issues with current data

alistairewj commented 7 years ago

eICU

eICU BMI is all zeros
eicu po2 max accidentally coded as min?
eicu maximum diasbp is very high OR anzics diasbp is very low
eicu missing pao2:fio2 max ratio

anzics

anzics minimum diasbp usually lower
missing d1 glucose for anzics
hco3 anzics units are wrong
hct for anzics is missing
anzics potassium min/max units are wrong
anzics sodium min/max units are wrong
temperature source in eicu vs anzics?
fio2 present in anzics?
apache glucose missing from anzics
anzics missing hematocrit_apache
anzics sodium_apache units wrong

mimic

inr mimic max/min switched?
mimic ph_apache is it arterial? has high pHs

MIMIC in general: how to define apache variables if score=0? e.g. messes up temperature

worth comment

hgb difference in mimic
hct difference in mimic
mbp high in anzics

alistairewj commented 7 years ago

Hospital death in eICU needs to be 0/1

alistairewj commented 7 years ago

all above fixed, with caveats below:

eicu maximum diasbp is very high OR anzics diasbp is very low

nothing wrong here.. seems to be a part of the data

apache glucose missing from anzics

seemed to be present, should double check

temperature source in eicu vs anzics?

haven't checked

MIMIC in general: how to define apache variables if score=0? e.g. messes up temperature

hard in general D:

alistairewj commented 7 years ago

eicu

[x] Eicu map female to F, male to M, other/unknown to null
[x] Readmission status in eicu is "eventually readmitted" but other data has it as "current admit is readmit"

anzics

[ ] 23 anzics patients survive hospital but die in ICU.. inter hospital transfer?
[x] Anzics pregnancy - apply dictionary to map 0/1

orchestra

[x] Orchestra map gender I to null
[ ] Orchestra ICU type numbers instead of words
[x] Bed size collapse into categories for orchestra

mimic

[x] Mimic map 0/1/2 for smoker to anzics text categories
[x] Mimic gcs verbal of 0 needs to be mapped to 5 or something
[x] Mimic ethnicity combination (use sepsis 3 code)

all

[ ] Hospital admit source and discharge location needs to be merged for all data sources (same for ICU admit and discharge)
[ ] ICU admit type in eicu and mimic are different (ICU type versus elect/emergency, prob incorrect for eicu).. missing anzics and orchestra?
[ ] ICU stay type has no definition and is inconsistent
[ ] ICU type is not well collapsed
[ ] add column for raw bed size number

hospital admit source

eicu Emergency Department 49735 Operating Room 12102 Floor 10443 Direct Admit 8418 Recovery Room 4162 Other Hospital 2424 Acute Care/Floor 2105 Step-Down Unit (SDU) 1481 PACU 1004 Chest Pain Center 274 Other ICU 265 ICU to SDU 56 ICU 40 Observation 12 Other 7 Name: hospital_admit_source, dtype: int64

anzics Home 204177 Other Acute Hospital 43118 Other hospital ICU 2134 Chronic Care Hospital (including nursing homes) 1960 Name: hospital_admit_source, dtype: int64

mimic Emergency Department 16526 Direct Admit 14361 Other Hospital 7246 Other 4 Acute Care/Floor 2 Name: hospital_admit_source, dtype: int64

icu_admit_source

eicu Emergency Department 64152 Operating Room 17734 Floor 13736 Direct Admit 9925 Recovery Room 5508 Acute Care/Floor 3467 Other Hospital 3389 Step-Down Unit (SDU) 2537 PACU 1315 Other ICU 581 Chest Pain Center 305 ICU 82 ICU to SDU 28 Observation 12 Other 8 Name: icu_admit_source, dtype: int64

anzics OT/Recovery 141593 Accident & Emergency 71713 Ward 35750 Other Hospital 14292 Other Hospital ICU 2196 Other ICU, same Hospital 440 Name: icu_admit_source, dtype: int64

mimic Series([], Name: icu_admit_source, dtype: int64)

orchestra Accident & Emergency 31348 OT/Recovery 17319 Ward 4474 Other Hospital 2401 Intervention room 2129 Other ICU, same Hospital 1105 Intermediate care unit 469 Other unkown 231 Home-care 217 Name: icu_admit_source, dtype: int64

nicst Series([], Name: icu_admit_source, dtype: int64)

satiq 7.0 259 10.0 104 2.0 104 1.0 65 0.0 26 11.0 11 12.0 3 9.0 2 5.0 2 14.0 1 6.0 1 4.0 1 Name: icu_admit_source, dtype: int64

icu_disch_location

eicu Floor 60655 Step-Down Unit (SDU) 16615 Acute Care/Floor 13549 Home 11358 Death 6874 Telemetry 6096 Other Hospital 2490 Other External 1779 Skilled Nursing Facility 1206 Other 920 Other ICU 520 Rehabilitation 451 Other Internal 183 Nursing Home 117 ICU 51 Operating Room 7 Other ICU (CABG) 5 Name: icu_disch_location, dtype: int64

anzics Survived ICU 242400 Died in ICU 13945 Transferred to another hospital 5205 Transferred to another ICU 3997 Name: icu_disch_location, dtype: int64

mimic Series([], Name: icu_disch_location, dtype: int64)

orchestra Survived ICU 40952 Transferred to another ICU 12334 Died in ICU 5665 Transferred to another hospital 469 Name: icu_disch_location, dtype: int64

nicst Series([], Name: icu_disch_location, dtype: int64)

satiq 1.0 490 5.0 54 2.0 30 4.0 5 Name: icu_disch_location, dtype: int64

arf_apache

eicu 0 118544 1 4349 Name: arf_apache, dtype: int64

anzics 0 253476 1 12660 Name: arf_apache, dtype: int64

mimic Series([], Name: arf_apache, dtype: int64)

orchestra Series([], Name: arf_apache, dtype: int64)

nicst Series([], Name: arf_apache, dtype: int64)

satiq N 505 S 74 Name: arf_apache, dtype: int64

alistairewj commented 7 years ago

add output of notebooks to GitHub with detail on exclusions (#/percent)
Why are there non unique patient ID in all databases?
Why are there non unique encounter ID for anzics?
Hospital death prob appears to be -1 for readmits, but readmission status is 0. What does readmission status = 0 mean?
Verify readmissions are excluded for all databases

alistairewj commented 7 years ago

current numbers on duplicate encounters

273 in ANZICS
MIMIC has repeated hospital admissions due to https://github.com/MIT-LCP/gosiss/blob/dc57a49d37ce7ab087791e8bbde654460741c343/mimic-iii/sql/cohort.sql#L68

need to decide with GOSSIS what the inclusion criteria are

is ventilated_apache == mechvent_24 (particularly for orchestra)? if so we can map orchestra's mechvent_24 to ventilated_apache
is pao2_apache == min(pao2) for first day? if so we can compare those plots

run notebooks and commit output to github so exclusion criteria are observed

alistairewj commented 7 years ago

[ ] add vitalAperiodic in the has_vit check
[ ] Columns (1,6,8,10,11,17,19,20,24,25,193) have mixed types

MIT-LCP / gossis