Sensitivity of Surveillance Testing for Multidrug-Resistant Gram-Negative Bacteria in the Intensive Care Unit

Abstract

We tested intensive care unit patients for colonization with multidrug-resistant Gram-negative bacilli (MDR GNB) and compared the results with those of concurrent clinical cultures. The sensitivity of the surveillance test for detecting MDR GNB was 58.8% (95% confidence interval, 48.6 to 68.5%). Among 133 patients with positive surveillance tests, 61% had no prior clinical culture with MDR GNB.

TEXT

There is little evidence regarding the utility of inpatient surveillance for multidrug-resistant Gram-negative bacilli (MDR GNB) as it pertains to infection control. This study aimed to (i) determine the sensitivity of surveillance testing for MDR GNB in the intensive care unit (ICU), (ii) identify the factors (e.g., antibiotic use) that affect the sensitivity of surveillance testing for MDR GNB, and (iii) determine the undetected ratio (i.e., the proportion of patients with a positive surveillance test for MDR GNB who did not have a prior positive clinical culture for MDR GNB) (1), which is a measure of how often the surveillance test provides new information regarding a patient's colonization status.

This study took place in the adult ICUs of four affiliated hospitals from 3 August 2009 through 31 July 2013. Approval was granted by the NorthShore University HealthSystem institutional review board. Once a month, each ICU patient was tested for MDR GNB colonization with three swabs taken from the upper respiratory tract (throat or endotracheal tube), the axilla, and the perirectal area. Samples were collected with double-headed culture swabs with liquid Amies medium (BD, Sparks, MD). The swabs were then collectively inoculated onto vancomycin, amphotericin B, ceftazidime, and clindamycin (VACC) agar (Remel, Lenexa, KS). After 24 h of inoculation, colonies resembling GNB were subcultured on MacConkey agar. Colonies then underwent identification via the Vitek 2 system (bioMérieux, Durham, NC). Susceptibility testing was performed using the Kirby-Bauer disk diffusion method.

An organism was defined as MDR if it was susceptible to ≤2 tested antibiotic classes. Pseudomonas aeruginosa and Acinetobacter baumannii isolates were also considered MDR if they were resistant to meropenem. Patients who were identified as being colonized with MDR GNB were placed in contact isolation to prevent transmission.

Some patients for whom surveillance testing was done had incidentally undergone clinical cultures (i.e., cultures obtained owing to a clinical indication) at a facility within our health care network around the time of the surveillance test, and those clinical cultures had been positive for MDR GNR; the test's sensitivity was determined using data from these patients.

The sensitivity of the surveillance test was defined as the proportion of all patients known to be colonized with MDR GNR (denominator) in whom surveillance testing demonstrated these organisms (numerator). Patients were known to be colonized (denominator population) when a clinical culture was positive for MDR GNR in the period from 30 days before to 30 days after the surveillance test. The undetected ratio was determined as follows: among patients who had a positive surveillance test, the proportion of those who had no previous corresponding positive clinical culture for MDR GNB was calculated.

Over the study period, 3,197 surveillance tests were performed on 2,164 patients. In 102 instances, a patient who had a positive clinical culture for MDR GNB had undergone surveillance testing within 30 days of the clinical culture. The sensitivity of the surveillance test for detecting any MDR GNB (not necessarily the same organism that grew in clinical culture) was 58.8% (95% confidence interval [CI], 48.6 to 68.5%) (60/102). The sensitivity of the surveillance test for detecting the same species of MDR GNB that grew in clinical culture was 43.1% (95% CI, 33.4 to 53.3%) (44/102). Table 1 lists the surveillance test sensitivities based on the organism that grew in clinical culture.

TABLE 1.

Sensitivities of surveillance test

MDR organisma	No. of patients detected with indicated organism/total no.	Sensitivity (%)	95% CI (%)
Acinetobacter baumannii	4/8	50.0	15.7–84.3
Enterobacter species	4/5	80.0	28.4–99.5
Escherichia coli	11/30	36.7	19.9–56.1
Klebsiella species	8/12	66.7	34.9–90.1
Proteus mirabilis	1/9	11.1	0.3–48.2
Pseudomonas aeruginosa	16/37	43.2	27.1–60.5

^aAmong organisms with ≥5 positive clinical cultures (P = 0.08).

Receipt of at least one dose of a fluoroquinolone was associated with a significant decrease in the sensitivity of the surveillance test (49.0% versus 68.6%, P = 0.04), but receipt of at least one dose of piperacillin-tazobactam was associated with an increased sensitivity (67.6% versus 41.2%, P = 0.01) (Table 2). Other classes of antibiotics were not significantly associated with test sensitivity.

TABLE 2.

Associations between antibiotic exposure and sensitivity of the surveillance test

Antibiotic	Sensitivity among patients who received the indicated antibiotic (no. of patients with positive test result/total no. [%])	Sensitivity among patients who did not receive the indicated antibiotic (no. of patients with positive test result/total no. [%])	P value
Aminoglycosides	16/22 (72.7)	44/80 (55.0)	0.13
Antipseudomonal cephalosporins	16/29 (55.2)	44/73 (60.3)	0.63
Carbapenems	18/34 (52.9)	42/68 (61.8)	0.39
Fluoroquinolones	25/51 (49.0)	35/51 (68.6)	0.04
Piperacillin-tazobactam	46/68 (67.6)	14/34 (41.2)	0.01
Tigecycline	6/11 (54.5)	54/91 (59.3)	0.76
Trimethoprim-sulfamethoxazole	5/8 (62.5)	55/94 (58.5)	0.82

Over the 4-year study period, 6.1% (133/2,164) of the patients had a positive surveillance test. Of these patients, 50% (67/133) had no positive clinical culture for MDR GNB, and 11% (14/133) had a positive clinical culture for MDR GNB only after the surveillance test was performed. Thus, the undetected ratio was 61% (81/133).

This study demonstrates that routine surveillance testing for MDR GNB in the ICU is fairly sensitive. Nearly 60% of the surveillance tests performed within 30 days of a positive clinical culture for MDR GNB were positive. The true sensitivity of the surveillance test is likely even higher. In the study, the sensitivity was determined by using the gold standard of a positive clinical culture within 30 days of the surveillance test. Ideally, had the information been available, we would have used a positive clinical culture collected on the same day as the surveillance test. Some patients may not have actually been colonized or infected with MDR GNB at the time of the test, meaning that our false-negative rate might be inflated. If so, then the calculated sensitivity is an underestimate. Thus, the results of this study represent a “floor” estimate of sensitivity.

In addition to being fairly sensitive, surveillance testing for MDR GNB in the ICU revealed a relatively high undetected ratio; more than 60% of the patients had no previous positive clinical culture for MDR GNB.

This study has limitations. Patients may have been colonized with MDR GNB at body sites other than those sampled during the surveillance test. Furthermore, MDR GNB were identified phenotypically. This precludes definite conclusions about surveillance test sensitivity for organisms with specific mechanisms of resistance (e.g., specific carbapenemases) (2). Another limitation is that our data are based on culture testing. Others have reported that using PCR-based surveillance for MDR GNB enhances the sensitivity of surveillance testing (2, 3).

This work suggests that routine surveillance has a fair sensitivity for identifying patients harboring MDR GNB. Given that the majority of colonized patients would have gone undetected without this testing, our findings suggest that such a surveillance method may be a beneficial component of an MDR GNB infection control program in the ICU. Further work is needed to evaluate the impact of MDR GNB detection on transmission and the cost-effectiveness of such a surveillance strategy (4, 5).