We observed an incidence density of ICU-acquired UTI of 9.6 per 1000 ICU days that is comparable to that observed in other studies that evaluated nosocomial UTIs in ICUs [2
]. However, an important strength of this study is that all patients admitted to adult ICUs (both academic-based and community-based) in a large region were included. As a result, this study should be representative of many critically ill populations at large and the results more widely generalizable. Previous studies have been limited to single specialized medical, surgical, or combined medical–surgical ICUs [9
] or in series of selected ICUs participating in surveillance systems [1
]. Our previous study, which included all multidisciplinary ICUs, was limited in part because we failed to include many cardiovascular surgical patients [2
]. As demonstrated by our observation of a significant difference in risk of acquiring ICU-acquired UTI between cardiac surgical, non-cardiac surgical, and medical patients, care must be paid to patient 'case-mix' in comparing between studies of these infections. It is noteworthy that we did not exclude non-residents of the CHR in this study despite the fact that we have previously argued for such a practice [18
]. Given that we did not observe any significant rate differences among CHR residents and non-residents (data not shown), that the population at risk was restricted to those admitted to CHR ICU (and not the entire base population of the CHR), and that the mortality outcome for those with a ICU-acquired UTI was not related to residency status, we pooled our entire patient cohort for analysis.
There are several possible explanations for our important observation of a lower rate of ICU-acquired UTIs in the latter 2 years of the study. The first possibility is that heightened awareness from our first report [2
] or concomitant preventive efforts (such as a large regional quality improvement initiative to reduce ventilator-associated pneumonia) with increased attention to the use of medical devices and attention to hand washing among staff could have had a role. Anecdotally, we feel it is unlikely to be related to a decreased use of urinary catheters because we estimate that nearly all (more than 90%) of our patients ill enough to require ICU admission for 2 or more days have an indwelling urinary catheter. A second possibility for the reduced rate in the latter years of the study is that there might have been increased use of systemic antimicrobials active against urinary pathogens. This is only speculative because we do not have actual data to support this possibility. A third consideration is that physicians less frequently ordered urine cultures in the second and third years of the study such that the overall culture positivity rate was less. Unlike in our first study, in which we collected data on negative cultures [2
], we did not have access to such results in the present study and are therefore unable to assess this. The fourth possibility, and the one that we suspect might be the most important reason for the reduced rate of ICU-acquired UTIs being diagnosed in the latter part of the study, is that we changed our laboratory testing practice in June 2001. At that point a bacteriuria screening assay was implemented regionally after demonstrating its utility in outpatients [16
]. Since that time only urine samples either positive by that assay or by specific physician request are cultured. On the basis of the sensitivity of the assay of 86%, it is expected that a 10–15% reduction in the rate of culture positivity would occur with its implementation. However, a more important influence is that this assay does not detect yeast because it is based on the specific detection of bacterial ATP [16
species are among the most important causes of ICU-acquired UTI and a reduced rate of ICU-acquired UTI is expected if these organisms are not routinely cultured. We are currently planning a study to evaluate the optimal laboratory means of identifying ICU-acquired UTIs in our region.
The most clinically important and novel finding of this study was that ICU-acquired UTIs do not independently increase the risk for death among patients admitted to ICUs. Unlike in all previous studies potentially able to investigate this question, the present study was adequately powered to detect a clinically significant increased mortality risk [2
]. Although we did observe that these infections increased the crude mortality risk, once confounding for measures of severity of disease, diagnostic category, and length of ICU stay were controlled for, ICU-acquired UTI was not significantly associated with death. This contrasts with the findings of Platt and colleagues, which showed that nosocomial UTIs were associated with a significant attributable mortality in a general hospital population [5
]. It is clinically important to ascertain whether ICU-acquired UTIs are associated with attributable mortality because there may be implications for treatment. Although practice variation among different intensivists and between ICUs in our region probably exists, we commonly withhold antimicrobial therapy for bacteriuria or funguria in the absence of an associated clinical infective syndrome. Although a randomized, prospective, clinical trial is required to address optimal practice, our current observations of a low rate of bacteremic/fungemic ICU-acquired UTI and lack of attributable mortality suggests that a clinical judgment-based approach to treatment may be reasonable.