El Niño/Southern Oscillation (ENSO) and stunting in children under 5 years in Peru: a double-difference analysis

Abstract

Background: The 1997–98 El Niño/Southern Oscillation (ENSO) was one of the most important natural disasters in 20th century Peru, associated with heavy rainfall and flooding. It affected 502 000 people, ruined 944 km of roads, and damaged 580 health centres and 956 schools. Almost a quarter of all people affected (23%) came from Piura, in the northwest of Peru. In this study, we assessed the association between the 1997–98 ENSO and growth stunting in children younger than 5 years on Peru's north coast. Methods: We analysed Peruvian Demographic and Health Survey (DHS) data, from 1996 and 2000, using a double-difference approach. Our primary outcome was childhood growth stunting, and our exposure was the ENSO phenomenon. The exposed group was from Peru's Piura coast, and the unexposed group was from southern coastal regions, which were less affected by the events associated with ENSO 1997–98. First, we conducted a regional comparison. Secondly, we used propensity score matching (PSM) to explore similarities between the exposed and non-exposed groups. We then estimated adjusted prevalence ratios (PR) of childhood growth stunting using generalised linear models. Finally, we conducted a subanalysis by childhood age. Findings: We studied 2270 and 1349 children younger than 5 years from 1996 and 2000, respectively. Regression without matching showed a PR of 1·37 (95% CI 0·93–2·01), and a PSM regression 1-to-2 PR of 1·45 (0·66–3·18), obtaining a PR greater than 1 but with a confidence interval that included the null value. Furthermore, we found a possible effect on children aged 0–11 months. In the regressions without PSM, analysis for this age group showed a PR of 2·48 (1·11–5·54); whereas, after PSM 1-to-2, the PR was 2·69 (1·03–7·03); alternative PSM was also consistent with these results. Analysis of the other age groups yielded results that were not statistically significant. Interpretation: Our results suggest that the ENSO phenomenon could have increased childhood stunting on the coast of Piura. More specifically, children who were aged 0 to 11 months during the worst and most disruptive period of the El Niño phenomenon were at increased risk of growth stunting. Our results were consistent with and without PSM. Factors that might have contributed to our results include increased childhood morbidity with concurrent reductions in maternal and familial access to health care, and diminished household incomes due to El Niño's devastating effects during the disaster period. Our findings outline a need to prevent harm to people and their health during weather and natural events associated with ENSO. Specifically, there is a need for childhood-focused social protection systems, especially during critical stages of growth and development. Childhood-focused social protection and delivery systems will be instrumental in preventing downstream effects of similar climatic events. Funding: Fogarty 2D43 TW007393.