CSF specimens and labs before hospital admission

[BoyceLab]

Hello,

First, I would like to thank you for making this beautiful data set. I am in awe of this accomplishment, and it is a pleasure to analyze it.

I am currently interested in meningitis, associated microbiology, labs, procedures, and antibiotics. My cohort definition is based on the presence of CSF sample and related lab results, not ICD code alone. I constructed a data set that joined on hadm_id. Then I noticed that there were a lot of patients who had a meningitis-related diagnosis in the diagnosis file that didn't make it into the data set. Upon investigating, it was clear that these patients were dropped because their CSF;fluid specimen types did not have a hadm_id.

I read in another post the reason that some labs might not have hadm_ids. It makes perfect sense that they would have their lumbar puncture and associated labs in the ER, or even at another hospital before transfer and that's why these hadm_ids would be missing at the time of specimen collection. I have spent a lot of hours this week trying to come up with the best approach to these non-hadm_id CSF folks, and to join all of the data from various tables based on the proximity of the CSF sample date to the dates of the other items. My solutions have all been somewhat messy and complicated, mostly due to the number of data elements that I am attempting to pull together based on dates alone.

I am looking for any suggestions on best approach other than what I am doing now, which is quite clunky and iterative. It could very well be that I am too close to it and overthinking!

Thank you again for this wonderful resource!
Danielle