Propensity Score Matching

Propensity scores were calculated as the probability of undergoing early (<24 hours) compared with late (≥24 hours) surgery using the logit method with data source (categorical), age (continuous), mechanism of injury (categorical), and baseline AMS (continuous), AIS grade (categorical), and neurologic level (continuous) as covariates. These covariates were selected based on hypothesized influence on choice of treatment (ie, early vs late surgical decompression) and also potential impact on neurologic recovery. Age at and measures of severity of injury (ie, AMS, AIS, neurologic level) factor into clinical decision-making in patients with acute SCI, and it is also recognized that both factors have an important effect on neurologic recovery, particularly in the setting of CCS.^{12,23,24,25,26,27} Similarly, the phenotype of patients with a low- vs high-energy trauma and resultant CCS is different, and the treatment considerations are varied.^6,15 Further, we hypothesized that there may be differences between data sources with regard to propensity for early surgical decompression, especially given these data sets spanned multiple decades, and treatment paradigms for acute SCI have evolved over time.⁸ Propensity score matching was performed in a 1:1 ratio using the optimal-matching technique to minimize the average absolute distance across all matched pairs. This resulted in 2 treatment groups (early surgery vs late surgery) adjusted for the baseline variables specified previously. Baseline characteristics were compared between treatment groups by t test for continuous variables and χ² test for categorical variables.