Multisource and temporal variability in Portuguese hospital administrative datasets: Data quality implications

Area of Science:

• Health Informatics
• Data Quality Assessment
• Healthcare Data Analysis

Background:

• Healthcare datasets exhibit variability, potentially compromising data quality and reusability.
• Lack of standardized methods and reference datasets hinders variability assessment.
• This study addresses the need for systematic approaches to identify data quality issues in large hospital databases.

Purpose of the Study:

• To describe a process for discovering data quality implications in healthcare data.
• To apply methods for assessing variability between sources and over time in a large hospital database.
• To enhance the credibility and reusability of healthcare data.

Main Methods:

• Applied multisource and temporal variability assessment methods to a Portuguese hospitalization database.
• Utilized Clinical Classification Software (CCS) for condition-specific admissions.
• Employed Statistical Process Control (SPC) with funnel plots for multisource variability and temporal heat maps/Information-Geometric Temporal (IGT) plots for temporal variability.

Main Results:

• Detected outlying hospitals with significantly different standardized hospitalization ratios (SHR).
• Found that adjusting SHR for hospital characteristics impacted multisource variability.
• Identified abrupt temporal changes in data distributions coinciding with coding system transitions (ICD-10-CM) and software updates (DRGs).

Conclusions:

• Successfully applied reproducible and generalizable methods for healthcare data variability assessment.
• Highlighted the importance of controlling for hospital characteristics and case-mix when estimating SHR.
• Demonstrated the utility of SPC, funnel plots, heat maps, and IGT plots for detecting abnormal patterns and data quality insights in healthcare data.