Key points are not available for this paper at this time.
OBJECTIVE: To explore epidemiological patterns of the incidence of Clostridium difficile infection (CDI) and hospital characteristics associated with increased incidence during nonepidemic and epidemic years. DESIGN: Retrospective and prospective ecological study. SETTING: Eighty-three acute care hospitals participating in CDI surveillance in the province of Quebec, Canada. METHODS: A Serfling-type regression model applied to data obtained from an administrative database (1998-2006) and prospective Quebec CDI surveillance (2004-2006) was used to calculate expected CDI baseline incidence and to detect incidence exceeding the defined epidemic threshold at the provincial and hospital level. Multivariable Poisson regression was used to determine hospital characteristics associated with increased incidence during nonepidemic (1998-2001) and epidemic (2003-2005) periods. RESULTS: During the study period (1998-2006), 4,525,847 discharges, including 45,508 with a CDI in any diagnosis field, were reported by 83 hospitals. During 1998-2001, the average Quebec incidence of CDI was 10,304 cases in 1,775,822 discharges (5.8 cases per 1,000 discharges) and presented a pattern of seasonality, with similar patterns at the hospital level for some hospitals. The Quebec epidemic started in October-November 2002 and peaked in March 2004 at 845 cases in 40,852 discharges (20.7 cases per 1,000 discharges). In multivariable analysis, higher incidence was associated with location in Montreal and surrounding regions, greater hospital size, larger proportion of hospitalized elderly patients, longer length of stay, and greater proportion of comorbidities in patients, whereas teaching profile was associated with decreased incidence during both nonepidemic and epidemic periods. The effect of geographical location on incidence was greater during the epidemic. CONCLUSION: Baseline incidence from nonepidemic years and hospital characteristics associated with CDI incidence should be taken into account when estimating the efficacy of interventions.
Gilca et al. (2010) studied this question.