Finding central line placement date on MIMIC-III Dataset

Question

I'm trying to find a record on MIMIC dataset showing that a patient being placed with a central line (aka central venous catheter) and I need this information only for CLABSI patients (i.e. diagnosed with ICD 9 codes '99931', '99932', '99933', '99662', '9993’). Here are the tables I looked and the problematic things about them:

-Procedure_icd (Dictionary table: D_icd_procedures): Contains the most through information. I checked the icd 9 code 38.93 (Venous catheterization, not elsewhere classified). Among 1300 total CLABSI patients, 1039 of them has a record in this table. However, there is no date and caregiver information (who placed the central line) in this table, so we cannot use this table.

-CptEvents table: I found out that the CPT codes for central line placement are 36555-36771. Among 1300 total CLABSI patients, only 261 has a record in this table.

-XEvents tables which uses D_items table as their dictionary table (ChartEvents, DatetimeEvents, InputEvents, etc.): I searched D_items table to find the codes for central line placement. I found 158 different codes. (Eg: 117;"Catheter Insert Date", 5442;"pic line placement", 4476;"PICC line placement”). Among 1300 total CLABSI patients, chartEvents and DatetimeEvents tables have the most record with 429 total unique patients, but none of these tables have more than ~250 records.

Why most of the patients do not have a record about being placed with a central line? Am I missing some codes/not looking at the correct place, or is it just missing data?

Thanks

Answer 1

It's most likely missing data - since in the US ICD codes are used for billing, codes are often absent if they don't result in a higher bill. Placing a central line / catheter is almost never a primary procedure, so often not billable under bundled payments, hence missing when clinical coding is done.

You'll need to mine this data from the free-text clinical notes, which you can do either manually or using a healthcare-specific NLP library.