Anomaly Detection and Correction in Dense Functional Data Within Electronic Medical Records.

Area of Science:

• Medical Informatics
• Data Science
• Statistical Modeling

Background:

• Accurate electronic medical record (EMR) data is crucial for medical research, especially with dense functional data.
• Anomalies in EMRs, often due to human error, can compromise research integrity and increase with data volume.
• Existing anomaly detection methods are underdeveloped for medical applications.

Purpose of the Study:

• To introduce a novel tool for identifying and correcting anomalies in dense functional EMR data.
• To enhance the reliability and quality of medical data analysis by addressing data errors.
• To provide a practical solution for anomaly detection in medical research settings.

Main Methods:

• Utilizes studentized residuals from a mean-shift model, assuming smooth functional data trajectories.
• Employs a conservative approach to focus on actual data collection errors.
• Controls for false discovery rates and type II errors to ensure precision.

Main Results:

• Validated methodology through rigorous simulation studies and real-world applications.
• Demonstrated accurate identification and correction of errors in dense functional EMR data.
• Confirmed enhancement of medical data analysis reliability and quality.

Conclusions:

• The novel tool effectively identifies and corrects anomalies in dense functional EMR data.
• The R package facilitates widespread implementation and application in diverse medical research settings.
• Improved data accuracy leads to more reliable and trustworthy medical research outcomes.