Using LQAS to assess changes in prevalence of global acute malnutrition within longitudinal surveillance systems

It is extremely important to accurately classify the prevalence of acute malnutrition in children as high and to detect sudden rises in malnutrition prevalence using cost-effective surveys in order to inform when aid should be sent to a region and how resources should be allocated to reduce malnutrition. We discuss challenges in quantifying the prevalence of acute malnutrition, as well as why it may be more meaningful to track trends in malnutrition and classify changes in prevalence over time. LQAS cluster surveys have recently been implemented in the malnutrition setting. We overcome several limitations of the current LQAS survey designs, developing a new LQAS classification procedure for classifying changes in prevalence (or coverage) across time within longitudinal surveillance systems, which we call longitudinal-LQAS, or L-LQAS. Using L-LQAS, we exploit information from multiple surveys to assess whether prevalence is rising (or falling) in a region. We also discuss simple ways to incorporate intracluster correlation into cluster sampling L-LQAS designs. Lastly, we extend this new survey design to classify changes in multiple indicators, as malnutrition is often quantified using two distinct indicators and there is no consensus on which indicator is superior. We apply L-LQAS to assess trends in acute malnutrition in Kenya and Sudan over 3 different time points spanning approximately 1.5 years, and compare the conclusions that we draw using L-LQAS to the previously proposed LQAS malnutrition survey designs. We develop free software which will aid in the implementation of these survey designs in the field.