Citation

Kump K, Whalen C, Tishler PV, Browner I, Ferrette V, Strohl KP, Rosenberg C, Redline S. Am. J. Respir. Crit. Care Med. 1994; 150(3): 735-741.

Affiliation

Department of Medicine, Case Western Reserve University, Cleveland, Ohio.

Copyright

(Copyright © 1994, American Thoracic Society)

DOI

unavailable

PMID

8087345

Abstract

Although questionnaires have been developed to assess symptoms of obstructive sleep apnea (OSA), their overall reliability and utility have not been established. We have evaluated the ability of a questionnaire to identify increased apnea activity (IAA) in 465 participants in an epidemiologic study of OSA. Subjects and their roommates each completed a questionnaire and underwent in-home sleep studies. Responses to 56 questions about sleep habits, sleepiness, and daytime performance were analyzed with factor analysis, logistic regression, and receiver-operator curves (ROCs). Factor analysis demonstrated that 16 questions, grouped into five factors (functional impact of sleepiness, self-reported breathing disturbances, roommate-observed breathing disturbances, driving impairment, and insomnia) explained 67% of the variance in the questionnaire data. Symptom questions demonstrated internal consistency (Cronbach correlations: 0.91 to 0.98). Moderate levels of agreement were observed between self- and roommate-reported responses for nine of ten questions asked of both the subject and his/her partner (kappa statistics: 0.34 to 0.57). Logistic regression analysis demonstrated that IAA could be best predicted by three questions about intensity of snoring, roommate-observed choking, and having fallen asleep while driving (ROC area: 0.78). Use of symptoms with data on gender and body mass index (BMI) improved predictive ability by 10% (ROC area: 0.87). Thus, questionnaire data provide a valid means of characterizing symptom distributions in population surveys of OSA. Predictive ability is not significantly improved with multiple questions or a separate roommate questionnaire, but is improved with consideration of data on BMI and gender.

Language: en