Limitations:
Our study has the following key limitations: (1) The NIS is an administrative claims-based database that utilized ICD codes which may be prone to errors and could introduce information bias due to outcome misclassification. The hard clinical end points, however, are less subjected to error. Additionally, AHRQ quality control measures are routinely instituted that guarantee data integrity (12). Additionally, the ICD-9 code utilized in this study was not specific to the Watchman device and could be referred for any LAA occlusion procedure. Due to the limited magnitude of other research studies of endocardial devices and any epicardial LAA occlusion procedures performed in US during the study period (13), we believe that application of this code for the purpose of our study was able to mostly characterize Watchman implants. (2) The NIS only captures inpatient admissions and does not provide any information on outpatient encounters. This limitation may result in selection bias; however, our data is well representative of national utilization of Watchman devices performed during in-patient settings; in fact since inpatient hospitalization is often required for reimbursement for the procedure, our results may be more indicative of widespread practice (19). (3) The NIS censors data gathering at discharge so long-term outcomes could not be ascertained from the present dataset. (4) Specific data on potential confounders including medications, as well as operator and intra-procedural characteristics could not be examined from the NIS.