ThirtyDayReadmission and LongLengthofStay tasks yield lower than expected AUC

[conorkcorbin]

We see AUC in the high 0.6's, should see in mid 0.8's based on other groups using STARR reporting performance on these tasks.

• May need to dig into cohort definitions and see where things could be going wrong.
• Try increasing # of examples sampled per year.