Which patients have missing data? An analysis of missingness in a trauma registry

Area of Science:

• Traumatology
• Data Science
• Medical Informatics

Background:

• Trauma registry data frequently suffer from missing information, potentially biasing analyses.
• Multiple imputation is a statistical method to address missing data, but improving data capture is the ideal solution.
• This study aimed to identify patient characteristics associated with incomplete data in trauma registries.

Purpose of the Study:

• To identify patient types most likely to have incomplete data in a regional trauma registry.
• To understand predictors of missing physiological variables crucial for trauma outcome analysis.
• To inform strategies for enhancing data quality in trauma registries.

Main Methods:

• Analysis of prospectively collected regional trauma registry data over one year.
• Multiple imputation techniques were used to estimate complete data.
• Logistic regression identified predictors of missingness for key variables like respiratory rate, GCS, and systolic blood pressure.

Main Results:

• Respiratory rate, Glasgow Coma Scale (GCS), GCS Qualifier, and systolic blood pressure were the most frequent variables with missing observations.
• Missing GCS and respiratory rate were associated with abnormal patient status and intubation (Qualifier).
• Death in hospital, abnormal pre-hospital GCS, severe chest injury, and inter-hospital transfer predicted missing data.

Conclusions:

• Death in hospital is the primary predictor of missing physiological data in trauma registries.
• Abnormal GCS and respiratory rate values are more likely to be missing.
• Improving data collection for severely injured patients, especially those intubated or with severe injuries, is crucial for data quality.