Background: Econometric techniques have been used to examine the relationship between costs of provision, case mix and unit size for health care providers. Estimation of cost functions in a health care context is complicated by poor understanding of the underlying production relationship and the constraints under which production takes place. Different results and policy implications can follow from different model specifications. This underscores the need for care in the construction of such functions and the interpretation of their results.
Methods: Cost and activity data from a study of neonatal care for an English Regional Health Authority are re-examined. Cost functions are estimated using alternative functional forms, and average cost per day is estimated and compared for two of these functions [one produced by Fordham et al. (J Publ Hlth Med 1992; 14(2): 127-130) and the other based more explicitly in economic and econometric theory].
Results: It is shown that estimates of average cost per day are sensitive to model specification. Such are the differences in cost that significantly different policy implications could follow from the different models.
Conclusion: We conclude that care must be taken in the construction and estimation of cost functions and that the assumptions upon which they rest be made explicit so that results can be properly interpreted.