RESEARCH ARTICLES CURRENT SCIENCE, VOL. 121, NO. 4, 25 AUGUST 2021 502 *For correspondence. (e-mail: sri@chg.res.in) † Equally contributed. Risk factors in childhood stunting in Karnataka, India, vary by geography Srilakshmi M. Raj 1,2, * ,† , Ruwanthi Ekanayake 3,† , Kiera Crowley 1 , Meenakshi Bhat 2 , Jayarama Kadandale 2 and Prabhu L. Pingali 1 1 Tata Cornell Institute, Cornell University, Ithaca, NY 14853, USA 2 Centre for Human Genetics, Electronic City, Bengaluru 560 100, India 3 Department of Biology and Society, Cornell University, Ithaca, NY 14853, USA Childhood stunting remains a public health concern in India. In Karnataka, the districts vary substantially in stunting prevalence. Using the NFHS-4 and AidData GEO datasets, we tested the hypothesis that ‘wet’ and ‘dry’ districts in Karnataka show different contribu- tions to stunting. We found that for 30 environmental and health factors, Bengaluru appears to be distinct from the other districts. Using a mixed linear model approach, we found that for the entire state, and in both wet and dry districts, preceding birth interval, altitude-adjusted haemoglobin level and child age showed significant correlations with height for age (HFA). The wet districts showed an additional associa- tion between maternal age and child HFA. However, interaction effects also differed among the three con- ditions. Our results suggest that subtle variations should not be ignored when considering factors im- pacting child health outcomes. Keywords: Childhood stunting in Karnataka, environ- ment, genetics, nutrition, public health. CHILDHOOD stunting, or reduced linear growth in chil- dren aged 5 years and under, has been correlated with both short-term and long-term adverse health and socio- economic outcomes. With international efforts to reduce malnutrition only purporting to lead to a partial reduction in childhood stunting prevalence in India and elsewhere 1 , it is important that region- and community-specific appro- aches are considered to reduce the burden further and contribute to improvements in healthy growth trajectories. Across India, 46.8 million children under 5 years of age are stunted, constituting one-third of all stunted chil- dren under 5 years globally 2 . This prevalence varies across states: North East India and the central states have prevalence >30%, while a majority of the remaining states have at least 20% stunting 3 . With increased consumption of convenience foods, Indian children may also have accelerated ‘catch-up growth’ following stunting, which results in rising national obesity rates. Therefore, India is considered to have a ‘triple burden of malnutrition’ characterized by calorically undernourished and overn- ourished individuals, with both groups at risk for disor- ders of micronutrient deficiencies such as anaemia, scurvy, night-blindness and rickets 4 . The triple burden of malnutrition has broadly resulted in several hallmarks of child growth failure (CGF), including stunting, under- weight and wasting. Most recently, the three conditions have been explored jointly across India and other coun- tries, in an effort to capture the full scope and context of CGF 5,6 . These analyses point to high variation within In- dia and among countries globally in the relationship be- tween the CGF indicators, depending on geography, environment (e.g. sanitation), socio-economic conditions and parental nutrition status. Therefore, stunting and oth- er indicators of CGF must be studied in their local con- text, often at the sub-state level in India. The importance of studying variation in stunting presentation leads us to Karnataka, India. Among the middle socio-demographic index (SDI) states, Karnataka has the greatest coefficient of variation for childhood stunting 6 . For example, 9 out of its 30 districts have stunting prevalence >40%, with Koppal and Yadgir dis- tricts having >55.5% stunting prevalence 7 . In contrast, Mandya district has a stunting prevalence of only 18%, half of the state-wide average and three times lower than in the hardest-hit states of India. Given this variability, we examine conditions within Karnataka that could lead to such vastly different stunting prevalence across the 30 districts, focusing on rainfall variation across the state. Shedding light on region-speci- fic risk factors and their interactions can help make future public health efforts more targeted and effective, both in Karnataka and across India. Data and methods All of the data used in this study were taken from the Natio- nal Family Health Survey-4 (NFHS-4) and the AidData GEO database, and data for Karnataka were extracted. First, 30 variables from the NFHS-4 and AidData GEO datasets were collapsed into the 30 districts of Karnataka