Feasibility of a simple method for identifying sleep periods from Holter recordings

To determine whether sleep periods can be identified from Holter-derived data using a combination of hourly average HR patterns and hourly spectral plots, Holter data from N=77 randomly-selected subjects with self-reported bed:wake (B:W) times were examined. N=50 were patients with a recent MI (C), 12 were younger subjects from a Holter quality validation study (Y), and 15 were subjects>65 years old from the Cardiovascular Health Study (O). N=18 (C) were excluded for a combination of abnormal circadian rhythm and abnormal HRV plots. The accuracy of hourly HR + HRV patterns to identify bedtime ± 1hr was: 100% (O), 84% (C), and 100% (Y). Hourly HR+HRV patterns identified wake-times ± 1hr 100% of the time for (O), 91% for (C) and 100% (Y). Identification of sleep periods from Holter data appears feasible in different groups using a simple algorithm. This method could permit detailed study of sleep periods in Holter cohorts where diaries are unavailable.