GCS refactor

mimic-iv/concepts/measurement/gcs.sql

@@ -105,26 +105,6 @@ with base as
   , EndoTrachFlag
   from gcs gs
 )
--- priority is:
---  (i) complete data, (ii) non-sedated GCS, (iii) lowest GCS, (iv) charttime
-, gcs_priority as
-(
-  select
-      subject_id
-    , stay_id
-    , charttime
-    , gcs
-    , gcsmotor
-    , gcsverbal
-    , gcseyes
-    , EndoTrachFlag
-    , ROW_NUMBER() over
-      (
-        PARTITION BY stay_id, charttime
-        ORDER BY components_measured DESC, endotrachflag, gcs, charttime DESC
-      ) as rn
-  from gcs_stg
-)
 select
   gs.subject_id
   , gs.stay_id
@@ -134,6 +114,5 @@ select
   , GCSVerbal AS gcs_verbal
   , GCSEyes AS gcs_eyes
   , EndoTrachFlag AS gcs_unable
-from gcs_priority gs
-where rn = 1
+from gcs_stg gs
 ;

mimic-iv/tests/test_measurement.py

@@ -69,3 +69,64 @@ def test_common_bg_exist(dataset, project_id):
             missing_observations[c] = df.loc[0, c]
 
     assert len(missing_observations) == 0, f'columns in bg missing data'
+
+
+def test_gcs_score_calculated_correctly(dataset, project_id):
+    """Verifies common blood gases occur > 50% of the time"""
+    # has verbal prev of 1 -> 11365767, 30015010, 2154-07-25T20:00:00
+    # has verbal prev of 0 -> 13182319, 30159144, 2161-06-19T00:16:00
+    query = f"""
+    SELECT g.stay_id
+    , g.charttime
+    , g.gcs
+    , g.gcs_motor
+    , g.gcs_verbal
+    , g.gcs_eyes
+    , g.gcs_unable
+    , gcs_v.valuenum AS gcs_verbal_numeric
+    , gcs_v.value AS gcs_verbal_text
+    FROM  {dataset}.gcs g
+    LEFT JOIN (
+        SELECT stay_id, charttime, value, valuenum
+        FROM `physionet-data.mimiciv_icu.chartevents`
+        WHERE itemid = 223900 AND stay_id IN (30015010, 30159144)
+    ) gcs_v
+        ON g.stay_id = gcs_v.stay_id
+        AND g.charttime = gcs_v.charttime
+    WHERE g.stay_id IN
+    (
+        30015010, -- subject_id: 11365767
+        30159144 -- subject_id: 13182319
+    )
+    """
+    df = gbq.read_gbq(query, project_id=project_id, dialect="standard")
+    df = df.sort_values(['stay_id', 'charttime'])
+    df['charttime'] = pd.to_datetime(df['charttime'])
+    df['charttime_lag'] = df.groupby('stay_id')['charttime'].shift(1)
+    df['gcs_verbal_text_lag'] = df.groupby('stay_id')['gcs_verbal_text'].shift(1)
+    df['gcs_verbal_numeric_lag'] = df.groupby('stay_id')['gcs_verbal_numeric'].shift(1)
+
+    idxTime = (df['charttime'] - df['charttime_lag']).astype('timedelta64[h]') <= 6
+    # remove verbal value if occurring more than 6 hr later
+    df.loc[~idxTime, 'gcs_verbal_text_lag'] = None
+    df.loc[~idxTime, 'gcs_verbal_numeric_lag'] = None
+
+    # verify GCS logic:
+    # (1) verbal score is correctly carried forward if 0
+    # (2) verbal score is imputed at 5 if nothing to carry forward
+
+    # verbal score for this row is "unable"
+    idxETT = (df['gcs_verbal_text'] == 'No Response-ETT')
+    # and the previous row was not
+    idxETT &= (df['gcs_verbal_text_lag'] != 'No Response-ETT')
+
+    assert idxETT.sum() > 0, 'expected rows with gcs imputed, check stay_id/subject_id data'
+    assert (df.loc[idxETT, 'gcs_verbal'] > 0).all(), 'expected no rows with verbal of 0 when prev val available'
+
+    # verbal score for this row is "unable"
+    idxETT = (df['gcs_verbal_text'] == 'No Response-ETT')
+    # and the previous row was not
+    idxETT &= (df['gcs_verbal_text_lag'].isnull())
+
+    assert idxETT.sum() > 0, 'expected rows without prior GCS in 6 hours'
+    assert (df.loc[idxETT, 'gcs_verbal'] == 0).all(), 'found rows without imputed verbal score'