Skip to main content

Extracting a Prediction Task Cohort from MEDS Data

To easily extract prediction tasks from MEDS Data, we will use the ACES package. This package allows you to define simple configuration files that specify the inclusion/exclusion criteria for tasks you want to extract and have them be automatically extractable from MEDS data via a command line interface. See the ACES documentation for more information.

In this tutorial, we'll run an end-to-end extraction of a prediction task from the MIMIC-IV Demo dataset.

Tutorial Set-up
[10]: 
1 
2
3
4
import os
from pathlib import Path

DEMO_DIR = Path(os.getenv("MEDS_DEMO_DIR", "./demo_output"))
MIMIC-IV Demo Dataset
You can use the MIMIC_IV_MEDS package to easily download and automatically transform the MIMIC-IV-Demo dataset into MEDS:
[22]: 
1 
2
3
4
OUTPUT_DIR = DEMO_DIR / "meds/"
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)

!MEDS_extract-MIMIC_IV root_output_dir=$OUTPUT_DIR do_demo=True do_copy=True hydra/job_logging=disabled
Extracting the Prediction Task
In-ICU Mortality


For this tutorial, we'll extract a cohort for a basic in-ICU mortality prediction task. Let's first define our
task parameters.


Our goal is to predict the mortality outcome for a patient's entire ICU admission using historical patient
data plus the initial 24 hours of data after the patient was first admitted into the ICU.


Suppose we only want to consider patients whose ICU admission was longer than 48 hours. As such, patients who
died or are discharged from the ICU within 48 hours of being admitted are excluded.


Note: This task is distinct from in-hospital mortality,  30-day mortality, or imminent
mortality.
We can visualize this task as a series of inter-related windows using the below timeline:


ICU Mortality
    

  • The "blue" region represents the "input" data window. All historical patient data up to and including the first 24 hours of a patient's ICU admission will serve as input into a downstream model.
    • 

  • The "red" bar represents the ICU admission, and would "trigger" the start of our prediction task.
    • 

  • The "yellow" region represents the first 48 hours of a patient's ICU admission, and we stipulate a "gap" whereby the patient must not have died or been discharged during that period.
    • 

  • The "magenta" region represents our prediction "target" window which can be of varying length, as it ends whenever a patient dies or is discharged.
    • 

ACES Configuration File
ACES uses a configuration language to capture cohort and task definitions. Details about this configuration language is available in the ACES Documentation.


Let's walkthrough the construction of an ACES configuration file. At a minimum, the most basic configuration file would contain the predicates, trigger, and windows sections.
Predicates


To capture our task definition, we must first define at least three simple concepts, or ACES plain predicates, for our demo dataset.


For starters, as we are specifically interested in mortality "in the ICU", an ICU admission and a ICU discharge predicate would be needed to represent events where patients are officially admitted to the ICU and where patients are officially discharged. We also need the death predicate to capture death events so we can accurately capture the mortality component.


To define predicates, we would need to find how these concepts are represented in our dataset. For MIMIC-IV-Demo in MEDS format, these concepts can be found using simple regular expressions:


predicates:
  icu_admission:
    code: { regex: "^ICU_ADMISSION//.*" }
  icu_discharge:
    code: { regex: "^ICU_DISCHARGE//.*" }
  death:
    code: { regex: "MEDS_DEATH.*" }



Since patients can either die or be discharged from the ICU, we may also create a more complex concept, or an ACES derived predicate, by joining the above simple concepts using an OR relationship:


predicates:
  discharge_or_death:
    expr: or(icu_discharge, death)

Trigger


As mentioned above, a patient's admission into the ICU triggers our prediction task. A designated field defines this ACES trigger, and its value must always be one of the specified predicates. For our task, this predicate would be icu_admission:


trigger: icu_admission

Windows


The windows section contains the remaining three windows we defined previously - input, gap, and target.


For details on the configuration language syntax for windows, please see the documentation. Briefly, certain fields are present in all windows:


windows:
  window_name:
    start:
    end:
    start_inclusive:
    end_inclusive:



However, some windows also have optional parameters, such as the has field, which captures predicate count criteria for that particular window:


windows:
  window_name:
    ...
    has:
      predicate_a: (min, max)
      predicate_b: (min, max)
      ...

For our in-ICU mortality prediction task, we can define:


    

  1. 

    input, which begins at the start of a patient's record (ie., null), and ends 24 hours past trigger (ie., icu_admission). As we'd like to include the events specified at both the start and end of input, if present, we can set both start_inclusive and end_inclusive as True.
    

    Note: Since we'd like our model to make a prediction at the end of input, we can set index_timestamp to be end, which corresponds to the timestamp of trigger + 24h.
    

    windows:
  input:
    start: null
    end: trigger + 24h
    start_inclusive: True
    end_inclusive: True
    index_timestamp: end

    

    2. 

  2. 

    gap, which also begins at trigger, and ends 48 hours after. As we have included the left boundary event in trigger (ie., icu_admission), it would be reasonable to not include it again as it should not play a role in gap. As such, we set start_inclusive to False. As we'd like our ICU admission to be at least 48 hours long, we can place constraints specifying that there cannot be any additional icu_admission, icu_discharge, or death in gap.
    

    windows:
  gap:
    start: trigger
    end: start + 48h
    start_inclusive: False
    end_inclusive: True
    has:
      icu_admission: (None, 0)
      discharge_or_death: (None, 0)

    

    3. 

  3. 

    target, which begins at the end of gap, and ends at the next discharge or death event (ie., discharge_or_death predicate). We can use this arrow notation which ACES recognizes as event references (ie., -> and <-; see Time Range Fields). In our case, we end target at the next discharge_or_death. Similarly, as we included the event at the end of gap, if any, already in gap, we can set start_inclusive to False.
    

    Note: Since we'd like to make a binary mortality prediction, we can extract the death predicate as a label from target, by specifying the label field to be death.
    

    windows:
  target:
    start: gap.end
    end: start -> discharge_or_death
    start_inclusive: False
    end_inclusive: True
    label: death

    

    4. 

Now, we can put all the components together to form a complete ACES configuration file that captures everything we need for our cohort and task:


predicates:
  icu_admission:
    code: { regex: "^ICU_ADMISSION//.*" }
  icu_discharge:
    code: { regex: "^ICU_DISCHARGE//.*" }
  death:
    code: { regex: "MEDS_DEATH.*" }
  discharge_or_death:
    expr: or(icu_discharge, death)

trigger: icu_admission

windows:
  input:
    start: null
    end: trigger + 24h
    start_inclusive: True
    end_inclusive: True
    index_timestamp: end
  gap:
    start: trigger
    end: start + 48h
    start_inclusive: False
    end_inclusive: True
    has:
      icu_admission: (None, 0)
      discharge_or_death: (None, 0)
  target:
    start: gap.end
    end: start -> discharge_or_death
    start_inclusive: False
    end_inclusive: True
    label: death

End-to-End Extraction using the ACES CLI
With the configuration file ready, extracting the cohort from our demo dataset is extremely straightforward. All we need to do is run a simple command-line tool.
[3]: 
1 
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
in_icu = """
predicates:
  icu_admission:
    code: { regex: "^ICU_ADMISSION//.*" }
  icu_discharge:
    code: { regex: "^ICU_DISCHARGE//.*" }
  death:
    code: { regex: "MEDS_DEATH.*" }
  discharge_or_death:
    expr: or(icu_discharge, death)

trigger: icu_admission

windows:
  input:
    start: null
    end: trigger + 24h
    start_inclusive: True
    end_inclusive: True
    index_timestamp: end
  gap:
    start: trigger
    end: start + 48h
    start_inclusive: False
    end_inclusive: True
    has:
      icu_admission: (None, 0)
      discharge_or_death: (None, 0)
  target:
    start: gap.end
    end: start -> discharge_or_death
    start_inclusive: False
    end_inclusive: True
    label: death
"""
Let's save the final configuration file in a YAML file in our demo directory:
[4]: 
1 
2
3
4
5
6
COHORT_NAME = "in_icu"
COHORT_DIR = DEMO_DIR / "cohorts"
COHORT_DIR.mkdir(parents=True, exist_ok=True)

with open(COHORT_DIR / f"{COHORT_NAME}.yaml", "w") as f:
    f.write(in_icu)
We can now set some variables for CLI parameters. For more information on CLI arguments, please see the documentation, including instructions for using expand_shards for simultaneous extraction of cohorts over multiple MEDS shards.
[5]: 
1 
2
3
DATA_STANDARD = "meds"
DATA_ROOT = OUTPUT_DIR / "MEDS_cohort/data/"
DATA_SHARD = "$(expand_shards train/1 tuning/1 held_out/1)"
[6]: 
1 
2
3
4
5
6
7
!aces-cli \
    cohort_name=$COHORT_NAME \
    cohort_dir=$COHORT_DIR \
    data=sharded \
    data.standard=$DATA_STANDARD \
    data.root=$DATA_ROOT \
    data.shard=$DATA_SHARD -m
[2025-03-25 02:15:42,893][HYDRA] Launching 3 jobs locally
[2025-03-25 02:15:42,893][HYDRA] 	#0 : cohort_name=in_icu cohort_dir=demo_output/cohorts data=sharded data.standard=meds data.root=demo_output/meds/MEDS_cohort/data data.shard=train/0
2025-03-25 02:15:43.313 | INFO     | aces.__main__:main:149 - Loading config from 'demo_output/cohorts/in_icu.yaml'
2025-03-25 02:15:43.318 | INFO     | aces.config:load:1341 - Parsing windows...
2025-03-25 02:15:43.318 | INFO     | aces.config:load:1350 - Parsing trigger event...
2025-03-25 02:15:43.318 | INFO     | aces.config:load:1392 - Parsing predicates...
2025-03-25 02:15:43.323 | INFO     | aces.__main__:main:159 - Attempting to get predicates dataframe given:
standard: meds
ts_format: '%m/%d/%Y %H:%M'
root: demo_output/meds/MEDS_cohort/data
shard: train/0
path: ${data.root}/${data.shard}.parquet
_prefix: /${data.shard}

2025-03-25 02:15:43.324 | INFO     | aces.predicates:generate_plain_predicates_from_meds:269 - Loading MEDS data...
2025-03-25 02:15:43.367 | INFO     | aces.predicates:generate_plain_predicates_from_meds:273 - Generating plain predicate columns...
2025-03-25 02:15:43.395 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_admission'.
2025-03-25 02:15:43.424 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_discharge'.
2025-03-25 02:15:43.464 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'death'.
2025-03-25 02:15:43.464 | INFO     | aces.predicates:generate_plain_predicates_from_meds:280 - Cleaning up predicates dataframe...
2025-03-25 02:15:43.505 | INFO     | aces.predicates:get_predicates_df:703 - Loaded plain predicates. Generating derived predicate columns...
2025-03-25 02:15:43.506 | INFO     | aces.predicates:get_predicates_df:717 - Added predicate column 'discharge_or_death'.
2025-03-25 02:15:43.506 | INFO     | aces.predicates:get_predicates_df:724 - Generating special predicate columns...
2025-03-25 02:15:43.506 | INFO     | aces.query:query:76 - Checking if '(subject_id, timestamp)' columns are unique...
2025-03-25 02:15:43.509 | INFO     | aces.utils:log_tree:67 - 
trigger
┣━━ input.end
┃   ┗━━ input.start
┗━━ gap.end
    ┗━━ target.end

2025-03-25 02:15:43.510 | INFO     | aces.query:query:85 - Beginning query...
2025-03-25 02:15:43.510 | INFO     | aces.query:query:92 - No static variable criteria specified, removing all rows with null timestamps...
2025-03-25 02:15:43.511 | INFO     | aces.query:query:99 - Identifying possible trigger nodes based on the specified trigger event...
2025-03-25 02:15:43.511 | INFO     | aces.constraints:check_constraints:110 - Excluding 72,774 rows as they failed to satisfy '1 <= icu_admission <= None'.
2025-03-25 02:15:43.512 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.end'...
2025-03-25 02:15:43.596 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.start'...
2025-03-25 02:15:43.707 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'gap.end'...
2025-03-25 02:15:43.770 | INFO     | aces.constraints:check_constraints:110 - Excluding 2 rows as they failed to satisfy 'None <= icu_admission <= 0'.
2025-03-25 02:15:43.771 | INFO     | aces.constraints:check_constraints:110 - Excluding 53 rows as they failed to satisfy 'None <= discharge_or_death <= 0'.
2025-03-25 02:15:43.772 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'target.end'...
2025-03-25 02:15:43.858 | INFO     | aces.query:query:114 - Done. 60 valid rows returned corresponding to 47 subjects.
2025-03-25 02:15:43.858 | INFO     | aces.query:query:129 - Extracting label 'death' from window 'target'...
2025-03-25 02:15:43.858 | INFO     | aces.query:query:150 - Setting index timestamp as 'end' of window 'input'...
2025-03-25 02:15:43.862 | WARNING  | aces.__main__:get_and_validate_label_schema:114 - Output contains columns that are not valid MEDS label columns. For now, we are dropping them.
If you need these columns, please comment on https://github.com/justin13601/ACES/issues/97
Columns:
  - trigger
  - input.end_summary
  - input.start_summary
  - gap.end_summary
  - target.end_summary
2025-03-25 02:15:43.870 | INFO     | aces.__main__:main:191 - Completed in 0:00:00.555944. Results saved to 'demo_output/cohorts/in_icu/train/0.parquet'.
[2025-03-25 02:15:43,871][HYDRA] 	#1 : cohort_name=in_icu cohort_dir=demo_output/cohorts data=sharded data.standard=meds data.root=demo_output/meds/MEDS_cohort/data data.shard=tuning/0
2025-03-25 02:15:44.054 | INFO     | aces.__main__:main:149 - Loading config from 'demo_output/cohorts/in_icu.yaml'
2025-03-25 02:15:44.058 | INFO     | aces.config:load:1341 - Parsing windows...
2025-03-25 02:15:44.059 | INFO     | aces.config:load:1350 - Parsing trigger event...
2025-03-25 02:15:44.059 | INFO     | aces.config:load:1392 - Parsing predicates...
2025-03-25 02:15:44.060 | INFO     | aces.__main__:main:159 - Attempting to get predicates dataframe given:
standard: meds
ts_format: '%m/%d/%Y %H:%M'
root: demo_output/meds/MEDS_cohort/data
shard: tuning/0
path: ${data.root}/${data.shard}.parquet
_prefix: /${data.shard}

2025-03-25 02:15:44.060 | INFO     | aces.predicates:generate_plain_predicates_from_meds:269 - Loading MEDS data...
2025-03-25 02:15:44.064 | INFO     | aces.predicates:generate_plain_predicates_from_meds:273 - Generating plain predicate columns...
2025-03-25 02:15:44.067 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_admission'.
2025-03-25 02:15:44.070 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_discharge'.
2025-03-25 02:15:44.074 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'death'.
2025-03-25 02:15:44.074 | INFO     | aces.predicates:generate_plain_predicates_from_meds:280 - Cleaning up predicates dataframe...
2025-03-25 02:15:44.080 | INFO     | aces.predicates:get_predicates_df:703 - Loaded plain predicates. Generating derived predicate columns...
2025-03-25 02:15:44.081 | INFO     | aces.predicates:get_predicates_df:717 - Added predicate column 'discharge_or_death'.
2025-03-25 02:15:44.081 | INFO     | aces.predicates:get_predicates_df:724 - Generating special predicate columns...
2025-03-25 02:15:44.081 | INFO     | aces.query:query:76 - Checking if '(subject_id, timestamp)' columns are unique...
2025-03-25 02:15:44.082 | INFO     | aces.utils:log_tree:67 - 
trigger
┣━━ input.end
┃   ┗━━ input.start
┗━━ gap.end
    ┗━━ target.end

2025-03-25 02:15:44.082 | INFO     | aces.query:query:85 - Beginning query...
2025-03-25 02:15:44.083 | INFO     | aces.query:query:92 - No static variable criteria specified, removing all rows with null timestamps...
2025-03-25 02:15:44.083 | INFO     | aces.query:query:99 - Identifying possible trigger nodes based on the specified trigger event...
2025-03-25 02:15:44.083 | INFO     | aces.constraints:check_constraints:110 - Excluding 6,242 rows as they failed to satisfy '1 <= icu_admission <= None'.
2025-03-25 02:15:44.084 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.end'...
2025-03-25 02:15:44.097 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.start'...
2025-03-25 02:15:44.123 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'gap.end'...
2025-03-25 02:15:44.136 | INFO     | aces.constraints:check_constraints:110 - Excluding 0 rows as they failed to satisfy 'None <= icu_admission <= 0'.
2025-03-25 02:15:44.136 | INFO     | aces.constraints:check_constraints:110 - Excluding 7 rows as they failed to satisfy 'None <= discharge_or_death <= 0'.
2025-03-25 02:15:44.137 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'target.end'...
2025-03-25 02:15:44.165 | INFO     | aces.query:query:114 - Done. 8 valid rows returned corresponding to 5 subjects.
2025-03-25 02:15:44.165 | INFO     | aces.query:query:129 - Extracting label 'death' from window 'target'...
2025-03-25 02:15:44.166 | INFO     | aces.query:query:150 - Setting index timestamp as 'end' of window 'input'...
2025-03-25 02:15:44.168 | WARNING  | aces.__main__:get_and_validate_label_schema:114 - Output contains columns that are not valid MEDS label columns. For now, we are dropping them.
If you need these columns, please comment on https://github.com/justin13601/ACES/issues/97
Columns:
  - trigger
  - input.end_summary
  - input.start_summary
  - gap.end_summary
  - target.end_summary
2025-03-25 02:15:44.173 | INFO     | aces.__main__:main:191 - Completed in 0:00:00.117831. Results saved to 'demo_output/cohorts/in_icu/tuning/0.parquet'.
[2025-03-25 02:15:44,174][HYDRA] 	#2 : cohort_name=in_icu cohort_dir=demo_output/cohorts data=sharded data.standard=meds data.root=demo_output/meds/MEDS_cohort/data data.shard=held_out/0
2025-03-25 02:15:44.267 | INFO     | aces.__main__:main:149 - Loading config from 'demo_output/cohorts/in_icu.yaml'
2025-03-25 02:15:44.271 | INFO     | aces.config:load:1341 - Parsing windows...
2025-03-25 02:15:44.271 | INFO     | aces.config:load:1350 - Parsing trigger event...
2025-03-25 02:15:44.271 | INFO     | aces.config:load:1392 - Parsing predicates...
2025-03-25 02:15:44.272 | INFO     | aces.__main__:main:159 - Attempting to get predicates dataframe given:
standard: meds
ts_format: '%m/%d/%Y %H:%M'
root: demo_output/meds/MEDS_cohort/data
shard: held_out/0
path: ${data.root}/${data.shard}.parquet
_prefix: /${data.shard}

2025-03-25 02:15:44.273 | INFO     | aces.predicates:generate_plain_predicates_from_meds:269 - Loading MEDS data...
2025-03-25 02:15:44.275 | INFO     | aces.predicates:generate_plain_predicates_from_meds:273 - Generating plain predicate columns...
2025-03-25 02:15:44.277 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_admission'.
2025-03-25 02:15:44.279 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'icu_discharge'.
2025-03-25 02:15:44.281 | INFO     | aces.predicates:generate_plain_predicates_from_meds:277 - Added predicate column 'death'.
2025-03-25 02:15:44.282 | INFO     | aces.predicates:generate_plain_predicates_from_meds:280 - Cleaning up predicates dataframe...
2025-03-25 02:15:44.285 | INFO     | aces.predicates:get_predicates_df:703 - Loaded plain predicates. Generating derived predicate columns...
2025-03-25 02:15:44.286 | INFO     | aces.predicates:get_predicates_df:717 - Added predicate column 'discharge_or_death'.
2025-03-25 02:15:44.286 | INFO     | aces.predicates:get_predicates_df:724 - Generating special predicate columns...
2025-03-25 02:15:44.286 | INFO     | aces.query:query:76 - Checking if '(subject_id, timestamp)' columns are unique...
2025-03-25 02:15:44.287 | INFO     | aces.utils:log_tree:67 - 
trigger
┣━━ input.end
┃   ┗━━ input.start
┗━━ gap.end
    ┗━━ target.end

2025-03-25 02:15:44.287 | INFO     | aces.query:query:85 - Beginning query...
2025-03-25 02:15:44.288 | INFO     | aces.query:query:92 - No static variable criteria specified, removing all rows with null timestamps...
2025-03-25 02:15:44.288 | INFO     | aces.query:query:99 - Identifying possible trigger nodes based on the specified trigger event...
2025-03-25 02:15:44.288 | INFO     | aces.constraints:check_constraints:110 - Excluding 4,163 rows as they failed to satisfy '1 <= icu_admission <= None'.
2025-03-25 02:15:44.289 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.end'...
2025-03-25 02:15:44.298 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'input.start'...
2025-03-25 02:15:44.322 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'gap.end'...
2025-03-25 02:15:44.333 | INFO     | aces.constraints:check_constraints:110 - Excluding 0 rows as they failed to satisfy 'None <= icu_admission <= 0'.
2025-03-25 02:15:44.333 | INFO     | aces.constraints:check_constraints:110 - Excluding 6 rows as they failed to satisfy 'None <= discharge_or_death <= 0'.
2025-03-25 02:15:44.334 | INFO     | aces.extract_subtree:extract_subtree:252 - Summarizing subtree rooted at 'target.end'...
2025-03-25 02:15:44.359 | INFO     | aces.query:query:114 - Done. 6 valid rows returned corresponding to 4 subjects.
2025-03-25 02:15:44.359 | INFO     | aces.query:query:129 - Extracting label 'death' from window 'target'...
2025-03-25 02:15:44.360 | WARNING  | aces.query:query:142 - All labels in the extracted cohort are the same: '0'. This may indicate an issue with the task logic. Please double-check your configuration file if this is not expected.
2025-03-25 02:15:44.360 | INFO     | aces.query:query:150 - Setting index timestamp as 'end' of window 'input'...
2025-03-25 02:15:44.363 | WARNING  | aces.__main__:get_and_validate_label_schema:114 - Output contains columns that are not valid MEDS label columns. For now, we are dropping them.
If you need these columns, please comment on https://github.com/justin13601/ACES/issues/97
Columns:
  - trigger
  - input.end_summary
  - input.start_summary
  - gap.end_summary
  - target.end_summary
2025-03-25 02:15:44.367 | INFO     | aces.__main__:main:191 - Completed in 0:00:00.099429. Results saved to 'demo_output/cohorts/in_icu/held_out/0.parquet'.

Inspecting Results
The CLI would output a parquet file for every shard in each corresponding data split. Let's examine the output for the first shard of the train split (this demo dataset only has one shard in the train split):
[7]: 
1 
2
import pandas as pd
results = pd.read_parquet(COHORT_DIR / COHORT_NAME / "train" / "0.parquet")
Here is the label distribution of the patients that met our cohort criteria:
[8]: 
1 
results['boolean_value'].value_counts()
[8]: 
boolean_value
False    52
True      8
Name: count, dtype: int64
Recall that our label predicate was set to death in our configuration file. Thus, we can interpret this as 8 patients who died, and 52 patients who were discharged from the ICU. The full results provide the exact subject_id for all patients that meet our cohort criteria, and prediction_time corresponds to the index_timestamp defined in our configuration file.
[9]: 
1 
results
[9]: 




  
    

      
      subject_id
      prediction_time
      boolean_value
      integer_value
      float_value
      categorical_value
    

  
  
    

      0
      10002428
      2156-04-13 16:24:18
      False
      NaN
      NaN
      None
    

    

      1
      10002428
      2156-04-20 18:11:19
      False
      NaN
      NaN
      None
    

    

      2
      10002428
      2156-05-01 21:53:00
      False
      NaN
      NaN
      None
    

    

      3
      10002428
      2156-05-12 14:49:34
      False
      NaN
      NaN
      None
    

    

      4
      10002495
      2141-05-23 20:18:01
      False
      NaN
      NaN
      None
    

    

      5
      10003400
      2137-02-26 23:37:19
      False
      NaN
      NaN
      None
    

    

      6
      10003400
      2137-08-11 19:54:51
      False
      NaN
      NaN
      None
    

    

      7
      10003400
      2137-08-18 17:36:37
      True
      NaN
      NaN
      None
    

    

      8
      10004235
      2196-02-25 17:07:00
      False
      NaN
      NaN
      None
    

    

      9
      10004422
      2111-01-18 09:44:50
      False
      NaN
      NaN
      None
    

    

      10
      10004720
      2186-11-13 19:55:00
      True
      NaN
      NaN
      None
    

    

      11
      10004733
      2174-12-05 11:28:24
      False
      NaN
      NaN
      None
    

    

      12
      10005817
      2132-12-16 09:29:01
      False
      NaN
      NaN
      None
    

    

      13
      10005817
      2135-01-04 21:55:32
      True
      NaN
      NaN
      None
    

    

      14
      10005866
      2149-10-03 12:48:08
      False
      NaN
      NaN
      None
    

    

      15
      10005909
      2144-10-30 23:09:03
      False
      NaN
      NaN
      None
    

    

      16
      10007058
      2167-11-08 20:22:00
      False
      NaN
      NaN
      None
    

    

      17
      10007818
      2146-06-23 11:46:29
      True
      NaN
      NaN
      None
    

    

      18
      10007928
      2129-04-07 00:25:00
      False
      NaN
      NaN
      None
    

    

      19
      10008454
      2110-12-01 17:11:36
      False
      NaN
      NaN
      None
    

    

      20
      10009628
      2153-09-20 09:54:49
      False
      NaN
      NaN
      None
    

    

      21
      10010471
      2155-12-03 20:33:00
      True
      NaN
      NaN
      None
    

    

      22
      10010867
      2147-12-31 09:33:00
      False
      NaN
      NaN
      None
    

    

      23
      10012552
      2140-03-26 14:37:26
      False
      NaN
      NaN
      None
    

    

      24
      10014078
      2166-08-23 00:36:00
      False
      NaN
      NaN
      None
    

    

      25
      10014354
      2148-07-08 15:48:09
      False
      NaN
      NaN
      None
    

    

      26
      10015272
      2137-06-13 18:37:22
      False
      NaN
      NaN
      None
    

    

      27
      10015931
      2177-03-25 21:48:07
      True
      NaN
      NaN
      None
    

    

      28
      10018081
      2133-12-19 17:10:00
      False
      NaN
      NaN
      None
    

    

      29
      10018081
      2134-08-06 14:53:33
      False
      NaN
      NaN
      None
    

    

      30
      10018328
      2154-04-25 23:03:44
      False
      NaN
      NaN
      None
    

    

      31
      10019003
      2153-03-29 02:21:00
      False
      NaN
      NaN
      None
    

    

      32
      10019003
      2153-04-14 19:45:30
      False
      NaN
      NaN
      None
    

    

      33
      10019003
      2155-07-11 17:48:57
      False
      NaN
      NaN
      None
    

    

      34
      10020187
      2169-01-16 04:56:00
      False
      NaN
      NaN
      None
    

    

      35
      10020306
      2135-01-22 17:01:57
      False
      NaN
      NaN
      None
    

    

      36
      10020640
      2153-02-14 01:38:00
      False
      NaN
      NaN
      None
    

    

      37
      10020944
      2131-02-28 16:40:00
      False
      NaN
      NaN
      None
    

    

      38
      10021487
      2116-12-04 01:02:00
      False
      NaN
      NaN
      None
    

    

      39
      10022017
      2189-09-11 10:05:24
      False
      NaN
      NaN
      None
    

    

      40
      10023117
      2171-11-15 10:06:41
      False
      NaN
      NaN
      None
    

    

      41
      10023117
      2175-03-22 03:20:53
      False
      NaN
      NaN
      None
    

    

      42
      10023117
      2175-07-07 17:41:00
      False
      NaN
      NaN
      None
    

    

      43
      10023239
      2137-06-20 19:09:00
      False
      NaN
      NaN
      None
    

    

      44
      10023239
      2140-10-04 09:07:56
      False
      NaN
      NaN
      None
    

    

      45
      10024043
      2117-04-12 22:05:00
      False
      NaN
      NaN
      None
    

    

      46
      10025612
      2125-09-26 13:23:24
      False
      NaN
      NaN
      None
    

    

      47
      10027445
      2142-08-01 01:41:00
      False
      NaN
      NaN
      None
    

    

      48
      10027602
      2201-10-31 12:25:00
      False
      NaN
      NaN
      None
    

    

      49
      10029291
      2123-02-21 04:13:00
      False
      NaN
      NaN
      None
    

    

      50
      10029291
      2123-02-27 12:12:32
      False
      NaN
      NaN
      None
    

    

      51
      10031757
      2137-10-16 17:29:21
      False
      NaN
      NaN
      None
    

    

      52
      10032725
      2143-03-23 06:42:00
      False
      NaN
      NaN
      None
    

    

      53
      10035631
      2116-02-29 18:43:20
      False
      NaN
      NaN
      None
    

    

      54
      10037861
      2117-03-15 16:34:58
      True
      NaN
      NaN
      None
    

    

      55
      10037975
      2185-01-18 19:12:12
      True
      NaN
      NaN
      None
    

    

      56
      10038933
      2148-09-11 13:19:00
      False
      NaN
      NaN
      None
    

    

      57
      10038999
      2131-05-23 21:50:33
      False
      NaN
      NaN
      None
    

    

      58
      10039708
      2140-01-24 18:08:00
      False
      NaN
      NaN
      None
    

    

      59
      10040025
      2148-01-25 04:50:17
      False
      NaN
      NaN
      None