Heart rate variability as a prognostic marker in critically ill patients [version 1; peer review: awaiting peer review]

Background: Heart rate variability (HRV) can be used to assess cardiac autonomic activity in critically ill patients. Heart rate variability is termed as fluctuation in the time interval between adjacent heartbeats. The equilibrium among the sympathetic and parasympathetic subgroups of the autonomic nervous system (ANS) is essential for the maintenance of systemic homeostasis and effective response to external stressors. Hence we aimed this study to determine whether heart rate variability can be used as a prognostic marker in critically ill patients. Methods: A cross-sectional study was conducted among 225 consecutive critically ill patients admitted to the medicine Intensive care unit (ICU) of AVBRH, Sawangi (Meghe) based on the inclusion and exclusion criteria. The selected participants were evaluated for 24 hours Heart Rate Variability (HRV) and APACHE 4 score. Outcomes like mortality and survival were corelated with 24 hours Heart Rate Variability and APACHE 4 score. Results: The variables were significantly associated (p<0.05) with Standard deviation of the average Normal-to normal HRV intervals (SDANN) and Standard deviation of the NN intervals (SDNN). The variables were also significantly associated (p<0.05) with the variable High frequency


Introduction
The fluctuation in the time between contiguous heartbeats is referred to as heart rate variability. For the maintaining of systemic homeostasis and an adequate response to external stimuli, the sympathetic and parasympathetic subgroups of the autonomic nervous system (ANS) must be in balance. 1 The autonomic nervous system (ANS) is a part of the peripheral nervous system (PNS) that is responsible for coordinating involuntary physiological functions that include respiration, heart rate, blood pressure, digestion, etc. These functions are further controlled by the sympathetic and parasympathetic systems. Dysfunction of ANS (one or more subdivisions) along with a pre-existing disease has been seen to be been associated with a poor outcome of the disease. Cardiac autonomic dysfunction can result from any disease that affects any of the various components of the ANS. 2 Disorders such as sepsis, coronary artery disease, strokes, chronic liver disease, chronic kidney disease, malignancy, hypertension, diabetes mellitus, etc, with autonomic dysfunction, often require management in intensive care units (ICUs). In the ICU, prompt and proper assessment of patients' prognosis, outcome, and mortality risk is of the most importance. It has been speculated that early recognition of autonomic dysfunction can help predict prognosis and direct appropriate and timely treatment. The available scoring systems such as APACHE 4, though competent, lack feasibility of application in the ICU setting due to lack of required information for scoring. 3 Therefore, over the years, many methods have been used in assessing autonomic function such as the baroreflex sensitivity, cardiac-chemoreflex sensitivity, and the heart rate variability (HRV). 4 Heart rate variability calculation is one of popular approaches to estimate autonomic nervous system (ANS) dysregulation. It is the variability between heart beats or more scientifically, variability between successive R waves on electrocardiogram (ECG) or the R-R interval (RR). It is a non-invasive method increasingly used in medicine. 3 The clinical potential of HRV recognized by Hon and Lee in 1965, when they discovered that the acute changes in HRV was linked to foetal distress and foetal hypoxia could be predicted from it before any appreciable changes occurred in the heart rate itself. 5 This finding was paramount in the development of the standard of care in monitoring fetal distress and therefore, significantly reduced incidences of morbidity and mortality. 6 Further, in the survivors of myocardial infarction, an association of reduced HRV with increased risk of mortality and subsequent events of arrhythmic events was noted. 7 In recent times, it has been found that reduced variability of the heart rate could prove to be an indicator of risks of adverse events and mortality in patients with several diseases that affects the autonomic nervous system more so in the critically ill. 8,9 A healthy individual with normal functioning autonomic nervous system usually has a high variability in the heart rate whereas a lower variability often indicates dysfunction and inadequate adaptability of the autonomic nervous system. 10 Patients who are critically ill may experience an altered sympathetic-parasympathetic balance and a key task of the ANS is to regulate the cardio-respiratory interaction to ensure an optimal oxygen supply. 4 Loss of this balance results in variability of the heart rate. It has been suggested that in the ICU and emergency settings, the heart rate variability analysis can help predict at risk patients and those who can be benefitted from early admissions and management. 11,12 Like many other physiological phenomena, HRV demonstrates intricate relationships between organs, tissues, and cells. The phenomena of HRV is based on the concept that the state of organs, tissues, and cells and the strength of interaction among them are affected by illness and treatment which in turn affects the HRV in an individual. 13 Heart rate variability can arise from both cardiac and noncardiac events, frequently reflecting overall/global rather than local processes.
Many methods for HRV analysis have been developed, some more accurate than others. All the methods essentially are derived from three general domains, namely, the time domain, frequency domain, the regularity domain. Continuous ECG recordings that are digitally processed at a frequency of at least 125 Hertz (Hz) are the source of HRV data. Several more adult studies have demonstrated that decreases in HRV are correlated with clinical outcomes and precede clinical decline. 11 As a result, HRV can be useful in the ICU as a gauge of clinical prospects. It is not yet possible to translate these findings into useful information that emergency room doctors can use right now. Use of heart rate variability is an uncommon clinical practice due to its limited applicability in acute care. Determining the function of heart rate variability as a predictive factor in critically ill patients admitted to the ICU was the goal of this study. The additional goals were to determine the mortality rate, the length of ICU/hospital stays, and how those factors are related to HRV.

Study design and ethical considerations
This cross-sectional study took place at the Department of Medicine at a rural tertiary care teaching hospital in Maharashtra, India. After receiving approval from the Institutional Ethics Committee of Datta Meghe Institute of Medical Sciences, the study was started. Studying took place from December 2020 until November 2022 with IEC number DMIMS (DU)/IEC/2022/1231. Dated 08/07/2022. On admission all consenting patients/patient relatives were asked to sign a written informed consent form (in the language best understood by them). The information regarding each patient was kept confidential & was not revealed at any point in time.

Sample
The study participants were critically ill patients admitted to the Medicine intensive care unit (ICU) of Acharya Vinoba Bhave rural hospital of Jawaharlal Nehru Medical College Sawangi, Wardha.

Inclusion criteria were
• Patient admitted to the Intensive Care Units (ICUs) for more than 24 hours • Patients or their relatives who gave consent to participate in study.
Exclusion criteria were as follows: • Patients who did not give consent • Patients who were on any sedative medications, drowsy state, or disoriented at a time of admission • Individuals who were currently taking medications that may interact with the HRV analysis. drugs like beta blockers • Age < 18 years

Sample size calculation
The minimum total sample size needed was calculated to be 215 using the below equation. We used a sample size of 225 for our study considering a drop out of 5%.

Procedure
The selected participants were evaluated for 24 hours. The principle investigator carried out all measures. Heart Rate Variability analysis was assessed for 24 hours and the APACHE 4 score was calculated within 48 hours of admission.
The following data were collected using structure pilot tested study proforma which includes demographic profile, co-morbidity status, Glasgow coma scale score, vitals (Systolic and Diastolic blood pressure: Blood pressure was measured twice using mercury sphygmomanometer, 5 minutes apart and average of both the reading was taken. Mean arterial pressure: It was calculated by formula Diastolic pressure + 1/3 (Systolic blood pressure -Diastolic blood pressure), Respiratory rate), biochemical parameters (Total leucocytes count, Hematocrit, Sodium, Creatinine, Urea, Serum lactate, Fasting blood sugar, Post-meal blood sugar, Hba1c, Albumin, Bilirubin, Total cholesterol, Low density lipid, High density lipid, Triglycerides) was processed in automated chemical analyzer (VITROS 5600) and heart rate variability parameters.
Holter machine: Holter system helps to record, analyze, display ECG signal, edit and create HRV analysis. The recorder was used under the direction of be doctors and trained healthcare professionals. The analysis result is for clinical reference only and the final diagnosis must be made by physicians.
On the first day of enrolment, each participant in the research was assessed with a 3-channel Holter (Cardios ® CardioLight model, So Paulo, Brazil). The 24-hour measure was implemented without interfering with ICU care as usual. Using a system created especially for this purpose (Cardios ® ), data analysis to determine HRV was carried out. This system automatically produces the following indices of HRV in the time domain: Normal-to-Normal (NN) average interval, standard deviation of the NN interval (SDNN), square root of the squared mean of the difference between subsequent NN-intervals (r-MSSD), and percentage of NN intervals that deviated by more than 50 ms from adjacent NN-intervals (pNN50); frequency domain with fast Fourier Transform (FFT) method: Power in total, Very Low Frequency (VLF) power, Low Frequency (LF) power, and High Frequency (HF) power.
APACHE score was calculated with the help of clinical history, vitals and biochemical analysis within 48 hours of admission. After assessing HRV and APACHE4 Score, participants were moniterd till the discharge from hospital and death. Participants were also monitored for use of vasopressor, Ventilator.

Statistical analysis
We performed statistical analysis using the Statistical Software STATA MP Version 14.0. For categorical variables, descriptive statistics, frequency analysis, and percentage analysis were used to describe the data, while mean and SD were used for continuous variables. The Chi-square test was used to determine the significance of categorical data (that is, to test the difference in proportions between two groups). The probability value of 0.05 is used as the significance level in all of the above statistical tools. Figure 1 shows the patients flow diagram. The total number of ICU patients admitted during the study was N = 1640. Individuals were excluded from the study for the following reasons: patients who are currently taking medications that may interact with HRV analysis drugs like beta blocker (N = 330), patients who are on sedatives (N = 490), patients who are drowsy and disoriented at the time of admission (N = 550), patients who did not give consent (N = 6) and due to technical error of hrv analysis (N = 19) and loss to follow up due to leaving against medical advice (N = 20) excluded from the study. A total of 225 subjects were eligible for study. 21 Table 1 shows a mean age of 54.0 years (16.3), systolic blood pressure of 124.4 mmHg (16.9), diastolic blood pressure of 83.2 mmHg (14.7), and respiratory rate of 26.6 (6.2) respectively. The most common co-morbidity affecting the study population was hypertension which was seen in 146 (64.9%) patients. The other co-morbidities were sepsis 82 (36.4), stroke 63 (28.4), diabetes 67 (29.8), chronic liver disease 20 (8.9). Male study participants had significantly higher no. of sepsis 65 (43.0%) as compared to in females 17(23.0%) and Chronic liver disease was seen in more males 18 (11.9%) as compared to females 2 (2.7%). 40% of female participants had a stroke which was significantly higher than males 21.9%. Table 2 shows the distribution of biochemical parameters among the study population. There was no significant difference between males and females in biomedical parameters.  Table 4 shows the variables High frequency, Low frequency/High frequency ratio were significantly associated (p < 0.05) with SDANN and SDNN. Table 5 shows the variables High frequency, Low frequency/High frequency ratio and Outcome were significantly associated (p < 0.05) with APACHE 4 score. HF parameter was abnormal in 83% of patients who died as compared to 44% of the survived group (p-value < 0.001). LF/HF parameter was abnormal in 86% of patients who died as compared to 54% of the survived group (p-value < 0.001). Table 6 shows the variables low frequency, High frequency, Low frequency/High frequency ratio, APACHE 4 score, GCS score were significantly associated (p < 0.05) with the variable's outcome. Table 7 shows the variables Urea, Low density lipid, Use of mechanical Ventilator, FiO 2 , PaO 2 significantly associated with survival were (p < 0.05) with the Sepsis, SDANN, and SDNN, APACHE 4, GCS score, LF, HF, and LF/HF ratio.        predicts mortality: Death with a sensitivity of 86%, and specificity of 31%. At a cutoff of RMSSD was ≥31 by ROC it predicts mortality: Death with a sensitivity of 67%, and as and specificity of 44%. At a cutoff of LF ≤9.2 by ROC it predicts mortality: Death with a sensitivity of 47%, and Specificity of 81%. At a cutoff of HF ≤39.5 by ROC it predicts mortality: Death with a sensitivity of 47%, and a specificity of 69%. At a cutoff of LF/HF ≤0.3 by ROC it predicts mortality: Death with a sensitivity of 60%, and a specificity of 64%. At a cutoff of APACHE 4 score was ≥71 by ROC it predicts mortality: Death with a sensitivity of 98% and specificity of 88%. Table 9 shows the coefficients representing the log-odds of the outcome variable for each predictor variable. APACHE 4 score shows the increase in score with the odds of the outcome variable while age and LF shows negative association.

Discussion
In this study, we show that heart rate variability is clinically useful for critically ill patients' severity rating (APACHE 4) and prognosis. When compared to patients who failed to improve and had to be brought to the ICU, individuals whose critical illness stabilised tended to have larger heart rate variability and showed bigger hour-by-hour rises. Mean and hourly heart rate variability also predicted variations in survival.
The gold standard for clinical HRV measurement is HRV recordings, which are influenced by circadian rhythms, core body temperature, metabolism, the sleep cycle, and the renin-angiotensin system. 14 varying heart rate The quantity of HRV that was detected during monitoring intervals that could last from 1 min to >24 hours is quantified by time   In the prospective, observational cohort study conducted by Bishop DG et al., 17 out of the total 55 patients, 35 required invasive ventilation (55%), 19 required inotropic support (35%) and 2 needed renal replacement therapy (4%). The median duration of stay in the ICU was 2.6 days (1-22 days). In the retrospective cohort study conducted by Liu N et al. 18 on 342 patients, 19% had 30 days of in-house mortality and the remaining 81% survived. In our study 225 patients, mean duration of hospital stay. 10.5 % died, 10.1% discharged. A significant association was observed between GCS score and patient outcome as the mean GCS score was significantly higher for the group who got discharged (13.7 AE 2.8) in comparison to the group who died (7.4 AE 1.4).
Studies by Maheshwari A et al., 19 Graff B et al., 20 Hadase M et al. 21 and Ong ME et al. 22 also reported an association between low heart rate variability with sudden cardiac death, stroke outcome, prognosis in heart failure and risk of cardiac arrest respectively. In our study, factors associated with mortality were low HF, low LF, low LF/HF ratio, high APACHE IV Score and low GCS score.
Takase et al. 23 demonstrated that SDANN lower than 30 ms had greater sensitivity and specificity (92%) than SDANN higher than 20 ms (31% sensitivity and 100% specificity) to detect autonomic dysfunction and cardiac events in cardiac autonomic neuropathy. In our study diagnostic accuracy of HRV was analyzed and it was higher in APACHE 4 (89.8%) and LF (74.1%).

Limitations of the study
This was a single centre study. Results from multi centre studies will help in generalizing the results. The results would have been more robust with a larger cohort. All the HRV were not taken into consideration.