Association between heart rate and mortality in patients with septic shock: an analysis revealed by time series data

Abstract

Background

Heart rate is crucial for patients with septic shock, but there are few studies on the scope of heart rate. Therefore, we studied the relationship between different heart rates and mortality of critically ill patients with septic shock, and explored the optimal heart rate range, in order to provide new insights for clinical treatment of septic shock.

Methods

This retrospective study utilized time-series heart rate data from the Medical Information Mart for Intensive Care (MIMIC) IV database. Patients with septic shock were identified as the Sepsis 3.0 criteria and received vasopressor therapy in the first 24 h since ICU admission. We calculated the time-weighted average heart rate (TWA-HR) based on the time-series data. The restricted cubic spline (RCS) analysis was employed to investigate the nonlinear relationship between heart rate and 28-day mortality, aiming to explore the optimal heart rate control target for septic patients and using this target as the exposure factor. The primary outcome was 28-day mortality, and the secondary outcome were ICU and in-hospital mortality. For the original cohort, we applied the log-rank test to infer the relationship between heart rate and mortality. To control for bias introduced by confounders, we utilized propensity score matching (PSM) to reduce imbalances between normal TWA-HR and high TWA-HR groups, and we established a series of models [the multivariable Cox model, matching weight (MW)-adjusted Cox model, multivariable logistic regression, MW-adjusted logistic regression, and doubly robust model] as sensitivity analyses and subgroup analyses to demonstrate the robustness of our findings.

Results

A total of 13492 patients were included in our study. The RCS analysis based on Cox and logistic regression showed increased risk of mortality (P < 0.001, non-linear P < 0.001) when TWA-HR > 85 beats per minute (bpm). The log-rank test revealed in terms of the 28-day mortality, the hazard ratio (HR) (95% confidence interval [CI]) was 1.92 (1.78—2.06, P < 0.001) for patients with high TWA-HR compared to normal TWA-HR group. Similarly, for the ICU mortality, the HR (95% CI) was 1.64 (1.52—1.78, P < 0.001), and for the in-hospital mortality, the HR (95% CI) was 1.61 (1.48—1.76, P < 0.001). Collectively, the sensitivity analysis consistently demonstrated higher 28-day mortality, ICU mortality, and in-hospital mortality in patients with TWA-HR > 85 bpm.

Conclusion

Patients with septic shock whose heart rate was controlled no more than 85 bpm during ICU stay received survival benefit in terms of 28-day, ICU and in-hospital mortality. 

Introduction

Heart rate is one of the vital signs of the human body, regulated by the autonomic nervous system, and is easily influenced by various diseases. Therefore, it has gradually become an important indicator for predicting the prognosis of certain diseases. However, it can also serve as a risk factor for certain diseases and is closely associated with the occurrence and progression of these diseases. Tachycardia, in particular, has been identified as an independent factor contributing to mortality in septic patients and has received significant attention in recent years.

Sepsis is defined as a syndrome of multiple organ dysfunction caused by a dysregulated host response to infection and is one of the leading causes of mortality in the intensive care unit (ICU). Studies have shown that more than 50% of sepsis patients experience sepsis-induced diastolic cardiac dysfunction, leading to inadequate oxygen and substrate supply to meet the metabolic demands of the body [1]. In the state of sepsis, the body may produce excessive catecholamines, resulting in tachycardia and increased myocardial oxygen consumption. Additionally, the shortened diastolic filling time due to increased heart rate may affect coronary perfusion, further contributing to myocardial dysfunction [2]. Ultimately, these factors contribute to poorer clinical outcomes in sepsis patients. Therefore, heart rate serves as an important indicator in sepsis, as an elevated heart rate during sepsis represents a compensatory mechanism in response to myocardial injury and sympathetic nervous system activation, while also reflecting the circulatory function status. Studies have shown that the use of β-blockers during sepsis can slow down heart rate, reduce myocardial oxygen consumption, provide cardiac protection, exert anti-inflammatory effects, decrease mortality, and improve prognosis [3].

Although numerous studies have shown that the paramount importance of heart rate control in septic shock patients, and heart rate is a dynamically changing index for these individuals, few investigations have employed time series heart rate data to assess the impact of overall heart rate target on mortality during ICU stay. Therefore, this study aims to elucidate the influence of heart rate on mortality in septic patients, which provided novel insights into heart rate management in the clinical care of septic shock.

Methods

Data source and participants

This study is a retrospective analysis, and the data are obtained from the Medical Information Mart for Intensive Care (MIMIC)-IV V2.0 database, which included 76,943 patients at Beth Israel Deaconess Medical Center between 2008 and 2019. Access to the MIMIC-IV database was granted to Yi-Le Ning, a member of our research team (Record ID 40974208). The Institutional Review Board of Beth Israel Deaconess Medical Center (IRB #2001P001699) waived the requirement for informed consent, as the patient data had undergone de-identification.

The initial study population was selected based on the criteria defined by the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). According to the Sepsis-3 criteria, patients accompanied with suspected infection and Sequential Organ Failure Assessment (SOFA) score ≥ 2 were identified as sepsis. Suspected infection was defined as the presence of blood pathogen cultures or antibiotic use. The septic shock subjects were identified as the septic patients who were treated with at least one vasopressor (vasopressin, dopamine, dobutamine, epinephrine, milrinone, norepinephrine, phenylephrine, angiotensin II) therapy in the first 24 h since ICU admission.

To ensure the continuity and density of the time-series heart rate data, the patients were required to have at least one heart rate record per day and no fewer than 3 heart rate records in total. We employed the Stineman polynomial algorithm for missing value imputation and align the time-series heart rate data into 1-h resolution.

The following patients were excluded: (1) Vasopressors were not initiated in the first 24 h; (2) ICU length of stay (LOS) ≤ 24 h; (3) Patients younger than 18 years old; (4) Unavailability of ICD information; (5) Errors in the date of death outside the hospital; (6) Heart rate data was not recorded daily during the ICU stay or with total count of heart rate records less than 3. Finally, the eligible subjects meeting the inclusion criteria were included in this study.

Data extraction

To facilitate efficient data extraction and manipulation, we developed and thoroughly tested structured query language (SQL) codes using DBeaver Community version 23.0.0. This powerful database management tool allowed us to create and refine SQL queries that accurately captured the desired information from the MIMIC-IV database.

Once the SQL codes were optimized, we leveraged the capabilities of the DBI package to seamlessly execute these queries within the R programming environment. By utilizing this package, we were able to establish a direct connection between R and the MIMIC-IV database, enabling the creation of relevant tables and the import of associated variables into the R global environment. This streamlined process of data collection ensured that the necessary information was readily accessible for subsequent analysis and exploration. By combining the strengths of SQL, DBeaver, and the DBI package, we were able to efficiently gather and organize the data required for our study.

Exposure and outcomes

We calculated the patients' time-weighted average heart rate (TWA-HR) as an indicator of their overall heart rate control level during their ICU stay, which was determined by calculating the area under the dynamic heart rate-time curve and dividing the result by the total duration of the patient's ICU stay. The TWA-HR serves as a superior measure of heart rate control compared to traditional metrics, such as the mean or median heart rate, as it takes into account the duration and variability of heart rate measurements over time. By assigning greater weight to heart rate values that persist for longer periods, the TWA-HR provides a more accurate representation of the patient's overall heart rate control throughout their ICU stay. This approach enables a more comprehensive assessment of the relationship between heart rate and mortality in sepsis patients, offering valuable insights for optimizing clinical management strategies.

To explore the nonlinear relationship between heart rate and 28-day mortality, we utilized the restricted cubic spline (RCS) analysis, which can be viewed as a piecewise polynomial that requires a continuous second derivative at each knot [4]. The RCS is a commonly used method for studying nonlinear relationships. It allows for the fitting of curved relationships between the independent variable and the outcome, enabling an analysis of the association between continuous exposure and outcome. In the current study, RCS curve were fitted with 3 knots based on the Cox and logistic regression model utilizing the rms package [5]. This approach enabled us to identify the optimal threshold for TWA-HR, which served as the exposure factor in our study.

The primary outcome of interest in this investigation was 28-day mortality, while secondary outcomes included mortality within the specific context of the intensive care unit (ICU) and the hospital setting. By focusing on these outcomes, we aimed to provide a comprehensive assessment of the impact of TWA-HR on patient survival across different time points and clinical environments.

Covariates

Baseline characteristics: gender, age, weight, race. Vital signs: temperature, mean arterial pressure (MAP). Comorbidities: heart failure (HF), hypertension, atrial fibrillation (AFIB), diabetes mellitus type 2 (T2DM), chronic renal disease, liver disease, chronic obstructive pulmonary disease (COPD), coronary artery disease (CAD), stroke, malignancy. Severity scores: simplified acute physiology score II (SAPS II), SOFA score, Charlson comorbidity index and time-weighted average Richmond Agitation-Sedation Scale (TWA-RASS). Laboratory tests: white blood cell (WBC) count, hemoglobin, platelet, potential of hydrogen (pH), partial pressure of oxygen (PO2), partial pressure of carbon dioxide (PCO2), lactate, creatinine. Interventions within the first 24 h of ICU admission: tidal volume, plateau pressure, positive end-expiratory pressure (PEEP), pulse index continuous cardiac output (PiCCO), non-invasive cardiac output monitoring (NICOM), mechanical ventilation, sedative therapy, albumin use, beta blockers use, fluid amount, and max vasoactive-inotropic score (VIS) of dopamine(DA), dobutamine(DBA), epinephrine(EPI), norepinephrine(NE) and phenylephrine(PE). The formula for calculating VIS is as follows: VIS = 10,000 × Vasopressin dose (U/kg/min) + 100 × Epinephrine dose (μg/kg/min) + 100 × Norepinephrine dose (μg/kg/min) + 50 × Levosimendan dose (μg/kg/min) + 25 × Olprinone dose (μg/kg/min) + 20 × Methylene blue dose (mg/kg/h) + 10 × Milrinone dose (μg/kg/min) + 10 × Phenylephrine dose (μg/kg/min) + 10 × Terlipressin dose (μg/min) + 0.25 × Angiotensin II dose (ng/kg/min) + Dobutamine dose (μg/kg/min) + Dopamine dose (μg/kg/min) + Enoximone dose (μg/kg/min). We used part of this to calculate the maximum value of each vasoactive drug within the first 24 h to represent its peak dose.Variables with missing values surpassing 40% were excluded from the model as covariates for analysis, in accordance with the recommendations outlined by Jakobsen et al. [6]. For variables with missing data below the 40% threshold, we employed multiple imputation techniques, as implemented by the MICE package [7], to handle the missing values effectively. This approach ensures that the analysis is based on a comprehensive set of covariates while minimizing the potential biases introduced by excessive missingness in the data.

Statistical methods

To assess the normality of data distribution, we applied the Shapiro–Wilk normality test (N <  = 2000) or Kolmogorov–Smirnov test (N > 2000). The F-test was used to evaluate the equality of variances. When the data demonstrated normal distribution across groups and the homogeneity of variance test yielded no significant differences, the t-test was applied for continuous covariates. In cases where these conditions were not met, the Wilcoxon test was deemed appropriate. Categorical covariates were analyzed using the Chi-square test, while Fisher's exact test was utilized when the sample size was less than 40. Continuous variables were expressed as mean (standard deviation), and categorical variables were presented as numerical values (percentage).

To robustly adjust for covariates and enhance the reliability of our findings, we employed both propensity score matching (PSM) and matching weight (MW) based on the propensity score. MW is a weighting method that assigns weights to each observation based on the propensity score [8], allowing for the balancing of covariates between treatment groups. By incorporating MW, we aimed to further strengthen the robustness of our results and account for potential confounding factors. This weighting scheme helps to create a pseudo-population in which the distribution of covariates is balanced between the treatment groups, thereby reducing the impact of confounding factors on the estimated treatment effect. The propensity score, derived from the logistic regression model, served as the foundation for further propensity score-based analysis. PSM was conducted using the Matching package, with cohorts matched 1:1 using the Match function with a caliper width equal to 0.2 of the standard deviation of the logit of the propensity score. Covariate balance was evaluated by assessing whether the absolute values of standardized mean difference (SMD) of all covariates between groups exceeded the threshold of 0.1.

Kaplan–Meier curve was used to visually compare the survival of different groups. The log-rank test is then used to quantify the difference with the survival package. To ensure the robustness of our findings, we conducted sensitivity analyses for 28-day, ICU, and in-hospital mortality using a range of models. These included the Multivariable Cox model adjusted with all covariates, Multivariable Cox model adjusted with all covariates using MW, Multivariable logistic model adjusted with all covariates, Multivariable logistic model adjusted with all covariates using MW, and doubly robust estimation (survey-weighted generalised linear model) with all covariates using MW. The proportional hazards assumption was examined through the analysis of Schoenfeld residuals. When this assumption was found to be statistically significant, time-changing covariates with time-transform features were incorporated into the Cox regression model or survey-weighted Cox model using the tt function. By employing a comprehensive suite of statistical methods, rigorous testing, and robust adjustment techniques, including both PSM and MW, we aimed to provide a thorough and reliable assessment of the relationships between the variables of interest and the outcomes under investigation, while minimizing potential biases and ensuring the validity of our conclusions.

We conducted subgroup analyses to explore the relationship between heart rate and mortality rates in different populations. The subgroups analyzed included age, gender, SAPS II, the use of PiCCO, Albumin and NICOM within 24 h of admission, HF, Hypertension, AFIB, T2DM, and CAD. These analyses aimed to provide further insights into the association between heart rate and mortality rates in specific patient populations.

All statistical analyses were performed using R version 4.2.3, with statistical significance set at a threshold of P < 0.05.

Results

Baseline characters and grouping

We initially screened a total of 34,677 sepsis patients, and ultimately, 13,492 were included in the final cohort for analysis (Fig. 1). Supplementary Table S1 presents the baseline characteristics of the original cohort. In this cohort, the average age was 67.05 ± 14.42 years, with women accounting for 39.23% (n = 5,293). The mean SAPS-II score was 44.62 ± 14.37, the average SOFA score was 7.61 ± 3.68, the mean Charlson comorbidity index was 5.32 ± 2.84, and the TWA-RASS was -1.21 ± 1.24.

Fig. 1

The study flow chart

To explore and visualize the relationship between TWA-HR and the risk of 28-day mortality for the population, we utilized the RCS analyses fitted with the Cox and logistic regression, both of these 2 models revealed the L shape (P < 0.001, nonlinear P < 0.001 for both 2 models), and 85 beats per minute (bpm) as the cutoff value (Fig. 2). It was observed rapaidly increased risk of when TWA-HR exceeded 85 bpm. Thus, we divided the included patients into two groups: TWA-HR ≤ 85 bpm were classified as the normal TWA-HR group, while TWA-HR > 85 bpm were labeled as the high TWA-HR group. Among the original cohort, there were 6,798 patients in the normal TWA-HR group and 6,694 patients in the high TWA-HR group, as shown in flowchart (Fig. 1). Additional File 1 and Additional File 2 respectively illustrate the frequency distribution and range distribution of TWA-HR.

Fig. 2

The nonlinear relationship of TWA-SHR and the risk of 28-day mortality fit by Multivariable Cox regression and Multivariable logistic regression with RCS analyses. A The nonlinear relationship of TWA-SHR and the risk of 28-day mortality fit by Multivariable Cox regression with RCS analyses. B The nonlinear relationship of TWA-SHR and the risk of 28-day mortality fit by Multivariable logistic regression with RCS analyses

The baseline characteristics before and after PSM of the cohort were presented in Table 1. Before PSM, there were significant differences in baseline characteristics between the normal TWA-HR and high TWA-HR group. Patients in the normal TWA-HR group were generally older (69.23 ± 13.10 vs 64.83 ± 15.33) and had a higher proportion of male (63.06% vs 58.44%) compared to the high TWA-HR group. The normal TWA-HR group also had a higher prevalence of comorbidities, including hypertension (70.31% vs 63.4%), coronary artery disease (45.54% vs 36.18%), chronic kidney disease (26.14% vs 22.96%). While in terms of laboratory findings, the high TWA-HR group had higher WBC count (15.13 ± 9.79 vs 13.81 ± 8.86), lactate (2.91 ± 2.4 vs 2.48 ± 1.87) and creatinine (1.67 ± 1.61 vs 1.58 ± 1.62). Regarding treatment and outcome measures, patients in the high TWA-HR group were more likely to require hemodynamic monitoring (PiCCO or NICOM). They also had longer ICU and hospital LOS, and higher SAPS II and SOFA scores, indicating a higher severity of illness. These differences in baseline characteristics highlight the need for PSM to balance the covariates between the two groups and reduce potential confounding factors when evaluating the impact of TWA-HR on mortality outcomes.

Table 1 Baseline characteristics before and after propensity score matching of cohort

After PSM, the baseline characteristics between the normal TWA-HR and high TWA-HR were well-balanced. The SMDs for all covariates were less than 0.1 (Supplementary Tables S2-S4, Supplementary Fig. S1), indicating that the matching process effectively reduced the differences in observed baseline characteristics between the two groups. By achieving a balanced distribution of baseline characteristics between the normal TWA-HR and high TWA-HR groups, the PSM process minimized the potential confounding effects of these variables on the relationship between TWA-HR and mortality outcomes. This allows for a more reliable comparison of the impact of TWA-HR on 28-day, ICU, and in-hospital mortality in the matched cohort.

Primary and secondary outcome and sensitivity analysis

We investigated the impact of high TWA-HR on 28-day mortality, ICU mortality, and in-hospital mortality using various statistical models. The primary outcome, 28-day mortality, was significantly higher in the high TWA-HR group compared to the normal TWA-HR group across all models. The K-M survival curve suggested that high TWA-HR is associated with an increased risk of 28-day mortality (Fig. 3A), and unadjusted log-rank test estimated a hazard ratio (HR) of 1.92 [95% Confidence interval (CI): 1.78—2.06; P < 0.001] (Supplementary Table S5).

Fig. 3

A Unadjusted Kaplan–Meier survival curve for 28-day mortality of original cohort. B Unadjusted Kaplan–Meier survival curve for ICU mortality of original cohort. C Unadjusted Kaplan–Meier survival curve for In-hospital mortality of original cohort. The differences for original cohort were significant

Secondary outcomes including ICU mortality and in-hospital mortality were also significantly higher in the high TWA-HR group. For ICU mortality (Fig. 3B), the unadjusted log-rank test showed an HR of 1.61 (95% CI: 1.48—1.76; P < 0.001) (Supplementary Table S6). Similarly, for in-hospital mortality (Fig. 3C), the unadjusted log-rank test resulted in an HR of 1.61 (95% CI: 1.48—1.76; P < 0.001) (Supplementary Table S6).

Sensitivity analysis and subgroup analysis

We conducted sensitivity analyses using various models, including: Multivariable Cox model adjusted for all covariates, Cox model adjusted for all covariates using MW, Multivariable logistic model adjusted for all covariates, logistic model adjusted for all covariates using MW, and doubly robust estimation with all covariates. As shown in Table 2, all models had P-values < 0.001, highlighting the robustness of the study's findings. Compared to the normal TWA-HR group, the high TWA-HR group demonstrated a significantly higher risk of 28-day mortality, ICU mortality, and in-hospital mortality.

Table 2 Primary and secondary outcome analyses with different models for cohort

The subgroup analyses revealed consistent results across most subgroups (Fig. 4). The association was significant in both age groups (< 60 years: HR 1.6, 95% CI: 1.34—1.9; ≥ 60 years: HR 2.13, 95% CI: 1.95—2.32) and both genders (Female: HR 2.0, 95% CI: 1.77–2.25; Male: HR 1.83, 95% CI: 1.66–2.03). Similarly, the association remained significant regardless of SAPS II score (< 49: HR 1.89, 95% CI: 1.67—2.15; ≥ 49: HR 1.72, 95% CI: 1.56—1.89), albumin administration (Yes: HR 2.65, 95% CI: 2.22—3.16; No: HR 1.73, 95% CI: 1.59—1.89), and the presence or absence of comorbidities such as heart failure, hypertension, atrial fibrillation, coronary artery disease, and type 2 diabetes mellitus. The use of PiCCO or NICOM monitoring did not significantly modify the association between high TWA-HR and mortality risk (PiCCO: HR 1.48, 95% CI: 0.6–3.65, P = 0.395; NICOM: HR 1.15, 95% CI: 0.86–1.55, P = 0.344). However, it should be noted that the sample sizes for these subgroups were relatively small, which might have limited the power to detect significant interactions. Similar findings were observed for ICU and in-hospital mortality (Supplementary Fig. S2). Our subgroup analyses demonstrate that the association between high TWA-HR and increased mortality risk is consistent across various patient subgroups, reinforcing the robustness of the findings. These results suggest that high TWA-HR could serve as a valuable prognostic marker for critically ill patients, irrespective of age, gender, disease severity, and comorbidities.

Fig. 4

Forest plot of subgroup analysis for 28-day mortality

Discussion

The RCS analysis in this study revealed a non-linear relationship between heart rate and mortality in septic patients. When the heart rate of septic patients was less than or equal to 85 beats per minute, the mortality rate remained relatively stable. However, when the heart rate exceeded 85 beats per minute, the mortality rate sharply increased. Among septic patients with heart rates greater than 85 beats per minute, the 28-day mortality rate was 1.77 times higher than that of patients with heart rates less than or equal to 85 beats per minute. The ICU mortality rate was 1.86 times higher, and the in-hospital mortality rate was 1.93 times higher. Moreover, variations in mortality rates were observed across different subgroups. In patients aged 60 years or older, those with comorbid AFIB, hypertension, or CAD, higher heart rates increased their risk of mortality. However, in the subgroup of patients with diabetes, those without comorbid diabetes had a higher risk of mortality associated with elevated heart rates.

Sepsis, one of the leading causes of mortality in intensive care units, was clinically defined only in the late 20th century, despite being recognized earlier. With the advancement of time, experts in the field have conducted further and more comprehensive research on the risk factors, pathogenesis, and treatment of sepsis, leading to a decrease in its mortality rate. Multiple organ dysfunction syndrome lies at the core of sepsis, and the cardiovascular system, as a crucial component of organ function, is often affected, making it a significant risk factor for sepsis-related mortality. Tachycardia, a common occurrence in ICU patients, is associated with poor prognosis in sepsis cases. During sepsis, the sympathetic nervous system plays a critical role in maintaining cardiac output and blood pressure [9]. This is achieved through changes in heart rate, vascular contractility, and vascular tone. In the early stages of sepsis, tachycardia serves as an important compensatory mechanism for myocardial cell damage. Accordingly, appropriate fluid resuscitation often leads to a reduction in heart rate. However, compensatory tachycardia can result in arrhythmias, including atrial fibrillation. If hemodynamic instability persists, heart rate control with agents such as landiolol may be necessary.

Furthermore, in patients with septic shock, the persistence of non-compensatory tachycardia occurs when elevated levels of endogenous and exogenous catecholamines impair the baroreflex response, leading to a state of adrenergic excess. This represents a manifestation of sympathetic overstimulation [10]. Patients who develop tachycardia within 24 h of initiating norepinephrine infusion have a threefold increase in the risk of mortality compared to those who do not experience tachycardia, possibly due to exhaustion of compensatory reflex mechanisms [10]. Prolonged tachycardia increases myocardial oxygen consumption, diminishes diastolic filling, and induces cardiac toxicity, thereby exerting deleterious effects on the heart [10]. These adverse effects are detrimental to patient prognosis and may even increase the risk of mortality, necessitating the implementation of heart rate control to improve patient outcomes and reduce mortality risk. Research has shown that the use of β-blockers can slow down the heart rate, prolong diastolic filling time, and reduce myocardial oxygen consumption [11], this can have an inhibitory effect on catecholamine-mediated cardiac toxicity and increase the body's sensitivity to catecholamines, thereby providing protection and anti-inflammatory effects on the heart, reducing mortality rates, and improving prognosis [3].

Furthermore, advancing age has an adverse impact on the mortality risk of septic patients. As age increases, our immune function tends to decline, rendering older individuals at higher risk for infection, severe infection, and prolonged duration of infection [12, 13]. Additionally, age influences the pharmacological treatment of septic patients. Possibly due to reduced tolerance and impaired hepatic and renal function in the elderly, medications such as beta-blockers and ACE inhibitors are less commonly used in patients aged 75 years and above. Studies have indicated that in elderly patients with heart failure, post-treatment heart rate with beta-blockers can serve as a predictor of overall mortality risk [14].

Research has demonstrated that sepsis significantly affects heart rate and increased the potential development or exacerbation of AFIB. The underlying mechanisms for this association are complex and multifactorial. Inflammatory mediators and neurohormonal imbalances during sepsis can directly impact the cardiac conduction system, resulting in atrial remodeling and the occurrence or persistence of AFIB. Concurrently, AFIB itself can influence the heart rate of septic patients [15]. AFIB causes ineffective atrial contraction and irregular ventricular response, thereby inducing unstable and elevated heart rates. The elevated heart rate, in turn, leads to further hemodynamic instability and exacerbates systemic inflammatory responses [16].

Long-standing hypertension can lead to functional and structural damage in the cardiovascular system, such as decreased myocardial contractility, myocardial ischemia, and myocardial fibrosis. These impairments may result in cardiac electrophysiological disturbances, further contributing to abnormal heart rates. Additionally, hypertension can cause endothelial dysfunction, including endothelial injury and impaired vasodilation, leading to increased vascular tone and enhanced inflammatory responses [17]. Moreover, hypertension can induce abnormalities in the immune system, including chronic inflammatory states and alterations in immune cell function, thereby reducing the patient's resistance to infections [18]. These factors may collectively contribute to an increased mortality rate in septic patients with comorbid hypertension.

However, within the subgroup of patients with diabetes, those without comorbid diabetes had a higher risk of mortality with elevated heart rates. There is considerable debate regarding the impact of diabetes and hyperglycemia on septic patients. A study utilizing Multivariable Cox regression analysis indicated that compared to HbA1c 48–52 mmol/mol (6.5–6.9%), the hazard ratios were 0.93 (0.87–0.99) for HbA1c 53–62 mmol/mol (7.0–7.8%), 1.05 (0.97–1.13) for HbA1c 63–72 mmol/mol (7.9–8.7%), 1.14 (1.04–1.25) for HbA1c 73–82 mmol/mol (8.8%-9.7%), and 1.52 (1.37–1.68) for HbA1c > 82 mmol/mol (9.7%), suggesting that glycemic control within a certain range can reduce the mortality risk in septic patients with diabetes [19]. Simultaneously, a retrospective cohort study demonstrated that diabetes was not associated with worse clinical outcomes in septic patients [20]. In some studies, hyperglycemia has been considered an adaptive response in the context of stress, exerting a certain protective effect in reducing mortality rate in septic patients [21, 22].

Certainly, our study has certain limitations. Firstly, the data for patients with heart rates below 60 beats per minute included only 140 cases, while those with heart rates above 130 beats per minute comprised only 30 cases, indicating insufficient data within these two ranges. Secondly, once the infection in septic patients is adequately controlled, the heart rate tends to decrease significantly and stabilize. This leads to an overall lower time-weighted average heart rate in our study, which imposes certain limitations. Thirdly, although we incorporated as many covariates as possible that could affect heart rate—such as mechanical ventilation parameters, sedation levels, dosage of vasoactive drugs, resuscitation fluid volume, albumin use, beta-blocker use, and other interventions—into the model for analysis, there may still be potential unaccounted-for covariates that could influence the study results. Future research should involve prospective, randomized, large-scale experiments to validate our findings and explore the specific mechanisms involved.

Conclusion

The risk of mortality in septic shock patients with high TWA-HR (≥ 85 bpm) associated with higher 28-day, ICU and in-hospital mortality. These findings provide new insights for the clinical management of septic shock.