Hospitals that participate in the ICU surveillance component of the NNIS system were invited to participate in the second (January 1996 through December 1997) and third (April 1998 through July 1999) phases of Project ICARE; 55 ICUs from 21 hospitals reported the required data to both the second and third phase of Project ICARE. The surveillance methods and definitions of the NNIS system (14,15) and Project ICARE (13) have been previously described. As participants in the NNIS system, ICUs had been previously categorized by the types of patients served: coronary (CCU), medical (MICU), general surgical (SICU), cardiothoracic, combined medical-surgical (<80% of patients can be classified into a single ICU patient type), neurosurgical, respiratory, trauma, burn, or other.

Participating hospitals reported the grams of select antimicrobial agents administered to patients and the antimicrobial susceptibility results of isolates recovered from clinical specimens from hospitalized patients each month. Microbiologic data were aggregated for each ICU separately, all non-ICU inpatient wards combined, and all outpatient areas combined (e.g., units that perform same-day surgery, simple diagnostic procedures or therapy, urgent care, or emergency care). Pharmacy data were reported for the same hospital strata, except for outpatient areas for which pharmacy data were not available. Amounts of antimicrobial drugs reported were standardized by conversion to defined daily doses; for parenteral vancomycin, one daily dose was defined as 2 g. This analysis includes both parenteral and oral (2% of total vancomycin use) vancomycin.

The microbiology laboratory reported antimicrobial susceptibility results for all enterococci and Staphylococcus aureus isolates recovered from all clinical specimens, whether associated with hospital- or community-acquired infection or colonization. Duplicate isolates were excluded: these were defined as isolates of the same organism with the same antimicrobial resistance pattern recovered from the same patient, regardless of the site of isolation (e.g., blood, sputum, urine, or wound), during the same calendar month. Susceptibility reports from isolates obtained as part of infection-control surveillance were excluded. When excluding these surveillance isolates, VRE or methicillin-resistant S. aureus (MRSA) prevalence more closely reflects data routinely aggregated as part of the cumulative susceptibility report (i.e., cumulative antibiogram). The validity of the susceptibility data submitted by participating hospitals for VRE and MRSA has previously been confirmed through a proficiency testing program at these laboratories and by confirmatory testing at the ICARE reference laboratory of up to 20 VRE and 20 MRSA isolates from these hospitals (13,16). Enterococci were considered vancomycin resistant if the MIC of vancomycin was 32 µg/mL or if the zone diameter by disk diffusion was 14 mm. S. aureus were considered oxacillin (methicillin) resistant if the MIC of oxacillin was 4 µg/mL or if the zone diameter by disk diffusion was 10 mm (17).

In October 1997, a report of local monitoring data for each hospital area compared to the national benchmark (i.e., aggregate summary data from all 41 Phase 2 ICARE hospitals, including 113 ICUs) was disseminated to each participating hospital (18). The aggregate benchmark data included numeric presentation of pooled means, medians, and key percentile distributions of the prevalence of selected antimicrobial-resistant organisms, stratified by ICU areas combined, non-ICU–inpatient areas combined, and outpatient areas combined (18). In addition, the data for antimicrobial agent use were stratified by the specific type of ICU (e.g., general-surgical separate from cardiothoracic ICU, non-ICU–inpatient areas combined, and outpatient areas combined) (18). Each individual hospital’s report included raw data and pooled means of the same target rates for each hospital area. Stratification of use and resistance prevalence by different hospital areas, as described, provided a valid benchmark by which hospitals were able to determine how their usage and resistance prevalence compared with the aggregate, when the data were adjusted for different patient populations in these different hospital areas.

To ascertain how the hospital infection-control staff used the feedback report, they were surveyed in September 1999. Information was collected on the use of the feedback report, recognition of problem pathogens or excessive use of specific antimicrobial agents, and specific practice changes. Questions were open-ended to include any change in infection-control practice (including hygienic practices, barrier precautions, and antimicrobial control practices) from baseline practice (i.e., during pre-intervention period), rather than a description of specific practices already in use. The infection-control practitioner overseeing the surveillance activities responded to the survey, with input from the infection-control committee, based on recollection or meeting minutes.

For this analysis, monthly data from each ICU were pooled for the entire study period and for each period of the study (i.e., pre-intervention and postintervention) by each ICU (data from non-ICU–inpatient areas and outpatient areas are not shown because of low statistical power). Pooled rates were calculated for prevalence of VRE (percentage), MRSA (percentage), and vancomycin use (defined daily doses/1,000 patient-days). For example, the pooled mean rate of vancomycin use was calculated for each ICU by dividing the total number of defined daily doses by the total number of patient-days reported over the study period by that ICU, multiplied by 1,000, and thus expressed as defined daily doses per 1,000 patient-days. If <10 S. aureus or enterococci isolates were tested for antimicrobial susceptibility from a specific ICU during the study period, that ICU was excluded from further analysis.

Data were analyzed by SAS Release 6.12 Software (SAS Institute Inc., Cary, NC). To assess the change in ICU-specific prevalence of MRSA, VRE, and vancomycin use, the pre-intervention rate was subtracted from the postintervention rate (i.e., difference in rates). Differences in the percent VRE and vancomycin use rate were evaluated by the paired t-test and further compared by type of practice change by a paired t-test. Frequency of MRSA in a hospital has been shown to be independently associated with rates of vancomycin use (19). We used linear regression modeling to determine which types of practice changes were independently associated with changes in vancomycin use in ICUs, after the data were adjusted for each ICU and changes in MRSA prevalence by the GLM procedure (SAS Institute Inc.). Detection of potential influential data points and their influence on main effect factors were also assessed in the modeling process. All reported p values are two-tailed. Analyses were repeated by using the relative change of each parameter rather than difference in rates.

During the study period, 21 hospitals representing 55 ICUs followed the surveillance protocol and reported at least 6 months of data by the time of the intervention and a median of 32 months (range 18–45) of data during the study period. Twenty (95%) hospitals completed the intervention survey representing the 50 ICUs included in this analysis. These hospitals were from 18 states and had a median hospital bed size of 351 (range 147–1,022); 13 (65%) reported an affiliation with a teaching institution (i.e., major teaching centers), and 2 (10%) were Veterans Affairs Medical Centers. The ICUs included 14 combined medical-surgical ICUs, 7 cardiac care units, 7 MICUs, 8 general SICUs, 6 cardiothoracic ICUs, 4 neurosurgical ICUs, 3 pediatric ICUs, and 1 burn ICU.

Infection-control teams at all hospitals disseminated the benchmark data to a variety of hospital committees and personnel (e.g., pharmacy and therapeutics, ICU personnel). This feedback was usually in the form of a committee report or memo. In addition to reporting, 12 (60%) hospitals changed prescribing of vancomycin (i.e., prescriber practice change). Eight (40%) hospitals reported at least one prescriber practice change (many hospitals combined several hospitalwide change in prescriber practice) that was applied hospitalwide; these changes encompassed 22 ICUs (Table). The hospitalwide changes included evaluating periodic drug use (19 ICUs), redistributing guidelines on appropriate uses of vancomycin (20 ICUs) by newsletter or mail (9 ICUs), and requiring prior approval for use of vancomycin (3 ICUs). In contrast, four hospitals reported focused (i.e., ICU-specific) practice changes in 11 ICUs; these included ICU-specific education in-service sessions on appropriate vancomycin use (8 ICUs) and removing vancomycin as routine prophylaxis for cardiac surgery (2 ICUs). Both practice changes were reported in one ICU (Table).

In the 50 study ICUs, the rates of vancomycin use during the pre-intervention period (Figure 1, plotted circles) were similar in range to the 113 ICARE Phase 2 ICUs contributing data to the national aggregate benchmark report (Figure 1, box plots). The overall (pooled mean) ICU-specific use of vancomycin in the 50 ICUs at the 20 study hospitals after the intervention was 89.1 defined daily doses per 1,000 patient-days, a 2.8% increase over the pre-intervention rate of use (86.6 defined daily doses per 1,000 patient-days). Despite this increase in aggregate usage among all ICUs, most ICUs reported lower rates of vancomycin use after the intervention compared with the pre-intervention rates. The median difference was -3 defined daily doses per 1,000 patient days (range -138 to +196), but this difference was not statistically significant.

Differences in the rate of vancomycin use varied substantially by the type of practice change. ICUs, in which unit-specific programs were implemented, used significantly lower rates of vancomycin in the postintervention period than in the pre-intervention period (Table), including both ICU-specific educational in-service (mean difference of -35.8 vs. +7.6, defined daily doses per 1,000 patient-days, p=0.01) and removal of vancomycin as routine surgical prophylaxis for cardiac surgery (mean difference -66.9 vs. +4.2 defined daily doses per 1,000 patient-days, p=0.01). In the multivariable analysis in which data were adjusted for each ICU and changes in MRSA prevalence, ICUs in which unit-specific practices were identified for improvement used, on average, 49 fewer daily doses of vancomycin per 1,000 patient days than did the other ICUs (parameter estimate -48.5; 95% confidence limits -68.8, -28.22; p=0.0001), compared to pre-intervention levels. The ICUs reported a 35%–37% decrease in median vancomycin use (median 132 daily doses of vancomycin per 1,000 patient days for unit-specific education and 149 for removal of prophylaxis) (Table). Analyses were repeated by using the relative change of each parameter rather than difference in rates, with similar results of statistical significance.

Thirty-five (70%) of the 50 study ICUs tested at least 10 isolates of enterococci for vancomycin susceptibility and were included in the calculations of VRE prevalence during both pre- and postintervention periods. During the pre-intervention period, these ICUs reported a median VRE prevalence of 11.7%. Overall, VRE prevalence increased during the postintervention period compared with the pre-intervention period among all study ICUs (median difference +2.3%; range –41% to +32%), although this difference was not statistically significant. However, when compared by type of practice change, the difference in VRE prevalence was significantly lower in ICUs in which unit-specific practice changes occurred, compared with other ICUs (mean difference -7.5% vs. +5.7%, p<0.001). Although many of the ICUs with decreases in vancomycin use reported increases in percent VRE, all the ICUs noting a unit-specific practice change reported decreases in both percent VRE and vancomycin use (Figure 2). Analysis of these data by using either the relative change in percent VRE or vancomycin use obtained results of similar statistical significance. However, since the relative changes were commonly of extreme values (range 0–400%), these are not reported here.

Because vancomycin use is associated with prevalence of MRSA in ICUs (19), we also evaluated temporal trends in MRSA prevalence. During the pre-intervention period, these ICUs reported a median MRSA prevalence of 33.5%. Overall, prevalence of MRSA increased during the postintervention period compared with the pre-intervention period in all study ICUs (mean difference +5.5%; range –22% to +38%; p=0.02). The increase in MRSA prevalence was similar in ICUs reporting unit-specific practice change compared with other ICUs (mean difference +2.7% vs. +7.1%, p=0.39).