Prevalence and Risk Factors for Carriage of Methicillin-Resistant Staphylococcus aureus at Admission to the Intensive Care Unit:  Results of a Multicenter Study FREE

MULTIPLE-DRUG–resistant bacteria, particularly methicillin-resistant Staphylococcus aureus (MRSA), have become prevalent in many countries.^1- 4 In French hospitals, the proportion of S aureus isolates that are resistant to methicillin is 30% to 40% overall^1,4 and 78% in intensive care unit (ICU) patients with staphylococcal nosocomial infections.² In this nearly endemic setting, hospital-to-hospital and ward-to-ward transfer of MRSA carriers contributes substantially to the spread of MRSA.^4- 8 In a recent large French multicenter study,⁴ 43% of cases of MRSA infection or colonization were considered imported, and 53% of these imported cases were in patients transferred from another hospital.

Intensive care units are particularly likely to receive patients infected or colonized with MRSA. Several studies^9- 11 involving MRSA screening at ICU admission found that 8% to 10% of patients had positive findings. These patients can form a reservoir for subsequent dissemination in the ICU unless the organism is identified by routine screening at admission and early isolation precautions are taken.¹⁰ However, this strategy remains controversial because it is costly and its efficacy has not been convincingly established.¹²

A prospective observational multicenter study was conducted in patients at ICU admission to (1) determine the prevalence of MRSA carriage at admission, (2) define the best method of screening specimens, (3) identify factors associated with MRSA carriage, and (4) attempt to define a strategy for selective MRSA screening and preventive isolation at ICU admission.

The study was conducted from July 1, 1997, to December 31, 1997, in 14 ICUs in 11 French hospitals. Six ICUs were medical ICUs, 4 were surgical ICUs, and 4 were medical and surgical ICUs. Thirteen ICUs were in the Paris area. Nine were in teaching hospitals.

All patients were screened within 24 hours after ICU admission. At least 2 screening specimens were obtained, 1 from the nares and 1 from intact skin. For the intact skin specimen, a single swab was used to sample 4 different sites (the axilla and groin on both sides). Breaks in the skin were sampled also. Premoistened swabs were used to collect nasal and skin samples. In addition, clinical specimens were obtained as requested by the physician in charge of the patient.

Swabs were plated on various media, according to standard practice in each hospital: Chapman agar alone (4 ICUs) or with a disk of 5-µg oxacillin sodium (1 ICU); Mueller-Hinton agar supplemented with tobramycin sulfate, 10 mg/L (6 ICUs), or with a disk of 5-µg oxacillin (1 ICU); or blood agar with disks of 5-µg pefloxacin mesylate and 10-µg tobramycin sulfate (2 ICUs). Methicillin-resistant S aureus isolates were subsequently identified according to standard French recommendations.¹³

The following variables were recorded on a standard form within 24 hours after ICU admission: baseline characteristics (age and sex), presence of immunosuppression, previous or current hospital stays (including length of stay and type of units), history of surgery or antimicrobial therapy, date of hospital admission, date of ICU admission, severity at ICU admission as evaluated by the Simplified Acute Physiology Score II,¹⁴ chronic health evaluation score,¹⁵ underlying disease severity as evaluated by the McCabe score,¹⁶ presence at ICU admission of breaks in the skin, and history of invasive procedures.

An MRSA case was defined as the presence of an MRSA isolate in any of the screening specimens collected within 24 hours after ICU admission or in a clinical specimen collected within 48 hours after ICU admission.

Data were recorded prospectively on a standardized form. Microbial findings were routinely cross-checked by reviewing computerized databases in each microbiology laboratory. After double data entry with verification of discordant data, a validation procedure was carried out and aberrant data were corrected.

The management of missing data was given careful attention. In some patients, retrospectively collected data (eg, previous hospital admissions and history of antimicrobial treatment) could not be obtained during the first 24 hours after ICU admission; in this situation, these potential risk factors were considered not present if they were not recorded on the standardized data collection form. Data collected prospectively at ICU admission (eg, severity scores, invasive procedures, and presence of open skin lesions) were considered missing if they were not recorded on the standardized data collection form.

Variables associated with MRSA carriage at ICU admission were only analyzed for first admissions (vs repeat admissions) with screening and without previously known MRSA carriage. Categorical variables were compared using the χ² test or Fisher exact test, as appropriate, and continuous variables were compared using the t test. Categories were defined, and relative risks with their 95% confidence intervals were calculated by comparison to reference categories. Variables significant at P = .10 in the univariate analysis were entered into a forward stepwise logistic regression model. First-class interaction variables were also included in the model. All tests were 2-tailed. P<.05 was considered significant.

Variables found to be significant in the logistic regression model were used to develop a predictive score for MRSA carriage at ICU admission. Our goal was to develop a predictive score that could be easily determined by clinicians at ICU admission. Because all variables yielded comparable adjusted odds ratios in the multivariate model, independent variables were given the same weight. According to the number of patients without risk factors or with 1 or more risk factors for MRSA carriage at ICU admission, the percentage of patients to be screened was determined, and the sensitivity and positive predictive value of MRSA screening in each group were calculated.

A cost-benefit analysis was performed for the first ICU admissions only; repeat admissions of a patient to the same ICU were not included. The objective was to estimate the monetary costs and savings associated with use of the predictive score (see previous paragraph) to select patients for screening. We did not include medical outcomes, and we performed a cost-benefit analysis. Costs were computed from the viewpoint of the hospital, excluding nonmedical costs. We defined the reference strategy as routine screening and preventive isolation of all patients initially admitted to the ICU. Strategies based on selective screening and preventive isolation were compared with this strategy. French francs were converted to US dollars using the Organization of Economic Cooperation and Development, Paris, purchasing power parity index.

The following costs were included in the analysis: costs of tests, costs of patient isolation during the interval before laboratory results were available, and costs of false-negative test results, ie, of not diagnosing MRSA carriers who could disseminate the organism.

Cost estimates were derived from data in the ledger of one large teaching hospital participating in the study. The cost of isolation was determined by a time-motion study in the same hospital's ICU. Additional costs related to MRSA infection had been calculated in this ICU in an earlier study.¹⁷ The cost of a nasal or skin swab was estimated at $30 and that of a skin-break swab at $36. The additional cost of caregiver time (nurses, physicians, and nursing assistants) and supplies needed for the swabs was estimated at $9 per patient per day, and that of contact isolation (pending results of admission screening) at $48 for 2 days. The additional cost of an MRSA infection per ICU patient, including all hospital costs and overheads, was $9275.¹⁷

The hypotheses used in the cost-benefit analysis were as follows: (1) A nonisolated infected ICU patient may transmit the organism to 3 ICU patients, according to an earlier study¹⁸; however, this number undoubtedly varies with the compliance with standard precautions in the ICUs. (2) An ICU patient who has acquired an organism has a 0.2 to 0.4 probability of developing an infection with that organism. The probability used in our baseline analysis was 0.4.^9,19 The savings related to preventive isolation precautions were obtained by multiplying the additional cost of an MRSA infection by the prevalence of MRSA carriage, the number of patients who may acquire the organism, and the probability that infection will develop in each of these patients.

A sensitivity analysis was performed in the universal screening and preventive contact isolation strategy on the following variables: carriage rate at ICU admission from 2% to 20%, risk of transmission from an unisolated patient to 1 or 2 patients, 0.2% risk of becoming infected after transmission, additional cost of an infection in the ICU set at $4000 and $6000, and additional cost of caregiver time and supplies for isolation precautions ($57 per patient per day according to an earlier study²⁰).

During the 6-month study, 2475 admissions (2399 patients) were recorded in the 14 ICUs: 330 (13.3%) to the surgical ICUs, 1266 (51.2%) to the medical ICUs, and 879 (35.5%) to the medical and surgical ICUs. For 128 admissions (5.2%), screening for MRSA was not performed (Figure 1). Of the 2347 admissions with MRSA screening, 162 (6.9%; 95% confidence interval, 5.9%-8.0%) had positive test results for MRSA. This prevalence of MRSA carriage varied across ICUs from 3.7% to 20.0%: it was 10.3% in the surgical ICUs, 7.0% in the medical and surgical ICUs, and 6.1% in the medical ICUs. In 61 (37.7%) of the 162 admissions with positive screening test results, MRSA carriage had been detected before ICU admission. Therefore, in 101 admissions (62.3%), MRSA carriage was newly identified by screening at admission. As shown in Table 1, carriage was detected by nasal swabs in 77.8% of the 162 MRSA-positive admissions and by nasal or skin swabs in 92.0%. Clinical specimens detected only 18.5% of MRSA-positive admissions (including 13 [12.9%] of the 101 admissions with newly detected MRSA carriage). Combining nasal swabs and clinical specimens detected 87.0% of MRSA-positive admissions. Finally, 54.3% (88/162) of MRSA-positive admissions would have been missed had screening not been performed (ie, there was no previous detection of MRSA carriage and clinical specimens were negative for MRSA).

Of the 2347 admissions with MRSA screening, 2275 were first admissions to the ICU and 72 were repeat admissions. The prevalences of MRSA carriage were 6.4% and 22.2% in these 2 groups, respectively.

Variables associated with MRSA carriage at ICU admission were assessed in the 2189 patients initially admitted (vs repeat admissions) with screening performed but without previously known MRSA carriage. Methicillin-resistant S aureus was found in 96 (4.4%) of these 2189 patients.

We defined transferred patients as patients who had been in a hospital for longer than 48 hours at the time of ICU admission or who had been discharged within 48 hours before ICU admission. The prevalence of MRSA carriage was significantly different between the transferred patients (53/746 or 7.1%) and the directly admitted patients (43/1443 or 3.0%) (relative risk, 2.38; 95% confidence interval, 1.61-3.53; P<.001). Consequently, we considered that inclusion of this variable was critical in a predictive score for the risk of MRSA carriage at ICU admission. Furthermore, transferred status interacted with many other study variables. Separate analyses were therefore conducted in transferred patients and in directly admitted patients.

In the transferred patients, variables associated with MRSA carriage were older age, transfer from a rehabilitation or long-term care unit, prolonged hospital stay before ICU admission, and presence of open skin lesions or stomas at ICU admission (Table 2). Logistic regression modeling identified age older than 60 years and a previous hospital stay of 21 days or longer as independently associated with MRSA carriage at ICU admission (Table 3).

In the directly admitted patients, variables associated with MRSA carriage were older age, antimicrobial therapy within the last year, history of hospitalization or surgery, poor chronic health status, rapidly or ultimately fatal disease, severe illness at ICU admission, presence of open skin lesions, or presence of a central venous catheter at ICU admission (Table 4). In the logistic regression model, age older than 60 years, history of stay in an ICU during the last 5 years, history of surgery during the last 5 years, and presence of open skin lesions were independently associated with MRSA carriage at ICU admission (Table 3).

Two hundred twenty (29.5%) of the 746 transferred patients and 7 (13.2%) of the 53 MRSA carriers among transferred patients had none of the identified risk factors for MRSA carriage. Therefore, limiting screening at admission to patients with at least 1 risk factor would have included 70.4% of transferred patients and identified 86.8% of MRSA carriers (Table 5). Limiting screening at admission to patients with at least 2 risk factors would have included 15.7% of transferred patients and identified 43.4% of MRSA carriers.

In the 1443 directly admitted patients, limiting screening at admission to patients with at least 1 risk factor (902 [62.6%] patients) would have identified 38 (88.4%) of the 43 MRSA carriers. Limiting screening at admission to the patients with at least 2 risk factors (277 [19.2%] patients) would have identified 24 (55.8%) of the 43 MRSA carriers.

In the transferred and directly admitted patients combined, screening all transferred patients (n = 746) and directly admitted patients with at least 1 risk factor (n = 902) would have included 75.5% of patients and detected 91 (94.8%) of the 96 MRSA carriers. Screening all transferred patients (n = 746) and the 277 directly admitted patients with at least 2 risk factors would have included 46.8% of patients and detected 77 (80.2%) of the 96 MRSA carriers.

The economic analysis was conducted in the 2189 patients initially admitted to the ICU without previously known MRSA carriage. The prevalence of MRSA carriage in this population was 4.4%. Screening all patients for MRSA saved money: the cost of routine MRSA screening and preventive isolation was lower than the cost of treating the MRSA infections prevented by this strategy. Universal screening and preventive isolation was more beneficial than any other strategy in our baseline analysis (Figure 2). The second best strategy was limiting screening and preventive isolation to transferred patients or to directly admitted patients with at least 1 risk factor. By contrast, limiting this strategy to transferred patients, or to transferred and directly admitted patients with at least 2 risk factors, resulted in higher costs.

In the sensitivity analysis, universal screening and preventive isolation saved money when the prevalence of MRSA carriage varied from 2% to 20% at ICU admission (Figure 2). The universal screening and isolation strategy resulted in financial savings for prevalence of MRSA carriage at ICU admission lower than 5% for the following situations: risk of transmission from an unisolated patient lowered to 1 patient, risk of MRSA infection in MRSA carriers lowered to 20%, cost of preventive isolation increased at $57 per day, and cost of an MRSA infection lowered to $4000 (Table 6).

Our results confirm the high prevalence of MRSA carriage at ICU admission in France. The 6.9% prevalence found in this study is consistent with earlier studies,^9- 11 in which prevalences ranged from 8% to 10% at ICU admission. We found substantial variations, however, across ICUs, from 3.7% to 20.0%.

Whether MRSA screening at ICU admission is worthwhile remains a matter of debate. Most recommendations for controlling the spread of MRSA include screening at hospital admission^21- 23 in all patients or in selected patients according to the epidemiological conditions. Other recommendations¹² do not include MRSA screening at admission, arguing that this strategy is costly and has not been proven in controlled studies to curb the incidence of MRSA. Recent publications,^10,24- 25 however, suggest that MRSA screening may be crucial to the successful control of MRSA outbreaks. In our study, 37.6% of cases of MRSA carriage at ICU admission were known previously, and an additional 8.0% of cases were associated with positive clinical samples during the first 48 hours in the ICU. Therefore, screening tests at admission elicited the only positive test results for MRSA in 54.3% (88/162) of MRSA carriers. This proportion is higher than the 38% found in a recent study by Girou et al,¹⁰ in which half of the imported cases had positive clinical samples (vs 18.5% in our study). This difference can be ascribed to restriction of screening to patients considered at high risk of MRSA carriage (ie, patients transferred from other wards with a known high prevalence of MRSA), patients with a prolonged stay in another ward before ICU admission, or patients with a history of hospitalization in high-risk areas.¹⁰ Patients with these risk factors for MRSA carriage may be more likely to have MRSA-positive clinical specimens at ICU admission.

Nasal swabs at ICU admission were positive in 77.8% of admissions with MRSA carriage. Similarly, earlier studies^10,24 found yields of about 80%. Swabs of intact skin and any open skin lesions were also collected in our study. Nasal and skin swabs in combination detected 92.0% of admissions with MRSA carriage, and nasal and clinical samples in combination detected 87.0%. Our data did not allow differentiation of the results of intact skin and open skin lesion samples. They suggest, however, that combining nasal samples, clinical specimens, and routine samples of open skin lesions ensures detection of most MRSA carriers.

The prevalence of MRSA was different between first admissions (6.4%) and repeat admissions (22.2%). In addition, a history of a previous ICU stay was associated with MRSA carriage at ICU admission. This adds to the evidence that ICUs contribute substantially to the spread of MRSA. Consequently, in areas with a high MRSA prevalence, patients with a history of ICU management should be routinely screened for MRSA at ICU admission.

As in other recent publications,^7,26- 27 we found that the presence of open skin lesions was significantly associated with MRSA carriage in directly admitted patients and was nonsignificantly associated with MRSA carriage in transferred patients (P = .06, multivariate analysis). This result is not surprising, because chronic open skin lesions, particularly pressure ulcers, are probably a marker for a high level of medical care and, consequently, for frequent opportunities for MRSA acquisition. In addition, there is a well established association between presence of skin lesions and persistent MRSA carriage in hospitalized patients.^27- 29

As expected, a hospital stay longer than 3 weeks in the transferred patients and a previous ICU or surgical ward stay in the directly admitted patients were risk factors for MRSA carriage at ICU admission. This is consistent with previous reports^{3- 4,30- 31} of high prevalences of MRSA in ICUs and surgical wards. Our finding that a previous stay in a long- or intermediate-term facility was not significantly associated with MRSA carriage is probably ascribable to the small number of patients with such a history (2.2% of the overall population). In the univariate analysis, however, this variable was strongly associated with MRSA carriage. Consequently, we believe it should be included among risk factors for MRSA carriage.

In 7 (16.3%) of the 43 MRSA-positive admissions in patients coming from their homes, we found no history of hospitalization within the last 5 years. An explanation may be that it is difficult to obtain a full medical history from severely ill patients or their relatives during the first 24 hours after ICU admission, as required in the study protocol. It is therefore likely that several MRSA carriers actually had a history of previous hospitalization that was missed by the local investigator. In addition, MRSA can be community-acquired, as shown in recent studies.^32- 33 True community-acquired cases³⁴ should be distinguished from community-onset cases among health care providers.²⁶ Community-acquired MRSA could also account for some of the MRSA-positive patients with no history of hospitalization.

Age older than 60 years was strongly associated with MRSA carriage, independent of other risk factors, such as a history of hospitalization or presence of skin lesions. It is not clear why older age was associated with MRSA carriage. Older age may be a marker for other risk factors for MRSA acquisition or persistence not collected in our study.

Two studies^7,26 have evaluated factors associated with MRSA carriage at hospital admission. One⁷ compared MRSA and methicillin-susceptible S aureus carriers in 5 hospitals. Factors associated with MRSA carriage were pressure ulcers, use of a feeding tube, organ transplantation, hospitalization during the last year, and employment in a health care facility. Methicillin-resistant S aureus carriers in this study, however, were identified by clinical samples only; screening samples were not obtained at admission. In addition, transferred patients were not included. In the other study,²⁶ which was too small for a multivariate analysis, presence of chronic open skin lesions increased the risk of MRSA carriage. Our results may be more reliable because our study included many ICUs and involved routine screening at ICU admission.

However, our data were obtained at admission to an ICU. All participating ICUs were in large tertiary hospitals, and most were in teaching hospitals. In addition, MRSA was nearly endemic in most of the participating hospitals. Therefore, it is unclear whether our results can be extrapolated to other wards or to hospitals with lower prevalences for MRSA.

One goal of our study was to develop a score predicting a need for screening and preventive isolation precautions. Our objective was to develop a pragmatic score that could be used by physicians within hours after ICU admission. For that reason, only data collected within 24 hours after ICU admission were included. Furthermore, variables were dichotomized and were given the same weight in the predictive score. Only routine screening detected MRSA carriage with acceptable sensitivity. Therefore, our study indicates that patients admitted to ICUs in areas with a high prevalence of MRSA should be screened routinely, as recommended in French²¹ and British²² guidelines.

For our cost-benefit analysis, we used costs previously estimated in one of the ICUs that participated in the present study.¹⁷ These costs were similar to those obtained by other investigators.^20,35 Our cost-benefit analysis suggested that screening and preventive isolation was most beneficial when applied routinely or to a large proportion of the patients. This finding reflects the low cost of screening and preventive isolation. Furthermore, sensitivity analyses indicated that universal screening and preventive isolation was a beneficial strategy when the prevalence of MRSA at ICU admission was higher than 1%. Other sensitivity analyses suggested that universal screening and preventive isolation was beneficial in most situations.

The following conclusions can be drawn from our study: (1) Screening for MRSA at ICU admission is probably useful, as it detects more than half of the MRSA carriers. (2) Screening at ICU admission should include nasal swabs and sampling of skin lesions, if any, in addition to collection of clinical specimens. (3) Factors associated with MRSA carriage at ICU admission do not allow the development of a sensitivity score for predicting MRSA carriage; consequently, screening at admission should be performed in most or all ICU patients. (4) Routine screening at ICU admission may result in financial savings in areas with a high prevalence of MRSA. Further studies are needed to determine whether these findings apply to wards other than ICUs and to improve our ability to predict MRSA carriage at hospital admission.

Corresponding author and reprints: Jean-Christophe Lucet, MD, MPH, Unité d'Hygiène et de Lutte contre l'Infection Nosocomiale, Hôpital Bichat–Claude Bernard, 75877 Paris CEDEX 18, France (e-mail: jean-christophe.lucet@bch.ap-hop-paris.fr).

Accepted for publication May 16, 2002.

This study was supported by grant PHRC 1995, AOM 95 210 from the French Ministry of Health, Paris, France.

This study was presented as an abstract (1711) at the 39th Interscience Conference on Antimicrobial Agents and Chemotherapy, San Francisco, Calif, September 26-29, 1999.

We thank the microbiology departments of the participating hospitals for performing the microbial studies, the intensive care unit personnel who collected the screening samples, and Isabelle Lolom, MT, and Pascale Régnier, MT, for their technical assistance.

Study group members: Hôpital d'Orléans: N. Bercault, MD. Hôpital de la Pitié Salpetrière, Paris: L. Bodin, MD. Hôpital Henri Mondor, Créteil: C. Brun-Buisson, MD. Hôpital Raymond Poincaré, Garches: B. Clair, MD. Hôpital Bichat–Claude Bernard: J. L. Dumoulin, MD, M. Thuong, MD, J. L. Trouillet, MD. Hôpital Saint Antoine, Paris: B. Gentil, MD, E. Maury, MD. Hôpital Saint Louis, Paris: G. Leleu, MD, O. Marie, MD, J. R. Zahar, MD. Hôpital de Poissy: J. Merrer, MD. Hôpital Victor Dupouy, Argenteuil: J. P. Sollet, MD. Hôpital Saint Joseph, Paris: J. F. Timsit, MD.