How would you optimize a slow-running data step or query when working with millions of clinical patient records?
