Quantitative characterization of disease severity in diseases with complex symptom profiles

A number of clinical and research situations arise that require the integration of complex, multi-dimensional, performance-related information to determine a single quantity such as "disease severity" or "risk level" (for subsequent development of a disease). This process is generally carried out either by relying on a subjective gestalt approach or by using rating scales that combine scores for component measures using an additive process without justification. Concepts from general systems performance theory have provided insights into this problem, teaching that a fundamentally multiplicative method of integrating components is often appropriate. In this paper, concepts and previous supportive experimental work are reviewed in the context of the quantitative characterization of disease severity. A Parkinson's disease study (n=114) is presented that mimics the "two-condition" situation encountered in the clinical trial of a new drug or other therapy. Traditional and performance theory-based composite scores are computed for each condition ("on" and "off" medications) and compared, with emphasis on the different quantitative "pictures" conveyed by each approach. It is concluded that performance theory based composites are not only more sensitive to therapeutic agents experimentally, but also have superior conceptual validity compared to traditional forms of single-number composites.