an introductory sense of the data include examination of variable distributions and clustering of the outcome variable of interest (stunting or height-for-age). Exposure variables are assessed by stunting status; differences between stunted and non-stunted groups are assessed using Student's t test for continuous variables, and Pearson's chi squared test (Fisher's exact test for variables with cell counts <5) for categorical variables. Additionally, the classification and prevalence of stunting is assessed using two different references: the World Health 2007 reference and Centers for Disease Control and Prevention 2000 reference; agreement in classification between the two is calculated using Cohen's kappa.

The primary method of analysis is mixed effects Poisson regression, with stunting as the outcome of interest. Final models include all exposure variables of interest, in order to assess any independent associations between each exposure and stunting risk. All models are adjusted for clustering at the household level.

A number of secondary analyses are used in order to check the robustness and specificity of associations. These include:

1. Adding maternal age as a covariate in models, examining associations of maternal age with stunting, and the effect of its addition on other associations
2. Adding in maternal education, paternal education or head of household's education, and assessing subsequent effects as described in (1)
3. Adding in paternal height, and assessing subsequent effects as described in (1)
4. Adding in a number of theoretically uncorrelated variables to test the specificity of associations, and assessing subsequent effects as described in (1)
5. Using mixed effects linear regression, with height-for-age z-score as the outcome of interest. Final models include all exposure variables of interest, in order to assess any independent associations between each exposure and stunting risk. All models are adjusted for clustering at the household level.

Both the primary and secondary analyses are run with stunting or height-for-age expressed according to (1) the Centers for Disease Control and Prevention 2000 reference and (2) the World Health Organization 2007 reference.

Inclusion Criteria:

1. Age >= 6 years and <= 19 years
2. Whether or not the potential participant has information on all exposure and outcome variables of interest.

Exclusion Criteria None

This study is based on existing observational data collected by the the South East Asia Community Observatory (SEACO), a health and demographic surveillance system (HDSS) covering approximately 45 000 individuals in Segamat, Malaysia. The HDSS regularly (annually) enumerates all consenting households and individuals within its catchment area, and has also conducted a health survey to date during which basic sociodemographic, lifestyle-related and anthropometric data were collected from individuals aged 6 years and above. Further information on the SEACO HDSS can be found at: https://academic.oup.com/ije/article/46/5/1370/4037470.This study uses cross-sectional data on children aged 6-19 years, collected between 2012-2014.