The validity of the MIMIC (Multiple Indicators/MultIple Causes) health index--some empirical evidence.

Abstract

This study evaluates the potential of econometric models with latent (unobservable) variables for measuring health or health impairment due to a specific disease. A MIMIC disability index is estimated for a sample of 145 adults with chronic bronchitis, expressing their self-reported disability caused by the disease on a one-dimensional scale. The index is determined up to a linear transformation. Disability is thus measured on an interval scale. The data were collected by interviews. The questionnaire used for this purpose is based on a number of in-depth interviews with selected bronchitis patients conducted beforehand. The study therefore focuses directly on the patients' perceptions of their disease. The validity of the index is evaluated in three different ways. First, construct validity is assessed performing groupwise analysis and testing for differences in the index values by subgroup. To a large extent, the index is consistent with a priori expectations. Therefore, we conclude that it has high construct validity. Second, validity of the index is assessed by comparing its results to a direct rating scale produced by 21 physicians with various medical backgrounds. The MIMIC index turns out to be related in a systematic, but nonlinear way to this direct rating scale. This can be interpreted in two different ways. If one accepts the preferences of health providers as the ultimate yardstick when it comes to ranking health or chronic states the result suggests that the MIMIC index estimated in this way is not a valid measure of treatment success. By contrast, if patients' preferences are considered to be decisive, it suggests that physician-based ratings should be substituted for or at least complemented with patient-based indices (such as the MIMIC disability index estimated here) when evaluating medical services in terms of cost-effectiveness. Third we explore the extent to which the MIMIC index reflects utility associated with different states of disability, using a modified Torrance Standard Gamble approach. The above-mentioned physicians are used as experts in this procedure. The results indicate that the MIMIC index as estimated here is related in a systematic, but nonlinear way to the Standard Gamble risk index as well. The fact that this relationship is nonlinear indicates that the MIMIC index does not measure utility as derived from the experts' preferences directly. How this index would fare compared to a Standard Gamble risk index provided by patients (bronchitis subjects) is a question which remains open.(ABSTRACT TRUNCATED AT 400 WORDS)