While the method presented in Query Set #6 solves the missing data problems for labs, all of the information contained in the actual lab values is discarded. For BNP, for example, only two of the patients don't have a value, and for the temperature vital sign, only one patient is missing. 

Some previous studies have experimented with this principle and have obtained good results with predictive models while using it. In (Donze et al., 2013), some of the patient discharges (around 1%) had missing data. This data was filled in by assuming it was in the normal range.

In SQL, single imputation can easily be done. We demonstrate this here.