Clustered and missing data in the US National Trauma Data Bank: implications for analysis

Area of Science:

• Trauma research
• Database methodology
• Statistical analysis

Background:

• The US National Trauma Data Bank (NTDB) is widely used in injury research.
• Methodological challenges in the NTDB may compromise study validity.
• Key issues include patient clustering within trauma centers and missing data.

Purpose of the Study:

• To evaluate the impact of missing data and patient clustering on injury research using the NTDB.
• To compare different analytical approaches for handling these methodological issues.

Main Methods:

• Four analytical approaches were compared to assess the association between blood alcohol concentration (BAC), resource utilization, and patient disposition.
• Approaches varied in how they handled missing BAC data (deletion vs. imputation) and patient clustering (ignored vs. adjusted).

Main Results:

• Imputing missing BAC data revealed biases in complete case analyses.
• Adjusting for patient clustering widened confidence intervals (CIs).
• An example showed a significant shift in the odds ratio for head CT scan use when accounting for clustering and imputation.

Conclusions:

• Excluding patients with missing BAC data in NTDB studies leads to biased estimates.
• Ignoring patient clustering results in artificially narrow CIs, potentially misrepresenting statistical significance.
• Proper statistical methods are crucial for valid injury research using the NTDB.