Abstract

A rigorous quasi-experimental technique for evaluating the causal effects of interventions on outcomes is regression discontinuity (RD) designs. RD can be used to estimate the causal effect of the treatment on health and other outcomes whenever a decision rule as- signs treatment, such as antihypertensive or antiretroviral therapies, to patients who score higher (or lower) than a specific cutoff value on a continuously measured variable, such as blood pressure or CD4 count. Similar to randomization, RD can address issues with confounding caused by unobserved factors and produce estimates of a treatment's causal effects that are free from bias. Due to the prevalence of treatments assigned using a cutoff rule, RD is a particularly helpful study design for medicine, epidemiology, and public health. Statins are prescribed by doctors using blood pressure cutoffs to decide how to manage hypertension, using mole size cutoffs as guidelines for mole removal, and recommending surgery for scoliosis when spinal curvature surpasses a specific threshold of severity. Additionally, RD possesses desirable practical traits. It may not be possible to conduct a randomised controlled trial (RCT) when a treatment has already become the norm, but RD can provide robust causal evidence on treatment efficacy when there is scant or no experimental evidence or when the evidence that is available has dubious internal or external validity.