A rigorous algorithm to detect and clean inaccurate adult height records within EHR systems

Area of Science:

• Biomedical Informatics
• Health Data Science
• Clinical Research

Background:

• Accurate patient height is crucial for biomedical analyses, including Body Mass Index (BMI) calculations.
• Electronic Health Record (EHR) height data often requires cleaning for research use, with limited established methods.
• Standardizing height data is essential for reliable epidemiological and clinical studies.

Purpose of the Study:

• To develop and validate an algorithm for cleaning adult height measurements from EHR.
• The algorithm uses only height values and associated ages for data correction.
• To provide a freely available, open-source tool for improving EHR height data quality.

Main Methods:

• Developed an algorithm using height and age data from EHR.
• Validated the algorithm on datasets from Northwestern and Marshfield Clinic biobanks.
• Assessed algorithm performance by comparing cleaned data to observer-measured heights.

Main Results:

• The algorithm identified and corrected 1262 erroneous height values in a sample of 33937 records.
• Cleaning resulted in significant median changes: 7.6 cm for height and 2.9 kg/m² for BMI.
• 94.5% of cleaned EHR height values were within 3.5 cm of observer-measured values.

Conclusions:

• A novel, open-source algorithm effectively cleans EHR adult height data using only height and age.
• This tool enhances the usability of EHR height and BMI data for biomedical research.
• The algorithm's availability and ease of modification benefit research groups working with clinical data.