Relationships among quality of life, severity, and control measures in asthma: an evaluation using factor analysis

Validated psychometric tools measuring quality of life, asthma control, and asthma severity have been developed, but their relationships with each other and with other important patient-centered outcomes have not been rigorously assessed.

To use factor analysis to evaluate the relationships of these validated tools with each other and with other patient-centered outcomes.

Surveys were completed by a random sample of 2854 Health Maintenance Organization members age 18 to 56 years with persistent asthma. Surveys included demographic information; validated tools measuring generic (Short Form-12; SF-12) and asthma-specific (Juniper Mini Asthma Quality of Life Questionnaire; AQLQ) quality of life, asthma control (Asthma Therapy Assessment Questionnaire), and asthma symptom severity (Asthma Outcomes Monitoring System); self-described severity, control, and course over time; and history of acute exacerbations.

Principal component analysis suggested a 5-factor model that accounted for approximately 59% of the variability. The most prominent rotated factor reflected asthma symptom frequency (19.4% of variability), was measured by the symptom subscale of the AQLQ, and was the only factor significantly related to the Asthma Therapy Assessment Questionnaire, Asthma Outcomes Monitoring System, or the self-reported assessments of severity, control, or course. Other factors included symptom bother (12.1% of variability), reflected by the environment and emotion AQLQ subscales; activity limitation (13.9% of variability), reflected by the activity AQLQ subscale and the SF-12 physical component scale; mental health (8.3% of variability), reflected by the SF-12 mental component scale; and acute exacerbations (5.0% of variability), not measured by any of the validated scales.

Distinct components of patient-reported asthma health status can be identified by factor analysis. Distinct constructs of severity versus control cannot be identified by the use of these tools alone.