Comparing health measures through a model for health care use

We consider two different methods of obtaining health measures. The first method consists of a health measurement model with a latent variable for general health. This model includes 23 self-assessed indicators of individual health, as well as the more objective health measure of grip strength to construct a health index that is corrected for differences in reporting styles across countries. The second method uses six subjective self-assessment questions on health problems in six different health domains, where responses are corrected for reporting style by making use of vignettes. To assess the usefulness of both methods, the alternative health measures are included in a model for health care use. The data that is used for implementation comes from the Survey of Health, Ageing and Retirement in Europe (SHARE). From this dataset, we use information from the Netherlands and Spain to perform all analyses. We find that both methods contain useful information on individuals' latent (unobserved) health, and that the health measures obtained from both methods are more useful in explaining health behavior than the uncorrected health indicators that are used to construct the health measures are. Once the six ill-health measures from the vignette method are included in the model for health care use, the latent health index does not improve explanatory power. Although the method using the health measurement model with a latent variable for health is found to be more helpful in finding relations between health indicators and ``true" health, the method using vignettes seems to explain health-related behavior better.