They didn't look at age of initiation, probably because the data aren't available every year. That would be one thing I might add to their analysis.