Identifying and categorizing spurious weight data in electronic medical records

Area of Science:

• Medical Informatics
• Data Quality Management
• Electronic Health Records

Background:

• Spurious weight measurements can invalidate summary statistics like averages and trends.
• Even infrequent errors in weight records can significantly reduce the usefulness of electronic medical record (EMR) data.

Purpose of the Study:

• To estimate the prevalence of spurious weight values within a large EMR system.
• To identify potential causes of erroneous weight data.
• To develop and validate simple algorithms for detecting spurious weight entries.

Main Methods:

• Analysis of weight change over time intervals (1-3000 days) in EMR data from 10,000 randomly selected patients aged 65+ in the VA system.
• Development of three algorithms to categorize weight change and identify likely spurious entries based on data distributions.
• Manual review and categorization of identified spurious weight errors.

Main Results:

• Approximately 0.8% of all recorded weights were identified as spurious.
• Around 20% of patient charts contained at least one spurious weight value.
• The most frequent error type was single-digit misentry (e.g., 148 instead of 178).

Conclusions:

• Spurious weight data is a prevalent issue in EMRs.
• The developed algorithms effectively identify and can help remove spurious weight data.
• Implementing these algorithms can enhance the overall reliability and utility of EMR data.