Introduction: Although it is generally appreciated that heart rate var
iability is low during severe illness, the extent, time course, and ma
thematical characteristics of heart rate variability during transition
s between health and illness have not been systematically examined. Th
e purpose of this study was to analyze heart rate variability in newbo
rn infants during a rapid recovery from severe respiratory and circula
tory failure. Methods and Results: From prolonged ECG recordings, we e
valuated heart rate variability in the time domain (mean, relative cha
nge, and coefficient of variation of RR intervals),in the frequency do
main(using power spectra of the time series of RR intervals), and usin
g a neural network. Qualitatively, RR interval plots showed little hea
rt rate variability during severe illness but became ''noisier'' durin
g recovery. Quantitatively, recovery was marked by twofold to threefol
d increases in time-domain parameters, by eightfold increases in frequ
ency-domain parameters, and by more than 20-fold increases in a neural
network measure. Time-domain and frequency-domain measures were corre
lated, but not strongly. Heart rate variability reached stable levels
by 4 to 5 days. Heart rate did not change dramatically. Conclusion: Re
covery from severe neonatal illness is accompanied by large and rapid
increases in heart rate variability, but not by large changes in heart
rate. This increase can be effectively assessed in the time domain, i
n the frequency domain, and by using a neural network.