Methods for Obtaining and Analyzing Unattended Polysomnography Data for a Multicenter Study

Abstract

THE SLEEP HEART HEALTH STUDY (SHHS) is the first large-scale multicenter study of sleep-disordered breathing. The objectives and general study design have been described in detail previously. 1 Briefly, the SHHS is a prospective cohort study designed to investigate the relationship between sleep-disordered breathing and cardiovascular disease. Participants were recruited from eight existing epidemiological studies in which data on cardiovascular risk factors had been collected previously. Although the infrastructure for obtaining cardiovascular risk factor data and for ascertaining morbid and mortal events was available through the parent cohorts, the SHHS developed pro-SLEEP, Vol. 21, No. 7, 1998 Summary: This paper reviews the data collection, processing, and analysis approaches developed to obtain comprehensive unattended polysomnographic data for the Sleep Heart Health Study, a multicenter study of the cardiovascular consequences of sleep-disordered breathing. Protocols were developed and implemented to standardize in-home data collection procedures and to perform centralized sleep scoring. Of 7027 studies performed on 6697 participants, 5534 studies were determined to be technically acceptable (failure rate 5.3%). Quality grades varied over time, reflecting the influences of variable technician experience, and equipment aging and modifications. Eighty-seven percent of studies were judged to be of good quality or better, and 75% were judged to be of sufficient quality to provide reliable sleep staging and arousal data. Poor submental EMG (electromyogram) accounted for the largest proportion of poor signal grades (9% of studies had <2 hours artifact free EMG signal). These data suggest that with rigorous training and clear protocols for data collection and processing, good-quality multichannel polysomnography data can be obtained for a majority of unattended studies performed in a research setting. Data most susceptible to poor signal quality are sleep staging and arousal data that require clear EEG (electroencephalograph) and EMG signals.