Methods of Constructing Health Measures

Published in: Measuring Functioning and Well-Being: The Medical Outcomes Study Approach / edited by Anita L. Stewart and John E. Ware, Jr., (Durham, N.C.: Duke University Press, 1992), Chapter 5, p. 67-85

Posted on RAND.org on January 01, 1992

by Anita Stewart, Ron D. Hays, John E. Ware

In research on health and health-related matters, the concepts are complex and difficult, if not impossible, to define by a single item. Rather, light must be cast at different angles. In constructing measures, the MOS approach was to write questionnaire items; pretest the measures; construct multi-item scales; evaluate scale variability, reliability, and stability; and label measures. This chapter offers criteria for each element of this process. For example, writing questionnaire items required operationalizing the concept being measured; that is, for the MOS, closed-ended questions with a specific set of responses were used, because in large-scale studies, open-ended questions are more burdensome and often yield uninterpretable answers, especially from the less educated. Response options consisted of choices on endorsement (how true/false), frequency, and intensity. A time frame also needed to be specified for most questions. Pretesting was conducted to judge clarity of instructions, determine whether questions made sense, and estimate respondent burden. MOS conducted nine full-scale pilot studies of various measures with varying purposes, spending more than one year on this phase. The chapter includes mathematical formulas for such factors as item-scale correlations uncorrected for overlapping items, scale transformations, reliability, and intra-class correlations.

This report is part of the RAND Corporation external publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.