JSTOR

Methods

This study took place in the ICU of Alfred Hospital (Melbourne, Australia) from December 16, 2002, through September 30, 2003. The Alfred Hospital is a 350‐bed tertiary referral hospital with a 28‐bed combined medical‐surgical ICU that treats large numbers of trauma patients and patients who have undergone cardiothoracic surgery, including heart‐lung transplantation. Only standard precautions were used for infection‐control management of patients with MRSA. There was no routine use of gowns, gloves, or masks, and affected patients were not isolated, although most beds in the ICU were in single‐bed cubicles. Patients were screened for MRSA at ICU admission, on Mondays and Thursdays, and at ICU discharge through swabbing of nose, throat, groin, and axilla (these swab specimens constituted a “swab set”). Swab specimens were processed using selective mannitol‐salt agar with 5 mg/L methicillin (MS5) plates, as described elsewhere.⁸ S. aureus was identified by latex agglutination or coagulase tests. Methicillin resistance was confirmed with disk susceptibility testing.

Only patients who had 2 sets of swabs taken, the first of which was negative for MRSA, were included in this study. A patient was considered to have acquired MRSA if culture of the second or subsequent swab set was positive for MRSA after culture of the initial swab set yielded negative results. For patients who did not acquire MRSA, censoring took place at the time that culture of the last swab set was determined to have negative results.

We analyzed the hazard of MRSA acquisition in relation to the length of time spent in the ICU, with day 0 as the time of admission to the ICU. The hazard is the daily risk of acquisition of MRSA as a proportion of patients at risk. It is calculated by using the number of patients who acquired MRSA on a given day of the ICU stay and dividing this by the number of people in the observational study who were in the ICU at least until this time and who remained free of MRSA colonization. The data were collected on discrete days, so this calculation can be expressed as where is the hazard on day k, is the number of cases of acquisition on day k, and is the number of people in the study who remained in the ICU and remained uncolonized until at least day k—that is, patients for whom at least 2 swab sets were obtained during the study, all of which were culture‐negative for MRSA before day k. In this study, we assumed that the day of acquisition was the day on which the first swab specimen with positive culture results was obtained. Sensitivity analyses were also performed with the assumption that the day of acquisition was the same as the day the last swab specimen with negative culture results was obtained, as well as with the assumption that the day of acquisition was half way between the days on which the last negative and first positive cultures were obtained. Patients were also stratified according to age, sex, admission category (trauma vs. nontrauma), and length of ICU stay.

In this study, we used an Epanechnikov kernel for hazard smoothing, with a width of 5 days, using Stata, version 9.0 (StataCorp).

Results

During the project period, 1,414 patients were admitted to the ICU, of whom 987 (69.8%) had at least 2 swab sets obtained. The rate of compliance with admission screening was 84%, and MRSA prevalence among those who had swab specimens obtained at admission was 6.9% (75 patients). There were 904 patients who had swab specimens obtained on at least 2 occasions and had culture results of the first set that were negative for MRSA. Of these 904, there were 116 (12.8%) who had positive results for the second swab set—that is, 12.8% of patients acquired MRSA. The mean length of stay was 7.2 days (range, less than 1 to 160 days). The length of stay was less than 1 day for 7.7% of patients; 62.8% stayed 1–7 days, 14.3% stayed 8–14 days, 7.2% stayed 15–21 days, and 8% stayed more than 21 days. Basic demographic data are shown in the Table.

Table.  Demographic Characteristics of Study Patients (n = 904)

Open New Window

Figure 1 shows the daily hazard of MRSA acquisition, with the assumption that the day of acquisition was the day that the first swab specimen with positive culture results was obtained. The hazard began below 1% and increased linearly to more than 2% risk per day by day 12. This was followed by a definite leveling off of risk, with a suggestion of decrease. Because very few people stayed for more than 3 weeks, the confidence intervals for hazard become broad at this point.

(15 KB)

Figure 1.  Variation in the hazard of acquisition of methicillin‐resistant Staphylococcus aureus (MRSA) colonization relative to time since admission to the intensive care unit (ICU). Black line, Smoothed hazard function. Gray lines, 95% Confidence interval.

Open New Window

We stratified the patients according to age, sex, medical unit, and length of stay. Figure 2 shows that, whereas the profiles of hazard differ over time, the increase in hazard in the early days of the ICU stay was demonstrable across all groups.

(24 KB)

Figure 2.  Hazard of acquisition from methicillin‐resistant Staphylococcus aureus (MRSA) colonization stratified by age, sex, medical unit, and length of stay in the intensive care unit. Black line, Smoothed hazard function. Gray lines, 95% Confidence interval.

Open New Window

Discussion

This analysis shows that the daily hazard of acquiring MRSA is not constant. The hazard increased for the first part of the patient’s ICU stay and then leveled off, with a possible decrease. This means that the common association of MRSA acquisition with increased length of stay is not accounted for merely by increased exposure; rather, there is an actual increased risk with time. This change in hazard could be explained in several ways, and this study investigated these possibilities further. First, the risk profile of remaining patients may have changed as time since admission changed. In other words, each individual’s risk profile remained constant with time but the type of person who remained in the ICU changed as low‐risk patients were discharged. The alternative explanation is that each individual’s risk changed as the time spent in the ICU changed.

In this study, we stratified patients by sex, age, type of unit (trauma and other), and length of stay, and we found that, in all strata, the hazard of MRSA colonization increased as time since ICU admission increased. This suggests that the observed increase in MRSA acquisition hazard was not due merely to a changing risk profile of those who remained in the ICU; rather, each individual had an increased risk over time. We postulate that this could be due to exposures that increase risk of MRSA colonization—for example, antibiotics or invasive procedures. Alternatively, there could be a nonlinear relationship between MRSA exposure and risk of acquisition, whereby small exposures rarely lead to colonization risk, whereas greater exposures confer much higher risk.

Another possibility for the increase in risk of detected new MRSA colonization over the early part of the ICU stay is that there may be a lag between acquisition of MRSA and ability to detect MRSA by culture of swab samples. This may be a result of the time needed for MRSA to multiply to levels at which it is detectable by current culture techniques. We do not fully understand the dynamics of MRSA behavior in a patient, and it could be speculated that MRSA acquisition may occur in one part of the body, such as from contaminated hands during skin care, but may require a period of time to spread from one body part to the part where a swab sample is obtained. The requirement that 2 sets of swab specimens be cultured to establish that new acquisition has occurred could also contribute to the findings of change in hazard, because there is an inevitable lag between the true time of acquisition and the time of detection. However, if this were the case, the analysis that assumed that acquisition occurred on the day that the last culture‐negative swab sample was obtained should have reflected this, and the resulting daily hazard in the analysis with that assumption also increased for the first week of the ICU stay.

Ismail and Pettitt,⁶ using a nonparametric Bayesian smoothing technique, also showed an increasing hazard for MRSA acquisition up to day 13, then an approximate leveling of hazard. These findings are consistent with those of the current study, with one main difference. The hazard on the first day of admission was 0 in the study of Ismail and Pettitt,⁶ and it increased over the next few days. In the current study, the hazard at day 1 was approximately 0.01, less than one‐half that of the peak hazard seen after day 12. This may reflect a different patient mix between the 2 studies or difference in the frequency of swab sampling, because in the current study not infrequently patients underwent swab sampling twice on day 1, which allowed for earlier detection of MRSA acquisition.

Our findings of an early increase in hazard of MRSA acquisition are consistent with another study that examined the percentage of patients in the ICU who acquired MRSA each week.⁷ The later falling‐off acquisition risk was not demonstrated in that study; however, the swab sampling regimen and analytic methods differed between the 2 studies. In that study, patients had swab samples obtained weekly (from the nose and groin), and only the percentage of patients who acquired MRSA per week was presented, rather than a hazard being determined. That study indicated that the risk of acquiring MRSA in the first week was 7.5% and was 20.3% per week thereafter.

These results have implications for studies that implicitly assume that the risk of MRSA acquisition is constant, such as risk‐factor analysis through linear regression. That type of analysis runs the risk of confounding with other factors that are associated with increased length of stay, such as antibiotic and device use. A more appropriate choice of analytical technique would include nonparametric or semiparametric statistical models, such as a Cox proportional hazards regression model, which allow examination of risk factors in the context of time to acquisition.

The present study took place in an ICU in which patient standard precautions were used for MRSA‐colonized and MRSA‐infected patients. Neither isolation in a single room nor cohorting with contact precautions was used. This may limit the ability to generalize results to other ICUs in which different infection‐control precautions are used. In this study, we adjusted for or stratified by age, duration of stay, sex, and patient's medical unit. We did not consider Acute Physiology and Chronic Health Evaluation or any direct measure of comorbidity. Therefore, we cannot exclude the possibility that the observed increasing hazard of MRSA acquisition was due to earlier discharge of healthier patients. Additionally, the sensitivity of swab‐specimen culture is not 100%, which could influence results, particularly if this sensitivity is associated with the length of stay.

It would be interesting to determine whether more‐sensitive molecular detection methods would have an impact on the shape of the acquisition curves (Figures 1 and 2) by allowing lower concentrations of MRSA to be detected, possibly earlier than with use of current detection techniques. However, the sensitivity analysis that assumes acquisition on the day that the last swab with negative culture results was obtained may represent this situation.