The authors have declared that no competing interests exist.
Reviewed and revised the manuscript critically for important intellectual content and approved of the final version: KG RM JS MY KC EL.. Conceived and designed the experiments: KG RM JS KC EL. Analyzed the data: MY KC EL. Wrote the paper: KG.
Current address: Generation Health, Inc., Boston, Massachusetts, United States of America
Patterns of methicillin-resistant
We conducted a retrospective cohort study including all patients admitted to 5 acute care VA hospitals between 2008–2010 who had nasal MRSA PCR testing within 48 hours of admission and repeat testing within 30 days. The PCR results were used to define a baseline nasal carriage pattern of never, intermittently, or always colonized at 30 days from admission. Follow-up was up to two years and included acute, long-term, and outpatient care visits. Among 18,038 patients, 91.1%, 4.4%, and 4.6% were never, intermittently, or always colonized at the 30-day baseline. Compared to non-colonized patients, those who were persistently colonized had an increased risk of death (HR 2.58; 95% CI 2.18;3.05) and MRSA infection (HR 10.89; 95% CI 8.6;13.7). Being in the non-colonized group at 30 days had a predictive value of 87% for being non-colonized at 1 year. Conversion to MRSA colonized at 6 months occurred in 11.8% of initially non-colonized patients. Age >70 years, long-term care, antibiotic exposure, and diabetes identified >95% of converters.
The vast majority of patients are not nasally colonized with MRSA at 30 days from acute hospital admission. Conversion from non-carriage is infrequent and can be risk-stratified. A positive carriage pattern is strongly associated with infection and death. Active surveillance programs in the year following carriage pattern designation could be tailored to focus on non-colonized patients who are at high risk for conversion, reducing universal screening burden.
Nasal colonization with
The Department of Veterans Affairs (VA) has been performing active surveillance for nasal MRSA carriage since late 2007. Unique features of this program include multiple assessments during a single hospitalization (admission, unit transfer, and discharge) and a high patient retention rate, as the majority of veterans return to the VA integrated health-care system for care. A recent report demonstrated reductions in MRSA infections on a national level in association with this program, but did not report individual risk-adjusted patient-level outcomes
VA Boston IRB approval was obtained prior to data extraction and analysis. As per IRB approval, written informed consent was waived as this was a retrospective database only study utilizing the existing electronic health record and the research would not be practicable without the waiver.
We conducted a retrospective cohort study of patients admitted to any of the 5 acute care hospitals in the New England Veterans Integrated Service Network (VISN 1) between January 2008 and December 2010. VISN 1 consists of 1,084 inpatient beds and cares for approximately 240,000 unique patients annually in the states of Maine, New Hampshire, Vermont, Rhode Island, Massachusetts, and Connecticut. Data from the acute care hospitals, 3 long term care facilities, and all associated VA outpatient community clinics in VISN 1 were captured.
All patients who had an acute care admission, a nasal PCR screening test for MRSA performed within 48 hours of the admission (index admission), and at least one more nasal screen performed within the next 30 days were eligible for study inclusion. Index admissions were censored at 6/30/2010 to allow a minimum of 6 months of follow-up time. Patients were followed for a maximum of 2 years, or until the date of death, a gap of 18 months without VA health care use, or the study end date (12/31/2010).
The GeneXpert MRSA PCR assay (Cepheid, Sunnyvale, CA) is used to test for nasal MRSA carriage at each hospital. Nasal carriage patterns were designated as never (no nasal swabs positive), always (>80% of nasal swabs positive), or intermittent (>1 but less than 80% of swabs positive) and were determined on the basis of all available nasal PCR test results within the first 30 day period of the index admission. The carriage patterns were determined during two different time periods. The first used all screen results from the first 30-days. In order to account for the changes in carriage pattern over the study period, a second carriage pattern designation was determined using all screen results from the entire follow-up window (dynamic carriage pattern). Patients were removed from the never colonized group and excluded from the study if they had a positive clinical culture for MRSA in the previous year. Outcomes were assessed for both the 30-day carriage pattern as well as the dynamic carriage pattern. We performed a validation of the 30-day PCR pattern rule by evaluating concordance with carriage patterns in patients who had 5 or more swabs within a 1 year period (including the swabs performed in the first 30 days) because a rule for determining
The electronic health record was used to capture antibiotic and health-care exposures, co-morbidities, and outcomes. Clinical cultures for MRSA and MSSA were extracted from the VISN 1 data warehouse. Any positive culture from blood or bone was coded as infection. Positive cultures from wounds, skin, sputum, urine or other non-sterile sites were coded as infection if there was an active antimicrobial agent administered within 5 days of the culture result. Data from all inpatient and outpatient encounters were included.
Baseline characteristics are described by frequencies and proportions stratified by the 30-day carriage pattern designation. Chi-squared tests were used to evaluate clinical parameters among patients in each of the three carriage pattern groups. A multi-variable logistic regression was used to assess the effects of multiple potential risk factors on the three categories of colonization status, as well as to determine what risk factors predicted conversion to MRSA colonized status during the six month period following initial screening.
Cox proportional hazards models were used to assess time to each of the three outcomes (death, first MRSA infection, and first MSSA infection) among the three carriage groups. The models were run twice for each outcome to allow a comparison of the 30-day carriage pattern and the dynamic carriage pattern over the study period. Cumulative incidence curves were generated to plot the resulting survival for each outcome. SAS statistical software was used for all programming with p-values of < 0.05 as statistically significant.
A total of 31,662 unique patients had 66,804 hospitalizations to at least one of the 5 acute care hospitals over the 3-year study period. A nasal MRSA PCR result was available within 2 days of the index admission in 24,101 (76%) patients, and two or more MRSA surveillance results were available over a 30 day period in 18,100 patients (
Derivation of the study cohort.
Baseline Characteristic | All | Carriage Pattern During First 30-Days | ||
N = 18,038n (%) | Nevern = 16,425 (91.1)n (%) | Intermittentn = 793 (4.4)n (%) | Alwaysn = 820 (4.6)n (%) | |
|
||||
Male Sex | 17,259 (95.7) | 15,690 (95.5) | 771 (97.2) | 798 (97.3) |
Age, Years, 71+ | 7,211 (40.0) | 6,397 (39.0) | 374 (47.2) | 440 (53.7) |
Race: White | 12,466 (69.1) | 11,273 (68.6) | 561 (70.7) | 632 (77.1) |
Black | 1,078 (6.0) | 984 (6.0) | 49 (6.2) | 45 (5.5) |
Other | 120 (0.67) | 108 (0.66) | <11 (0.76) | <11 (0.73) |
Missing | 4,374 (24.3) | 4,060 (24.7) | 177 (22.3) | 137 (16.7) |
|
||||
Acute Care Admission | 3,348 (18.6) | 2,820 (17.2) | 208 (26.2) | 320 (39.0) |
Long Term Care Admission | 570 (3.2) | 410 (2.5) | 53 (6.7) | 107 (13.1) |
Outpatient Procedure | 8,949 (49.6) | 8,116 (49.4) | 400 (50.4) | 433 (52.8) |
Inpatient Procedure | 309 (1.7) | 250 (1.5) | 17 (2.1) | 42 (5.1) |
Dialysis | 83 (0.46) | 71 (0.43) | <11 (0.25) | <11 (1.2) |
|
||||
MRSA Clinical Culture | 104 (0.58) | 0 (0.0) | 28 (3.5) | 76 (9.3) |
MSSA Clinical Culture | 224 (1.2) | 190 (1.2) | 16 (2.0) | 18 (2.2) |
|
||||
Malignancy | 3,276 (18.2) | 2,940 (17.9) | 170 (21.4) | 166 (20.2) |
Chronic lung disease | 437 (2.4) | 379 (2.3) | 24 (3.0) | 34 (4.2) |
Diabetes mellitus | 5,019 (27.8) | 4,465 (27.2) | 252 (31.8) | 302 (36.8) |
Renal disease | 1,058 (5.9) | 878 (5.4) | 70 (8.8) | 110 (13.4) |
Liver disease | 454 (2.5) | 395 (2.4) | 25 (3.2) | 34 (4.2) |
Skin disease | 902 (5.0) | 806 (4.9) | 36 (4.5) | 60 (7.3) |
Decubitus Ulcer | 137 (0.76) | 72 (0.44) | 24 (3.0) | 41 (5.0) |
|
||||
Beta-lactams | 2,498 (13.9) | 2,144 (13.1) | 135 (17.0) | 219 (26.7) |
Fluoroquinolone | 1,899 (10.5) | 1,623 (9.9) | 119 (15.0) | 157 (19.2) |
Topical (mupirocin) | 135 (0.75) | 95 (0.58) | 12 (1.5) | 28 (3.4) |
MRSA Active Antibiotic | 1,177 (6.5) | 982 (6.0) | 60 (7.6) | 135 (16.5) |
Any Antibiotic | 5,311 (29.4) | 4,678 (28.5) | 260 (32.8) | 373 (45.5) |
The proportions of patients categorized as never, intermittently, or always colonized at the 30-day baseline were 91.1%, 4.4%, and 4.6%, respectively (
In multivariate models, independent variables associated with MRSA nasal carriage (intermittent or always) compared to non-carriage at the 30-day baseline included advancing age, acute or long-term care admission in the past year, diabetes mellitus, renal disease (independent of dialysis), decubitus ulcer, and any antibiotic use in the past 6 months (
|
Odds Ratio | 95% CI | P Value |
|
0.61 | (0.54; 0.68) | <.001 |
|
1.77 | (1.56; 2.01) | <.001 |
|
2.82 | (2.28;3.49) | <.001 |
|
1.22 | (1.09;1.37) | <.001 |
|
1.47 | (1.22;1.77) | <.001 |
|
5.33 | (3.71;7.66) | <.001 |
|
1.35 | (1.21;1.51) | <.001 |
There were 4,972 patients with a carriage pattern designated at 30 days who had 5 or more nasal screens for MRSA performed during the one year period from the index admission. Among the 4,415 patients with an initial non-colonized carriage pattern, 3,857 (87.4%) remained negative for nasal MRSA, 553 (12.5%) were intermittently MRSA nasal positive, and 5 (0.1%) developed persistent MRSA colonization. Thus, the 30-day non-colonized carriage pattern had a specificity of 50% (558/1115) compared to a 1-year pattern, but because of a high prevalence of the non-colonized state, the positive predictive value was 87% (3857/4415) for being non-colonized at one year. Movement from an intermittent carriage pattern to an always pattern occurred in 39/304 (12.8%) patients and from always (more than 80% of screens positive) to intermittent (less than 80% screens positive) in 80/253 (31.6%) of patients.
Among 4597 patients who were non-colonized at 30 days and had at least one re-screening performed in the following 6 months, 509 (11%) had a positive nasal PCR for MRSA. One to 25 screening tests were performed per person during the 6 months, with no difference in the mean number of tests between carriage patterns. An additional 33 (0.8%) patients had a conversion based on a positive culture for MRSA during the same time period. During the subsequent 6 months, an additional 8.5% of 2,496 patients with an initial non-colonized carriage pattern who were not identified as converting in the first 6 months became positive for MRSA. After adjusting for the number of swabs performed, number of long term care admissions, days of inpatient care, and co-morbidities, the predictors of conversion from never to a colonized pattern included age greater than 70 years, long term care admission in the previous year, and antibiotic use in the follow-up period (MRSA-active as well as other) (
|
Odds Ratio | 95% CI | P value |
|
0.66 | (0.54;0.80) | <.001 |
|
1.10 | (1.06;1.14) | <.001 |
|
2.64 | (1.73;4.04) | <.001 |
|
1.20 | (0.98;1.46) | .08 |
|
1.35 | (1.11;1.65) | .003 |
|
1.57 | (1.19;2.07) | .002 |
Applying the 4 clinical criteria to the 16,425 patients who were not colonized at baseline identified 5,850 patients who comprised a low risk for conversion group (
Number of nasal MRSA screens avoided by application of the initial carriage pattern and clinical criteria identifying patients at low risk for conversion.
The outcomes of death and infection (MRSA and MSSA) were evaluated based on the 30 day carriage pattern as well as the dynamic carriage pattern over the study period. Death occurred in 1500 patients (8.4% of the study cohort), with an increasing rate with higher carriage burden at the 30-day baseline; 7.6% of never colonized, 15.0% of intermittently colonized, and 18.6% of always colonized patients died during follow-up. The mean follow-up time was 17.5, 16.5, and 17.2 months and the mean time to death from initial screening was 4.7, 4.5, and 4.9 months, respectively, in the never, intermittent and always colonized groups. Compared to non-colonized patients, those with intermittent or always colonized carriage patterns had a greater than 2-fold increased risk of death (HR 2.09; 95% CI 1.73;2.53 and HR 2.58; 95% CI 2.18;3.05, respectively) (
The time to (a) death, (b) MRSA infection, and (c) MSSA infection in patients who had a non-colonized, intermittently colonized, or always colonized nasal carriage pattern at 30 days. The risk of death and of MRSA infection, adjusted for age, number of screening tests in the follow-up period, number of acute care and long term care admissions, total acute care and long term care hospital days, diabetes, renal disease, HIV infection, decubitus ulcer, eczema, and antibiotic exposure, was significantly higher among patients with a colonized compared to non-colonized nasal carriage pattern.
There were 400 patients with an MRSA infection over the study follow-up period, with 221 (1.36%), 73 (9.48%), and 106 (13.43%) occurring in the non, intermittent, and always colonized 30-day carriage pattern groups. The 400 MRSA infections in unique patients occurred in blood (14.8%), bone (2.8%), abscess/fluid (26.5%), respiratory (38%), skin and soft tissue (16.5%), and other sites (1.3%). The mean follow-up time was 17.5, 15.2, and 15.6 months and the mean time to infection was 8.4, 5.0, and 6.1 months respectively among never, intermittently, and always colonized groups. The risk of infection was more than 10 fold greater among always colonized carriage pattern patients compared to the non-colonized carriage pattern (HR 10.89; 95% CI 8.6; 13.72) (
MSSA infections occurred in 536 patients and were most common in the intermittent MRSA carriage pattern group (6.3%) compared to the always (3.2%) and non-colonized (2.9%) groups. The risk of infection with MSSA was significantly higher in intermittently colonized as compared to non-colonized patients (HR 2.36; 95% CI 1.76;3.16) but not in patients who were persistently colonized with MRSA (
The VA MRSA Prevention Initiative, a bundle which includes both horizontal and vertical measures, has been temporally associated with significant reductions in hospital-associated MRSA infections
We employed a novel approach by identifying initial patterns of nasal MRSA carriage and then developing an algorithm to reduce screening based on eliminating further screens on patients already colonized as well as those at low risk for conversion. Given that the average prevalence of MRSA colonization on hospital admission to an acute care VA hospital is 13%, the non-colonized group represents the vast majority of hospitalized veterans
Patients who have a positive carriage pattern on the initial hospitalization continue to carry MRSA either always or intermittently over long periods of time
Future screening and infection prevention efforts should focus on the group of patients with an initial negative MRSA carriage pattern who do not meet criteria for being at low-risk for MRSA conversion. Based on our results, patients who are in long term care, greater than 70 years of age, have diabetes or antibiotic use are at highest risk of conversion to a MRSA colonized pattern over a 6 month period. These criteria identified 95% of patients who converted from non-colonized to colonized in our cohort and could be applied at the time of admission to help direct screening. Although antibiotic use may be more reliably captured in health care systems with robust and integrated electronic health records, a four point algorithm is less complex than what other studies have proposed and could be applied at the bedside
Our study demonstrates that establishing a carriage pattern of MRSA colonization is useful not only for developing targeted screening programs, but also for predicting the risk of death and MRSA infection. Previous studies have found that MRSA carriage is a risk factor for subsequent infection, both in the immediate period after acquisition of MRSA as well as among patients who are carriers for over a year
The main limitation of our study is the potential misclassification of patients who had fewer screening tests (less chance of converting carriage pattern). To minimize bias in our findings, our models are adjusted for the number of screening tests performed, and we also did not perform a time to event analysis of conversion since it greatly depended on repeat testing rather than a clinically apparent event such as death or infection. The findings are derived from a demographic representative of the Veteran population and thus may not translate fully to other populations. The performance of the reduced screening strategy based on carriage pattern and risk for conversion merits continued evaluation over time and in additional populations. We used an electronic definition of infection that may have misclassified some positive cultures in either direction. The infection rates generated using this electronic definition, however, are similar to those reported by another study that used a chart review-based infection definition
The greatest strength of our study compared to previous studies of infection risk in colonized patients is the ability to capture data from all inpatient and outpatient encounters at multiple institutions over a broad geographic region over a 3 year period and define patterns of nasal carriage rather than rely on a single assessment. The VA is the largest integrated health care system in the country with a robust electronic health record system, enhancing the detection of events occurring over the continuum of care, including all VA acute care hospitals in the region, long-term care facilities, and the outpatient setting. Our data also have the advantage of representing the real-practice setting, without experimental interventions that may or may not be feasible within current health care delivery systems. Also, we were not reliant on ICD-9 coding for MRSA infections, which has been shown to be a poor measure in VA-based studies
The MRSA nasal carriage pattern based on at least 2 nasal PCR screens within 30 days in acutely hospitalized patients is a strong predictor of ongoing carriage (and non-carriage), MRSA infection, and all-cause mortality. Our findings have significant implications for future policy considerations regarding the highly successful VA MRSA Prevention Initiative, providing evidence that one of the most costly components, universal active surveillance, could be targeted towards the group of patients most likely to convert from negative to positive over a one year period. The effectiveness and optimal implementation of this strategy needs to be studied along with continued assessment of additional activities aimed at reducing subsequent infections.
We appreciate the assistance of David Burkhardt and Dr. Joseph Erdos with data acquisition, Galina Sokolovskaya with data analysis, the VA Boston Infection Prevention Team for dedication to MRSA prevention, and the VHA MDRO Program Office for critical review of the manuscript draft.
Dr. Kalpana Gupta had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.