MIT-LCP / vancomycin-nephrotoxicity

Study in eICU-CRD about vancomycin and associated nephrotoxicity
5 stars 4 forks source link

Routes in eICU-CRD #1

Closed alistairewj closed 4 years ago

alistairewj commented 4 years ago

The routes in eICU-CRD that we decided on during the project do not exhaustively cover the routes in the full version of eICU. e.g. running this query on hera:

WITH ce AS
(
SELECT 
  patientunitstayid,
  CASE 
      WHEN drugstopoffset = 0 THEN drugstartoffset
      WHEN drugstartoffset <= drugstopoffset THEN drugstartoffset
      WHEN drugstopoffset < drugstartoffset THEN drugstopoffset
  END AS drugstartoffset,
  CASE 
      WHEN drugstopoffset = 0 THEN NULL
      WHEN drugstartoffset <= drugstopoffset THEN drugstopoffset
      WHEN drugstopoffset < drugstartoffset THEN drugstartoffset
  END AS drugstopoffset,
  drugorderoffset,
  routeadmin,
  m.frequency,
  map.classification,
  dosage
FROM eicu_full_phi.medication m
LEFT JOIN vanco.medication_frequency_map map
  on m.frequency = map.frequency
WHERE 
  (drughiclseqno IN (10132,35848,37021) OR LOWER(drugname) LIKE '%cefepim%' OR LOWER(drugname) LIKE '%maxipim%'
  OR drughiclseqno IN (4042, 10093, 8466, 37442) OR LOWER(drugname) LIKE '%vanco%'
  OR drughiclseqno IN (8738, 32900, 33427) OR lower(drugname) like '%zosyn%' OR lower(drugname) like '%piperacillin%tazo%')
AND drugordercancelled = 'No'
AND prn = 'No'
AND COALESCE(map.classification, '') NOT IN
(
  'TPN', 'dialysis', 'prophylactic', 'prn'
)
AND lower(m.frequency) NOT LIKE '%dialysis%'
AND lower(m.frequency) NOT LIKE '%prn%'
AND lower(m.frequency) NOT LIKE '%tpn%'
)
select ce.routeadmin,
abx.code,
-- below was the code used on eICU-CRD that covered all IV routes in the public data
CASE WHEN ce.routeadmin IN (
  'IV', 
  'Intravenous', 
  'INTRAVENOU', 
  'INTRAVEN', 
  'IntraVENOUS', 
  'IV (intravenous)                                                                                    ', 
  'INTRAVENOUS',
  'IV - brief infusion (intravenous)                                                                   ',
  'PERIPH IV',
  'IV Push'
) THEN 1 ELSE 0 END AS in_original_query,
count(*) as n, count(distinct patientunitstayid) as n_pt
from ce
LEFT JOIN vanco.abx_route abx
ON ce.routeadmin = abx.routeadmin
group by 1, 2
order by 2, 3 desc;
routeadmin code in_original_query n n_pt
IV (intravenous) IV 1 66981 43208
PERIPH IV IV 1 27 15
INTRAVEN IV 1 73825 18837
INTRAVENOU IV 1 61584 16123
Intravenous IV 1 106668 31549
IntraVENOUS IV 1 6757 2887
INTRAVENOUS IV 1 5902 1329
IV IV 1 1446437 538850
IV - brief infusion (intravenous) IV 1 7522 5214
IV (intravenous) IV 0 1896 1305
INTRA-CATHET IV 0 1 1
Intercatheter IV 0 9 3
CATH IV 0 34 9
EXTENDEDIV IV 0 368 302
PEDIV IV 0 3 3
Catheter IV 0 3 2
CATHETER FLU IV 0 18 9
IV Sedation IV 0 1 1
Catheter lock IV 0 12 5
*IV IV 0 108 94
IV* IV 0 2 2
IV112 IV 0 3 3
IVNTRC IV 0 6 2
IV Continuous IV 0 11 11
IVD IV 0 1 1
IVDRIP IV 0 4 3
IVENTRIC IV 0 11 9
IVI IV 0 962 259
IV INFUSION IV 0 29270 5466
IVNT IV 0 3 1
IN VITRO IV 0 1 1
INTRAVEN. IV 0 12764 3028
Intracatheter IV 0 2 1

This seems to have increased cohort sizes by 30% across the board.

Old code missing IV administrations:

262498 patients.
Antibiotic use for 48 hours in the ICU:
  23814 ( 9.1%) - vanco
   4756 ( 1.8%) - cefepime
  24190 ( 9.2%) - zosyn

Concurrent antibiotic use:
   1958 ( 0.7%) - vanco & cefepime
   9424 ( 3.6%) - vanco & zosyn

New code with all IV administrations:

262498 patients.
Antibiotic use for 48 hours in the ICU:
  30694 (11.7%) - vanco
   7119 ( 2.7%) - cefepime
  31377 (12.0%) - zosyn

Concurrent antibiotic use:
   3208 ( 1.2%) - vanco & cefepime
  12718 ( 4.8%) - vanco & zosyn