Discussion

Analysis of 30-day all-cause readmissions after the LAAC procedure revealed several important findings. First of all, 8.2% of patients who received the LAAC device were readmitted within 30 days. Independent predictors of readmission included non-routine disposition such as discharge to Home Health Care (HHC) or a SNF, chronic pulmonary disease, renal disease, and anemia.
Both cardiac and non-cardiac causes accounted for readmission, with atrial arrhythmias and congestive heart failure being the most common. Charges varied significantly between readmitted and non-readmitted patients on index admissions. The readmission rates of LAAC procedures seem to be lower when compared with those of other cardiac procedures and diseases, including readmissions for congestive heart failure, TAVR, or PCI15–17. This may reflect the fact that procedural success has increased since FDA approval, while complication rates have decreased8. Interestingly, neither hospital procedural volume nor peri-procedural complications correlated significantly with readmission rate, save for peri-procedural acute renal failure and venous thromboembolism, though this may be a function of the low number of reported complication rates (Table 3).
Since 30-day readmissions are viewed as a quality performance measure, this study is an important one in identifying patients at risk for readmission and directing resources towards them. Our results suggest that patients with kidney disease, COPD, CHF, ischemic heart disease, and anemia are at elevated risk for readmission within 30 days. Interestingly, though the readmission rate for LAAC devices was lower than in TAVR, the predictors for readmission are strikingly similar, likely reflecting the fact that patients with greater disease burden in general are at elevated risk of readmission16. Our multivariable model had modest discriminatory ability in predicting these readmission (ROC 0.70), which is on par or better than many rehospitalization prediction models18.
These predictors of readmission reflect the fact that patients with increased comorbidities are at higher risk of readmission. Patients who are discharged to HHC or SNF often require greater care, have a greater number of comorbidities, incur higher inpatient charges, and are generally frailer than patients who are discharged home—a common pattern among post-surgical, acute care, trauma, stroke, post-cardiac surgery, and heart failure patients19–23. Medicare data examining characteristics predisposing SNF residents to readmission found that ~21% of residents had been readmitted within 30 days, with a significant percentage of residents presenting with an admission diagnosis of cardiac conditions 24,25. Therefore, it is not surprising that these patients are more likely to be readmitted after receiving a LAAC device.
Anemia is an especially common comorbidity encountered in LAAC patients, as the primary criterion for LAAC selection is the desire to avoid indefinite anticoagulation, generally due to
bleeding risk26. Current guidelines recommend that patients who undergo WATCHMAN continue on anticoagulation for at least 45 days5. Speculatively, it is possible that patients who are anemic during their index hospitalization may be anemic because of an increased predisposition towards bleeding and would have greater difficulties tolerating post-procedural anticoagulation, leading to increased readmission. Further study regarding the possible utility of post-procedural hemoglobin monitoring or alternative post-procedural anticoagulation strategies is warranted.
Major causes of readmissions included arrhythmia (atrial fibrillation and flutter), heart failure, gastrointestinal disease, and sepsis. Readmissions for cerebral infarction made up only 1.63% of readmissions (n=3). One of the major causes of rehospitalization is congestive heart failure, and numerous studies are underway to investigate strategies to aid in reducing overall readmissions for heart failure27,28. Interestingly, however, neither congestive heart failure nor cardiomyopathy during the index admission were predictive of readmissions on multivariate analysis. This discrepancy may reflect hemodynamic changes caused by LAA occlusion itself. In animal studies, occlusion of the left atrial appendage results in increases in left atrial filling pressures, which results in diastolic dysfunction29. One single center study reported increases in mitral E/E’ ratios three months after left atrial occlusion on echocardiography, indicative of increases in left- sided filling pressures.30 Further study is needed in this area.
Interestingly, gastrointestinal disease was not an independent risk factor for readmission but was a top cause of readmission. The subcategories of the general ICD-10 code for gastrointestinal disease include hematemesis, melena, hemorrhage, mucositis, and unspecified. Therefore, the gastrointestinal disease code likely captured many patients with bleeding complications after LAAC. While gastrointestinal disease was not an independent risk factor for readmission, anemia was an independent risk factor. Patients who undergo LAAC usually have contraindications to anticoagulation, and therefore the required 45 days of anticoagulation afterwards may predispose to bleeding complications, particularly gastrointestinal bleeding.
Further studies examining anticoagulation strategies after WATCHMAN are needed, given the significant portion of readmissions post-WATCHMAN implantation that are related to gastrointestinal hemorrhage.
Atrial fibrillation and atrial flutter as causes of readmission may reflect the arrhythmic burden in patients who receive the procedure, as all patients who receive the device have atrial fibrillation. When compared with 30-day readmission rates in patients who are admitted for atrial fibrillation, readmissions for atrial fibrillation after LAAC appear lower. A 2017 study of 388,340 patients admitted for atrial fibrillation and subsequent 30-day readmissions found that atrial fibrillation was the most common readmission diagnosis and accounted for 27.1% of readmissions, followed by heart failure, accounting for 11.4%. In contrast, in the current study, only 8.23% of readmissions after LAAC were due to atrial fibrillation, and 8.23% of readmissions were secondary to CHF 31.
The average length of stay during index admission for readmitted patients was nearly twice the length of stay for non-readmitted patients (2.8 days versus 1.5 days). This difference may explain the discrepancy in accrued charges observed between both groups ($139,869 versus $115,258).
Unfortunately, given the constraints of the NRD, it was not possible to elucidate the reason for the prolonged stay during the index readmission. To our knowledge, there is no standardized guideline for length of stay for LAAC procedures. This is considerably less than the average length of stay for TAVR, percutaneous ASD closure, and transcatheter mitral valve repair, which are all reported to be greater than 5 days based on analysis of national databases32–35.
Given the relative lack of medical guidelines for defining an appropriate length of stay post LAAC, further investigation may be warranted. Standardization of post-LAAC length of stay will result in significant charge savings, given that the average charge of the readmitted hospital stay was nearly 50% of the cost of the index admission ($62,577). Similar efforts are underway for elective percutaneous coronary intervention (PCI), as evidenced by a recent study highlighting roughly $5,000 worth of savings from employing a same-day discharge strategy, rather than one involving overnight stay, without a corresponding increase in adverse events36.
The role of 30-day readmissions as a quality performance measure underscores the importance of this analysis as a vital first step in elucidating risk factors for readmission after the LAAC procedure. Recent data suggest that quality of life improves after the WATCHMAN procedure, when compared with quality of life of patients reliant on lifelong warfarin37. Atrial fibrillation diagnoses are expected to double by 2050, and it is likely that left atrial appendage occlusion will continue to gain favor as a method of stroke prevention, due to decreasing rates of periprocedural complications, overall safety, and patient preference1,38. The results of this study will be useful in identifying patients at high-risk of 30-day readmission after LAAC and developing strategies to lower readmission rates.
Readmission scoring systems are of particular interest since CMS implemented the Hospital Readmissions Reduction Program (HRRP), and robust models are scarce. Mortality models fare reasonably well, but readmissions models are less discriminatory with an average c statistic of 0.6339. The CMS-endorsed Readmission Risk Score (RRS), a composite of 37 variables, has a c statistic of 0.6 for 30 day all-cause readmissions for CHF40. Similarly, the HOSPITAL score, designed to predict avoidable all-cause 30 day readmission, has a c statistic of 0.7141. Countless additional scores exist, and there is considerable overlap between variables used in mortality scores with high predictive ability and readmissions scores that do not fare as well.
Our readmission scoring system had modest discriminatory ability with a C-statistic of 0.67 in derivation and 0.63 in validation. Our scoring system is on par with most readmission models and may be more clinically useful given its simplicity. We suggest designating patients with a score of 1 or less as “low” risk, 2 to 3 as “moderate” risk, and 4 and above as “high” risk.This may aid clinicians in risk stratifying patients at greater risk for being readmitted post-LAAC

Limitations

There are few limitations to this study. This analysis is based solely on diagnostic and administrative variables, and there is a noted absence of procedural or detailed clinical data. Details on LAAC device size, type, echocardiographic variables, such as LAA dimensions and
the degree of para-device leak, were not available. Additionally, the cause of readmission was elucidated through discharge diagnosis codes, an approach that is utilized extensively in 30-day readmissions analyses. Furthermore, the NRD does not contain data on patients who are readmitted to a hospital in another state, which may result in an underestimation of the true readmission rate. For the scoring system, ICD9 and ICD10 codes are not 1:1, so potential data points may have been either not counted for or incorrectly included. Efforts were made to minimize this. Finally, the NRD does not track mortality, and so the influence of mortality on the rates of readmission cannot be accounted for.