abstract
- Global indices, which aggregate multiple health or function attributes into a single summary indicator, are useful measures in health research. Two key issues must be addressed in the initial stages of index construction from the universe of possible health and function attributes, which ones should be included in a new index? and how simple can the statistical model be to combine attributes into a single numeric index value? Factorial experimental designs were used in the initial stages of developing a function index for evaluating a program for the care of young handicapped children. Beginning with eight attributes judged important to the goals of the program by clinicians, social preference values for different function states were obtained from 32 parents of handicapped children and 32 members of the community. Using category rating methods each rater scored 16 written multi-attribute case descriptions which contained information about a child's status for all eight attributes. Either a good or poor level of each function attribute and age 3 or 5 years were described in each case. Thus, 2(8) = 256 different cases were rated. Two factorial design plans were selected and used to allocate case descriptions to raters. Analysis of variance determined that seven of the eight clinician selected attributes were required in a social value based index for handicapped children. Most importantly, the subsequent steps of index construction could be greatly simplified by the finding that a simple additive statistical model without complex attribute interaction terms was adequate for the index. We conclude that factorial experimental designs are an efficient, feasible and powerful tool for the initial stages of constructing a multi-attribute health index.