Evaluating neurological group homogeneity in assessing the mortality risk for people with spinal cord injuries

A study of 3178 individuals injured in Britain between 1943 and 1990 and surviving the first year post-injury was conducted to evaluate the homogeneity of mortality risk ratios within groups based on varying degrees of neurological injury level and completeness of the injury. The study shows that it is less than optimal to combine individuals into neurological groupings of C1-C4 ABC, C5-C8 ABC and T1-S5 ABC since the risk ratios are not homogeneous within these groups. Similarly, combining individuals into neurological groupings of tetraplegia complete, tetraplegia incomplete, paraplegia complete and paraplegia incomplete may not be appropriate for the same reasons. The consequence of performing a survival analysis using either of the traditional sets of groups is to dilute the risk ratios for a subset of individuals within a particular group, thereby providing less discrimination between neurological groups. Cox proportional hazards regression was employed to determine a set of neurological groupings with homogeneous risk ratios within a group while providing better differentiation between groups.