Development of a scalar hospital-specific severity of illness measure.

Cost-function analysis of hospitals has been criticized for not including severity adjustments. We tested a scalar hospital-specific severity index, derived from Admission MedisGroup Scores. Alternative versions (i.e., linear/nonlinear) of the index were evaluated by estimating cost functions on a sample of 201 Pennsylvania hospitals. The scalar index was a strong predictor of costs. The results also suggest that the omission of a severity variable in a hospital cost function may cause a specification error.