OBJECTIVE: To determine the prevalence and effect on traditional heart rate variability (HRV) indices of abnormal HRV patterns in the elderly. METHODS: Hourly Poincare plots and plots of spectral HRV from normal-to-normal interbeat intervals and hourly nonlinear HRV values were examined in a subset of 290 consecutive participants in the Cardiovascular Health Study. Only subjects in normal sinus rhythm with > or = 18 hours of usable data were included. Eligible subjects were 71 +/- 5 years. During 7 years of follow-up, 21.7% had died. Hours were scored as normal (0), borderline (0.5), or abnormal (1) from a combination of plot appearance and HRV. Summed scores were normalized to 100% to create an abnormality score (ABN). Short-term HRV versus each 5th percentile of ABN was plotted and a cutpoint for markedly increased HRV identified. The t-tests compared HRV for subjects above and below this cutpoint. Cox regression evaluated the association of ABN and mortality. RESULTS: Of 5,815 eligible hourly plots, 64.4% were normal, 14.5% borderline, and 21.1% abnormal. HR, SDNN, SDNNIDX, ln VLF and LF power, and power law slope did not differ by the cutpoint for increased short-term HRV, while SDANN and ln ULF power were significantly lower for those above the cutpoint. However, many HRV indices including LF/HF ratio and normalized LF and HF power were significantly different between groups (P < 0.001). Increased ABN was significantly associated with mortality (P = 0.019). Despite similar values for many HRV indices, being in the group above the cutpoint was significantly associated with mortality (P = 0.04). CONCLUSIONS: Abnormal HR patterns that elevate many HRV indices are prevalent among the elderly and associated with higher risk of mortality. Consideration of abnormal HRV may improve HRV-based risk stratification.