The associations between variables in large databases

A database of 2126 consecutive cases where an ultrasound device was used to aid healing of bone fracture cases, was examined. All fractures had been fixed surgically first but appeared to exhibit delayed healing. After treatment with ultrasound, 7% of the fractures had still failed to heal. The aim was to predict those fractures that would not heal. Data collected from the treating physicians about the cases, was used. Although individual variables were associated with failure to heal, only four variables predicted the failure with statistical significance. The variables were patient age, time since fracture, fracture type and whether the fracture was in the humerus or not.