[Comparison of simple and multiple imputation methods using a risk model for surgical mortality as example].

Area of Science:

• Biostatistics
• Health Research Methodology

Background:

• Missing data is a prevalent challenge in health studies.
• Data imputation techniques artificially complete datasets for analysis.
• This study evaluates imputation methods using real-world data.

Purpose of the Study:

• To compare the performance of three imputation methods.
• To assess the impact of imputation on risk model development.
• To analyze imputation effects on surgical mortality models.

Main Methods:

• Utilized data from a surgical mortality risk model study (n=450).
• Applied two single imputation methods and one multiple imputation method.
• Assumed data followed a Missing at Random (MAR) mechanism.

Main Results:

• Serum albumin exhibited a 27.1% missing rate.
• Single imputation logistic models were similar.
• Multiple imputation yielded different models regarding variable inclusion.

Conclusions:

• Albumin's relationship with other variables is crucial for model accuracy.
• Single imputation can underestimate variability and confidence intervals.
• Multiple imputation is recommended for handling missing data, especially in risk modeling, due to its consideration of imputation variability.