The Distribution Of Short-Term Heart Rate Variability Indices In Long-Term Series And Its Influence Factors

Background:Heart rate variability(HRV)is a non-invasive tool reflecting the function of the cardiac autonomic nervous system(ANS).Under normal conditions,the sympathetic and vagal nerves of the autonomic nervous system are in a dynamic equilibrium process,which may lead to life-threatening adverse cardiac events when the balance is destroyed.As the "gold standard" of clinical HRV evaluation,the concept of predictive utility of HRV comes mainly from the time-domain and frequency-domain studies of 24-hour dynamic monitoring.24-hour HRV records can better reflect the slower fluctuation process(such as circadian rhythm),while the sequence analysis of short RR interval is more consistent with the non-stationary characteristics of heart rate dynamics,but they are not interchangeable due to different physiological meanings.Analyzing the distribution characteristics of short-term HRV indices on long-term heart rate sequence not only reflects the non-stationarity of RRI sequence,but also takes advantage of the data quantity advantage of long-term sequence.In order to improve the accuracy of 24-hour HRV prediction,it is essential to further explore the influence of physiological and pathological factors.Method:In our study,part of the data were provided by THEW.The 24-hour Holter data of normal people aged over 18 years old(n=177)in the database Normal were divided into 5 control groups according to age(18≤y≤25,n=35;25<y≤35,n=44;35<y≤45,n=41;45<y≤55,n=34;y>55,n=23).The Holter data in database ESRD were selected as the group with high risk of arrhythmia and sudden cardiac death(n=43),and the Holter recording of congestive heart failure patients provided by PhysioNet were selected as the CHF mild group(n=12)and CHF severe group(n=32).Using sliding windows,taking 5 min as window width and 2.5 min as step length,the short-term HRV linear and non-linear indexes in each window are calculated according to time.Spearman correlation coefficients(Spearman CC)between RRI mean(MRRI),LF/HF and α1 were calculated respectively,and the proportion of people with good correlation in each group was counted.Then,24-hour HRV time-domain indices of different age groups were calculated.93 normal persons aged 25-65 with normal work-rest time and sufficient record length were selected and divided into 4 groups.The mean values of 5 min sliding windows were calculated for each 2 h period(EM MRRI,EM LF/HF,EM_α1).The data of 8 healthy subjects was selected from database Physiologic Response to Changes in Posture(PRCP)published on PhysioNet,providing documentary ECG during slow tilt stimulation(ST,75°HUT over 50 s).Short term measures,such as a geometric measure for HRV(rrHRV),RMSSD,SD1/SD2,were applied to analyze RRI behaviors under steady and dynamic situations.For evaluating the ability in tracking heart rate dynamics caused by HUT,Spearman correlation coefficient(Spearman CC)was used between RRI series and each of the above three measures with the moving window of 10,20,30,60 successive RRIs.Comparison of steady state was made before and after HUT(S1and S2).Result:The proportion of people with good correlation maintained a high level among the normal people aged 18-55,which was more than 94%,and declined sharply after the age of 56(MRRI vs LF/HF:78.26%;MRRI vs α1:65.22%).But up to the age of 65,the proportion of normal people was still higher than that of the patients(less than 60%).In 24-hour time domain indices of normal people of different ages,RRI mean(RRImean)did not differ significantly among different age groups,while the other indices(pNN50,RMSSD,SDNN,RMSSD/SDNN,pNN50/SDNN)decreased with aging.For the lowest EM_MRRI period in the morning,there were no significant differences in EM_MRRI,EM_LF/HF and EM_α1 among different age groups(P>0.05),although there might be age-related differences in other periods.While for Spearman CC,though there was tendency of decrease with the reduction of the number of successive RRIs,no significant differences were found among them(60 and 10 successive RRIs,Spearman CCs are 0.744±0.070 and 0.695±0.077 for rrHRV,0.753±0.087 and 0.732±0.073 for RMSSD,0.666±0.083 and 0.587±0.119 for SD1/SD2).By comparing the steady state befor and after HUT,it was found that there existed significant differences between S1 and S2(S1 vs S2:6.09(4.10)vs 2.56(1.04)for rrHRV,50(45)vs 18(10)for RMSSD(ms),0.505(0.168)vs 0.233(0.047)for SD1/SD2).Conclusions:In this study,we used the long-term RRI sequences provided by Holter records to analyze the correlation between the short-term RRI mean and the corresponding HRV indices on the long-term sequence,and to explore the characteristics of HRV indices in the morning characteristic period.The results suggest the existence of age-related inflection points and the consistency of HRV indicators in different age groups during the morning characteristic period.The distribution characteristics of short-term HRV indices in long-term series are studied,which not only reflects the no-stationary characteristics of RRI series,but also takes advantage of the data quantity of long-term series.Recently,with the development of wearable monitoring technology,the possibility of obtaining long-term heart rate data of normal people has been greatly improved.The methods and results of this study provide a new idea for the development of HRV analysis methods for long-term sequences.