Handling missing data in a rheumatoid arthritis registry using random forest approach.

Area of Science:

• Clinical Epidemiology
• Biostatistics
• Data Science

Background:

• Missing data in clinical research can compromise study integrity by violating the intention-to-treat principle and introducing bias.
• Accurate imputation of missing patient data is crucial for reliable statistical analysis and valid research conclusions.

Purpose of the Study:

• To estimate and impute missing values in patient records from the Kuwait Registry for Rheumatic Diseases.
• To evaluate the accuracy of different imputation methods for handling missing data in rheumatoid arthritis research.

Main Methods:

• Utilized data from 1735 rheumatoid arthritis patients.
• Implemented and compared multiple imputation techniques, selecting the best based on the lowest Root Mean Square Error (RMSE).
• Evaluated the sequential random forest method (missForest) for its imputation accuracy across various missing data rates (10%, 20%, 30%).

Main Results:

• Missing data ranged from 5% to 65.5% across patient records.
• The sequential random forest method (missForest) demonstrated high accuracy in estimating missing values.
• missForest achieved the lowest imputation error for both continuous and categorical variables, outperforming other methods.

Conclusions:

• Sequential random forest (missForest) is a highly accurate method for imputing missing data in rheumatoid arthritis patient records.
• This imputation approach minimizes prediction error, enhancing the reliability of analyses using imputed laboratory values.
• Accurate data imputation is vital for robust clinical epidemiological research.