Missing Data in Surgical Data Sets: A Review of Pertinent Issues and Solutions

Area of Science:

• Biostatistics
• Data Science
• Medical Research

Background:

• Incomplete data is a pervasive challenge in research studies, impacting data analysis and results.
• Effective methods for handling missing observations are essential for maintaining statistical power and data integrity.
• Disseminating these statistical methodologies to the broader research community is an ongoing necessity.

Purpose of the Study:

• To elucidate fundamental principles of missing data in research.
• To identify and explain practical, commonly employed adjustment methods for surgical datasets.
• To compare the performance of complete case analysis, single imputation (SI), and multiple imputation (MI) using an example dataset.

Main Methods:

• Description of basic principles of missing data.
• Identification of practical adjustment methods for surgical data.
• Comparative analysis of complete case analysis, SI, and MI models using an example dataset.
• Guidance on conducting MI using Stata IC.

Main Results:

• Complete case analysis showed the greatest differences in odds ratios compared to SI and MI, indicating a significant impact on parameter estimates due to reduced statistical power.
• Odds ratio estimates from SI and MI methods were generally similar.
• SI tended to overestimate effect sizes compared to MI in certain instances.
• The study found clear indications favoring the use of MI over SI.

Conclusions:

• Robust imputation methods, particularly multiple imputation (MI), are recommended for handling missing observations in research.
• MI provides more reliable parameter estimates and preserves statistical power better than complete case analysis.
• Encouraging the adoption of advanced imputation techniques like MI can enhance the quality and validity of research findings.