Back to Annual Meeting Page		American Public Health Association 133rd Annual Meeting & Exposition December 10-14, 2005 Philadelphia, PA

Abstract #114343

John Ware, PhD, Mark R. Kosinski, MA, Jakob B. Bjorner, MD PhD, and Xiaowu Sun, PhD. QualityMetric Incorporated, 640 George Washington Highway, Lincoln, RI 02865, (401) 334-8800, ext. 226, jware@qualitymetric.com

Background: The “evidence” in evidenced-based health care is increasingly based on information about self-reported health status and the standardization of measures in population surveys, clinical trials and individual patient-level assessments is greatly facilitating the comparison of results from health services research. To achieve measures that better meet the requirements of each research application (e.g. population survey versus individual patient-level assessment), we developed a “family” of static short-form (SF) surveys (SF-36, SF-12 and SF-8) and dynamic assessments using item response theory (IRT) and computerized adaptive testing (CAT) software all standardized to estimate familiar SF-36 profiles and summary physical and mental health measures. While having the same mean and variance for each health domain, each survey method varies in terms of the precision and range of measurement from the most coarse SF-8 single-item estimates suitable for large population surveys to the most precise CAT-based estimates required for studies of individual patients. Methods: A series of general population surveys (N=12,050) were conducted to evaluate and compare each survey method for each of the eight SF-36 health domains. The item banks included the SF-36, SF-12, and SF-8 Health Surveys, as well as items from 52 other surveys measuring the same health domains. IRT methods were used to cross-calibrate and score the items in each item bank on a common scale and t-score transformations standardized the scores to have the same means and variances in the general population (norm-based scoring). The comparability of scores from the SF-8, SF-12, and SF-36 and CAT administration of items was assessed and respondent burden was compared. Results: For each domain, scale scores from the SF-8, SF-12, SF-36 and CAT were highly correlated (r>0.75) with one another and mean scores never differed by more than 1-2 points in the general population or across self-reported condition groups. Score estimates mainly differed in terms of their precision, with CAT scores being the most precise and SF-8 scores being the least precise. Precise scores from CAT assessments were achieved with a 70% to 80% reduction in respondent burden. Conclusion: IRT-based cross-calibration and SF-36 norm-based scoring methods enable the comparison of scores across the SF-8, SF-12 and SF-36 and CAT surveys. We discuss implications for health service research including large population surveys, clinical trials and applications requiring individual patient–level assessments in clinical practice.

Learning Objectives:

• At the conclusion of this session, participants will be able to identify the methods used to cross-calibrate item responses from various self-reported health status surveys measuring similar health domains, understand that the outcomes from health status surveys used in population studies can be made comparable to those used in clinical practice and other research settings, and discuss new standards for the interpetation and scoring of scales from a family of short-form measures centered around the SF-36 Health Survey.

Keywords: Health Assessment, Outcome Measures

Presenting author's disclosure statement:

I wish to disclose that I have NO financial interests or other relationship with the manufactures of commercial products, suppliers of commercial services or commercial supporters.

Recorded presentation