Studying missingness in spinal cord injury data: challenges and impact of data imputation

Medical research often struggles with missing data, especially in studies of rare conditions like spinal cord injury (SCI). Researchers commonly address this by removing patients with incomplete data, but this can lead to biased results. This study investigates the impact of different ways of handling missing data in SCI research to provide guidelines for future studies. The study uses data from the Sygen clinical trial to simulate missing data and test different imputation methods (ways to fill in the gaps). They look at how the type of missing data, the pattern of missingness, and the imputation strategy affect the results. The research shows that simply replacing missing values with the average (mean imputation) can significantly distort the results. For repeated measurements, carrying the last known value forward is often a better approach. The study emphasizes that a one-size-fits-all approach to data imputation doesn't work for SCI data.

• 1
  Mean imputation can lead to results that strongly deviate from expected results.
• 2
  For repeated measures missing at late stages, carrying the last observation forward is a preferable imputation option.
• 3
  A one-size-fits-all imputation strategy falls short in SCI datasets; data-tailored imputation strategies are required.