SQL issue in reason_for_admission.ipynb about amsterdamumcdb

[patrickthoral]

Thanks for pointing this out! Indeed, the intended way was to use the last documented diagnosis, because typically that was a correction on the initial documentation (often by more 'junior' physicians) by staff (intensivists). However, I incorrectly used measuredat instead of updatedat for sorting. This, as you found out, lead to unpredictable sorting since measuredat is not unique for a given admission diagnosis , while updatedat should always be unique. I'll fix this shortly.

[tedinburgh]

Great, that makes sense! Is it better to use updatedat instead of measuredat in other situations too?

[patrickthoral]

Not necessarily. For most time series the time the measurement belongs to (measuredat) is the most relevant one, because that allows us to see the correct order of sequential measurements. The updatedat only documents when the record was last updated (or created), which is often slightly later, because nurses, performing most documentation and data validation, will document when they are less busy, but try to make sure that they put the values in the right ‘column’ (i.e. time @ 12:00 hours) of the flowsheet they are documenting on. However, for measurements/variables that are not really time series (such as the reason for admission) the measuredat does not really have a meaning other than the time of the session in which it was created. In those cases the highest measuredat (for the same admissionid) suggests updates that improved or corrected previous documentation. Especially since APACHE II and IV reasons for admission are used in calculating standardized mortality ratios for national benchmarking, I am quite confident that the most recent value is the ‘correct’ one. Hope this makes the distinction between the two times more clear.