5.1 Quick Start

There really is no quick start to exploring the data in MIMIC II DB. First you need to decide which type of data you want to process:

1. `waveform' data
2. high (temporal) resolution trend data (numerics)
3. `clinical' data - low temporal resolution data and/or categorical data.

You may also want to decide if you want to use all of the available patients, or just narrow your focus on a subset cohort. For example, you may wish to concentrate on patients who were admitted with a particular diagnosis, or who were administered with a particular drug. In order to find patient cohorts, you may wish to use our prototype search engine, which can be found at:
http://mimic.physionet.org.

Currently, with this interface you can search in the following ways:

• Locate free text strings the nursing progress notes and discharge summaries.
• Locate patients with particular combinations of waveforms.
• Locate patients with particular demographics (age and gender).

Each of these queries can be 'ANDed' together to construct more restrictive subsets.

Once you have found your subset of patients on which you want to work, you will want to download the data you wish to work on, or perform more detailed queries. Methods for doing this are described in chapter 3. Sample queries can be found in chapter 4. We strongly suggest you work through these before attempting to construct patient cohorts and analyze data. We also strongly advise that you take note of the different definitions of patient identifiers, as described in section 1.4.2. Note that you can analyze patients by individual (subject) ID, hospital admission, ICU stay, or by waveform Case ID. We also suggest that you read the frequently asked questions (FAQs) below.