We will start with NHANES 3, one of my favorite datasets. The database has been aggregated with all data and linked to the national death index from 2011. All of this is possible via the CDC website that has guidelines on linkage, but can be a little onerous. There is a good NHANES R package that is a very easy tool to create the dataset.
We have used NHANES 3 for a variety of projects. The two that were published and influenced guidelines focused on simple clinical questions that arose when I was a resident.
One project examined the role of fasting versus non-fasting lipids – our findings changed guidelines in the United States and abroad. We challenged the need for patients to fast (not eat) before their lab draws, and did not find a difference between mortality in groups that had fasted with similar lipids versus those who had not.
Tips for analysis
01 : The power of the dataset is large, but the need for precise methods of analysis is also important.
02 : Data from NHANES 3 was collected from ‘88-’94 in a cross sectional survey.
03 : The National Death Index from 2006 and 2011 was linked to the data, which gives us our outcomes.
04 : If you would like to generalize to the entire US population, it is possible but survey weights must be used.