mimic

Imported from mkurman/zorai (skills/scientific-skills/mimic/SKILL.md). Install upstream with npx skills add mkurman/zorai --skill mimic. Copyright stays with the author.

Overview

MIMIC (Medical Information Mart for Intensive Care) provides ICU data: vitals, labs, medications, notes, diagnoses. Tools for querying MIMIC-III/IV, building ML features, and reproducing clinical benchmarks.

Access

Apply for access at https://physionet.org/content/mimiciv/ -- requires CITI data use training.

Installation

uv pip install psycopg2 pandas

Python Analysis

import pandas as pd
from sqlalchemy import create_engine

engine = create_engine("postgresql://user:pass@localhost:5432/mimiciv")

# First 24h vitals
query = """
SELECT subject_id, charttime, valuenum
FROM mimiciv_icu.chartevents
WHERE itemid = 220045
AND valuenum IS NOT NULL
LIMIT 100
"""
hr = pd.read_sql(query, engine)

Workflow

1. Apply for MIMIC access (physionet.org)
2. Load MIMIC-IV into PostgreSQL
3. Query ICU stays, diagnoses, labs, medications, notes
4. Extract features (vitals over time, labs at admission, comorbidity scores)
5. Build ML benchmarks (in-hospital mortality, LOS prediction, sepsis detection)