Summarization of the results was done using descriptive statistics and cross-tabulation. Continuous variables were explained by using mean and standard deviation (SD). In order to denote categorical data and continuous data, we used a two-sided Pearson’s Chi-squared test and a two-tailed Independent Sample t-test, respectively. A multinomial logistic regression model was used to measure the likelihood of associations using odds ratios (OR), p-value and 95% Confidence Interval (95% CI) in terms of LOS, total charges, severity of illness, comorbidities, suicidal ideations, and utilization of ECT. Discharge weights within the NIS database [19] were applied in all regression models to obtain a nationally representative inpatient sample. Age, race, gender and median household income of the patient were adjusted. A p-value <0.001 was chosen to be the reference value to measure statistical significance. SPSS version 23 (IBM Corp., Armonk, NY, USA) was used to do all the statistical analysis in this study.
