All hospitals that had purchased the fluoroquinolones investigated in this study and that had admitted patients with community-acquired pneumonia were included in the study. All statistical analyses were restricted to hospitals with at least 10 acute care beds (to reduce statistical bias). Provincial drug purchasing data were aggregated using the WHO ATC/DDD classification system (2003 edition) for the fiscal years 1997/1998 to 2002/2003. Raw data were coded and computed using a standard spreadsheet program (Office 97 Excel, Microsoft, Redmond, Washington). Drugs were classified and volume data were transformed to ATC/DDD values. Values for utilization of fluoroquinolones were expressed as number of DDDs per 1000 (acute care) bed-days per year and number of
DDDs per 100 cases of community-acquired pneumonia with admission to hospital per year.
Use of fluoroquinolones was determined for each hospital, and the results were also aggregated for each of the 9 DHAs in the province. Hospital size was categorized according to the number of acute care beds: small = 10—15 beds, medium = 16-100 beds, and large = more than 100 beds. Each hospital- size category contained approximately the same number of hospital-year observations. Annual periods were related to fiscal years (i.e., from April 1 to March 31). vardenafil 20 mg
Survey data from the director of pharmacy in each DHA were used to determine whether or not a given hospital had antimicrobial policies and care pathways for fluoroquinolones. It was assumed that policies applied across each district, but respondents were asked to complete multiple surveys if one or more hospitals within their respective districts had policies that differed from the district policies. These data were incorporated into a dichotomous variable indicating whether or not a hospital had a policy. Hospitals were also coded according to the presence of more than one utilization policy or no utilization policies. Because the survey was administered during the last year of the study period, data for only the 2002/2003 year were used to examine the impact of hospital pharmacy policies (number and type) on fluoroquinolone use. Student t tests were performed to determine statistical differences in fluoroquinolone use between hospitals with and without antimicrobial policies and therapeutic pathways. Trends in the use of specific fluoroquinolones and percentage of ciprofloxacin use relative to respiratory fluoroquinolone use over time were examined in a post hoc analysis.
The results were aggregated and combined, using a standard statistical software package (STATA, version 7, STATA Corporation, College Station, Texas), to produce simple descriptive statistics of fluoroquinolone use. Data are reported as means and 95% confidence intervals. The relationship among the 3 predictor variables (DHA, year, and hospital size) and DDD values was assessed. A regression approach that accounted for clustered data was employed because hospital- year data were clustered within years. Regressions were estimated using a generalized estimating equation and employing a log correlation structure between study years. Because of the limited sample size, only unadjusted associations were examined, and the statistical level of significance was a = 0.05. canadian antibiotics
The institution variable was nested within the DHA variable, and a standard univariable regression analysis was therefore not appropriate (because of violation of independence of observations). Thus, a multilevel linear model was constructed to account for the natural hierarchical structure of the data. The use of a multilevel linear model is often suggested when data are nested naturally, as in this study.