Session S53.5

Detecting OSAHS From Patterns Seen On Heart-Rate Tachograms

P.K. Stein, P.P. Domitrovich

Washington University School of Medicine
St. Louis, MO, USA

Objective: The purpose of this study was to determine the accuracy of a simple, visual heart rate(HR) tachogram-based method for identifying significant obstructive sleep apnea hypopnea syndrome (OSAHS).
Method: N=35 HR tachograms were generated from beat-to-beat RR interval data provided by PhysioNet, with each point plotted being the instantaneous HR. Each page of the tachogram (landscape mode) contained 1 hour of data in 6 parallel, sequential 10-min plots (length 24 cm),plotted with customized software written in UNIX/C and using gnuplot. The x-axis (time within segment, 0-10 mins) is drawn at the mean heart rate for that segment. HR, plotted on the y-axis, was on a scale of 0-100 beats in 5 cm for each 10-min segment. Each tachogram was analyzed for the presence of cyclic variation of HR (CVHR), i.e., visible, rapid increases and subsequent decreases in HR. CVHR was considered to exist if there were >=3 cycles of generally regular (in amplitude, duration and frequency) cyclic heart rate changes (of at least 6 bpm) or if there were >=3 cycles of irregular cyclic heart rate changes (of at least 6 bpm) during a period of <=3 minutes. CVHR at a frequency of >2/min or a duration of <10 s was excluded. Studies were scored either as "severe" if there were more than 15 CVHR events in one hour and at least 100 minutes of CVHR during the entire night, or "none" if they did not meet these criteria.
Results: Of the 30 studies that were clearly identified as having either severe OSAHS or none, 93.3% were correctly scored using this simple method.
Conclusions: Results suggest that HR tachograms which can easily be generated as a part of routine Holter scanning may identify patients with previously undetected OSAHS, permitting two diagnostic tests for the effort and cost of one. Also, although RR interval data were used to generate tachograms in the current study, we plan to determine if normal-to-normal interval data, which include only beat-to-beat intervals under autonomic control, will provide improved accuracy.