Abstract
Atrial fibrillation (AF) is a common arrhythmia associated with increased morbidity and mortality among intensive care unit (ICU) patients. This study aimed to evaluate heart rate fluctuation (HRF) and its association with in-hospital mortality among ICU patients with AF. Atrial fibrillation (AF) is a common arrhythmia associated with increased morbidity and mortality among intensive care unit (ICU) patients. This study aimed to evaluate heart rate fluctuation (HRF) and its association with in-hospital mortality among ICU patients with AF. Atrial fibrillation (AF) is a common arrhythmia associated with increased morbidity and mortality among intensive care unit (ICU) patients. This study aimed to evaluate heart rate fluctuation (HRF) and its association with in-hospital mortality among ICU patients with AF. This study utilized the Medical Information Mart for Intensive Care (MIMIC)-III and MIMIC-IV databases. Patients with recorded AF at ICU admission were included. The primary exposure variables were heart rate change, measured using the initial heart rate at ICU admission and the maximum heart rate within the first six hours, and HRF, assessed by the median absolute deviation (MAD) of all heart rate measurements within the first 24 h. Logistic regression was used to assess the association with in-hospital mortality. A total of 13,475 patients were included in the analysis. For heart rate change, compared to the high to high group (initial and maximum heart rate > 110), the low to high group (initial heart rate ≤ 110 and maximum heart rate > 110) showed a higher mortality risk (Odds ratio [OR]: 1.30, 95% confidence interval [CI] 1.09–1.54; p = 0.003). Higher HRF was associated with increased mortality risk (p for trend = 0.002). Patients in the highest HRF quintile had a higher risk of in-hospital mortality (OR: 1.25, 95% CI 1.07–1.46; p = 0.004). Subgroup analyses revealed significant interactions between HRF and both rate control and initial heart rate. However, in patients receiving rate control therapy, HRF was not associated with mortality. Changes in heart rate and HRF were associated with in-hospital mortality in ICU patients with AF. However, in patients receiving rate control therapy, the impact of HRF on mortality is less significant and may not warrant as much clinical attention.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-025-15875-z.
Keywords: Atrial fibrillation, Heart rate fluctuation, Heart rate change, Mortality, Intensive care unit
Subject terms: Diseases, Medical research, Risk factors
Introduction
Atrial fibrillation (AF) is the most prevalent persistent cardiac arrhythmia, contributing to considerable morbidity and mortality1. AF detected in the ICU setting has been highlighted in recent studies for its critical importance in clinical management, given its significant impact on treatment strategies2. While AF may have been present but undiagnosed before the acute illness, its detection during such periods is associated with increased hospital stay duration3–5greater morbidity6–9and increased mortality3,9. Particularly in intensive care unit (ICU) patients, both new-onset and pre-existing AF are associated with higher mortality rates10,11. AF occurrence in ICU patients is frequent, driven by factors such as acute illness, electrolyte imbalances, sepsis, postoperative status, and cardiovascular diseases12,13. AF episodes also increase the risk of cardiovascular complications, stroke, and readmission4,14,15.
Heart rate fluctuation (HRF) has been identified as an easily accessible prognostic predictor in chronic heart failure, potentially linked to autonomic tone and exercise capacity16. In general ICU cohorts, 24-hour HRF has been recognized as a predictor of mortality17,18. In ambulatory AF populations, visit-to-visit heart-rate variability has also been independently associated with mortality19. However, the prognostic significance of short-term (24-hour) HRF in critically ill ICU patients with AF remains underexplored. Given the pronounced beat-to-beat variability inherent to AF and the unique hemodynamic stresses of critical illness, evaluating 24-hour HRF in this setting is particularly important.
We hypothesize that HRF may be positively correlated with the risk of mortality in ICU patients with AF. The aim of this study was to evaluate the variability of HRF and analyze its association with in-hospital mortality among ICU patients with AF. This study aims to demonstrate that HRF can serve as an important prognostic indicator in ICU AF patients, thereby underscoring the need for early prognostic evaluation.
Methods
Study population
This study utilized the Medical Information Mart for Intensive Care (MIMIC)-III and MIMIC-IV databases20,21. MIMIC-III and MIMIC-IV are large, publicly available databases containing de-identified patient data collected from the ICU at Beth Israel Deaconess Medical Center. We used the MIMIC-III CareVue subset version 1.4, which includes data from 2001 to 2008 and comprises over 27,000 ICU patient records. Additionally, we used MIMIC-IV version 3.0, which encompasses data from 2008 to 2022 and includes over 94,000 ICU patient records. MIMIC-III CareVue and MIMIC-IV employ distinct de-identification workflows and unique patient identifier schemes, ensuring that no patient record is duplicated across the two datasets21.
The study population consisted of all patients whose cardiac rhythm was recorded as AF at the time of ICU admission. AF at admission was defined as any recorded instance of AF from 24 h before ICU admission to 24 h after ICU admission. Patients who stayed in the ICU for less than 24 h were excluded from the analysis. Additionally, patients with fewer than four heart rate measurements within the first 24 h of ICU admission were excluded.
Exposures and outcome
This study analyzed two primary exposure variables. The first exposure variable was the change in heart rate following ICU admission. We defined heart-rate change by comparing the admission heart rate with the maximum rate during the first six hours in the ICU. We selected the first six hours after ICU admission to define maximum heart rate, as this interval corresponds to the early assessment and management phase in critical care and aligns with established resuscitation protocols22,23. Patients were categorized into four distinct groups based on these measurements: (1) patients whose initial heart rate was > 110 bpm and whose maximum heart rate within the first six hours was also > 110 bpm (high to high group), (2) patients whose initial heart rate was > 110 bpm but whose maximum heart rate within the first six hours was ≤ 110 bpm (High to Low group), (3) patients whose initial heart rate was ≤ 110 bpm but whose maximum heart rate within the first six hours was > 110 bpm (Low to High group), and (4) patients whose initial heart rate was ≤ 110 bpm and whose maximum heart rate within the first six hours was also ≤ 110 bpm (Low to Low group). This threshold of 110 bpm, used to define heart rate change, aligns with existing guidelines that recommend maintaining heart rate control below this level in AF patients24–28. The high to high group served as the reference group against which the outcomes of the other groups were compared.
The second exposure variable was HRF, which was measured by calculating the median absolute deviation (MAD) of all heart rate measurements taken within the first 24 h of ICU admission. HRF was quantified as the median absolute deviation (MAD) of all 24-hour heart-rate measurements, calculated as the median of |HRi − median(HR)|, where HRi denotes each individual heart rate observation. Based on the MAD values, patients were divided into five quintiles. The quintile with the lowest MAD values was used as the reference group for comparison with the other quintiles. The primary outcome variable of this study was in-hospital mortality, defined as death occurring at any point during the hospital stay. By evaluating these exposure variables and their association with in-hospital mortality, the study aimed to identify significant predictors and provide insights into the management of AF patients in the ICU setting.
Variables
In this study, several covariates were adjusted to analyze the association between the exposure variables and the outcome. Demographic variables included the patient’s age and sex. Vital signs included systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate, respiratory rate, body temperature, and oxygen saturation (SpO2) at the time of ICU admission. Comorbidities included congestive heart failure, ischemic heart disease, peripheral vascular disease, chronic pulmonary disease, hypertension, diabetes, chronic kidney disease, end-stage kidney disease, chronic liver disease, and metastatic cancer, with the presence of each comorbidity defined by the respective diagnostic codes. Additionally, the study included variables for cardiac surgery and aorta surgery at the time of admission, defined by the presence of relevant ICD procedure codes. Major interventions and procedures performed within the first six hours of ICU admission included continuous kidney replacement therapy (CKRT) and mechanical ventilation. We extracted the infusion rate of norepinephrine, dopamine, dobutamine, epinephrine, and vasopressin recorded closest to ICU admission (within the first six hours) to reflect the initial hemodynamic support phase and included the Glasgow Coma Scale (GCS) score as a covariate. For each vasoactive agent, the infusion rate was set to zero in patients without any administration, such that cohort-level mean dose rates reflect both recipients and non‐recipients. The study considered the use of rate control and rhythm control medications. Rate control was defined as the use of beta-blockers, calcium channel blockers, or digoxin2. Rhythm control medications included the use of amiodarone29,30. Furthermore, the study included the presence of positive blood culture results from 48 h before to 6 h after ICU admission. New-onset AF was defined in patients who had only sinus rhythm recorded on all chart-event heart-rhythm entries prior to ICU admission, with no AF documented in that pre-admission period, and who then experienced their first AF episode between ICU admission and 24 h thereafter.
Statistical analysis
Continuous variables were presented as mean ± standard deviation, and categorical variables were presented as counts and percentages. Logistic regression models were utilized to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between heart rate change or HRF and the risk of in-hospital mortality. In Model 1, adjustments were made for age, sex, and the presence of new-onset atrial fibrillation. Model 2 included the variables in Model 1 with additional adjustments for heart rate, SBP, DBP, respiratory rate, and SpO2. Model 3 further adjusted for congestive heart failure, hypertension, diabetes, chronic kidney disease, end-stage kidney disease, chronic liver disease, metastatic cancer, intubation, continuous kidney replacement therapy, positive blood culture, cardiac surgery, aorta surgery, and the GCS. Model 4 included the variables in Model 3 with additional adjustments for norepinephrine, dopamine, dobutamine, epinephrine, vasopressin administration, rhythm control, and rate control. For each model, trends were analyzed using p-values by treating the groups as continuous variables. Potential effect modifications by age, sex, congestive heart failure, new-onset AF, rhythm control, rate control, inotrope use, heart rate, and SBP were assessed through subgroup analyses and tests for interaction terms, based on previous research indicating these factors’ relevance to mortality risk31–33. Additionally, we explored the potential non-linear relationship between continuous HRF (MAD) and in-hospital mortality by fitting restricted cubic splines within the fully adjusted logistic regression model. To account for potential bias from unequal sampling frequency, we first quantified the total number of heart-rate measurements per patient. As a sensitivity analysis, we included the total number of heart-rate measurements per patient as an additional covariate in our multivariable logistic regression models. Multicollinearity was assessed by calculating variance inflation factors for all independent variables, with rescaled values reported. All statistical analyses were conducted using the latest version of R software (R version 4.5.0). P-values were two-sided, and a significance level of 0.05 was used.
Results
Baseline characteristics
A total of 16,118 patients were admitted to the ICU with AF. Of these, 62 patients were excluded due to having fewer than four heart rate measurements within the first 24 h, 169 were excluded due to missing demographic or vital sign data, and 2,412 were excluded because their ICU stay was less than 24 h. Ultimately, 13,475 patients were included in the analysis (Fig. 1).
Fig. 1.
Flow diagram of the study population. ICU intensive care unit, MIMIC Medical Information Mart for Intensive Care.
In-hospital mortality occurred in 2,707 patients, representing 20.1% of the total population. The mean age of the included patients was 74.31 ± 11.6 years, and 57.8% of the patients were male (Table 1). The average heart rate at admission was 97.0 ± 23.8 beats per minute (bpm), and the average maximum heart rate within the first six hours was 106.7 ± 25.1 bpm. New-onset AF was observed in 6.7% of the patients. Rhythm control medications were used by 10.2% of the patients, and rate control medications were used by 25.3%. The average length of ICU stay was 121.5 ± 152.1 h.
Table 1.
Baseline characteristics.
| Variable | Total (N = 13475) |
In-hospital mortality (N = 2707) |
No in-hospital mortality (N = 10768) |
p-value |
|---|---|---|---|---|
| Age (years) | 74.314 ± 11.641 | 75.928 ± 11.29 | 73.909 ± 11.693 | < 0.001 |
| Male | 57.8% (7784) | 56.5% (1530) | 58.1% (6254) | 0.148 |
| Heart rate (/min) | 96.963 ± 23.841 | 99.549 ± 23.892 | 96.313 ± 23.785 | < 0.001 |
| Systolic blood pressure (mmHg) | 120.478 ± 25.186 | 117.457 ± 25.863 | 121.238 ± 24.957 | < 0.001 |
| Diastolic blood pressure (mmHg) | 73.772 ± 581.315 | 65.808 ± 19.587 | 75.775 ± 650.208 | 0.112 |
| Respiratory rate (/min) | 20.535 ± 6.591 | 21.669 ± 6.944 | 20.25 ± 6.468 | < 0.001 |
| SpO2 (%) | 96.2 ± 4.636 | 95.654 ± 5.759 | 96.338 ± 4.298 | < 0.001 |
| Maximum heart rate within 6 h (/min) | 106.707 ± 25.104 | 110.156 ± 25.212 | 105.84 ± 25.003 | < 0.001 |
| New onset | 6.7% (906) | 8.8% (237) | 6.2% (669) | < 0.001 |
| Congestive heart failure | 55.9% (7531) | 57.7% (1562) | 55.4% (5969) | 0.035 |
| Ischemic heart disease | 43.3% (5841) | 44.4% (1203) | 43.1% (4638) | 0.207 |
| Heart surgery | 8% (1077) | 1.4% (38) | 9.6% (1039) | < 0.001 |
| Aorta surgery | 0.8% (102) | 0.5% (14) | 0.8% (88) | 0.137 |
| Hypertension | 49.2% (6624) | 43.5% (1177) | 50.6% (5447) | < 0.001 |
| Diabetes | 33.1% (4457) | 32% (865) | 33.4% (3592) | 0.172 |
| Chronic kidney disease | 24% (3230) | 25% (676) | 23.7% (2554) | 0.180 |
| End-stage kidney disease | 5.6% (752) | 6.9% (188) | 5.2% (564) | < 0.001 |
| Chronic liver disease | 9.6% (1292) | 15.7% (424) | 8.1% (868) | < 0.001 |
| Metastatic cancer | 5.9% (796) | 8.9% (242) | 5.1% (554) | < 0.001 |
| Intubation | 23.8% (3213) | 31.9% (863) | 21.8% (2350) | < 0.001 |
| CKRT | 0.6% (79) | 1.3% (36) | 0.4% (43) | < 0.001 |
| Blood culture positive | 3.2% (434) | 6.2% (167) | 2.5% (267) | < 0.001 |
| Glasgow coma scale | 13.439 ± 2.736 | 12.704 ± 3.438 | 13.623 ± 2.495 | < 0.001 |
| Norepinephrine (mcg/kg/min) | 0.022 ± 0.082 | 0.044 ± 0.118 | 0.016 ± 0.069 | < 0.001 |
| Dopamine (mcg/kg/min) | 0.251 ± 1.651 | 0.447 ± 2.248 | 0.202 ± 1.46 | < 0.001 |
| Dobutamine (mcg/kg/min) | 0.052 ± 0.567 | 0.089 ± 0.745 | 0.043 ± 0.513 | 0.003 |
| Epinephrine (mcg/kg/min) | 0.002 ± 0.026 | 0.004 ± 0.041 | 0.002 ± 0.02 | 0.003 |
| Vasopressin (units/hours) | 0.065 ± 0.606 | 0.142 ± 0.707 | 0.045 ± 0.577 | < 0.001 |
| Length of stay hours (hours) | 121.488 ± 152.143 | 155.695 ± 182.967 | 112.889 ± 142.075 | < 0.001 |
| Rhythm control | 10.2% (1373) | 11.8% (319) | 9.8% (1054) | 0.002 |
| Rate control | 25.3% (3407) | 21.9% (592) | 26.1% (2815) | < 0.001 |
| Heart rate MAD | 9.383 ± 5.381 | 9.74 ± 5.325 | 9.293 ± 5.391 | < 0.001 |
| Number of heart rate measurements | 29.505 ± 9.128 | 30.976 ± 10.858 | 29.136 ± 8.6 | < 0.001 |
SpO2saturation of percutaneous oxygen, CKRT continuous kidney replacement therapy, MAD median absolute deviation.
Association between heart rate change and risk of in-hospital mortality
Rescaled generalized variance inflation factors for all covariates ranged from 1.009 to 1.247, confirming the absence of significant multicollinearity (Supplementary Table S1). In Model 1, compared to the high to high group (both initial and maximum heart rate > 110 bpm), only the low to low group (both initial and maximum heart rate ≤ 110 bpm) demonstrated a significantly lower mortality risk (Table 2). However, in the fully adjusted Model 4, the association between the low to low group and mortality risk was no longer statistically significant. Instead, only the low to high group (initial heart rate ≤ 110 bpm and maximum heart rate > 110 bpm) showed a significantly higher mortality risk (OR: 1.30, 95% CI: 1.09–1.54). The other groups did not show statistically significant associations with mortality. Similar trends were observed in Models 2 and 3. While the p for trend was statistically significant in Model 1, it did not reach statistical significance in Models 2, 3, and 4.
Table 2.
Odds ratio for In-hospital mortality according to heart rate change.
| Group | No. of group | No. of mortality | Model 1a | Model 2b | Model 3c | Model 4d | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | |||
| high to high | 3597 | 795 | reference | < 0.001 | reference | 0.650 | reference | 0.291 | reference | 0.311 | ||||
| high to low | 49 | 10 | 0.810 (0.380–1.572) | 0.557 | 0.848 (0.395–1.659) | 0.649 | 0.989 (0.455–1.962) | 0.9759 | 0.966 (0.443–1.920) | 0.925 | ||||
| low to high | 2008 | 475 | 1.039 (0.911–1.184) | 0.564 | 1.336 (1.135–1.573) | < 0.001 | 1.295 (1.093–1.533) | 0.003 | 1.297 (1.094–1.538) | 0.003 | ||||
| low to low | 7821 | 1427 | 0.730 (0.661–0.806) | < 0.001 | 1.050 (0.886–1.244) | 0.574 | 1.104 (0.927–1.315) | 0.268 | 1.099 (0.921–1.311) | 0.296 | ||||
a: adjusted for age, sex, new onset atrial fibrillation.
b: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen.
c: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen, congestive heart failure, hypertension, diabetes, chronic kidney disease, chronic liver disease, metastatic cancer, intubation, continuous kidney replacement therapy, blood culture positive, heart surgery, aorta surgery, glasgow coma scale.
d: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen, congestive heart failure, hypertension, diabetes, chronic kidney disease, chronic liver disease, metastatic cancer, intubation, continuous kidney replacement therapy, blood culture positive, heart surgery, aorta surgery, glasgow coma scale, norepinephrine, dopamine, dobutamine, epinephrine, vasopressin, rhythm control, rate control.
Association between heart rate fluctuation and risk of in-hospital mortality
When comparing HRF to the lowest quintile (quintile 1), higher HRF was associated with an increased risk of in-hospital mortality across all models (p for trend < 0.001 for model 1 and 2, 0.003 for model 3, and 0.002 for model 4) (Table 3). In Model 4, patients in the highest HRF quintile (quintile 5) had a significantly higher risk of in-hospital mortality compared to those in the lowest HRF quintile (OR: 1.25, 95% CI: 1.07–1.46). This trend was consistent across all models, indicating a robust association between increased HRF and higher mortality risk. A restricted cubic spline analysis further illustrated a monotonic relationship, with higher MAD values associated with progressively increased odds of in-hospital mortality (Fig. 2). Measurement frequency rose across MAD quintiles—from 28.0 ± 7.4 in Q1 to 31.8 ± 10.9 in Q5 (p < 0.001; Supplementary Table S2). In a sensitivity analysis adjusting for measurement frequency, the highest HRF quintile remained associated with increased in-hospital mortality (OR 1.18; 95% CI 1.013–1.383), and the trend across quintiles persisted (p for trend = 0.021) (Supplementary Table S3).
Table 3.
Odds ratio for In-hospital mortality according to heart rate fluctuation.
| Group | No. of group | No. of mortality | Model 1 | Model 2 | Model 3 | Model 4 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | Odds ratio | p-value | p for trend | |||
| Quantile 1 | 2695 | 471 | reference | < 0.001 | reference | < 0.001 | reference | 0.003 | reference | 0.002 | ||||
| Quantile 2 | 2695 | 494 | 1.067 (0.928–1.227) | 0.366 | 1.044 (0.906–1.202) | 0.554 | 1.047 (0.904–1.212) | 0.541 | 1.056 (0.912–1.224) | 0.466 | ||||
| Quantile 3 | 2695 | 545 | 1.223 (1.066–1.403) | 0.004 | 1.173 (1.02–1.35) | 0.025 | 1.165 (1.007–1.347) | 0.040 | 1.190 (1.028–1.378) | 0.020 | ||||
| Quantile 4 | 2695 | 577 | 1.32 (1.152–1.513) | < 0.001 | 1.225 (1.065–1.411) | 0.005 | 1.151 (0.994–1.333) | 0.060 | 1.174 (1.012–1.362) | 0.034 | ||||
| Quantile 5 | 2695 | 620 | 1.482 (1.295–1.698) | < 0.001 | 1.293 (1.121–1.493) | < 0.001 | 1.230 (1.059–1.429) | 0.007 | 1.254 (1.074–1.463) | 0.004 | ||||
a: adjusted for age, sex, new onset atrial fibrillation.
b: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen.
c: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen, congestive heart failure, hypertension, diabetes, chronic kidney disease, chronic liver disease, metastatic cancer, intubation, continuous kidney replacement therapy, blood culture positive, heart surgery, aorta surgery, glasgow coma scale.
d: adjusted for age, sex, new onset atrial fibrillation, heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, saturation of percutaneous oxygen, congestive heart failure, hypertension, diabetes, chronic kidney disease, chronic liver disease, metastatic cancer, intubation, continuous kidney replacement therapy, blood culture positive, heart surgery, aorta surgery, glasgow coma scale, norepinephrine, dopamine, dobutamine, epinephrine, vasopressin, rhythm control, rate control.
Fig. 2.
Restricted cubic spline of the association between heart rate median absolute deviation and adjusted odds of in-hospital mortality. OR, odds ratio; MAD, median absolute deviation.
Subgroup analysis and interaction term
In the analysis of the association between heart rate change and mortality, no significant interactions were observed with any of the subgroup variables (Fig. 3). In patients without new-onset AF, the low to high group exhibited a higher mortality risk compared to the high to high group. However, this association was not significant in patients with new-onset AF. Additionally, among patients who did not receive rate control, the low to high group showed a higher mortality risk compared to the high to high group, whereas this association was not significant in patients who were under rate control.
Fig. 3.
Forest plot of subgroup analysis based on heart rate change groups. CI, confidence interval; SBP, systolic blood pressure.
For the analysis of the association between HRF and mortality, in most subgroups, mortality odds increased progressively from the second through the fifth HRF quintiles, with the fifth quintile exhibiting the greatest risk compared to the reference first quintile (Fig. 4). For this association, significant interactions were observed with rate control therapy and initial heart rate. In patients not receiving rate control, higher HRF was associated with increased mortality risk, while this association was not significant in patients receiving rate control. Furthermore, in patients with an initial heart rate ≤ 110 bpm, higher HRF was linked to a higher mortality risk, but this association was not observed in patients with an initial heart rate > 110 bpm.
Fig. 4.
Forest plot of subgroup analysis based on heart rate fluctuation quintile groups: Q1–Q5; Q1 = reference, Q2–Q5 = second through fifth quintiles of HRF, with Q5 indicating highest fluctuation. CI, confidence interval, Q, quintile, SBP, systolic blood pressure, HR, heart rate.
Discussion
We investigated how heart-rate trajectories and variability (HRF) relate to in-hospital mortality in ICU patients with AF. The main findings indicated that patients with an initially low heart rate that increased significantly within the first six hours of ICU admission (low to high group) had a higher risk of in-hospital mortality. Additionally, higher HRF was associated with an increased mortality risk.
The inclusion of new-onset AF as a variable is justified by its frequent occurrence in ICU patients and its association with increased mortality and cardiovascular events34. Cardiac surgery was also included as a variable because AF frequently occurs postoperatively and is known to increase mortality risk35–38. Furthermore, HRF can be affected by medications used for rate and rhythm control, and factors like norepinephrine, dopamine, dobutamine, epinephrine, vasopressin administration, and sepsis, which was included as bacteremia in the analysis, were also considered due to their potential to trigger AF and influence HRF9,39.
Our findings revealed that patients in the low to high group had a higher risk of in-hospital mortality compared to those in the high to high group, indicating that heart rate changes after ICU admission are a critical predictor of mortality. Factors such as intubation, CKRT, and infections can cause rapid increases in heart rate during ICU stay40–42. Even after adjusting for these factors, the mortality risk remained significantly high, suggesting that an increase in heart rate is an independent risk factor. Therefore, physicians should remain vigilant even with a low initial heart rate in ICU patients with AF, as rising heart rates were associated with increased mortality risk, especially when the increase follows an initially low rate.
We also confirmed that higher HRF is associated with an increased risk of in-hospital mortality. Severe diseases, ICU environments, and acute stress responses can contribute to significant vital sign fluctuations. Patients with significant HRF in the ICU are at higher risk for both short-term and long-term mortality17,43. Although previous studies have linked higher heart rate in AF patients to increased mortality, the specific association between HRF and mortality had not been well-established until now. These findings underscore the need for physicians managing ICU patients with AF to focus not only on managing elevated heart rates but also on monitoring HRF.
In subgroup analyses, despite the lack of significant interaction, low to high heart rate changes were significantly associated with higher mortality in patients with pre-existing AF, but this was not observed in patients with new-onset AF. Moreover, the lack of significant sex–exposure interactions suggests that sex is unlikely to influence the relationships between heart rate change or HRF and in-hospital mortality. This difference may be attributed to the potential heart remodeling in pre-existing AF patients, making their hearts more vulnerable to rapid increases in heart rate44,45. Among patients receiving rate-control therapy, we did not observe a statistically significant association between HRF and in-hospital mortality. While HRF may reflect underlying disease severity or acute hemodynamic changes, fluctuations occurring under rate-control regimens appeared to carry distinct prognostic implications. Accordingly, the presence of marked HRF in this setting alone should not be taken as a reason to withhold or discontinue established rate-control treatments. Nevertheless, these observations require cautious interpretation: rate-control agents can directly influence HRF, and residual confounding—such as indication bias or unmeasured factors—may affect both HRF and outcomes. Further research, incorporating high-resolution physiologic monitoring and rigorous causal inference methods, is needed to clarify the interplay between pharmacologic rate control, HRF, and patient prognosis. Furthermore, a significant association between HRF and mortality was observed in patients with an initial heart rate ≤ 110 bpm, suggesting that even when the initial heart rate is low, attention to HRF is crucial. These findings emphasize the importance of careful monitoring of HRF, particularly in patients without rate control medication and those with initially low heart rates.
This study has several limitations. First, it is a single-center retrospective cohort study, which may introduce bias. Second, we did not consider the existing AF treatment or anticoagulation therapy status. Third, the potential confounding effects of other medications and interventions not accounted for in the models might have influenced the heart rate and HRF. Fourth, drug infusion rates are calculated per kilogram of body weight; however, when a patient’s weight was missing, the weight used to derive these per-kilogram rates is not documented in the data source. It is therefore likely that a default or imputed weight was applied, introducing potential inaccuracy into our infusion-rate estimates. Fifth, differences in the number and timing of heart-rate recordings across patients may have influenced our HRF estimates. Although we adjusted for total measurement count in sensitivity analyses, residual bias arising from unmeasured variation in monitoring practices cannot be excluded. Lastly, validation of these findings in other populations is necessary to ensure generalizability.
In conclusion, this study demonstrated that changes in heart rate and HRF were associated with in-hospital mortality in ICU patients with AF. However, in patients receiving rate control therapy, the association between HRF and mortality was attenuated, suggesting a more limited prognostic role in this subgroup.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors declare that no funding was received for conducting this study.
Author contributions
MWK and YK designed the study. MWK collected, generated, and analyzed the data. MWK, SYA and YK interpreted the data. MWK wrote the manuscript.
Data availability
The datasets generated and/or analyzed during the current study are available in the MIMIC-III and MIMIC-IV database from PhysioNet repository, https://physionet.org/content/mimiciii/1.4/ and https://physionet.org/content/mimiciv/3.1/.
Declarations
Competing interests
The authors declare no competing interests.
Ethical approval and consent to participate
As this study was an analysis of a public database, this study was exempt from approval by the institutional review board of Seoul National University Hospital (no. 2405-061-1535).
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Brundel, B. J. J. M. et al. Atrial fibrillation. Nat. Reviews Disease Primers. 810.1038/s41572-022-00347-9 (2022).
- 2.Chyou, J. Y. et al. Atrial fibrillation occurring during acute hospitalization: A scientific statement from the American heart association. Circulation147, e676–e698. 10.1161/cir.0000000000001133 (2023). [DOI] [PubMed] [Google Scholar]
- 3.Klein Klouwenberg, P. M. et al. Incidence, predictors, and outcomes of new-onset atrial fibrillation in critically ill patients with sepsis. A cohort study. Am. J. Respir Crit. Care Med.195, 205–211. 10.1164/rccm.201603-0618OC (2017). [DOI] [PubMed] [Google Scholar]
- 4.Kuipers, S., Klein Klouwenberg, P. M. & Cremer, O. L. Incidence, risk factors and outcomes of new-onset atrial fibrillation in patients with sepsis: a systematic review. Crit. Care. 18, 688. 10.1186/s13054-014-0688-5 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Moss, T. J. et al. New-onset atrial fibrillation in the critically ill. Crit. Care Med.45, 790–797. 10.1097/ccm.0000000000002325 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bedford, J. et al. Pharmacological and non-pharmacological treatments and outcomes for new-onset atrial fibrillation in ICU patients: the CAFE scoping review and database analyses. Health Technol. Assess.25, 1–174. 10.3310/hta25710 (2021). [DOI] [PubMed] [Google Scholar]
- 7.Massera, D., Wang, D., Vorchheimer, D. A., Negassa, A. & Garcia, M. J. Increased risk of stroke and mortality following new-onset atrial fibrillation during hospitalization. Europace19, 929–936. 10.1093/europace/euw110 (2017). [DOI] [PubMed] [Google Scholar]
- 8.Søgaard, M. et al. Thromboembolic risk in patients with pneumonia and new-onset atrial fibrillation not receiving anticoagulation therapy. JAMA Netw. Open.5, e2213945. 10.1001/jamanetworkopen.2022.13945 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Walkey, A. J., Wiener, R. S., Ghobrial, J. M., Curtis, L. H. & Benjamin, E. J. Incident stroke and mortality associated with new-onset atrial fibrillation in patients hospitalized with severe sepsis. Jama306, 2248–2254. 10.1001/jama.2011.1615 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mehreen, T. et al. In-hospital mortality among critically Ill patients with atrial fibrillation (AF) versus patients without AF. Cureus13 e18761, (2021). 10.7759/cureus.18761 [DOI] [PMC free article] [PubMed]
- 11.Shaver, C. M. et al. Atrial fibrillation is an independent predictor of mortality in critically ill patients. Crit. Care Med.43, 2104–2111. 10.1097/ccm.0000000000001166 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bedford, J., Harford, M., Petrinic, T., Young, J. D. & Watkinson, P. J. Risk factors for new-onset atrial fibrillation on the general adult ICU: protocol for a systematic review. BMJ Open.8, e024640. 10.1136/bmjopen-2018-024640 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Bedford, J. P. et al. New-onset atrial fibrillation in intensive care: epidemiology and outcomes. Eur. Heart J. Acute Cardiovasc. Care. 11, 620–628. 10.1093/ehjacc/zuac080 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Qian, J., Kuang, L., Chen, F., Liu, X. & Che, L. Prognosis and management of new-onset atrial fibrillation in critically ill patients. BMC Cardiovasc. Disord. 21, 231. 10.1186/s12872-021-02039-w (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Seguin, P. & Launey, Y. Atrial fibrillation is not just an artefact in the ICU. Crit. Care. 14, 182. 10.1186/cc9093 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cubbon, R. M. et al. Ambulatory heart rate range predicts mode-specific mortality and hospitalisation in chronic heart failure. Heart102, 223. 10.1136/heartjnl-2015-308428 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Guo, Q. et al. Heart rate fluctuation and mortality in critically ill myocardial infarction patients: A retrospective cohort study. Front. Cardiovasc. Med.8, 577742. 10.3389/fcvm.2021.577742 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Guo, Q. et al. Heart rate fluctuation predicts mortality in critically ill patients in the intensive care unit: a retrospective cohort study. Ann. Transl Med.9, 334. 10.21037/atm-20-7897 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zhou, X. et al. The impact of visit-to-visit heart rate variability on all-cause mortality in atrial fibrillation. Ann. Noninvasive Electrocardiol.29, e13094. 10.1111/anec.13094 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Johnson, A. E. W. et al. MIMIC-III, a freely accessible critical care database. Sci. Data. 3, 160035. 10.1038/sdata.2016.35 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Johnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci. Data. 10, 1. 10.1038/s41597-022-01899-x (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rivers, E. et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl. J. Med.345, 1368–1377. 10.1056/NEJMoa010307 (2001). [DOI] [PubMed] [Google Scholar]
- 23.Evans, L. et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med.47, 1181–1247. 10.1007/s00134-021-06506-y (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Stratton, T. & Nasser, L. BET 1: lenient or strict rate control for atrial fibrillation. Emerg. Med. J.35, 765–768. 10.1136/emermed-2018-208261.1 (2018). [DOI] [PubMed] [Google Scholar]
- 25.Wyse, D. G. et al. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl. J. Med.347, 1825–1833. 10.1056/NEJMoa021328 (2002). [DOI] [PubMed] [Google Scholar]
- 26.January, C. T. et al. AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines and the Heart Rhythm Society. Circulation 130, e199-267, (2014). 10.1161/cir.0000000000000041 (2014). [DOI] [PMC free article] [PubMed]
- 27.Hindricks, G. et al. ESC Guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS): The Task Force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC) Developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC. Eur Heart J 42, 373–498, (2020). 10.1093/eurheartj/ehaa612 (2021). [DOI] [PubMed]
- 28.Correction 2023 ACC/AHA/ACCP/HRS guideline for the diagnosis and management of atrial fibrillation: A report of the American college of cardiology/american heart association joint committee on clinical practice guidelines. Circulation149, e167. 10.1161/cir.0000000000001207 (2024). [DOI] [PubMed] [Google Scholar]
- 29.Balik, M. et al. Propafenone for supraventricular arrhythmias in septic shock-Comparison to Amiodarone and Metoprolol. J. Crit. Care. 41, 16–23. 10.1016/j.jcrc.2017.04.027 (2017). [DOI] [PubMed] [Google Scholar]
- 30.Singh, B. N. et al. Amiodarone versus Sotalol for atrial fibrillation. N Engl. J. Med.352, 1861–1872. 10.1056/NEJMoa041705 (2005). [DOI] [PubMed] [Google Scholar]
- 31.Lee, E. et al. Mortality and causes of death in patients with atrial fibrillation: A nationwide population-based study. PLoS One. 13, e0209687. 10.1371/journal.pone.0209687 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Cavallari, I. & Patti, G. Early risk of mortality, cardiovascular events, and bleeding in patients with newly diagnosed atrial fibrillation. Eur. Heart J. Supplements. 22, L110–L113. 10.1093/eurheartj/suaa147 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Jacobs, M. S. et al. Incidence, treatment and mortality of new-onset atrial fibrillation patients at the intensive care unit. Open. Heart. 7, e001226. 10.1136/openhrt-2019-001226 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Bosch, N. A., Cimini, J. & Walkey, A. J. Atrial fibrillation in the ICU. Chest154, 1424–1434. 10.1016/j.chest.2018.03.040 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.El-Chami, M. F. et al. New-onset atrial fibrillation predicts long-term mortality after coronary artery bypass graft. J. Am. Coll. Cardiol.55, 1370–1376. 10.1016/j.jacc.2009.10.058 (2010). [DOI] [PubMed] [Google Scholar]
- 36.Villareal, R. P. et al. Postoperative atrial fibrillation and mortality after coronary artery bypass surgery. J. Am. Coll. Cardiol.43, 742–748. 10.1016/j.jacc.2003.11.023 (2004). [DOI] [PubMed] [Google Scholar]
- 37.Saxena, A. et al. Usefulness of postoperative atrial fibrillation as an independent predictor for worse early and late outcomes after isolated coronary artery bypass grafting (multicenter Australian study of 19,497 patients). Am. J. Cardiol.109, 219–225. 10.1016/j.amjcard.2011.08.033 (2012). [DOI] [PubMed] [Google Scholar]
- 38.Creswell, L. L., Schuessler, R. B., Rosenbloom, M. & Cox, J. L. Hazards of postoperative atrial arrhythmias. Ann. Thorac. Surg.56, 539–549. 10.1016/0003-4975(93)90894-n (1993). [DOI] [PubMed] [Google Scholar]
- 39.Suita, K. et al. Norepinephrine-Induced adrenergic activation strikingly increased the atrial fibrillation duration through β1- and α1-Adrenergic Receptor-Mediated signaling in mice. PLoS One. 10, e0133664. 10.1371/journal.pone.0133664 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Prys-Roberts, C., Greene, L. T., Meloche, R. & Foëx, P. Studies of anaesthesia in relation to hypertension. II. Haemodynamic consequences of induction and endotracheal intubation. Br. J. Anaesth.43, 531–547. 10.1093/bja/43.6.531 (1971). [DOI] [PubMed] [Google Scholar]
- 41.Shribman, A. J., Smith, G. & Achola, K. J. Cardiovascular and catecholamine responses to laryngoscopy with and without tracheal intubation. Br. J. Anaesth.59, 295–299. 10.1093/bja/59.3.295 (1987). [DOI] [PubMed] [Google Scholar]
- 42.Douvris, A. et al. Mechanisms for hemodynamic instability related to renal replacement therapy: a narrative review. Intensive Care Med.45, 1333–1346. 10.1007/s00134-019-05707-w (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Fukui, S. et al. Association between respiratory and heart rate fluctuations and death occurrence in dying cancer patients: continuous measurement with a non-wearable monitor. Support Care Cancer. 30, 77–86. 10.1007/s00520-021-06346-y (2022). [DOI] [PubMed] [Google Scholar]
- 44.Nattel, S., Burstein, B. & Dobrev, D. Atrial remodeling and atrial fibrillation. Circulation: Arrhythmia Electrophysiol.1, 62–73. 10.1161/CIRCEP.107.754564 (2008). [DOI] [PubMed] [Google Scholar]
- 45.Rolfes, C. D., Howard, S. A., Goff, R. P. & Iaizzo, P. A. Cardiac remodeling as a consequence of atrial fibrillation: an anatomical study of perfusion-fixed human heart specimens. J. Geriatr. Cardiol.8, 141–146. 10.3724/sp.J.1263.2011.00141 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analyzed during the current study are available in the MIMIC-III and MIMIC-IV database from PhysioNet repository, https://physionet.org/content/mimiciii/1.4/ and https://physionet.org/content/mimiciv/3.1/.