Assessing the Readability of Freely Available ICU Notes

Area of Science:

• Medical Informatics
• Natural Language Processing in Healthcare
• Clinical Data Analysis

Background:

• Unstructured data in electronic health records (EHRs) contain valuable clinical information.
• Accessing clinical data for research is challenging due to privacy and proprietary restrictions.
• The MIMIC-III database was created to provide accessible ICU data for research.

Purpose of the Study:

• To summarize the text characteristics and readability of clinical notes within the MIMIC-III database.
• To compare these characteristics with those of proprietary EHR notes from a previous study.
• To assess the viability of MIMIC-III for research on clinician language use.

Main Methods:

• Analysis of text characteristics (e.g., word count, vocabulary richness) of MIMIC-III ICU notes.
• Calculation of readability indices for the MIMIC-III notes.
• Comparative analysis against previously published data from proprietary EHR notes.

Main Results:

• Text characteristics of MIMIC-III notes were found to be comparable to proprietary EHR notes.
• The readability index of MIMIC-III notes was slightly lower than that of proprietary EHR notes.
• MIMIC-III notes present a viable alternative for researchers lacking access to proprietary EHR systems.

Conclusions:

• The MIMIC-III database provides a valuable, accessible resource for clinical informatics research.
• Researchers can effectively study clinician language patterns using MIMIC-III data.
• The findings support the use of MIMIC-III for studies focusing on the nuances of clinical documentation.