An Exploratory Classification Framework for Poststroke Dysphagia Severity Using Administrative Data: Administrative...

Area of Science:

• Neurology
• Medical Informatics
• Data Science

Background:

• Accurate classification of big data is vital for poststroke dysphagia research and healthcare decisions.
• Current medical coding lacks an administrative tool for classifying poststroke dysphagia severity.
• A novel classification framework for poststroke dysphagia severity was needed.

Purpose of the Study:

• To develop a novel classification framework for poststroke dysphagia severity.
• To utilize administrative data for classifying dysphagia severity in stroke patients.
• To explore the relationship between stroke severity and dysphagia severity.

Main Methods:

• Utilized data from 445 acute ischemic stroke patients with dysphagia from the 2017 Medicare 5% Limited Data Set.
• Employed unsupervised k-means clustering on dysphagia indicators derived from International Classification of Diseases, Tenth Revision (ICD-10) codes.
• Validated cluster stability by applying the algorithm to three random 60% data samples.

Main Results:

• A three-cluster algorithm was developed, categorizing dysphagia severity into mild, moderate, and severe.
• Analysis revealed that the identified dysphagia severity clusters were not analogous to stroke severity categories.
• The classification framework demonstrated stability across multiple data samples.

Conclusions:

• A novel administrative data-driven framework for classifying poststroke dysphagia severity has been developed.
• A significant discordance was observed between stroke severity and dysphagia severity.
• The findings have implications for classification methodologies in administrative poststroke dysphagia research and suggest the need for future validation studies.