Handling and reporting missing data in training load and injury risk research

Area of Science:

• Sports Science
• Injury Epidemiology
• Data Science

Background:

• Handling missing data in training load and injury risk studies is crucial for accurate analysis.
• Current practices for managing missing training load data are not well-documented.
• Inconsistent methods can lead to biased results in sports injury research.

Approach:

• Conducted a systematic review of 108 studies on training load and injury risk to assess missing data handling.
• Performed simulations to compare various missing data imputation methods.
• Utilized a Norwegian Premier League football dataset (n=39) with session Rating of Perceived Exertion (sRPE) and global positioning system (GPS) data.

Key Points:

• Only 34% of reviewed studies reported on missing training load data.
• Multiple Imputation using Predicted Mean Matching demonstrated superior accuracy across different scenarios.
• This method proved effective for both internal (sRPE) and external (GPS) training load variables.

Conclusions:

• Future research must transparently report missing data extent and handling methods.
• Recommends Multiple Imputation with Predicted Mean Matching for imputing sRPE and GPS data in training load and injury risk studies.
• Emphasizes the importance of robust data handling for reliable sports injury prediction models.