The main limitation of this study is the lack of patient-level data since NIS is an administrative database. NIS data regarding bipolar disorder and BPD diagnoses are limited to inpatient hospitalization only and do not include data from outpatient settings. Re-hospitalizations, which add to the total inpatient burden, are not accounted for in our study, given the nature of the data. Selection bias is of concern, as this is a retrospective study. In addition, due to the lack of ICD-9-CM codes for psychotropic medications, we could not evaluate other treatment measures if given during inpatient management. However, despite these limitations, NIS is still an excellent population-based representation of disease associations with systematic and temporal factors. However, the main strength of this study is the national representation provided by the dataset, as well as a uniform collection of data obtained over five years through ICD-9-CM codes. Another strength of our study is its large sample size of 510,611 and the reliability of the data, given that the information is coded independently of the individual practitioner and therefore it is subjected to
