Comparing discriminative validity between a disease-specific and a generalhealth scale in patients with moderate asthma

Health-related quality of life scales such as the Asthma Quality of Life Qu estionnaire and the Medical Outcomes Study Short-form General Health Survey SF-36 have become important measures of health status in clinical asthma t rials. The discriminative properties of these scales, however, have not bee n extensively evaluated and compared. The purposes of this study were to as sess and compare scale and discriminative properties of the Asthma Quality of Life Questionnaire (AQLQ) and the SF-36 in a group of patients with mode rate asthma using a patient-rated global measure of disease activity as the criterion variable. Patients were interviewed in-person with a series of q uestionnaires including the AQLQ and the SF-36, and were also asked the glo bal question "How active is your asthma now?" with possible responses of "e xtremely," "very," "moderately," "mildly" or "not active." Discriminative p roperties were determined using receiver operating characteristic (ROC) cur ves with responses to the global question as the criterion variable and mea n domain scale scores as the independent variables. Relative validities for the AQLQ and SF-36 domains were also compared. A total of 230 patients, me an age of 41 years, were enrolled. Scores were lower and ranges were narrow er for the AQLQ compared to the SF-36. In general, the AQLQ and the SF-36 w ere highly correlated, with r = 0.69 for the AQLQ overall score and the SF- 36 Physical Component Summary (PCS) score. According to ROC analyses, both scales had excellent discriminative properties; however the area under the ROC curve was higher for the AQLQ overall score (0.81) than for the PCS (0. 75). When ranked according to ROC area, the symptoms domain (0.83) had the greatest area under the ROC curve, followed by the emotional (0.76) and act ivities (0.76) domains of the AQLQ. However, in some cases, the area under the curve was less for an AQLQ domain (for example, 0.71 for the environmen tal domain) than for SF-36 domains (for example, 0.75 for the role physicia l, and 0.75 for the social domain). Similarly, the AQLQ overall had a highe r relative validity (5.2) compared to the PCS (2.2), and the symptoms domai n of the AQLQ had the highest relative validity (6.0). Thus, both the Asthm a Quality of Life Questionnaire and the SF-36 were able to characterize pat ients with moderate asthma in our cross-sectional study. In addition, both scales had strong discriminative properties when assessed with a global pat ient rating of current disease activity. (C) 2001 Elsevier Science Inc. All rights reserved.