© 2020 IEEE. In heart failure, early detection of exacerbation is required to avoid unnecessary hospitalization. Although respiration instability detected by a respiration sensor attached to the bed at home has been proposed, detection by a more popular general sensor is desired. Using the fact that the frequency of respiratory sinus arrhythmia ({F}-{RSA}) reflects the frequency of breathing ({F}-{B}), we developed a method to estimate the instability of {F}-{B} by analyzing the instability of {F}-{RSA} only using R-R interval time series obtained from nighttime ECG. In 69 healthy control subjects, the fluctuation of {F}-{RSA} reduced to 6.8 \pm 3.9{\%} of local mean during sleep, while it remained 16.4\pm 7.7{\%} and 17.5 \pm 5.8{\%} in sinus rhythm patients with heart failure who survived for 38-mo follow-up period after recording (n = 52) and those died during the follow-up ({n}=29), respectively. {F}-{RSA} instability detected by ECG during sleep may be used as a feature to screen patients with heart failure.