Longitudinal Functional Data Analysis Of Heart Rate Rhythm In The Elderly

BACKGROUND:There is a distinct circadian rhythm of heart rate under the regulation of the autonomic nervous system.The abnormal function of the autonomic nervous system in the elderly leads to changes in the circadian rhythm of heart rate and the increase in the proportion of non-dipping heart rate.Abnormal circadian rhythm of heart rate is an important risk factor for cardiovascular disease.Existing studies focused on the relationship between resting heart rate and cardiovascular diseases,but the relationship between circadian rhythm of heart rate and cardiovascular diseases is still unclear.Smart wearable devices have developed rapidly in recent years,enabling economical and convenient heart rate monitoring.Along with heart rate monitoring,smart wearable devices such as smart bracelets can collect a large amount of heart rate data at the minute level.This kind of data is functional and longitudinal in nature,and is called longitudinal functional data.For this kind of data,researchers have proposed a series of methods,forming a complete analysis system of longitudinal functional data,including longitudinal functional principal component analysis,multilevel functional principal component analysis,multilevel functional cluster analysis,multilevel mixed effects regression and generalized multilevel functional regression.However,these methods have not been applied in the field of medicine or heart rate monitoring.In view of the research status mentioned above,this study aimed to explore the association of cardiovascular disease and circadian rhythm of heart rate using heart rate monitoring data from population wearing smart bracelet for health management,adopting longitudinal functional analysis methods.In addition,this study aimed to identify different patterns of heart rate circadian rhythm in the elderly and explore the relationship between different patterns of heart rate circadian rhythm and cardiovascular disease.This study aimed to explore the correlation between cardiovascular disease and functional heart rate in the elderly using longitudinal functional data analysis based on the heart rate monitoring data collected by smart wearable devices.In addition,this study aimed to identify different circadian rhythm of heart rate,explore the relationship between different patterns of heart rate circadian rhythm and cardiovascular disease and provide a theoretical basis for the daily health monitoring of the elderly.METHODS:The data of this study came from a study of health management of the elderly based on smart wearable devices.The participants in the health management group participated in the baseline survey in January 2018.Demographic characteristics such as age,sex,height,education level were collected through questionnaires.The data of heart rate monitoring was collected by wearing smart bracelets.In September 2022,chronic cardiovascular diseases such as coronary heart disease,hypertension and diabetes were collected through the questionnaire of chronic disease.Multilevel functional principal component analysis was used to dimensionality reduction and feature extraction of the heart rate monitoring data,and generalized multilevel functional regression was used to explore the association between cardiovascular diseases and functional heart rate.Multilevel functional clustering was used to identify different patterns of heart rate rhythm in the elderly,and the relationship between patterns of heart rate circadian rhythm and cardiovascular disease was further determined by logistic regression.The variables of baseline nighttime heart rate,daytime heart rate and heart rate circadian rhythm values were defined and the relationship between these variables and cardiovascular disease was assessed by logistic regression.Baseline heart rate and baseline heart rate circadian rhythm were cross-grouped and the relationship between these cross-groups and cardiovascular disease was identified by logistic regression.RESULTS:A total of 1566 participants were included in this study.The cumulative variance explained by the first three principal components at level 1 was 93.53%in the multilevel functional principal component analysis.The results of the generalized multilevel functional regression showed that people with high levels of heart rate who have high first principal component scores had a significantly higher risk of hypertension,with an OR(95%CI)of 1.22(1.09,1.37)in model 3.Individuals with high scores of the second principal component had a lower daytime heart rate and a higher night heart rate,so they were likely to have non-dipping heart rate,and their risk of hypertension were significantly increased,with an OR(95%CI)of 1.34(1.19,1.52)in model 3.The risk of hypertension was significantly lower for those with a high third principal component score,with an OR(95%CI)of 0.84(0.75,0.94)in model 3.The relationship between coronary heart disease and diabetes and heart rate in the study population was similar to that for hypertension.The results of level 1 hard clustering showed that there were three clusters of heart rate circadian rhythm in the elderly,labeled as "high daytime heart rate-low nighttime heart rate","low daytime heart rate-low nighttime heart rate" and "high daytime heart rate-high nighttime heart rate"."The high daytime heart rate-low nighttime heart rate group" had 1254 participants,accounting for 80.08%of the participants.A total of 236 participants(15.07%)were in the "low daytime heart rate-low nighttime heart rate group" and 76 participants(4.85%)were in the "high daytime heart rate-high nighttime heart rate group".The results of relationship between hypertension and the clusters of heart rate circadian rhythm obtained by level 1 hard clustering showed that,compared to the"high daytime heart rate-low nighttime heart rate group",which was closest to dipping heart rate,the "high daytime heart rate-high nighttime heart rate group",which was closest to the non-dipping heart rate had a higher risk of hypertension,with an OR(95%CI)of 2.75(1.65,4.63)in model 3.No significant association of the"low daytime heart rate-low nighttime heart rate group" and hypertension was found.The results of relationship between coronary heart disease,diabetes and the clusters of heart rate circadian rhythm were similar to hypertension.Regression analysis of heart rate circadian rhythm value and hypertension showed that heart rate circadian rhythm value was a protective factor for hypertension,with OR(95%CI)of 0.84(0.75,0.94)in Model 3.Nighttime heart rate was a risk factor for hypertension,and OR(95%CI)was 1.18(1.04,1.33)after adjusting for covariates in Model 3.There was no significant increase in the risk of hypertension associated with daytime heart rate.In addition,regression analyses of baseline heart rate and heart rate circadian rhythm values with coronary heart disease and diabetes were similar to hypertension.CONCLUSIONS:The heart rate circadian rhythm of the elderly in this study showed a general trend of "two peaks and one valley".There are three clusters of heart rate circadian rhythms in the elderly,labelled as the "high daytime heart rate-low nighttime heart rate",the "low daytime heart rate-low nighttime heart rate" and "the high daytime heart rate-high nighttime heart rate".The risk of hypertension,coronary heart disease and diabetes was significantly increased in the "high daytime heart rate-high nighttime heart rate group",which was closest to the non-dipping heart rate.Heart rate circadian rhythm value is a protective factor and night heart rate is a risk factor for hypertension,coronary heart disease and diabetes,which can be used for cardiovascular disease risk prediction and daily cardiovascular health monitoring in the elderly.