Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 May 1.
Published in final edited form as: J Vasc Surg. 2016 Mar 19;63(5):1240–1247. doi: 10.1016/j.jvs.2015.12.046

Transient postoperative atrial fibrillation after abdominal aortic aneurysm repair increases mortality risk

Anai N Kothari a,b, Pegge M Halandras a,b, Max Drescher b,c, Robert H Blackwell b,d, Dawn M Graunke b,e, Stephanie Kliethermes f, Paul C Kuo a,b, Jae S Cho a,b
PMCID: PMC5110229  NIHMSID: NIHMS828144  PMID: 27005752

Abstract

Objective

The purpose of this study was to determine whether new-onset transient postoperative atrial fibrillation (TPAF) affects mortality rates after abdominal aortic aneurysm (AAA) repair and to identify predictors for the development of TPAF.

Methods

Patients who underwent open aortic repair or endovascular aortic repair for a principal diagnosis AAA were retrospectively identified using the Healthcare Cost and Utilization Project-State Inpatient Database (Florida) for 2007 to 2011 and monitored longitudinally for 1 year. Inpatient and 1-year mortality rates were compared between those with and without TPAF. TPAF was defined as new-onset atrial fibrillation that developed in the postoperative period and subsequently resolved in patients without a history of atrial fibrillation. Cox proportional hazards models, adjusted for age, gender, comorbidities, rupture status, and repair method, were used to assess 1-year survival. Predictive models were built with preoperative patient factors using Chi-squared Automatic Interaction Detector decision trees and externally validated on patients from California.

Results

A 3.7% incidence of TPAF was identified among 15,148 patients who underwent AAA repair. The overall mortality rate was 4.3%. The inpatient mortality rate was 12.3% in patients with TPAF vs 4.0% in those without TPAF. In the ruptured setting, the difference in mortality was similar between groups (33.7% vs 39.9%, P = .3). After controlling for age, gender, comorbid disease severity, urgency (ruptured vs nonruptured), and repair method, TPAF was associated with increased 1-year postoperative mortality (hazard ratio, 1.48; P < .001) and postdischarge mortality (hazard ratio, 1.56; P = .028). Chi-squared Automatic Interaction Detector-based models (C statistic = 0.70) were integrated into a Web-based application to predict an individual's probability of developing TPAF at the point of care.

Conclusions

The development of TPAF is associated with an increased risk of mortality in patients undergoing repair of nonruptured AAA. Predictive modeling can be used to identify those patients at highest risk for developing TPAF and guide interventions to improve outcomes. (J Vasc Surg 2016;63:1240-7.)

Atrial fibrillation (AF) is a common cardiac dysrhythmia and confers an elevated risk of myocardial infarction, stroke, and death.1-4 New-onset transient postoperative atrial fibrillation (TPAF) is a common complication after cardiac and noncardiac surgery.5 Several factors have been identified that may confer increased risk of TPAF, including inappropriate pain management, hypervolemia/hypovolemia, local and systemic inflammatory processes, intraoperative mechanical manipulation of the heart, and myocardial injury.6,7

Although studies have demonstrated increased cost of care, length of stay, inpatient mortality and morbidity, including increased postoperative stroke risk in cardiac surgery patients who develop TPAF,8-10 TPAF has been considered a benign condition after noncardiac surgery. However, the adverse effect of TPAF after noncardiac surgery has been increasingly recognized. In a study of 200 patients undergoing elective abdominal aortic aneurysm (AAA) repair, Noorani et al11 demonstrated an association between the development of TPAF and increased hospital length of stay but found no effect on long-term outcomes. Unlike cardiothoracic surgery, where thoracotomy and mechanical manipulation of the heart likely play a major causative role in the development of TPAF, AAA repair may be associated with development of new-onset supraventricular dysrhythmias through other mechanisms.12,13 Given this indirect relationship, identification of preoperative variables that predict the development of TPAF may be feasible.

The overall objective of this study was to measure the effect of TPAF after AAA repair on mortality and develop a clinically applicable predictive model to identify patients at risk for TPAF. We hypothesized that patients who develop TPAF are at increased risk of death. To test this, a population-based cohort review was performed to compare inpatient and 1-year survival in patients with and without postoperative TPAF after AAA repair. We also hypothesized that commonly collected individual patient data could predict patients at risk for developing TPAF, and a Web-based predictive analytic client was built to test this hypothesis.

Methods

Data source

This was a population-based, retrospective study with patient-level data from the Healthcare Cost and Utilization Project State Inpatient Database (HCUP-SID) for the states of Florida and California. HCUP-SID files were developed through sponsorship by the Agency of Health Research and Quality and include all patient discharge records, regardless of payer, for the states that participate in the project. Each SID is unique to its individual state. Data are deidentified, protected, and include >100 clinical and nonclinical variables.12,14 The study was exempt from Institutional Review Board approval based on the use of deidentified records. Owing to the nature of the administrative database, it is neither necessary nor possible to obtain informed consent from patients.

Patient inclusion

Clinical records were identified between 2007 and 2011 using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes. The study included patients with a principal diagnosis of nonruptured or ruptured AAA and who underwent open aortic repair (OAR) or endovascular aortic repair (EVAR). ICD-9-CM diagnosis codes used to define the study population were 441.4 or 441.3 in combination with procedure codes 38.44 or 39.71. Excluded were patients with a diagnosis of aortic dissection (441.0), thoracic or thoracoabdominal aortic aneurysm (441.1, 441.2, 441.6, 441.7), coarctation of the aorta (747.1), Marfan syndrome and other congenital anomalies (759.8), gonadal dysgenesis (758.6), and polyarteritis nodosa (446.0). Also excluded were patients with a known history of AF.

Patient characteristics

Patient demographic variables, based on availability in HCUP-SID, included gender and race (Caucasian, African American, Hispanic, other). Socioeconomic factors included median annual income by zip code ($0-$38,999, $39,000-$47,999, $48,000-$63,999, and ≥$64,000) and primary insurance type (Medicare, Medicaid, private, other).

Clinical characteristics included comorbidities assigned based on Agency of Health Research and Quality comorbidity software. This software uses ICD-9-CM codes and diagnosis-related groups to identify conditions not related to the principal diagnosis and group them into usable categories to describe patient comorbidity.15 The Deyo adaptation of the Charlson Comorbidity Index (CCI) was calculated using ICD-9-CM codes to assign disease severity.16

Outcomes

Development of TPAF was defined using three criteria: (1) AF was not present on admission, (2) no prior documented history of AF, and (3) of patients who developed AF, it was not present on subsequent encounters. AF was identified using the ICD-9-CM diagnosis codes 427.3, 427.31, or 427.32. This approach was based on prior literature identifying patients with AF and validating the use of indicators for AF present on admission.17-21

The primary study outcomes were inpatient and 1-year mortality. One-year mortality was determined using HCUP-SID supplemental revisit variables that allow individual patients to be monitored through multiple inpatient encounters.

Statistical analysis

Baseline patient characteristics and surgical outcomes in patients with and without TPAF were compared using independent t-tests for continuous variables and χ2 tests for categoric variables. Time-to-event analysis was conducted using the Kaplan-Meier survival function, with patient death as the measured event. Patients were right censored at their last follow-up date. The log-rank test was used to compare between-group differences in stratified measures. Multivariable Cox proportional hazards models were fit using demographic and clinical covariates to estimate the effect of TPAF on mortality over time. Statistical analyses were conducted using Stata 13 software (StataCorp LP, College Station, Tex).

Predictive modeling

Predictive models were built using the Chi-squared Automatic Interaction Detector (CHAID) for growing and pruning decision trees using the following settings: the maximum tree depth was seven, the minimum number of cases in parent node was set to 36, and the minimum number of cases in the child node was set to 20. Variables included primary payer, gender, race, income, patient comorbidities, operative approach, urgency of procedure (rupture vs nonrupture), and weekend admission. Training was performed on a balanced data set, and a randomly selected 10% hold-out sample was used for model validation. The finalized predictive model performance was assessed using the complete data set. External validation of models was done using patient encounters from the state of California in 2011, identified with the same inclusion and exclusion criteria as the study population. Predictive modeling was performed using R x64 3.0.2 software (The R Foundation for Statistical Computing, http://www.r-project.org/foundation/) and SPSS 21 software (IBM Corp, Armonk, NY).

Web-based interface development

A Web-based interface (regular computer display and mobile-friendly version) was built using Dreamweaver CS6 (Adobe Systems, San Jose, Calif) in HTML format. The interactive form and the subsequent predictions were coded using PHP server-side scripting language.

Results

Baseline characteristics of study population

We identified 15,148 patients in the state of Florida with a diagnosis of AAA who underwent repair and met the inclusion criteria. The incidence rate of TPAF was 3.7% (554 of 15,148). Patient demographic, socioeconomic, and operative characteristics of the study population are summarized in Table I.

Table I. Baseline demographic, socioeconomic, and clinical characteristics of study population.

Variablesa All patients (N = 15,148) No TPAF (n = 14,594) TPAF (n = 554) P value
Age, years 73.7 ± 8.5 73.7 ± 8.5 75.6 ± 7.7 <.001
Female gender 2827 (18.7) 2684 (18.4) 143 (25.8) <.001
Race .032
 White 13,401 (88.5) 12,893 (88.3) 508 (91.7)
 Black 502 (3.3) 494 (3.4) 8 (1.4)
 Hispanic 842 (5.6) 819 (5.6) 23 (4.2)
 Other 403 (2.7) 388 (2.7) 15 (2.7)
Insurance type .001
 Medicare 12,665 (83.6) 12,167 (83.4) 498 (89.9)
 Medicaid 195 (1.3) 189 (1.3) 6 (1.1)
 Private 1908 (12.6) 1868 (12.8) 40 (7.2)
 Other 380 (2.5) 370 (2.5) 10 (1.8)
Comorbidities
 Deficiency anemia 2367 (15.6) 2196 (15.1) 171 (30.8) <.001
 Congestive heart failure 61 (0.4) 48 (0.3) 13 (2.4) <.001
 Chronic lung disease 5245 (34.6) 4983 (34.1) 262 (47.3) <.001
 Coagulopathy 1486 (9.8) 1324 (9.1) 162 (29.2) <.001
 Diabetes Mellitus 2755 (18.2) 2664 (18.3) 91 (16.4) .27
 Chronic hypertension 11,218 (74.1) 10,801 (74.0) 417 (75.3) .51
 Hypothyroidism 1214 (8.0) 1166 (8.0) 48 (8.7) .57
 Electrolyte disorders 2499 (16.5) 2259 (15.5) 240 (43.3) <.001
 Neurologic disorder 533 (3.5) 503 (3.5) 30 (5.4) .014
 Obesity 1179 (7.8) 1136 (7.8) 43 (7.8) .99
 Peripheral vascular disease 5798 (38.3) 5538 (38.0) 260 (47.0) <.001
 Chronic renal disease 2001 (13.2) 1878 (12.9) 123 (22.2) <.001
 Malnutrition 418 (2.8) 349 (2.4) 69 (12.5) <.001
Age-adjusted CCI 2.2 (1.3) 2.2 (1.3) 2.6 (1.4) <.001
Operative characteristics <.001
 Nonruptured
  OAR 2512 (16.6) 2279 (15.6) 233 (42.1)
  EVAR 11,594 (76.5) 11,356 (77.8) 238 (43.0)
 Ruptured
  OAR 674 (4.5) 612 (4.2) 62 (11.2)
  EVAR 368 (2.4) 347 (2.4) 21 (3.8)
Open in a new tab

CCI, Charlson Comorbidity Index; EVAR, endovascular aortic repair; OAR, open aortic repair; TPAF, transient postoperative atrial fibrillation.

a

Continuous variables are shown as the mean ± standard deviation and categoric variables as number (%).

Bivariate analysis showed patients with TPAF were older (75.6 ± 7.7 vs 73.7 ± 8.5 years; P < .001) and had a higher preoperative prevalence of deficiency anemias (P < .001), congestive heart failure (P < .001), chronic lung disease (P < .001), coagulopathy (P < .001), electrolyte disorders (P < .001), peripheral vascular disease (P < .001), chronic renal insufficiency (P < .001), and malnutrition (P < .001). Patients who went on to develop TPAF also had a significantly greater baseline burden of comorbid disease severity on the CCI score (2.6 ± 1.4 vs 2.2 ± 1.3, P < .001) than those who did not.

Effect of TPAF on inpatient outcomes and 1-year survival

Crude inpatient and 1-year mortality rates were compared between patients with and without TPAF. Inpatient mortality for patients without TPAF was 8.3% less than those who developed TPAF (4.0% vs 12.3%; P < .001). Fig 1 details inpatient mortality rates stratified by operative approach (OAR vs EVAR) and urgency (rupture vs intact). Crude inpatient mortality rates were not significantly different between groups with and without TPAF after repair of ruptured AAA, regardless of approach. In contrast, inpatient mortality rates were higher in TPAF groups after intact AAA repair after OAR (9.4% and 4.5%; P < .001) and EVAR (7.6% vs 2.9%; P < .001). The incidence of TPAF was significantly higher after non-ruptured OAR (233 of 2512) than after EVAR (238 of 11,594), but the mortality rate was the same in those who developed TPAF regardless of approach (P = .47).

Fig 1.

Fig 1

Open in a new tab

Inpatient and 1-year mortality stratified by open aortic repair (OAR) vs endovascular aortic repair (EVAR) and urgency of repair (ruptured vs nonruptured). Mortality rates are reported as unadjusted percentages. TPAF, Transient postoperative atrial fibrillation.

Survival models were used to further study the effect of TPAF on 1-year outcomes (Fig 2). For patients who developed TPAF, 1-year postoperative survival was 75.8% (95% confidence interval [CI], 71.4%-79.7%) compared with 90.3% (95% CI, 89.7%-90.8%) for those without TPAF. From time of discharge (for those who survived the index AAA repair), 1-year survival in TPAF patients was 90.9% (95% CI, 87.3%-93.5%) compared with 96.0% (95% CI, 95.5%-96.4%). After adjustment for potential confounders, including baseline patient characteristics, operative approach, and urgency, the overall hazard ratio for 1-year postoperative mortality was 1.48 (95% CI, 1.19-1.84; P < .001) and 1.56 (95% CI, 1.05-2.33; P = .028) for 1-year postdischarge mortality.

Fig 2.

Fig 2

Open in a new tab

A, Survival analysis of patients with and without transient postoperative atrial fibrillation (TPAF) at 365 days postoperatively (P < .001 by log-rank). B, Survival analysis of patients with and without TPAF from time of discharge (P = .001 by log-rank). *Hazard ratio (HR) reported with 95% confidence interval (CI). †Cox proportional hazards model controlling for patient demographics (age, gender), operative approach, comorbid disease severity, and urgency of procedure.

Predicting the development of TPAF

Given the significant effect of TPAF on inpatient and 1-year mortality, a predictive model was built to estimate the individualized risk of developing TPAF. Using a CHAID-based decision tree trained on a 90% sample of a balanced data set (n = 1016), performance on a 10% training holdout (n = 111) was a correct classification rate of 70.3% with a sensitivity of 76.0% and specificity of 65.0%. Application of this model to the overall patient population of 15,148 resulted in a correct classification rate of 65.0% with a sensitivity of 76.0% and specificity of 65.0%. The overall model C statistic was 0.70 (Table II).

Table II. Confusion matrix of observed transient postoperative atrial fibrillation (TPAF) compared with predicted TPAF in balanced subset of 1127 patient encounters from Floridaa.

Sample Actual observed Predicted
No TPAF TPAF % Correct
Training (90%) No TPAF 361 157 69.7
TPAF 146 346 70.3
Test (10%) No TPAF 35 20 63.6
TPAF 14 48 77.4
Overall modelb: Sensitivity, 71%; specificity, 69%; ROC curve, 0.70
Open in a new tab

ROC, Receiver operating characteristic.

a

Results reported for model training (90%) and testing (10% hold-out).

b

Performance of model on predicting TPAF in entire population of 15,158.

External validation was used using this model on 12,919 patients from California in 2011 who were identified using the same inclusion and exclusion criteria. The model correctly classified 70% of patients in this cohort with a sensitivity of 74% and specificity of 70%.

Web-based client for predicting TPAF

The decision rules identified by this model were used to create an interactive tool for point-of-care prediction. With a simple, Web-based interface, clinicians would be able to enter baseline patient characteristics and obtain the predicted probability their patient would have for developing TPAF. A screenshot of the pilot tool is presented in Fig 3, A.

Fig 3.

Fig 3

Open in a new tab

A, Screenshot of the full Web-based application to predict transient postoperative atrial fibrillation (TPAF) in patients presenting for abdominal aortic aneurysm (AAA) repair. B, Screenshot of results screen displayed after the clinician has entered the patient's information into Web-based application.

Clinicians enter the patient's operative approach (OAR/EVAR), CCI score (number), age (number), history of peripheral vascular disease (yes/no), recent weight loss (yes/no), and history of fluid/electrolyte disorder (yes/no), deficiency anemia (yes/no), chronic pulmonary/lung disease (yes/no), or uncomplicated diabetes mellitus (yes/no). The tool then provides a response with the predicted probability of the patient developing TPAF and a summary of the data entered by the clinician (Fig 3, B).

Discussion

This study used a population-based database to demonstrate that TPAF confers increased risk of death after OAR and EVAR of nonruptured AAA. This study is also the first to develop a predictive model for identifying patients at risk for developing TPAF after AAA repair with adequate discriminatory ability using administrative clinical data. With a companion mobile-friendly Web-based interface, clinicians caring for patients before nonruptured AAA repair could use this resource at the bedside.

Although the literature is replete with studies of the effect of AF after cardiothoracic surgery, the data are sparse in the arena of vascular surgery. Only a handful of studies have examined this topic. Feringa et al22 found an overall 5% incidence rate of new-onset AF in a series of 175 infrarenal AAA repairs. OAR was associated with a higher but not significantly different incidence compared with EVAR (6.3% vs 2%).22 Valentine et al23 found an 11% incidence rate of TPAF in a review of 211 thoracic and abdominal aortic repairs for a variety of indications. In that study, no TPAF patients died, and TPAF had no effect on the mean length of stay in the hospital or intensive care unit or on hospital deaths. They concluded that TPAF does not affect the outcomes of aortic operations. Noorani et al11 observed a 10% incidence of AF after 200 elective OARs. They identified cerebrovascular disease and postoperative cardiac failure as independent predictors of AF.11 Similar to the study of Valentine et al,23 AF had no effect on early mortality or late survival.

The findings in the present study differ from earlier studies in several ways. The AF incidence of 3.7% in this study is lower than the incidence in the aforementioned single-center studies. This may be due to smaller sample sizes and heterogeneity of earlier studies that included both thoracic and abdominal aortic repairs and a mixture of indications. That thoracic aortic repair is associated with a fourfold increase in the development of postoperative AF has been well established.24 Limitation to surgery of the abdominal aorta for the indication of aneurysm alone may account for the differences in the incidence rates of TPAF. The present analysis used a population-level database to create a study sample that is not reflective of only a single-center's experience, and the reported incidence may be closer to the true population incidence.

That TPAF was associated with higher inpatient and 1-year mortality is the most salient finding in the present study. This is consistent with the findings of other reports, including several large epidemiologic studies demonstrating AF is an independent risk factor for death in the general population and in postsurgical patients.3,4,9,25 Although patients with TPAF in the present study were those with increased burden of comorbid disease, this body of literature supports the relationship between TPAF and mortality as being more than an epiphenomenon.

Also, multivariable analyses accounting for clinical confounders found TPAF to be independently associated with mortality. Although the inclusion of other perioperative variables, including acute blood loss, volume status, and postoperative ileus, could further reduce bias due to confounding, these elements are not present in HCUP-SID data. Also supporting TPAF as a primary factor for mortality were the similar mortality rates seen in those who developed TPAF after both OAR and EVAR for intact AAA. The incidence of TPAF was significantly lower after EVAR than after OAR, but the mortality rate was the same for both approaches after TPAF developed.

Interestingly, much of the difference in mortality risk during the year after discharge appears to occur within the first 40 to 60 days. Although this study was not designed to characterize the risk over time, others have studied postdischarge trajectories in the context of medical conditions to more precisely define how long the initial hospitalization affects outcomes.26 A similar study examining the rate of daily change in mortality risk over time and the causes of death after discharge in AAA repair is warranted. The specific cause of death could not be determined in the present study due to limitations of the data set, but the relationship between TPAF and mortality, may be at least partly explained by an increased risk of cardiac events associated with TPAF. In this study, patients with TPAF were 5.5-times more likely to have postoperative acute myocardial infarction than those without TPAF.

Ischemic cardiac complications remain a leading cause of morbidity and mortality after vascular surgical procedures. Increasingly, evidence supports ischemia as a mechanism for the development of AF. Animal studies have shown atrial ischemia can cause AF.27 Patients with perioperative cardiac arrhythmias are similarly more likely to have perioperative ischemia and troponin T leakage than those without arrhythmias.1 Further supporting the relationship between cardiac arrhythmia and cardiovascular events, Blackwell et al28 found a more than twofold increase in the incidence of cardiovascular events at 1 year in patients who developed TPAF compared with those who did not after radical cystectomy.

The findings in this study demonstrate that AF is not a benign condition that can be disregarded. Two potential areas for intervention are proposed: (1) preoperative risk stratification and (2) early initiation of preventative measures. Significant differences in baseline characteristics between patients who did and did not develop TPAF were identified in the current study. These data were leveraged to create a predictive tool to help clinicians identify those at highest risk for developing TPAF. The performance of four different modeling techniques (simple logistic regression, multilevel regression, artificial neural networks, tree-based models) was compared before selection of a CHAID-based decision tree. Model performance on a validation cohort, based on the C statistic, ranged from 0.55 to 0.70 depending on the approach. The finalized model performance (C statistic = 0.70) compared favorably with other registry and administrative data-based predictive models, including the Scientific Registry of Transplant Recipients adult survival models (C statistic = 0.72-0.75), The Society of Thoracic Surgeons coronary artery bypass major morbidity or mortality risk-adjusted models (C statistic = 0.73), and the initial version of the American College of Surgeons-National Surgical Quality Improvement Project Pancreatectomy Risk Calculator (C statistic = 0.74).29-31 Nevertheless, the local addition of more granular clinical elements could improve the accuracy of prediction, and this is currently being investigated with patients who underwent AAA repair at our institution.

The integration of this model into an easy-to-use clinical tool allows risk stratification at the point of care and aggressive modification of factors that can be controlled preoperatively, including electrolyte or fluid disorders and anemia, before elective repairs. Increasing screening intensity for high-risk individuals (ie, prolonged telemetry) or targeting prophylactic interventions at only high-risk individuals could also help prevent TPAF in a cost-effective manner. This outcome-specific predictive model offers the advantage of identifying patients at highest risk for TPAF to guide individualized interventions. For patients undergoing AAA repair, these could be based on strategies already in use in cardiac surgery, including prophylactic rate control with β-blockade or amiodarone, anti-inflammatory agents (colchicine), anti-ischemic (ranolazine), aggressive electrolyte replacement (magnesium), and optimization of fluid status.6,32-34 This approach follows how predictive analytics are used in many other nonmedical fields, including marketing and finance. Instead of targeting the entire population, focusing on specific and vulnerable individuals is cost-effective and efficient.35 Although this continues to be studied in clinical settings, early results with advanced modeling techniques applied to clinical trial data find a similar benefit for the use of predictive analytics in medicine.36 Importantly, when TPAF is detected, a thorough evaluation should be under taken to exclude cardiac ischemia as the cause.

This study has several limitations. The HCUP-SID allows the study of preoperative, operative, and postoperative admissions to precisely define TPAF, but these data cannot be used to study specifics regarding treatment interventions, duration of TPAF, and time to resolution cannot be studied. As discussed, the specific cause of death could not be determined.

Other limitations of administrative clinical data are well described by several authors.37,38 In particular, the under-reported incidence of low-revenue diagnoses is worth further discussion. A potential exists for systemic undercoding for diagnostic categories without a large financial incentive for accurate documentation.39 The true incidence of TPAF is possibly higher than observed in this study because some transient events that are poorly documented or unrecognized do not translate into administrative codes. To measure the consistency of the TPAF coding, the definition of TPAF was internally measured using an analytic sample by comparing hospital-to-hospital and year-to-year rates. A total of 144 hospitals contributed patients to the sample, with an interquartile range of TPAF between 1.4% and 5.2% with minimal variance (0.009) suggesting consistency between hospitals. Similarly, the range was 3.4% to 3.9% across years, indicating minimal variation over time.

Finally, the applicability of this model for predicting TPAF requires further validation. The model's performance on a population of patients from a different state was tested, but the ability of the tool to correctly classify patients at any hospital remains unknown. Presently, the Web application is being piloted at our institution for elective AAA repair to measure its accuracy in predicting TPAF before full-scale deployment. In addition, a standardized protocol is being developed in collaboration with cardiologists to preemptively treat high-risk patients, improve their inpatient monitoring, and ensure postdischarge follow-up.

Conclusions

TPAF is associated with an increased risk of death in patients undergoing repair of intact AAA. Predictive analytics were used to identify patient factors that increase the risk of developing TPAF were identified, and a Web-based client was created that can estimate individual patient risk for TPAF. These data support further study on the prevention and treatment of TPAF after major vascular surgery.

Footnotes

Author conflict of interest: none.

Presented in the ePoster session at the Thirty-ninth Annual Meeting of the Midwestern Vascular Surgical Society, Chicago, Ill, September 10-12, 2015.

The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a conflict of interest.

Author Contributions: Conception and design: AK, PH, JC

Analysis and interpretation: AK, PH, RB, DG, JC

Data collection: AK, MD

Writing the article: AK, PH, JC

Critical revision of the article: AK, PH, PK, JC

Final approval of the article: AK, PH, MD, RB, DG, SK, PK, JC

Statistical analysis: AK, PH, RB, DG, SK, JC

Obtained funding: Not applicable

Overall responsibility: JC

References

  • 1.Soliman EZ, Safford MM, Muntner P, Khodneva Y, Dawood FZ, Zakai NA, et al. Atrial fibrillation and the risk of myocardial infarction. JAMA Intern Med. 2014;174:107–14. doi: 10.1001/jamainternmed.2013.11912. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Wyse DG, Waldo AL, DiMarco JP, Domanski MJ, Rosenberg Y, Schron EB, et al. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl J Med. 2002;347:1825–33. doi: 10.1056/NEJMoa021328. [DOI] [PubMed] [Google Scholar]
  • 3.Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham Study. Stroke. 1991;22:983–8. doi: 10.1161/01.str.22.8.983. [DOI] [PubMed] [Google Scholar]
  • 4.Benjamin EJ, Wolf PA, D'Agostino RB, Silbershatz H, Kannel WB, Levy D. Impact of atrial fibrillation on the risk of death: the Framingham Heart Study. Circulation. 1998;98:946–52. doi: 10.1161/01.cir.98.10.946. [DOI] [PubMed] [Google Scholar]
  • 5.Mathew JP, Parks R, Savino JS, Friedman AS, Koch C, Mangano DT, et al. Atrial fibrillation following coronary artery bypass graft surgery: predictors, outcomes, and resource utilization. MultiCenter Study of Perioperative Ischemia Research Group. JAMA. 1996;276:300–6. [PubMed] [Google Scholar]
  • 6.Hogue CW, Jr, Creswell LL, Gutterman DD, Fleisher LA. Epidemiology, mechanisms, and risks: American College of Chest Physicians guidelines for the prevention and management of postoperative atrial fibrillation after cardiac surgery. Chest. 2005;128(2 Suppl):9S–16S. doi: 10.1378/chest.128.2_suppl.9s. [DOI] [PubMed] [Google Scholar]
  • 7.Ascione R, Caputo M, Calori G, Lloyd CT, Underwood MJ, Angelini GD. Predictors of atrial fibrillation after conventional and beating heart coronary surgery: a prospective, randomized study. Circulation. 2000;102:1530–5. doi: 10.1161/01.cir.102.13.1530. [DOI] [PubMed] [Google Scholar]
  • 8.Almassi GH, Schowalter T, Nicolosi AC, Aggarwal A, Moritz TE, Henderson WG, et al. Atrial fibrillation after cardiac surgery: a major morbid event? Ann Surg. 1997;226:501–11. doi: 10.1097/00000658-199710000-00011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Villareal RP, Hariharan R, Liu BC, Kar B, Lee VV, Elayda M, et al. Postoperative atrial fibrillation and mortality after coronary artery bypass surgery. J Am Coll Cardiol. 2004;43:742–8. doi: 10.1016/j.jacc.2003.11.023. [DOI] [PubMed] [Google Scholar]
  • 10.Hogue CW, Jr, Hyder ML. Atrial fibrillation after cardiac operation: risks, mechanisms, and treatment. Ann Thorac Surg. 2000;69:300–6. doi: 10.1016/s0003-4975(99)01267-9. [DOI] [PubMed] [Google Scholar]
  • 11.Noorani A, Walsh SR, Tang TY, Sadat U, Cooper DG, Callaghan CJ, et al. Atrial fibrillation following elective open abdominal aortic aneurysm repair. Int J Surg. 2009;7:24–7. doi: 10.1016/j.ijsu.2008.09.002. [DOI] [PubMed] [Google Scholar]
  • 12.Bessissow A, Khan J, Devereaux PJ, Alvarez-Garcia J, Alonso-Coello P. Postoperative atrial fibrillation in non-cardiac and cardiac surgery: an overview. J Thromb Haemost. 2015;13(Suppl 1):S304–12. doi: 10.1111/jth.12974. [DOI] [PubMed] [Google Scholar]
  • 13.Yap YG, Camm J. Essentials of atrial fibrillation. New York: Springer Healthcare; 2014. Pathogenesis of atrial fibrillation; pp. 7–10. [Google Scholar]
  • 14.Agency for Healthcare Research Quality RM. HCUP Databases. Healthcare Cost and Utiliation Project (HCUP) [Accessed April 1, 2015]; Available at: https://www.hcup-us.ahrq.gov/databases.jsp.
  • 15.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36:8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
  • 16.Stagg V. CHARLSON: stata module to calculate Charlson index of comorbidity. [Accessed March 17, 2016]; Available at: http://EconPapers.repec.org/RePEc:boc:bocode:s456719.
  • 17.Go AS, Hylek EM, Phillips KA, Chang Y, Henault LE, Selby JV, et al. Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the AnTi-coagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study. JAMA. 2001;285:2370–5. doi: 10.1001/jama.285.18.2370. [DOI] [PubMed] [Google Scholar]
  • 18.Wattigney WA, Mensah GA, Croft JB. Increasing trends in hospitalization for atrial fibrillation in the United States, 1985 through 1999: implications for primary prevention. Circulation. 2003;108:711–6. doi: 10.1161/01.CIR.0000083722.42033.0A. [DOI] [PubMed] [Google Scholar]
  • 19.Gialdini G, Nearing K, Bhave PD, Bonuccelli U, Iadecola C, Healey JS, et al. Perioperative atrial fibrillation and the long-term risk of ischemic stroke. JAMA. 2014;312:616–22. doi: 10.1001/jama.2014.9143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Birman-Deych E, Waterman AD, Yan Y, Nilasena DS, Radford MJ, Gage BF. Accuracy of ICD-9-CM codes for identifying cardiovascular and stroke risk factors. Med Care. 2005;43:480–5. doi: 10.1097/01.mlr.0000160417.39497.a9. [DOI] [PubMed] [Google Scholar]
  • 21.Walkey AJ, Wiener RS, Ghobrial JM, Curtis LH, Benjamin EJ. Incident stroke and mortality associated with new-onset atrial fibrillation in patients hospitalized with severe sepsis. JAMA. 2011;306:2248–54. doi: 10.1001/jama.2011.1615. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Feringa HH, Karagiannis S, Vidakovic R, Noordzij PG, Brugts JJ, Schouten O, et al. Comparison of the incidences of cardiac arrhythmias, myocardial ischemia, and cardiac events in patients treated with endovascular versus open surgical repair of abdominal aortic aneurysms. Am J Cardiol. 2007;100:1479–84. doi: 10.1016/j.amjcard.2007.06.043. [DOI] [PubMed] [Google Scholar]
  • 23.Valentine RJ, Rosen SF, Cigarroa JE, Jackson MR, Modrall JG, Clagett GP. The clinical course of new-onset atrial fibrillation after elective aortic operations. J Am Coll Surg. 2001;193:499–504. doi: 10.1016/s1072-7515(01)01028-6. [DOI] [PubMed] [Google Scholar]
  • 24.Karimi A, Walker KL, Martin TD, Hess PJ, Klodell CT, Feezor RJ, et al. Midterm cost and effectiveness of thoracic endovascular aortic repair versus open repair. Ann Thorac Surg. 2012;93:473–9. doi: 10.1016/j.athoracsur.2011.10.016. [DOI] [PubMed] [Google Scholar]
  • 25.Bhave PD, Goldman LE, Vittinghoff E, Maselli J, Auerbach A. Incidence, predictors, and outcomes associated with postoperative atrial fibrillation after major noncardiac surgery. Am Heart J. 2012;164:918–24. doi: 10.1016/j.ahj.2012.09.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Dharmarajan K, Hsieh AF, Kulkarni VT, Lin Z, Ross J, Horwitz LI, et al. Trajectories of risk after hospitalization for heart failure, acute myocardial infarction, or pneumonia: retrospective cohort study. BMJ. 2015;350:h411. doi: 10.1136/bmj.h411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Sinno H, Derakhchan K, Libersan D, Merhi Y, Leung TK, Nattel S. Atrial ischemia promotes atrial fibrillation in dogs. Circulation. 2003;107:1930–6. doi: 10.1161/01.CIR.0000058743.15215.03. [DOI] [PubMed] [Google Scholar]
  • 28.Blackwell RH, Ellimoottil C, Bajic P, Kothari A, Zapf M, Kliethermes S, et al. Postoperative atrial fibrillation predicts long-term cardiovascular events after radical cystectomy. J Urol. 2015;194:944–9. doi: 10.1016/j.juro.2015.03.109. [DOI] [PubMed] [Google Scholar]
  • 29.Shahian DM, O'Brien SM, Filardo G, Ferraris VA, Haan CK, Rich JB, et al. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 1–coronary artery bypass grafting surgery. Ann Thorac Surg. 2009;88(1 Suppl):S2–22. doi: 10.1016/j.athoracsur.2009.05.053. [DOI] [PubMed] [Google Scholar]
  • 30.Parikh P, Shiloach M, Cohen ME, Bilimoria KY, Ko CY, Hall BL, et al. Pancreatectomy risk calculator: an ACS-NSQIP resource. HPB (Oxford) 2010;12:488–97. doi: 10.1111/j.1477-2574.2010.00216.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Scientific Registry of Transplant Recipients. New SRTR risk-adjustment models for adult kidney transplant outcomes. [Accessed May 29, 2015]; Available at: http://srtr.org/publications/content/pres/2015/New_SRTR_Risk-Adjustment_Models_for_Adult_Kidney_Transplant_Outcomes.pdf.
  • 32.Frendl G, Sodickson AC, Chung MK, Waldo AL, Gersh BJ, Tisdale JE, 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:e153–93. doi: 10.1016/j.jtcvs.2014.06.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Tagarakis GI, Aidonidis I, Daskalopoulou SS, Simopoulos V, Liouras V, Daskalopoulos ME, et al. Effect of ranolazine in preventing postoperative atrial fibrillation in patients undergoing coronary revas-cularization surgery. Curr Vasc Pharmacol. 2013;11:988–91. doi: 10.2174/157016111106140128123506. [DOI] [PubMed] [Google Scholar]
  • 34.Imazio M, Brucato A, Ferrazzi P, Pullara A, Adler Y, Barosi A, et al. Colchicine for prevention of postpericardiotomy syndrome and postoperative atrial fibrillation: the COPPS-2 randomized clinical trial. JAMA. 2014;312:1016–23. doi: 10.1001/jama.2014.11026. [DOI] [PubMed] [Google Scholar]
  • 35.Bose R. Advanced analytics: opportunities and challenges. Ind Manage Data Syst. 2009;109:155–72. [Google Scholar]
  • 36.Jaskowski M, Jaroszewicz S. Uplift modeling for clinical trial data. Presented at: International Conference on Machine Learning for Clinical Data Analysis; Edinburgh, Scotland. June 26-July 1, 2012. [Google Scholar]
  • 37.Kothari AN, Zapf MA, Blackwell RH, Markossian T, Chang V, Mi Z, et al. Components of hospital perioperative infrastructure can overcome the weekend effect in urgent general surgery procedures. Ann Surg. 2015;262:683–91. doi: 10.1097/SLA.0000000000001436. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Grimes DA. Epidemiologic research with administrative databases: red herrings, false alarms and pseudo-epidemics. Hum Reprod. 2015;30:1749–52. doi: 10.1093/humrep/dev151. [DOI] [PubMed] [Google Scholar]
  • 39.Haut ER, Pronovost PJ, Schneider EB. Limitations of administrative databases. JAMA. 2012;307:2589–90. doi: 10.1001/jama.2012.6626. [DOI] [PubMed] [Google Scholar]

RESOURCES