Abstract
Background
Postoperative atrial fibrillation (POAF) in trauma patients is closely related to poor prognosis. This study aims to identify the risk factors of POAF and establish a predictive model.
Methods
We extracted data from the MIMIC-IV 2.2 database on ICU trauma patients who underwent surgery. The patients were randomly divided into a training set and a validation set at a ratio of 7:3. We used least absolute shrinkage and selection operator (LASSO) regression combined with multivariable logistic regression to select predictive factors. Receiver operating characteristic (ROC) curves, calibration curves, and decision curves were used to evaluate the developed nomogram model.
Results
Among 5170 included patients, POAF incidence was 9.15%. POAF was associated with significantly postoperative higher mortality at 7 days, 28 days, and 1 year, as well as prolonged hospital stays and ICU stays. Independent predictors of POAF included age (OR 1.06, 95% CI 1.04–1.07, P < 0.001), congestive heart failure history (CHF) (OR 1.85, 95% CI 1.35–2.54, P < 0.001), sequential organ failure assessment (SOFA) score on the first ICU day (OR 1.09, 95% CI 1.03–1.16, P = 0.002), and epinephrine use on the day of surgery (OR 1.95, 95% CI 1.14–3.28, P = 0.013). The nomogram, developed from age, CHF history, and ICU SOFA score, showed an area under the curve (AUC) of 0.791 (95% CI 0.768–0.814) in training set and 0.800 (95% CI 0.763–0.836) in validation set.
Conclusion
POAF significantly worsens outcomes in trauma patients. The developed nomogram provides effective risk stratification for early identification and clinical decision-making.
Highlights
POAF incidence was 9.15% and predicted higher mortality in trauma patients.
A novel nomogram was developed using age, congestive heart failure history, and SOFA score.
The model showed good predictive performance with an AUC of 0.800 upon validation.
This tool aids in early risk stratification for POAF to guide clinical management.
Supplementary Information
The online version contains supplementary material available at 10.1186/s40001-026-03841-y.
Keywords: Trauma, Surgery, Postoperative atrial fibrillation, Nomogram, Risk prediction
Background
Severe trauma is a major cause of global mortality and disability, significantly reducing life expectancy and imposing a substantial disease burden [1]. Surgery is one of the important components of trauma treatment, but trauma patients still face many challenges during the perioperative period [2]. Among them, atrial fibrillation (AF) is one of the most common and serious perioperative arrhythmias [2].
In trauma patients, including those with hip fractures, the incidence of postoperative atrial fibrillation (POAF) ranges from 5.6% to 13.59% [3–5]. The onset of POAF is associated with an increased risk of various complications such as myocardial infarction, stroke, infection, heart failure and thromboembolism [6–9]. AF often leads to prolonged hospitalization and ICU stay, along with higher readmission rates and medical costs [10, 11, 15, 16]. AF in ICU trauma patients is linked to a poor prognosis and frequently accompanied by multiple organ dysfunction, sepsis, shock, and systemic inflammation [17].
Therefore, it is important to develop a predictive model for POAF in trauma patients. A large-scale study established a POAF risk prediction model for patients undergoing non-cardiac surgery (including trauma patients), demonstrating the feasibility and clinical value of POAF risk prediction [9]. Previous studies have concentrated on specific types of trauma (such as hip fractures) [3, 5] or preoperative risk assessment [18]. There are still few predictive models specifically for POAF in trauma patients. This study is based on a cohort of trauma patients admitted to the ICU, aiming to identify independent predictors of POAF in trauma patients and develop a clinical predictive model.
Methods
Data resources
The data source for this retrospective research was the Medical Information Mart for Intensive Care (MIMIC) IV v2.2 database [19], which archives clinical information from all hospitalizations at the Beth Israel Deaconess Medical Center (BIDMC, USA) between 2008 and 2019. The MIMIC IV database has de-identified all patient personal information, replacing identifiable details with randomly generated identity codes. The establishment of the MIMIC-IV database was approved by the Massachusetts Institute of Technology (MIT) and the Institutional Review Boards (IRB) of BIDMC, which granted a waiver of informed consent. One of the authors of this study completed the Collaborative Institutional Training Initiative (CITI) program and was granted permission to extract data (Certification ID: 66716350). The reporting of this study follows the STROBE statement for cohort studies (Supplementary Material 1).
Inclusion and exclusion criteria
Inclusion criteria: adult trauma patients undergoing surgery in the MIMIC IV 2.2 database.
Exclusion criteria: (1) Patients with repeated surgical records; (2) Not admitted to ICU; (3) History of AF; (4) Preoperative AF after admission. (Fig. 1).
Fig. 1.
Flow chart. POAF: postoperative atrial fibrillation, MIMIC: Medical Information Mart for Intensive Care, ICU: Intensive Care Unit
Data collection
The identification of eligible adult trauma patients who underwent surgery was performed using SQL queries on the MIMIC-IV v2.2 database. The following information was extracted: (1) Demographic characteristics: age, sex, weight, height, and emergency admission status; (2) Mean vital signs and mean blood glucose on the first ICU day: glucose, heart rate, oxygen saturation, respiratory rate, temperature, systolic blood pressure, mean arterial pressure, diastolic blood pressure; (3) First day ICU disease severity scores: acute physiology score III (APS III), oxford acute severity of illness score (OASIS), simplified acute glasgow coma scale (GCS) score, sequential organ failure assessment (SOFA), and scorephysiology score II (SAPS II); (4) Comorbidities: diabetes, paraplegia, congestive heart failure (CHF), malignancy, chronic pulmonary disease, peptic ulcer disease, mild liver disease, peripheral vascular disease, dementia, metastatic solid tumor, cerebrovascular disease, renal disease, rheumatic disease, myocardial infarction, severe liver disease, history of AF prior to admission, and the Charlson comorbidity index.; (5) Lifestyle factors: alcohol abuse history; (6) Trauma characteristics: head, chest, abdomen, extremities, multiple injuries; (7) Initial admission laboratory parameters: sodium, mean corpuscular hemoglobin, anion gap, potassium, albumin, chloride, calcium, phosphate, creatinine, glucose, lactate, bicarbonate, mean corpuscular volume, indirect bilirubin, low-density lipoprotein cholesterol, direct bilirubin, white blood cell count, basophil count, blood urea nitrogen, aspartate aminotransferase, lactate dehydrogenase, C-reactive protein, platelet count, globulin, hematocrit, high-density lipoprotein cholesterol, prothrombin time, red cell distribution width, total bilirubin, alkaline phosphatase, neutrophil count, lymphocyte count, monocyte count, eosinophil percentage, international normalized ratio, red blood cell count, d-dimer, uric acid, gamma-glutamyl transferase, alanine aminotransferase, total cholesterol, hemoglobin a1c, triglycerides, red cell distribution width standard deviation, eosinophil count, activated partial thromboplastin time, Troponin T, N-terminal pro-B-type (NTproBNP); (8) Preoperative medication use: β blocker use, calcium channel blockers (CCB) use; (9) Interventions on the day of surgery: midazolam use, dexmedetomidine use, morphine use, epinephrine use, dopamine use, frozen plasma transfusion, platelets transfusion, red blood cells transfusion; (10) preoperative AF after admission; (11) Outcome measures: POAF, postoperative stroke, postoperative transient ischemic attack (TIA), ICU and hospital length of stay, in-hospital mortality, postoperative 7-day mortality, postoperative 28-day mortality, postoperative 1-year mortality.
The diagnosis of AF was extracted from the derived.rhythm entity view in the MIMIC-IV database (v2.2) [20], which contains rhythm classifications derived from bedside ICU monitoring data. The classifications in the derived.rhythm table were generated through a two-step process: an initial automated identification by bedside monitors, followed by clinical confirmation by ICU clinicians. Therefore, this study defined POAF as any new-onset AF event that occurred within 30 days after surgery [11], during the patient's stay in the ICU. This 30 day definition was adopted to align with previous studies on POAF [11–14] and to ensure the capture of clinically significant postoperative arrhythmias that may be associated with the physiological stress of surgery and hospitalization. This definition applied irrespective of whether the patient was admitted directly to the ICU or transferred from the post-anesthesia care unit (PACU), and it included all episodes ranging from single transient events to sustained arrhythmia.
Statistical analysis
All statistical computations were conducted with R software (v4.4.3). Patients were randomly divided into training and validation sets in a 7:3 ratio for model development and internal validation, respectively. The training set was utilized to construct a predictive model for POAF, whereas the validation set was used for internal validation. Variables exhibiting a missing data proportion greater than 20% were removed. For variables with a missing data rate less than 20%, the “missRanger” package in R was used to separately perform random forest imputation within the training and validation sets (Supplementary Material 2). The same model parameters were applied to both groups during imputation to ensure data consistency. Variable selection was conducted in two steps. First, least absolute shrinkage and selection operator (LASSO) regression was applied to all the candidate variables for feature selection, and those variables with nonzero coefficients were retained. The optimal λ value was determined through tenfold cross-validation with the one-standard-error rule (λ 1SE), which selects the largest λ value within one standard error of the minimum cross-validation error, thereby providing a more parsimonious model. Concurrently, univariable logistic regression was performed on all the variables to preliminarily screen for factors associated with POAF (P < 0.05). Variance inflation factor (VIF) analysis was subsequently used to exclude variables with significant multicollinearity (VIF > 10) among the preliminarily screened variables. After troubleshooting for multicollinearity, we constructed a multivariate logistic regression model to determine the independent predictors of POAF. Finally, the variables jointly determined by LASSO regression and multivariate logistic regression were incorporated into the nomogram model. LASSO regression is only used for variable selection to reduce overfitting, while multivariate logistic regression is subsequently used for statistical inference and effect estimation.
The predictive ability of the model is assessed by the receiver operating characteristic (ROC) curve. An area under the curve (AUC) value of ≥ 0.75 is considered satisfactory. We use calibration curves to test the consistency between the model's predictions and the observed results. Decision curve analysis (DCA) was applied to evaluate the clinical net benefit of the model.
Only patients with complete survival information (N = 5168) were included in the survival analysis. The Kaplan–Meier survival curve was plotted to visually compare the survival rates of POAF patients and Non-POAF patients at 7 days, 28 days and 1 year after surgery. The Log-rank test was employed to assess the statistical significance of differences between the survival curves. The effect of POAF on the risk of death was measured using the Cox regression model. The results are expressed as hazard ratio (HR) and 95% confidence interval (CI).
Categorical variables are presented as frequencies and percentages (%), and were compared between groups using the chi-square test. Normally distributed continuous variables are expressed as mean ± standard deviation and were compared using the independent samples t-test. Non-normally distributed continuous variables are expressed as median (interquartile range) and were compared using the Mann–Whitney U test. All statistical tests were two-sided, and a P-value < 0.05 was considered statistically significant.
Results
Baseline characteristics of patients
This study included 5170 adult trauma patients in the ICU who received surgical intervention (Fig. 1). The incidence of POAF was 9.15% (473/5170) (Table 1). Among patients who developed POAF, the median time from surgery to the first AF episode was 1 (IQR: 0–3) day. The baseline characteristics of the training and validation sets are compared in Table 1. Comparison of the baseline data demonstrated general comparability between the training and validation cohorts. Statistically significant differences were only noted for age (P = 0.045), admission lactate (P = 0.008) and admission BUN (P = 0.022). The study population had a mean age of 61.00 [45.00, 74.00] years (Table 1). The predominant type of trauma included multiple injuries (46.87%), followed by head trauma (22.07%), extremity trauma (14.66%), chest trauma (13.50%), and abdominal trauma (2.90%). On the first day in the ICU, the severity scores were as follows: APS III score of 40.00 [30.00, 56.00] points, SAPS II score of 32.00 [23.00, 41.00] points, GCS score of 15.00 [15.00, 15.00] points, OASIS score of 32.00 [26.00, 38.00] points, and SOFA score of 4.00 [2.00, 7.00] points. Regarding preoperative medication, 21.62% (1118/5170) of patients were on β blockers and 8.97% (464/5170) on CCB.
Table 1.
Demographic and clinical characteristics of patients
| Variables | All | Training cohort | Validation cohort | P value |
|---|---|---|---|---|
| N = 5170 | N = 3618 | N = 1552 | ||
| Age (median [IQR]) | 61.00 [45.00, 74.00] | 61.00 [45.00, 74.00] | 62.00 [46.00, 76.00] | 0.045 |
| Sex (%) | 0.775 | |||
| Female | 2029 (39.26) | 1425 (39.39) | 604 (38.92) | |
| Male | 3141 (60.74) | 2193 (60.61) | 948 (61.08) | |
| Weight (median [IQR]) | 78.36 [66.00, 92.20] | 78.35 [66.00, 92.50] | 78.36 [66.10, 92.00] | 0.711 |
| Admission emergency (%) | 0.285 | |||
| No | 1198 (23.17) | 823 (22.75) | 375 (24.16) | |
| Yes | 3972 (76.83) | 2795 (77.25) | 1177 (75.84) | |
| Trauma location (%) | 0.773 | |||
| Head | 1141 (22.07) | 801 (22.14) | 340 (21.91) | |
| Chest | 698 (13.50) | 481 (13.30) | 217 (13.98) | |
| Abdomen | 150 (2.90) | 110 (3.04) | 40 (2.58) | |
| Extremities | 758 (14.66) | 522 (14.43) | 236 (15.21) | |
| Multiple | 2423 (46.87) | 1704 (47.10) | 719 (46.33) | |
| Comorbidities | ||||
| Myocardial infarct (%) | 0.874 | |||
| No | 4620 (89.36) | 3231 (89.36) | 1389 (89.50) | |
| Yes | 550 (10.64) | 387 (10.70) | 163 (10.50) | |
| Congestive heart failure (%) | 0.134 | |||
| No | 4445 (85.98) | 3093 (85.49) | 1352 (87.11) | |
| Yes | 725 (14.02) | 525 (14.51) | 200 (12.89) | |
| Peripheral vascular disease (%) | 0.803 | |||
| No | 4832 (93.46) | 3384 (93.53) | 1448 (93.30) | |
| Yes | 338 (6.54) | 234 (6.47) | 104 (6.70) | |
| Cerebrovascular disease (%) | 0.112 | |||
| No | 4648 (89.90) | 3269 (90.35) | 1379 (88.85) | |
| Yes | 522 (10.10) | 349 (9.65) | 173 (11.15) | |
| Dementia (%) | 0.180 | |||
| No | 4992 (96.56) | 3502 (96.79) | 1490 (96.01) | |
| Yes | 178 (3.44) | 116 (3.21) | 62 (4.00) | |
| Chronic pulmonary disease (%) | 0.258 | |||
| No | 4282 (82.82) | 2982 (82.42) | 1300 (83.76) | |
| Yes | 888 (17.18) | 636 (17.58) | 252 (16.24) | |
| Rheumatic disease (%) | 0.346 | |||
| No | 5053 (97.74) | 3531 (97.60) | 1522 (98.07) | |
| Yes | 117 (2.26) | 87 (2.41) | 30 (1.93) | |
| Peptic ulcer disease (%) | 0.968 | |||
| No | 5069 (98.05) | 3548 (98.06) | 1521 (98.00) | |
| Yes | 101 (1.95) | 70 (1.94) | 31 (2.00) | |
| Mild liver disease (%) | 0.963 | |||
| No | 4750 (91.88) | 3325 (91.90) | 1425 (91.82) | |
| Yes | 420 (8.12) | 293 (8.10) | 127 (8.18) | |
| Paraplegia (%) | 0.322 | |||
| No | 4893 (94.64) | 3432 (94.86) | 1461 (94.14) | |
| Yes | 277 (5.36) | 186 (5.14) | 91 (5.86) | |
| Renal disease (%) | 0.961 | |||
| No | 4627 (89.50) | 3239 (89.52) | 1388 (89.43) | |
| Yes | 543 (10.50) | 379 (10.48) | 164 (10.57) | |
| Malignant cancer (%) | 0.854 | |||
| No | 4733 (91.55) | 3310 (91.49) | 1423 (91.69) | |
| Yes | 437 (8.45) | 308 (8.51) | 129 (8.31) | |
| Severe liver disease (%) | 0.690 | |||
| No | 4970 (96.13) | 3475 (96.05) | 1495 (96.33) | |
| Yes | 200 (3.87) | 143 (3.95) | 57 (3.67) | |
| Metastatic solid tumor (%) | 0.385 | |||
| No | 5012 (96.95) | 3502 (96.79) | 1510 (97.29) | |
| Yes | 158 (3.06) | 116 (3.21) | 42 (2.70) | |
| Diabetes (%) | 0.445 | |||
| No | 4186 (80.97) | 2919 (80.68) | 1267 (81.64) | |
| Yes | 984 (19.03) | 699 (19.32) | 285 (18.37) | |
| Charlson comorbidity index (median [IQR]) | 4.00 [2.00, 6.00] | 4.00 [2.00, 6.00] | 4.00 [2.00, 6.00] | 0.481 |
| Alcohol abuse history (%) | 0.785 | |||
| No | 4214 (81.51) | 2945 (81.40) | 1269 (81.77) | |
| Yes | 956 (18.49) | 673 (18.60) | 283 (18.23) | |
| Admission laboratory results | ||||
| Eosinophil percent (median [IQR]) | 0.80 [0.10, 1.50] | 0.80 [0.10, 1.50] | 0.89 [0.10, 1.50] | 0.757 |
| Lactate (median [IQR]) | 2.00 [1.40, 2.90] | 2.00 [1.40, 2.90] | 2.10 [1.50, 2.99] | 0.008 |
| APTT (median [IQR]) | 28.20 [25.40, 32.10] | 28.30 [25.40, 32.30] | 28.10 [25.30, 31.50] | 0.063 |
| PT (median [IQR]) | 12.40 [11.40, 13.90] | 12.50 [11.40, 13.90] | 12.40 [11.30, 13.70] | 0.175 |
| INR (median [IQR]) | 1.10 [1.00, 1.28] | 1.10 [1.00, 1.30] | 1.10 [1.00, 1.20] | 0.103 |
| Calcium (median [IQR]) | 8.40 [7.90, 8.90] | 8.40 [7.90, 8.90] | 8.40 [7.90, 8.90] | 0.804 |
| Phosphate (median [IQR]) | 3.50 [2.90, 4.10] | 3.50 [2.90, 4.10] | 3.50 [2.98, 4.10] | 0.532 |
| Anion gap (median [IQR]) | 14.00 [12.00, 17.00] | 14.00 [12.00, 17.00] | 14.00 [12.00, 17.00] | 0.877 |
| Bicarbonate (median [IQR]) | 23.00 [20.00, 25.00] | 23.00 [20.00, 25.00] | 23.00 [20.00, 25.00] | 0.492 |
| Glucose (median [IQR]) | 128.00 [105.00, 161.00] | 127.00 [104.00, 160.00] | 129.00 [107.00, 162.00] | 0.285 |
| Chloride (median [IQR]) | 104.00 [101.00, 108.00] | 104.00 [101.00, 108.00] | 104.00 [101.00, 108.00] | 0.513 |
| Sodium (median [IQR]) | 139.00 [137.00, 141.00] | 139.00 [137.00, 141.00] | 139.00 [137.00, 142.00] | 0.118 |
| Potassium (median [IQR]) | 4.10 [3.80, 4.50] | 4.10 [3.80, 4.50] | 4.10 [3.80, 4.50] | 0.278 |
| BUN (median [IQR]) | 16.00 [12.00, 23.00] | 16.00 [11.77, 22.00] | 16.00 [12.00, 23.00] | 0.022 |
| Creatinine (median [IQR]) | 0.90 [0.70, 1.20] | 0.90 [0.70, 1.10] | 0.90 [0.70, 1.20] | 0.200 |
| MCH (median [IQR]) | 30.40 [29.10, 31.80] | 30.40 [29.00, 31.80] | 30.40 [29.20, 31.70] | 0.790 |
| MCV (median [IQR]) | 91.00 [88.00, 95.00] | 91.00 [88.00, 95.00] | 91.00 [87.00, 95.00] | 0.950 |
| RBC (median [IQR]) | 3.80 [3.28, 4.30] | 3.82 [3.29, 4.32] | 3.78 [3.25, 4.27] | 0.122 |
| RDW (median [IQR]) | 13.80 [13.10, 14.90] | 13.80 [13.10, 14.90] | 13.80 [13.10, 14.90] | 0.904 |
| Hematocrit (median [IQR]) | 34.80 [30.10, 38.90] | 34.90 [30.20, 38.99] | 34.50 [29.80, 38.70] | 0.181 |
| Platelet (median [IQR]) | 199.00 [151.00, 252.00] | 199.50 [151.00, 254.00] | 198.00 [151.00, 248.00] | 0.382 |
| WBC (median [IQR]) | 11.00 [8.10, 14.90] | 11.00 [8.00, 14.97] | 11.00 [8.40, 14.62] | 0.606 |
| First day ICU vital signs | ||||
| SBP mean (median [IQR]) | 119.00 [108.00, 130.00] | 119.00 [108.00, 130.00] | 118.00 [108.00, 129.00] | 0.430 |
| DBP mean (median [IQR]) | 64.00 [57.00, 72.00] | 64.00 [57.00, 72.00] | 63.16 [57.00, 72.00] | 0.375 |
| RR mean (median [IQR]) | 18.00 [16.00, 21.00] | 18.00 [16.00, 21.00] | 18.00 [16.00, 21.00] | 0.993 |
| MBP mean (median [IQR]) | 79.00 [72.00, 87.00] | 79.00 [73.00, 87.00] | 79.00 [72.00, 87.00] | 0.394 |
| HR mean (median [IQR]) | 84.00 [74.00, 96.00] | 84.00 [74.00, 96.00] | 85.00 [74.00, 97.00] | 0.561 |
| Temperature mean (median [IQR]) | 36.90 [36.70, 37.20] | 36.90 [36.70, 37.20] | 36.90 [36.70, 37.20] | 0.288 |
| Glucose (median [IQR]) | 130.00 [111.32, 155.78] | 129.60 [111.00, 155.67] | 131.35 [112.00, 155.85] | 0.388 |
| SpO₂ (median [IQR]) | 97.00 [96.00, 99.00] | 97.00 [96.00, 99.00] | 97.00 [96.00, 99.00] | 0.984 |
| GCS score (median [IQR]) | 15.00 [15.00, 15.00] | 15.00 [15.00, 15.00] | 15.00 [15.00, 15.00] | 0.497 |
| SOFA score (median [IQR]) | 4.00 [2.00, 7.00] | 4.00 [2.00, 7.00] | 4.00 [2.00, 7.00] | 0.514 |
| SAPS II score (median [IQR]) | 32.00 [23.00, 41.00] | 31.00 [23.00, 41.00] | 32.00 [24.00, 41.00] | 0.061 |
| APS III score (median [IQR]) | 40.00 [30.00, 56.00] | 40.00 [29.00, 56.00] | 41.00 [30.00, 57.00] | 0.068 |
| OASIS score (median [IQR]) | 32.00 [26.00, 38.00] | 32.00 [26.00, 38.00] | 32.00 [26.00, 39.00] | 0.661 |
| Preoperative medication use | ||||
| β blocker use (%) | 0.265 | |||
| No | 4052 (78.38) | 2820 (77.94) | 1232 (79.38) | |
| Yes | 1118 (21.62) | 798 (22.06) | 320 (20.62) | |
| CCB use (%) | 0.068 | |||
| No | 4706 (91.03) | 3311 (91.51) | 1395 (89.88) | |
| Yes | 464 (8.97) | 307 (8.49) | 157 (10.12) | |
| Medication use on the day of surgery | ||||
| Midazolam use (%) | 0.637 | |||
| No | 4643 (89.81) | 3244 (89.66) | 1399 (90.14) | |
| Yes | 527 (10.19) | 374 (10.34) | 153 (9.86) | |
| Dexmedetomidine use (%) | 0.326 | |||
| No | 4880 (94.39) | 3423 (94.61) | 1457 (93.88) | |
| Yes | 290 (5.61) | 195 (5.39) | 95 (6.12) | |
| Morphine use (%) | 0.645 | |||
| No | 4854 (93.89) | 3401 (94.00) | 1453 (93.62) | |
| Yes | 316 (6.11) | 217 (6.00) | 99 (6.38) | |
| Epinephrine use (%) | 0.896 | |||
| No | 5001 (96.73) | 3501 (96.76) | 1500 (96.65) | |
| Yes | 169 (3.27) | 117 (3.24) | 52 (3.35) | |
| Dopamine use (%) | 1.000 | |||
| No | 5106 (98.76) | 3573 (98.76) | 1533 (98.78) | |
| Yes | 64 (1.24) | 45 (1.24) | 19 (1.22) | |
| Red blood cell transfusion (%) | 0.085 | |||
| No | 4499 (87.02) | 3168 (87.56) | 1331 (85.76) | |
| Yes | 671 (12.98) | 450 (12.44) | 221 (14.24) | |
| Platelet transfusion (%) | 0.145 | |||
| No | 4907 (94.91) | 3445 (95.22) | 1462 (94.20) | |
| Yes | 263 (5.09) | 173 (4.78) | 90 (5.80) | |
| Frozen plasma transfusion (%) | 0.682 | |||
| No | 4977 (96.27) | 3486 (96.35) | 1491 (96.07) | |
| Yes | 193 (3.73) | 132 (3.65) | 61 (3.93) | |
| POAF (%) | 0.636 | |||
| No | 4697 (90.85) | 3282 (90.71) | 1415 (91.17) | |
| Yes | 473 (9.15) | 336 (9.29) | 137 (8.83) | |
| Post-Stroke (%) | 0.433 | |||
| No | 4537 (87.76) | 3184 (88.00) | 1353 (87.18) | |
| Yes | 633 (12.24) | 434 (12.00) | 199 (12.82) | |
| Post-TIA (%) | 0.309 | |||
| No | 4869 (94.18) | 3399 (93.95) | 1470 (94.72) | |
| Yes | 301 (5.82) | 219 (6.05) | 82 (5.28) | |
APTT: activated partial thromboplastin time; APS III: acute physiology score III; BUN: blood urea nitrogen; DBP: diastolic blood pressure; GCS: glasgow coma scale; HCT: hematocrit; HR: heart rate; INR: international normalized ratio; IQR: interquartile range; MBP: mean blood pressure; MCH: mean corpuscular hemoglobin; MCV: mean corpuscular volume; OASIS: oxford acute severity of illness score; POAF: postoperative atrial fibrillation; PLT: platelet; PT: prothrombin time; RBC: red blood cell; RDW: red cell distribution width; RR: respiratory rate; SAPS II: simplified acute physiology score II; SBP: systolic blood pressure; SOFA: sequential organ failure assessment; SpO₂: peripheral oxygen saturation; TIA: transient ischemic attack; WBC: white blood cell
P-value < 0.05 was considered statistically significant
A comparative analysis of baseline characteristics between patients with and without POAF is detailed in Supplementary Material 3. Patients with POAF had a significantly higher incidence of postoperative stroke compared to those without POAF (15.86% vs. 11.88%, P = 0.015). No significant difference was observed in the incidence of postoperative TIA between the two groups (7.19% vs. 5.68%, P = 0.181).
Development and validation of a nomogram for POAF
First, LASSO regression was used to preliminarily screen for predictors of POAF in trauma patients. Variables were centered and normalized via tenfold cross-validation (Fig. 2). The predictors selected via LASSO regression included age, history of CHF, SOFA score and SAPS II score on the first day in the ICU. Multivariable logistic regression identified four factors as significant independent predictors of POAF: age (OR 1.06, 95% CI 1.04–1.07, P < 0.001), CHF history (OR 1.85, 95% CI 1.35–2.54, P < 0.001), SOFA score on the first ICU day (OR 1.09, 95% CI 1.03–1.16, P = 0.002), and epinephrine use on the day of surgery (OR 1.95, 95% CI 1.14–3.28, P = 0.013) (Table 2). The primary goal of our model was to establish a preoperative personalized prediction tool for early risk stratification. To ensure clinical practicality and model parsimony, we excluded epinephrine use on the day of surgery from the final nomogram. Epinephrine is a intraoperative intervention that cannot be anticipated during preoperative assessment. Despite its statistical significance as an independent predictor (Table 2), sensitivity analysis confirmed comparable performance between models with and without epinephrine (Supplementary Material 4, 5). We constructed a nomogram of POAF using the following predictive factors (Fig. 3A): age, history of CHF, and SOFA score on the first ICU day. To enhance clinical practicality and convenience, we further developed an interactive online prediction tool (https://mizhou.shinyapps.io/dynnomapp). Clinicians can receive real-time individualized risk estimations by entering basic patient parameters (Fig. 3B). In addition, we have provided the mathematical formula of the model. This facilitates manual calculation and improves clinical applicability. The calculation formula of the linear predictor (LP) is as follows: LP = –6.6217 + (0.1216 × SOFA score) + (0.0527 × Age) + (0.6713 × CHF). Among them, CHF is a binary variable (1 represents a history of CHF, and 0 represents no history of CHF). The final probability of POAF is P = 1/(1 + e–LP). For example, a 75 year-old patient with a SOFA score of 8 has a history of CHF (CHF = 1). Its LP is approximately − 1.027, corresponding to a probability of 26.4% for POAF.
Fig. 2.
Potential predictive variables selection by LASSO regression in the training cohort. LASSO: least absolute shrinkage and selection operator; SE: standard error. A Coefficient paths of candidate variables. The vertical dashed line (λ 1SE) indicates the optimal penalty value where 4 features were selected; B tenfold cross-validation curve. The left dashed line (λ Min) corresponds to the minimum error, and the right line (λ 1SE) represents the most parsimonious model within one standard error. The selected model (λ 1SE) achieved the balance between predictability and simplicity
Table 2.
Univariate and multivariable logistic regression analysis of risk factors for Postoperative Atrial Fibrillation (POAF) in trauma patients
| Variables | Univariable model | Multivariable model | ||||
|---|---|---|---|---|---|---|
| OR | 95%CI | P value | OR | 95%CI | P value | |
| Age | 1.06 | 1.05–1.06 | < 0.001 | 1.06 | 1.04–1.07 | < 0.001 |
| Congestive heart failure | 3.95 | 3.09–5.05 | < 0.001 | 1.85 | 1.35–2.54 | < 0.001 |
| SOFA score | 1.13 | 1.10–1.16 | < 0.001 | 1.09 | 1.03–1.16 | 0.002 |
| Epinephrine use | 3.60 | 2.34–5.54 | < 0.001 | 1.95 | 1.14–3.28 | 0.013 |
CI confidence interval; SOFA: Sequential Organ Failure Assessment; OR: Odds Ratio; POAF: postoperative atrial fibrillation
P-value < 0.05 was considered statistically significant
Fig. 3.
Nomogram for predicting POAF in adult trauma patients undergoing surgery. POAF: postoperative atrial fibrillation; SOFA: sequential organ failure assessment. A The nomogram for predicting POAF. B Screenshot of the interactive online prediction tool (https://mizhou.shinyapps.io/dynnomapp). Clinicians can input the required parameters in the web interface to obtain a real-time, quantitative risk estimation
The nomogram demonstrates excellent discriminative ability, surpassing individual predictive factors. In the training set, the AUC of the model is 0.791 (95% CI 0.768–0.814) (Fig. 4). The sensitivity is 80.4%, the specificity is 65.0%, the accuracy is 66.4%, the positive predictive value (PPV) is 19.0%, and the negative predictive value (NPV) is 97.0% (Supplementary Material 4). Similarly, in the internal validation set, the AUC is 0.800 (95% CI 0.763–0.836) (Fig. 4). PPV is 19.0%, NPV is 97.3%, accuracy is 67.8%, specificity is 66.6%, and sensitivity is 81.0% (Supplementary Material 4). The predictive accuracy of the nomogram was supported by the calibration curves, which showed close concordance between predictions and observations across both the training and validation cohorts (Fig. 5). Hosmer–Lemeshow tests indicated a good model fit (training cohort P = 0.641; validation cohort P = 0.344). The decision curve analysis for the nomogram in the training set and validation set is presented in Fig. 6.
Fig. 4.
Receiver operating characteristic curves of predictive models for POAF. AUC: area under the curve; CHF: congestive heart failure; POAF: postoperative atrial fibrillation; ROC: receiver operating characteristic; SOFA: sequential organ failure assessment. A ROC curves of the training cohort. B ROC curves of the validation cohort. The figure contains four curves corresponding to different predictors: age, CHF, the SOFA score, and the combined model
Fig. 5.
Calibration curves for predicting POAF. A Training cohort; (B) Validation cohort. The curves depict the agreement between predicted probabilities (x-axis) and observed frequencies (y-axis). The grey dashed line represents the ideal reference line where predictions perfectly match observations. The black solid line shows the logistic calibration curve, and the black dotted line represents the nonparametric calibration curve. Model performance metrics, including the C-index, intercept, slope, and Brier score, are displayed for each cohort. A curve closer to the ideal line indicates better calibration
Fig. 6.
Decision curve analysis in the prediction of POAF in trauma patients. POAF: postoperative atrial fibrillation. A Training cohort. B Validation cohort. The red curve represents the nomogram in the training cohort or the validation cohort. The grey lines represent the assumptions that all patients would develop POAF (“All”) or that no patients would develop POAF (“None”). The y-axis indicates the net benefit, and the x-axis represents the threshold probability
Comparison of clinical outcomes between trauma patients with and without POAF
Univariate logistic regression analysis further confirmed a significant association between POAF and postoperative stroke, with POAF being associated with an approximately 1.40-fold increased risk (95% CI 1.08–1.82, P = 0.012). In contrast, the association between POAF and postoperative TIA was not statistically significant (OR 1.29, 95% CI 0.89–1.86, P = 0.184).
The impact of POAF on mortality risk was assessed through survival analysis in the cohort of patients with complete survival data (N = 5168). Univariate Cox analysis revealed that POAF significantly increased the risk of mortality at 7 days (HR 2.84, 95% CI 2.25–3.57, P < 0.001), 28 days (HR 3.21, 95% CI 2.68–3.84, P < 0.001), and 1 year (HR 2.99, 95% CI 2.59–3.47, P < 0.001) postoperatively. After adjusting for potential confounding factors, multivariate analysis demonstrated that POAF remained an independent risk factor for mortality at postoperative 7 days (HR 1.35, 95% CI 1.04–1.76, P = 0.027), postoperative 28 days (HR 1.42, 95% CI 1.16–1.75, P < 0.001), and postoperative 1 year (HR 1.38, 95% CI 1.17–1.63, P < 0.001).
The Kaplan–Meier curve showed that the cumulative survival rates of POAF patients were significantly lower (Log-rank test, all P < 0.0001) (Fig. 7). Patients with POAF had a considerably higher in-hospital mortality rate than those without POAF (28.96% vs. 8.84%, P < 0.001) (Table 3). Additionally, hospital stays (8.00[4.00, 14.00] vs. 6.00 [3.00,11.00], P < 0.001) and ICU stays (3.00[2.00,8.00] vs. 2.00 [1.00,5.00], P < 0.001) were considerably longer in the POAF group (Table 3).
Fig. 7.
Kaplan–meier survival curves comparing POAF versus Non-POAF group in trauma patients. AF: atrial fibrillation; POAF: postoperative atrial fibrillation; P-value < 0.05 was considered statistically significant. A Postoperative 7 day survival; B Postoperative 28 day survival; C Postoperative 1 year survival. The survival probability of the POAF group was significantly lower than that of the Non-POAF group at all time points (Log-rank test, P < 0.0001 for all comparisons). The shaded areas represent the 95% confidence intervals. The number of patients at risk at specific time points is shown below each graph
Table 3.
Comparison of survival outcomes between patients with and without Postoperative Atrial Fibrillation (POAF) following trauma
| Variables | Non-POAF | POAF | P value |
|---|---|---|---|
| N = 4695 | N = 473 | ||
| In-hospital death (%) | 415 (8.84%) | 137 (28.96%) | < 0.001 |
| Hospital days (median [IQR]) | 6.00 [3.00, 11.00] | 8.00 [4.00, 14.00] | < 0.001 |
| Length of ICU stay (median [IQR]) | 2.00 [1.00, 5.00] | 3.00 [2.00, 8.00] | < 0.001 |
IQR: interquartile range; ICU: intensive care unit; POAF: postoperative atrial fibrillation
P-value < 0.05 was considered statistically significant
Discussion
To our knowledge, this study established the first predictive model for POAF in trauma patients. Survival analysis indicated that the cumulative survival rate of the POAF group was significantly lower. Multivariate Cox regression analysis indicated that POAF was significantly associated with death at 7 days, 28 days and 1 year after surgery. The independent risk factors for POAF include: age, history of CHF, SOFA score on the first day in the ICU, and the use of epinephrine on the day of surgery. We have developed a nomogram model, which has excellent discriminative ability, calibration ability. This provides a reliable tool for the early identification of high-risk patients with POAF.
Perioperative AF in trauma patients
Bae et al. reported that the incidence of POAF in patients undergoing hip fracture surgery was 8.16% [3]. The risk of congestive heart failure in patients with POAF increases by 4.86 times, and the ICU admission rate increases by 6.62 times [3]. However, the sample size of this study was only 245 cases, with only 20 cases of POAF [3]. In our study, the incidence of POAF in trauma patients was 9.15%, similar to that in Bae's study. Another study [21] confirmed that the incidence of new-onset AF during the perioperative period was 3.7% (15/410). Perioperative AF is the strongest independent predictor of long-term mortality, increasing the risk of death by 6.7 times [21]. Our research also found that both the short-term and long-term mortality rates of patients with POAF were significantly increased. However, the study populations included in these previous studies primarily consisted of elderly hip fracture patients and had relatively small sample sizes. In contrast, our study focused on adult trauma patients and had a relatively larger sample size, thus suggesting that our model may be applicable to a broader population. Furthermore, Lai et al. [16] reported that trauma patients involved in traffic accidents who developed AF exhibited significantly higher surgical resource utilization and medical costs but poorer ultimate outcomes, with a 131% increased risk of postoperative complications and a 66.5% increased risk of in-hospital mortality being observed. These findings align with our findings. We observed that POAF patients demonstrated significantly longer hospital days and length of ICU stay (Table 2). These results further confirm that POAF promotes poor prognosis in trauma patients.
Therefore, it is crucial to establish a predictive model for POAF in trauma patients. We use multivariate logistic regression to construct a nomogram prediction model. This model can transform complex statistical models into intuitive graphical tools, thereby promoting clinical applications [22]. Our nomogram shows good predictive performance for POAF. The AUC of the training set is 0.791 (95% CI 0.768–0.814). The AUC of the internal validation set reached 0.800 (95% CI 0.763–0.836). The nomogram constructed in this study provides a practical tool for clinical decision-making. Clinical doctors only need to input relevant variables to achieve personalized, rapid, and intuitive risk assessment. This helps to identify high-risk patients with POAF, prevent and intervene in a timely manner.
Possible mechanisms of POAF
The mechanism of POAF in trauma patients may involve multiple pathophysiological processes. This includes excessive activation of the sympathetic nerve, inflammation and oxidative stress, acute hypoxia, metabolic and internal environment disorders, etc. These factors jointly promote rapid electrical and structural remodeling of the atrium, ultimately inducing POAF.
Firstly, excessive activation of the sympathetic nervous system is the core link. Trauma itself, surgical intervention, pain, low blood volume, hypotension, hypoxia, and other factors can cause excessive activation of the sympathetic nervous system [18, 23–26]. Resulting in a significant release of catecholamines and glucocorticoids [18, 23, 25, 26]. The significant increase in sympathetic nervous tension increases the autonomy and contractility of myocardial cells, accelerating conduction [27]. This can cause abnormal rapid discharge of AF trigger lesions [27]. On the other hand, this excessive excitation promotes the influx of calcium ions and the release of sarcoplasmic reticulum calcium [25, 27, 28]. Resulting in intracellular calcium overload. The effective refractory period of the atrium is unevenly shortened, promoting the formation of reentry mechanisms and inducing AF [28].
Secondly, systemic inflammatory response and oxidative stress are another key mechanism. Severe trauma and surgery can trigger systemic inflammatory reactions and oxidative stress [28–30]. These two reactions further exacerbate atrial electrophysiology and structural remodeling [30]. The sharp increase in pro-inflammatory cytokines (such as IL-6, TNF-α) and oxidative stress products can directly cause damage to atrial myocytes, interstitial fibrosis, and slow conduction velocity [27, 30]. Thereby lowering the threshold for the occurrence of AF [27, 30]. Moreover, AF itself can further exacerbate the inflammatory state, forming a vicious cycle between inflammation and AF [27].
Additionly, acute hypoxia is common in trauma patients. Hypoxia directly exacerbates sympathetic nervous system excitation [25]. It also causes pulmonary blood vessels to contract, increasing right atrial pressure and tension [25]. These changes can cause improper myocardial conduction and arrhythmia [25]. Meanwhile, myocardial ischemia and hypoxia can lead to intracellular acidosis, adenosine triphosphate (ATP) depletion, and ion pump dysfunction [25]. Consequently, it affects the electrical activity of the heart and provides an electrophysiological basis for the occurrence of AF [25].
Finally, metabolic and internal environmental disturbances are also key links in POAF [31, 32]. Factors such as traumatic stress, pain, low blood volume, and intraoperative hypotension can lead to electrolyte disturbances (such as hypokalemia and hypomagnesemia) [25, 31]. This will directly lead to unstable myocardial electrical activity. Excessive fluid load during resuscitation may cause acute atrial dilation [25, 31]. The mechanical dilation of the atrium can alter its electrical conductivity and promote the maintenance of AF [25, 31]. As a result, POAF is caused by several interrelated causes.
Risk factors and underlying mechanisms for POAF in trauma patients
Increased age is a clear risk factor for POAF [2, 33–35]. Our research shows that for every one-year increase in age, the risk of developing POAF increases by 6%. This is consistent with the study by Amar et al. [37]. They found that for each decade increase in age during thoracic surgery, the probability of POAF increased by 1.8 times [37]. A meta-analysis of a large cohort showed a positive correlation between age and the risk of POAF in non-cardiac surgery [38]. Mechanistically, as age increases, atrial tissue gradually becomes fibrotic, and the ion channel function related to electrocardiographic activity also deteriorates [32, 36]. At the same time, the diastolic ability of the heart is weakened [32, 36]. These alterations may lead to conduction delay and reentry formation, increasing susceptibility to AF [32, 36].
In our study, for every 1-point increase in SOFA score, the risk of POAF increased by approximately 9% (95% CI 1.03–1.16, P = 0.002). This discovery is consistent with the research of Li Y et al. [36]. In this study, SOFA score was an independent predictor of AF occurrence in ICU patients with acute heart failure [36]. The study by Li Z et al. in sepsis patients also supports our conclusion [39]. They found that for every 1-point increase in SOFA score, the risk of AF increased by 31% [39]. The mechanism may be an increase in SOFA score, reflecting an exacerbation of systemic inflammatory response and multiple organ failure status [39]. As mentioned earlier, severe inflammation can release a large amount of pro-inflammatory cytokines (such as IL-6, TNF-α) [27, 30]. These substances directly disrupt the electrophysiological stability of the myocardium and promote the occurrence of AF [27, 30].
Perioperative epinephrine use is another independent risk factor for POAF. In our research, the use of epinephrine increased the risk of POAF by approximately 1.95-fold (95% CI 1.14–3.28, P = 0.013). This result is consistent with the study conducted by Seguin P et al. [17] on trauma patients admitted to the ICU. They reported a 5.7-fold increase in the risk of AF associated with catecholamine use [17]. The increased POAF risk associated with perioperative epinephrine use may be related to catecholamine-induced β-adrenergic receptor activation, thus leading to imbalance in intracellular calcium homeostasis, increased myocardial automaticity, and triggered activity [17, 27].
We also reported that a history of CHF increased the risk of POAF by approximately 1.85-fold (95% CI 1.35–2.54, P < 0.001). A history of CHF was identified as a significant risk factor for POAF, demonstrating a 2.51-fold increase in risk in a cohort of 2588 thoracic surgery patients from the study by Vaporciyan et al. [40] Similarly, the meta-analysis by Chebbout et al. [38] reported that CHF is typically linked to a significant risk of POAF. The mechanism of these results may involve atrial dilation, increased pressure load, and neurohormonal activation, which are common in heart failure patients [41]. These pathological changes promote atrial electrical and structural remodeling, thereby predisposing patients to AF and increasing POAF risk [41].
Limitations
This study, based on an analysis of the MIMIC-IV v2.2 database, identified independent predictors of POAF in trauma patients and developed the first nomogram for risk prediction. However, several limitations must be acknowledged. Firstly, generalizability may be limited by variations in trauma mechanisms, perioperative care, and ICU admission criteria across regions. The clinical applicability of our nomogram requires further external validation in multi-center cohorts. Secondly, the limitations of this study also stem from the data availability of the MIMIC-IV database. For variables missing less than 20%, we used Random Forest to fill in the gaps. Although this method is widely used. It still may bring potential estimation biases and affect the model's accuracy. Also, there is a lack of records of anesthesia types and postoperative myocardial injury markers (such as troponin T, NT proBNP) [42]. These potential unmeasured confounding factors may limit the practicality of the model. Finally, we were unable to compare the developed nomogram model with existing POAF risk scores [9]. This means we could not further validate its discriminative ability and demonstrate its potential clinical benefits. Moreover, the positive predictive value of the model suggests that it may be more appropriately applied as a screening tool in clinical practice. Additionally, the broad 30 day POAF definition may not distinguish between early- and late-onset cases. Our future research aims to develop separate prediction models for early (≤ 7 days) and late (8–30 days) POAF to guide stage-specific monitoring.
Conclusion
Our study reveals that POAF is associated with adverse clinical outcomes in trauma patients. Age, a history of CHF, and the first-day ICU SOFA score emerging as significant predictors of POAF. This study is the first to construct and validate a risk prediction model for POAF in trauma patients. Our nomogram exhibits good predictive performance.
Supplementary Information
Abbreviation
- AUC
Area under the curve
- AF
Atrial fibrillation
- ALB
Albumin
- ALP
Alkaline phosphatase
- ALT
Alanine aminotransferase
- APS III
Acute physiology score III
- APTT
Activated partial thromboplastin time
- AST
Aspartate aminotransferase
- ATP
Adenosine triphosphate
- BIDMC
Beth Israel Deaconess Medical Center
- BUN
Blood urea nitrogen
- CCB
Calcium channel blockers
- CHF
Congestive heart failure
- CI
Confidence interval
- CITI
Collaborative Institutional Training Initiative
- CRP
C-reactive protein
- DBP
Diastolic blood pressure
- DCA
Decision curve analysis
- GCS
Glasgow coma scale
- GLB
Globulin
- HCT
Hematocrit
- HDL
High-density lipoprotein
- HR
Hazard ratio
- IBIL
Indirect bilirubin
- ICU
Intensive Care Unit
- INR
International normalized ratio
- LDH
Lactate dehydrogenase
- LASSO
Least absolute shrinkage and selection operator
- LDL
Low-density lipoprotein
- LP
Linear predictor
- LOS
Length of stay
- LYMPH
Lymphocytes
- MBP
Mean blood pressure
- MCH
Mean corpuscular hemoglobin
- MCV
Mean corpuscular volume
- MIMIC
Medical Information Mart for Intensive Care
- MIT
Massachusetts Institute of Technology
- MONO
Monocytes
- NEUT
Neutrophils
- NPV
Negative predictive value
- OASIS
Oxford acute severity of illness score
- OR
Odds ratio
- PLT
Platelets
- POAF
Postoperative atrial fibrillation
- PP
Pulse pressure
- PPV
Positive predictive value
- PT
Prothrombin time
- RBC
Red blood cells
- RDW
Red cell distribution width
- ROC
Receiver operating characteristic
- RR
Respiratory rate
- SAPS II
Simplified acute physiology score II
- SBP
Systolic blood pressure
- SOFA
Sequential organ failure assessment
- SpO₂
Peripheral oxygen saturation
- TBIL
Total bilirubin
- TIA
Transient ischemic attack
- VIF
Variance inflation factor
- WBC
White blood cell
Author contributions
Conception and design: Q.M., X.X., Y.R., C.L and M.Z.; Data curation: P.H.; Data analysis and interpretation: X.X., M.Z., P.H.; Funding acquisition: Q.M., X.X.; Investigation: X.X., M.Z., P.H., Y.R., C.L.; Methodology: X.X., M.Z.; Project administration: Q.M., Y.R.,C.L.; Resources: Q.M., Y.R., X.X., C.L.; Supervision: Q.M., Y.R., C.L.; Validation: X.X., M.Z.; Visualization: X.X., M.Z., P.H., Y.R.; Writing-original draft: X.X.; Writing-review & editing: X.X. M.Z., P.H., Y.R., C.L., Q.M.
Funding
This work was supported by the Chongqing Talent Program for Leading Innovation (CSTC2024YCJH-BGZXM0011), the Joint Medical Research Project of Chongqing Science and Health Commission/Health Commission (2025MSXM042), the Daping Hospital Project (ZXAIYB020), and the China Scholarship Council (CSC).
Data availability
The data utilized in this study were obtained from the MIMIC-IV database v2.2, which is a publicly available but de-identified critical care database. Access to the MIMIC database requires completing a recognized course in the protection of human research participants and signing a data use agreement. Detailed instructions for gaining access can be found at: https://mimic.mit.edu.
Declarations
Ethics approval and consent to participate
The establishment of the MIMIC-IV database was approved by the Institutional Review Boards (IRB) of BIDMC and the Massachusetts Institute of Technology (MIT), which granted a waiver of informed consent. One of the authors of this study completed the Collaborative Institutional Training Initiative (CITI) program and was granted permission to extract data (Certification ID: 66716350).
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Xiaojuan Xiong and Mi Zhou have contributed equally to this work.
Qingxiang Mao and Chang Liu are co-corresponding authors.
Contributor Information
Chang Liu, Email: frog_0612@tmmu.edu.cn.
Qingxiang Mao, Email: qxmao@tmmu.edu.cn.
References
- 1.GBD 2021 Diseases and Injuries Collaborators. Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2024;403(10440):2133–61. 10.1016/S0140-6736(24)00757-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Haysley J, Soliman-Aboumarie H, Huang J, Kalra DK. Perioperative atrial fibrillation. BJA Educ. 2025;25(3):99–106. 10.1016/j.bjae.2024.10.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bae SJ, Kwon CH, Kim TY, et al. Predictors and prognostic impact of post-operative atrial fibrillation in patients with hip fracture surgery. World J Clin Cases. 2022;10(11):3379–88. 10.12998/wjcc.v10.i11.3379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Gupta BP, Steckelberg RC, Gullerud RE, et al. Incidence and 1-year outcomes of perioperative atrial arrhythmia in elderly adults after hip fracture surgery. J Am Geriatr Soc. 2015;63(11):2269–74. 10.1111/jgs.13789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Li W, Min A, Zhao W, et al. Predictors and prognosis of elderly hip fracture patients with perioperative atrial fibrillation: a nested case-control study. BMC Geriatr. 2025;25(1):4. 10.1186/s12877-024-05647-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Xiang K, Akram M, Elbossaty WF, Yang J, Fan C. Exosomes in atrial fibrillation: therapeutic potential and role as clinical biomarkers. Heart Fail Rev. 2022;27(4):1211–21. 10.1007/s10741-021-10142-5. [DOI] [PubMed] [Google Scholar]
- 7.Guo R, Fan C, Sun Z, et al. Clinical efficacy and safety of Cox-maze IV procedure for atrial fibrillation in patients with aortic valve calcification. Front Cardiovasc Med. 2023;10:1092068. 10.3389/fcvm.2023.1092068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sun Z, Fan C, Song L, et al. Effect of electrophysiological mapping on non-transmural annulus ablation and atrial fibrillation recurrence prediction after 6 months of Cox-Maze IV procedure. Front Cardiovasc Med. 2022;9:931845. 10.3389/fcvm.2022.931845. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Azimaraghi O, Rudolph MI, Wongtangman K, et al. Role of anticoagulation therapy in modifying stroke risk associated with new-onset atrial fibrillation after non-cardiac surgery. Nat Med. 2024;30(11):3310–7. 10.1038/s41591-024-03206-0. [DOI] [PubMed] [Google Scholar]
- 10.Jiang S, Liao X, Chen Y, Li B. Exploring postoperative atrial fibrillation following noncardiac surgery: mechanisms, risk factors, and prevention strategies. Front Cardiovasc Med. 2023;10:1273547. 10.3389/fcvm.2023.1273547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Conen D, Alonso-Coello P, Douketis J, et al. Risk of stroke and other adverse outcomes in patients with perioperative atrial fibrillation 1 year after noncardiac surgery. Eur Heart J. 2020;41(5):645–51. 10.1093/eurheartj/ehz431. [DOI] [PubMed] [Google Scholar]
- 12.Alonso-Coello P, Cook D, Xu SC, et al. Predictors, prognosis, and management of new clinically important atrial fibrillation after noncardiac surgery: a prospective cohort study. Anesth Analg. 2017;125(1):162–9. 10.1213/ANE.0000000000002111. [DOI] [PubMed] [Google Scholar]
- 13.Smith H, Li H, Brandts-Longtin O, et al. External validity of a model to predict postoperative atrial fibrillation after thoracic surgery. Eur J Cardiothorac Surg. 2020;57(5):874–80. 10.1093/ejcts/ezz341. [DOI] [PubMed] [Google Scholar]
- 14.William J, Rowe K, Hogarty J, et al. Predictors of late atrial fibrillation recurrence after cardiac surgery. JACC Clin Electrophysiol. 2024;10(7 Pt 2):1711–9. 10.1016/j.jacep.2024.05.030. [DOI] [PubMed] [Google Scholar]
- 15.Staerk L, Sherer JA, Ko D, Benjamin EJ, Helm RH. Atrial fibrillation: epidemiology, pathophysiology, and clinical outcomes. Circ Res. 2017;120(9):1501–17. 10.1161/CIRCRESAHA.117.309732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lai HC, Chien WC, Chung CH, et al. Atrial fibrillation increases medical cost and complicates hospital outcome of traffic accident-related physical trauma--a nationwide population-based study. Int J Cardiol. 2014;177(3):964–9. 10.1016/j.ijcard.2014.09.190. [DOI] [PubMed] [Google Scholar]
- 17.Seguin P, Laviolle B, Maurice A, Leclercq C, Mallédant Y. Atrial fibrillation in trauma patients requiring intensive care. Intensive Care Med. 2006;32(3):398–404. 10.1007/s00134-005-0032-2. [DOI] [PubMed] [Google Scholar]
- 18.Fu M, Zhang Y, Zhao Y, et al. Characteristics of preoperative atrial fibrillation in geriatric patients with hip fracture and construction of a clinical prediction model: a retrospective cohort study. BMC Geriatr. 2023;23(1):310. 10.1186/s12877-023-03936-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1. 10.1038/s41597-022-01899-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Zhou Z, Zhang H, Gu Y, Zhang K, Ouyang C. Relationship between glycemic variability and the incidence of postoperative atrial fibrillation following cardiac surgery: a retrospective study from MIMIC-IV database. Diabetes Res Clin Pract. 2025;219:111978. 10.1016/j.diabres.2024.111978. [DOI] [PubMed] [Google Scholar]
- 21.Leibowitz D, Abitbol C, Alcalai R, Rivkin G, Kandel L. Perioperative atrial fibrillation is associated with increased one-year mortality in elderly patients after hip fracture repair. Int J Cardiol. 2017;227:58–60. 10.1016/j.ijcard.2016.11.067. [DOI] [PubMed] [Google Scholar]
- 22.Wang X, Lu J, Song Z, Zhou Y, Liu T, Zhang D. From past to future: bibliometric analysis of global research productivity on nomogram (2000-2021). Front Public Health. 2022;10:997713. 10.3389/fpubh.2022.997713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Heintz KM, Hollenberg SM. Perioperative cardiac issues: postoperative arrhythmias. Surg Clin North Am. 2005;85(6):1103–viii. 10.1016/j.suc.2005.09.003. [DOI] [PubMed] [Google Scholar]
- 24.Lenz A, Franklin GA, Cheadle WG. Systemic inflammation after trauma. Injury. 2007;38(12):1336–45. 10.1016/j.injury.2007.10.003. [DOI] [PubMed] [Google Scholar]
- 25.Danelich IM, Lose JM, Wright SS, et al. Practical management of postoperative atrial fibrillation after noncardiac surgery. J Am Coll Surg. 2014;219(4):831–41. 10.1016/j.jamcollsurg.2014.02.038. [DOI] [PubMed] [Google Scholar]
- 26.Bjerrum E, Wahlstroem KL, Gögenur I, Burcharth J, Ekeloef S. Postoperative atrial fibrillation following emergency noncardiothoracic surgery: a systematic review. Eur J Anaesthesiol. 2020;37(8):671–9. 10.1097/EJA.0000000000001265. [DOI] [PubMed] [Google Scholar]
- 27.Han L, Cai W. Advances in the pathogenesis and treatment of atrial fibrillation. J Pract Electrocardiol. 2024;33(1):87–92. 10.13308/j.issn.2095-9354.2024.01.020. [Google Scholar]
- 28.Gaudino M, Di Franco A, Rong LQ, Piccini J, Mack M. Postoperative atrial fibrillation: from mechanisms to treatment. Eur Heart J. 2023;44(12):1020–39. 10.1093/eurheartj/ehad019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gaudino M, Andreotti F, Zamparelli R, et al. The -174G/C interleukin-6 polymorphism influences postoperative interleukin-6 levels and postoperative atrial fibrillation. Is atrial fibrillation an inflammatory complication? Circulation. 2003;108(1):II195–9. 10.1161/01.cir.0000087441.48566.0d. [DOI] [PubMed] [Google Scholar]
- 30.Zakkar M, Ascione R, James AF, Angelini GD, Suleiman MS. Inflammation, oxidative stress and postoperative atrial fibrillation in cardiac surgery. Pharmacol Ther. 2015;154:13–20. 10.1016/j.pharmthera.2015.06.009. [DOI] [PubMed] [Google Scholar]
- 31.Hadjizacharia P, O’Keeffe T, Brown CV, et al. Incidence, risk factors, and outcomes for atrial arrhythmias in trauma patients. Am Surg. 2011;77(5):634–9. 10.1177/000313481107700526. [DOI] [PubMed] [Google Scholar]
- 32.Prince-Wright LH, Akinyemi O, Nnorom SO, Bauer ES, Cornwell EE III, Fullum TM. Postoperative atrial fibrillation following noncardiac surgery: predictors and risk of mortality. Am J Surg. 2022;224(4):1062–7. 10.1016/j.amjsurg.2022.07.010. [DOI] [PubMed] [Google Scholar]
- 33.Lohani KR, Nandipati KC, Rollins SE, et al. Transthoracic approach is associated with increased incidence of atrial fibrillation after esophageal resection. Surg Endosc. 2015;29(7):2039–45. 10.1007/s00464-014-3908-9. [DOI] [PubMed] [Google Scholar]
- 34.Kazaure HS, Roman SA, Tyler D, Sosa JA. The significance of atrial fibrillation in patients aged ≥ 55 years undergoing abdominal surgery. World J Surg. 2015;39(1):113–20. 10.1007/s00268-014-2777-7. [DOI] [PubMed] [Google Scholar]
- 35.Frendl G, Sodickson AC, Chung MK, et al. 2014 AATS guidelines for the prevention and management of perioperative atrial fibrillation and flutter for thoracic surgical procedures. J Thorac Cardiovasc Surg. 2014;148(3):e153–93. 10.1016/j.jtcvs.2014.06.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Li Y, Cai Z, She Y, Shen W, Wang T, Luo L. Development and validation of a nomogram for predicting atrial fibrillation in patients with acute heart failure admitted to the ICU: a retrospective cohort study. BMC Cardiovasc Disord. 2022;22(1):528. 10.1186/s12872-022-02973-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Amar D, Goenka A, Zhang H, Park B, Thaler HT. Leukocytosis and increased risk of atrial fibrillation after general thoracic surgery. Ann Thorac Surg. 2006;82(3):1057–61. 10.1016/j.athoracsur.2006.03.103. [DOI] [PubMed] [Google Scholar]
- 38.Chebbout R, Heywood EG, Drake TM, et al. A systematic review of the incidence of and risk factors for postoperative atrial fibrillation following general surgery. Anaesthesia. 2018;73(4):490–8. 10.1111/anae.14118. [DOI] [PubMed] [Google Scholar]
- 39.Li Z, Pang M, Li Y, et al. Development and validation of a predictive model for new-onset atrial fibrillation in sepsis based on clinical risk factors. Front Cardiovasc Med. 2022;9:968615. 10.3389/fcvm.2022.968615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Vaporciyan AA, Correa AM, Rice DC, et al. Risk factors associated with atrial fibrillation after noncardiac thoracic surgery: analysis of 2588 patients. J Thorac Cardiovasc Surg. 2004;127(3):779–86. 10.1016/j.jtcvs.2003.07.011. [DOI] [PubMed] [Google Scholar]
- 41.Newman JD, O’Meara E, Böhm M, et al. Implications of atrial fibrillation for guideline-directed therapy in patients with heart failure: JACC state-of-the-art review. J Am Coll Cardiol. 2024;83(9):932–50. 10.1016/j.jacc.2023.12.033. [DOI] [PubMed] [Google Scholar]
- 42.Duchnowski P, Śmigielski W. Usefulness of myocardial damage biomarkers in predicting cardiogenic shock in patients undergoing heart valve surgery. Kardiol Pol. 2024;82(4):423–6. 10.33963/v.phj.99553. [DOI] [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 data utilized in this study were obtained from the MIMIC-IV database v2.2, which is a publicly available but de-identified critical care database. Access to the MIMIC database requires completing a recognized course in the protection of human research participants and signing a data use agreement. Detailed instructions for gaining access can be found at: https://mimic.mit.edu.