Resource Utilization of Adults Admitted to a Large Urban Hospital With Community-Acquired Pneumonia Caused by Streptococcus pneumoniae: Statistics
Economic analyses were conducted from the hospital perspective. The primary analysis separated the cohort into two groups based on penicillin susceptibility (susceptible vs nonsusceptible). For all comparisons, the nonsusceptible group included both intermediate-resistant and resistant S pneumoniae. Continuous data were compared using a Student t test for normally distributed data or Mann-Whitney U test for nonnormally distributed data (eg, costs and length of stay). x2 or Fisher exact test were used to compare proportions between the two groups. For cost comparison between susceptible, intermediate, and resistant groups, data were analyzed by the Kruskal-Wallis rank-sum test.
Multivariate linear regression was utilized to control for confounding variables and determine covariates that predicted total costs and length of stay for the entire cohort. For all regression analyses, total costs and length of stay were log-transformed so that parametric tests could be utilized. In both regression models, all variables (patient demographics, comorbidities, S pneumoniae susceptibility, ICU stay, admission year, presence of bacteremia, PSI score, mortality, and receipt of oral antibiotic therapy) were inserted into the model at once to control for covariance. An additional variable, termed unexplained delayed discharge, was added into the models to improve predictability. Knowing that at our hospital the ultimate decision regarding patient discharge is made by the attending physician, this variable was designed to capture prolonged observation of the patient. A patient was defined as having a delayed discharge if they remained in the hospital for > 48 h after they clinically met criteria for discharge. Criteria for discharge included stable normalization of temperature and WBC, and the receipt of oral antibiotics or the ability to receive oral antibiotic therapy, which ever came later. All patients who died in the hospital were automatically excluded from being defined as a delayed discharge.
A p value < 0.05 was considered significant during all statistical analysis. All statistical analysis was conducted using statistical software (SPSS/PC +; SPSS; Chicago, IL).