SLEEP 2011 Abstract Supplement

B. Clinical Sleep Science XIV. Instrumentation and Methodology
0961
VALIDATION OF THE SLEEP DISORDERS SCREENING
QUESTIONNAIRE
Jungquist CR 1,3 , Marcus JA 4 , Valerio TD 5 , White TV 5 , Perlis ML 2 ,
Pigeon WR 3 , Hill K 6
1
School of Nursing, University of Rochester, Rochester, NY, USA,
2
Psychiatry, University of Pennsylvania, Philadelphia, PA, USA,
3
Psychiatry, University of Rochester, Rochester, NY, USA, 4 Neurology,
University of Rochester, Rochester, NY, USA, 5 Illinois Neurological
Institute Sleep Center, OSF Saint Francis Medical Center, Peoria, IL,
USA, 6 Rochester Institute of Technology, Rochester, NY, USA
Introduction: According to the National Sleep Foundation, 11-14% of
Americans are at risk for insomnia, restless legs syndrome and/or obstructive
sleep apnea; yet sleep disorders are widely under diagnosed.
To address this problem, a Sleep Disorders Questionnaire (SDQ) was
developed for use in primary care practice.
Methods: To test the psychometric properties of the SDQ, 413 subjects
underwent overnight polysomnography (PSG) for diagnostic purposes
after filling out the Sleep Disorders Screening Questionnaire (SDQ), the
Multidimensional Fatigue Inventory (MFI) or the Fatigue Severity Scale
and the Epworth Sleepiness Scale (ESS). Sleep diagnoses were derived
from clinical interview and PSG findings using AASM diagnosis criteria
for insomnia, obstructive sleep apnea, central sleep apnea, restless
legs syndrome and narcolepsy. The SDQ was evaluated using Principal
Component Factor Analysis with Varimax rotation, Pearson’s correlations
using SPSS (version 18). SDQ subscales were extracted, assessed
for face validity, and then correlated with categorical outcomes representing
specific sleep diagnoses.
Results: 413 subjects (47% female), ages 18-83 (M 49), BMI 18-72
(M 35), ESS 0-24 (M11), MFI subscales General Fatigue M15.5, Physical
Fatigue M12.4, Reduced Activity M10.6, Reduced Motivation 10.3,
Mental Fatigue 11.7. Sleep Diagnosis: OSA (n=357), CSA (n=3), RLS
(n=28), Insomnia (n=81), Narcolepsy (n=5). Principle Component
Analysis extracted five subscales that accounted for 57.66% of variance.
Pearson correlation between SDQ subscales and diagnosis categories is
as follows: Subscale 1 - Insomnia (r = .379, p. = .000); Subscale 2 -
Narcolepsy (r = .112, p. = .02); Subscale 3 - OSA (r = .105, p. = .03);
Subscale 4 - RLS (r = .367, p. = .000).
Conclusion: The Sleep Disorders Screening Questionnaire shows some
preliminary promise as a screening tool for the major sleep disorders.
Subscales are correlated with matching diagnosis. Further work is ongoing
to establish other psychometric properties of the SDQ.
Support (If Any): T32, National Institute on Aging, #5T32AG020493-05
and the University of Rochester School of Nursing fellowship program
0962
INCLUSIVE MULTIPLE IMPUTATION FOR MISSING DATA
IN A VA LONGITUDINAL TRIAL ASSESSING COGNITIVE
BEHAVIORAL THERAPY IN PATIENTS WITH SLEEP
DISORDERS
Stechuchak KM 1 , Woolson R 2 , Ulmer CS 2 , Edinger JD 2 , Olsen MK 2
1
Health Services Research and Development, Department of Veterans
Affairs Medical Center, Durham, NC, USA, 2 Durham VA and Duke
University Medical Centers, Durham, NC, USA
Introduction: We describe and implement an inclusive multiple imputation
(MI) strategy for handling missing data in a randomized trial examining
sleep outcomes. This strategy is compared with last observation
carried forward (LOCF), a commonly used, single imputation strategy.
Methods: Eighty-one veterans with chronic primary or comorbid insomnia
were randomized to receive cognitive behavioral therapy (CBT)
or sleep hygiene (SH). Sleep measures, including the Pittsburgh Sleep
Quality Index (PSQI) and Dysfunctional Attitudes and Beliefs About
Sleep Scale (DBAS), were assessed at baseline, post-treatment, and follow-up.
Markov chain Monte Carlo methods were used to create multi-
ply-imputed datasets of the sleep measures at all time points. Auxiliary
variables predicting dropout were included in the imputation model.
General linear models were fit to analyze the treatment effect of CBT
compared to SH over time.
Results: Approximately 19% (n=15) of the sample was lost to attrition
with most dropout occurring before post-treatment. Dropout predictors
included insomnia type, therapist, medication use, comorbidities,
age, and full-time employment. Model-based standard errors (SE) with
LOCF were smaller than MI for all outcomes. Subjects randomized to
CBT had greater PSQI improvement at post-treatment compared to SH
using LOCF (-1.6, 95% CI:-3.1, -0.1; p=0.04); in contrast, MI results
were not statistically significant (-1.3; 95% CI:-2.8, 0.2; p=0.09). However,
CBT subjects had greater DBAS baseline to post-treatment improvement
compared to SH when using MI (-10.3; 95% CI:-20.2, -0.4;
p=0.04); LOCF results were not statistically significant (-6.8, 95% CI:-
15.0, 1.4; p=0.10).
Conclusion: Methodologies for accommodating missing data can produce
different results with regard to both direction and strength of treatment
effects. LOCF typically underestimates SE by not accounting for
the uncertainty attributable to missing data. MI provides a framework
to incorporate information from auxiliary variables predicting dropout
while preserving a parsimonious main treatment effect model, as well as
appropriately estimating SE.
Support (If Any): This material is based upon work supported (or supported
in part) by the Department of Veterans Affairs, Veterans Health
Administration, Health Services Research and Development Service.
0963
SLEEP DISORDERED BREATHING IN INSOMNIA PATIENTS
RECOGNIZED BY ECG ANALYSIS
Penzel T, Glos M, Schoebel C, Fietze I
Sleep Medicine Center, Depart. of Cardiology, Charite University
Hospital Berlin, Berlin, Germany
Introduction: Baseline recordings in insomnia patients in a sleep laboratory
are conducted in order to quantify the extend of insomnia in terms
of sleep efficiency and sleep onset latency. In addition the sleep study
is used to exclude other sleep disorders which may occur together with
primary insomnia.
Methods: We investigated 64 patients with insomnia with cardiorespiratory
polysomnography in our sleep center. Sleep stages, arousal and
respiratory events were scored according to AASM criteria by an experienced
sleep technician. Recorded ECG was analyzed by a software
(Hypnocore) which could provide a sleep evaluation and a respiratory
event score by a new automated analysis (denoted as ECG). All patients
were analyzed first. A second analysis was performed on 54 patients after
removing subjects with bad signal quality, arrhythmias and a total
sleep time below 3 hours.
Results: The analysis of respiratory events based on ECG in the group
of 64 subjects resulted in 52 subjects (48 true negative, 4 false negative)
with an RDI≤5/h. 12 subjects (10 true positive, 2 false positive) were
scored with an RDI>5/h. Agreement was 0.91. For the second analysis
agreement remained the same. Sleep stages in the second analysis were
scored surprisingly good: 48.9% (ECG) vs. 48.7% (PSG) for light sleep,
15.7% (ECG) vs. 15.8% (PSG) for slow wave sleep, 14.0% (ECG) vs.
23.4% (PSG) for wake, and 19.9% (ECG) vs. 12.2% (PSG) for REM
sleep. Sleep efficiency was 84% (ECG) vs. 77% (PSG).
Conclusion: Not too many respiratory events occur in insomnia patients.
These events are detected with a sufficient accuracy. Agreement
for classification of subjects is adequate. Sleep stage analysis based on
ECG reveals the very good ability to distinguish light sleep, slow wave
sleep, and wake / REM sleep. To distinguish wake and REM sleep by
ECG alone is possible moderately. The biggest effect of this uncertainly
is apparent in sleep efficiency. To distinuish wake and REM cannot be
perfect due to high sympathetic activity in both states.
A329
SLEEP, Volume 34, Abstract Supplement, 2011