Transfer from High-Acuity Long-Term Care Facilities Is Associated with Carriage of Klebsiella pneumoniae Carbapenemase–Producing Enterobacteriaceae: A Multihospital Study

Abstract

Objective.

To determine whether transfer from a long-term care facility (LTCF) is a risk factor for colonization with Klebsiella pneumoniae carbapenemase (KPC)–producing Enterobacteriaceae upon acute care hospital admission.

Design.

Microbiologic survey and nested case-control study.

Setting.

Four hospitals in a metropolitan area (Chicago) with an early KPC epidemic.

Patients.

Hospitalized adults.

Methods.

Patients transferred from LTCFs were matched 1 : 1 to patients admitted from the community by age ( ± 10 years), admitting clinical service, and admission date ( ± 2 weeks). Rectal swab specimens were collected within 3 days after admission and tested for KPC-producing Enterobacteriaceae. Demographic and clinical information was extracted from medical records.

Results.

One hundred eighty patients from LTCFs were matched to 180 community patients. KPC-producing Enterobacteriaceae colonization was detected in 15 (8.3%) of the LTCF patients and 0 (0%) of the community patients (P< .001). Prevalence of carriage differed by LTCF subtype: 2 of 135 (1.5%) patients from skilled nursing facilities without ventilator care (SNFs) were colonized upon admission, compared to 9 of 33 (27.3%) patients from skilled nursing facilities with ventilator care (VSNFs) and 4 of 12 (33.3%) patients from long-term acute care hospitals (LTACHs; P< .001). In a multivariable logistic regression model adjusted for a propensity score that predicted LTCF subtype, patients admitted from VSNFs or LTACHs had 7.0-fold greater odds of colonization (ie, odds ratio; 95% confidence interval, 1.3–42; P = .022) with KPC-producing Enterobacteriaceae than patients from an SNF.

Conclusions.

Patients admitted to acute care hospitals from high-acuity LTCFs (ie, VSNFs and LTACHs) were more likely to be colonized with KPC-producing Enterobacteriaceae than were patients admitted from the community. Identification of healthcare facilities with a high prevalence of colonized patients presents an opportunity for focused interventions that may aid regional control efforts.

Antimicrobial resistance among Enterobacteriaceae has emerged as a significant clinical and infection control problem. Resistance to carbapenem antibiotics is particularly worrisome because carbapenems have been the mainstay of therapy against extended-spectrum β-lactamase-producing bacteria, and there are few effective alternative antimicrobial treatments. The major mechanism of carbapenem resistance is direct hydrolysis by broad-spectrum β-lactamases known as carbapenemases; Klebsiella pneumoniae carbapenemase (KPC) is the most common of these enzymes.¹ Infections due to KPC-producing bacteria have been associated independently with increased mortality.^2-4

Since its initial description in 2001, KPC has spread globally.^5,6 In the United States, long-term care facilities (LTCFs) have been implicated in regional spread of KPC-producing bacteria.^7-10 Seventy-five percent of carbapenem-resistant K. pneumoniae isolates within a northeastern-Ohio healthcare system came from patients admitted to the system from LTCFs; 60.7% of these patients were transferred back to an LTCF upon hospital discharge.⁷ In Los Angeles County, 46% of 164 reported cases of carbapenem-resistant K. pneumoniae came from long-term acute care hospitals (LTACHs), which are healthcare facilities that care for patients who require continuing mechanical ventilation or complex wound care after transfer from a short-term acute care hospital,¹¹ and 38% of cases were reported from a skilled nursing facility.⁸ KPC outbreak investigations at hospitals in Chicago and West Virginia revealed that a high proportion of case patients (90% and 84%, respectively) had a recent LTCF stay and that all K. pneumoniae isolates were genetically related in each outbreak, suggesting interfacility spread.^9,10

Despite these studies, the precise risk of infection or colonization with KPC-producing bacteria that is attributable to prior LTCF stay is unknown. So we performed a microbiologic survey to determine the relative prevalence of KPC-producing Enterobacteriaceae carriage at the time of short-term acute care hospital (hereafter “hospital”) admission in LTCF and community patients in a region with an early KPC epidemic.¹² A nested case-control study was conducted in the LTCF cohort to further analyze the association of LTCF residence and colonization with KPC-producing Enterobacteriaceae.

METHODS

Setting

Study subjects were selected from patients admitted to 4 hospitals (hospitals A–D) in metropolitan Chicago, Illinois, from November 2010 to June 2011. Hospital A is a 676-bed tertiary care academic medical center, hospital B is a 264-bed private hospital, hospital C is a 645-bed private academic medical center, and hospital D is a 328-bed private teaching hospital. The study was reviewed by each hospital’s Institutional Review Board; it was approved with a requirement for written informed consent at hospital C and determined to be nonresearch at hospitals A, B, and D.

Study Participants

LTCF patients were identified prospectively and matched 1:1 to community patients. A sample size of 164 patients per group was chosen to achieve at least 80% power to detect a difference in prevalence of colonization with KPC-producing Enterobacteriaceae of 5% between LTCF patients and community patients, assuming 0.1% prevalence of colonization among community patients, with an α level of 0.05 in a 2-sided Fisher exact test for proportions. Patients with immediate prior residence in one of the following facility types were considered to be LTCF patients: skilled nursing facilities without ventilator care (SNFs), skilled nursing facilities with ventilator care (VSNFs), and LTACHs. At hospitals A–C, LTCF patients were selected from a list of all currently hospitalized patients at least 18 years old admitted in the previous 3 days who were transferred from an LTCF. Because a transfer list was not available at hospital D, admitting housestaff interviewed every patient admitted to medical, surgical, and intensive care unit (ICU) wards to determine immediate prior residence. Patients who were transferred from another hospital or emergency department were included in the study if the total time spent in a hospital (including the current facility) was no more than 3 days. Patient selection occurred on predetermined days based on investigator availability.

One community patient was matched to each LTCF patient by age (±10 years), admission date (±2 weeks), and admitting clinical service (medicine/neurology, surgery, or ICU). If more than one community patient met matching criteria, the patient whose admission date and time were closest to those of the LTCF patient was chosen. If no community patient met matching criteria, the LTCF patient was excluded from the study. No patient was included more than once.

Surveillance Cultures and Laboratory Methods

Selected patients were approached by an investigator, who obtained assent or consent for participation. Rectal swab specimens were obtained by inserting a sterile, dry polyester culture swab (BBL CultureSwab, Becton, Dickson) into the anal canal. Alternatively, stool was sampled for culture if visible on the perianal skin or in an ostomy or fecal incontinence bag. Specimens were plated within 4 hours of collection. If immediate plating was not possible, swabs were stored at 4°C for up to 3 days.

Specimens were screened for KPC-producing Enterobacteriaceae by an ertapenem disk method;¹³ bla_KPC was confirmed by polymerase chain reaction.¹⁴ Isolates that carried bla_KPC were identified to species and tested for antimicrobial susceptibility by the MicroScan Walkaway System (Siemens). Tigecycline and colistin minimum inhibitory concentrations (MICs) were determined by E test (bioMérieux). Isolates were considered susceptible to tigecycline and colistin if the MIC was no more than 2 μg/mL.¹⁵ Pulsed-field gel electrophoresis (PFGE)¹⁶ was performed on all bla_KPC-positive isolates. Strain relatedness was determined according to the criteria of Tenover et al.¹⁷

Medical Record Review

Upon enrollment, the following information was collected from each subject’s medical record: age; sex; contact isolation status; current Braden Scale score (a measure of decubitus ulcer risk);¹⁸ number of inpatient hospitalizations during the past 6 months; number of unique antibiotic exposures within the past 6 months; clinical comorbidities (cerebrovascular and cardiovascular disease, diabetes, chronic obstructive pulmonary disease, chronic kidney disease [CKD] or kidney failure, cirrhosis, dementia, malignancy, human immunodeficiency virus infection, history of solid organ transplantation, incontinence of urine or stool, or decubitus ulcer); presence of chronic indwelling devices such as a urinary catheter, gastrostomy tube, ostomy bag, or tracheostomy tube; history of infection with carbapenem-resistant Enterobacteriaceae (CRE); and in-hospital mortality. A unique antibiotic exposure was defined as receipt of 1 or more doses of a systemic antibiotic; repeat courses of the same antibiotic were counted only once. For LTCF patients, the name of the transferring facility and the length of stay prior to hospital admission were also recorded.

Statistical Analysis

Comparative analyses were performed with the χ² or Fisher exact test for categorical variables and the Student t test or Wilcoxon rank-sum test for continuous variables. A P value of less than .05 was considered significant.

In the nested case-control study of LTCF patients, relationships of the outcome (KPC carriage at the time of hospital admission), the primary exposure of interest (LTCF subtype, ie, SNF, VSNF, or LTACH), and other potential risk factors for colonization with KPC-producing Enterobacteriaceae were explored. Because there were few positive outcomes, and in order to adjust for multiple covariates, a propensity score¹⁹ was created to account for confounders of the association between LTCF subtype and KPC carriage. First, all 2-way interactions between covariates were analyzed; no interaction terms were identified that met a P< .1 standard. Next, bivariable tabular analysis was performed to identify potential confounders. Ten covariates were identified that met a P ≤ .2 standard: age, sex, Braden score, history of multidrug-resistant organism (MDRO) infection or carriage or Clostridium difficile infection, dementia, presence of a gastrostomy tube or indwelling urinary catheter, history of antibiotic receipt, CKD, cardiovascular disease, or diabetes. To generate a propensity score for LTCF subtype, these variables were analyzed by means of backward selection at a prespecified α (P< .1) for retaining terms, with LTCF subtype as the dependent variable. Three covariates were retained: presence of a gastrostomy tube, history of MDRO infection or carriage or C. difficile infection, and CKD. The propensity score was included as a covariate in a multivariable logistic regression model to estimate the association of LTCF subtype and colonization with KPC-producing Enterobacteriaceae upon hospital admission. All analyses were performed with SAS 9.1.3 or Stata/IC 12.1.

RESULTS

Of 442 patients meeting inclusion criteria, 360 (180 matched pairs) were enrolled: 138 from hospital A, 96 from hospital B, 84 from hospital C, and 42 from hospital D. The majority of nonparticipating patients (48 from an LTCF and 32 from the community) were excluded because they declined participation. Two additional LTCF patients were excluded because a matched community patient was not available.

The average number of days from admission to rectal surveillance culture was the same for LTCF and community patients, 1.3 ± 0.9 days (P = .99). Within matched pairs, LTCF patients were more likely than community patients to have clinical comorbidities, to be receiving antibiotics, to have 1 or more indwelling medical devices, or to have a history of MDRO infection or carriage or C. difficile infection (P< .001; Table 1).

TABLE 1.

Demographic and Clinical Characteristics of Patients Admitted to an Acute Care Hospital from an LTCF and Those Admitted from the Community

Characteristic	LTCF patients (n = 180)	Community patients (n = 180)	P value
Mean age ± SD, yearsa	71.6 ± 14	71.6 ± 14	.99
Clinical servicea
Medicine	258 (71.7)	258 (71.7)	.99
Surgery	20 (5.6)	20 (5.6)	.99
Intensive care	82 (22.8)	82 (22.8)	.99
Male sex	82 (45.6)	90 (50.0)	.40
Contact isolation	83 (46.1)	32 (17.8)	<.001
History of MDRO or CDI	79 (43.9)	31 (17.2)	<.001
History of CRE	6 (3.3)	1 (0.56)	.056
Dementia	63 (35.0)	21 (11.7)	<.001
Incontinence	115 (64.3)	43 (24.0)	<.001
Braden scoreb	14 ± 3.5	18 ± 3.3	<.001
Decubitus ulcer	86 (47.8)	18 (10.0)	<.001
Gastrostomy tube	58 (32.2)	4 (2.2)	<.001
Chronic indwelling urinary catheter	32 (17.8)	1 (0.56)	<.001
Tracheostomy	30 (16.7)	0 (0)	<.001
Current receipt of antibioticsc	139 (77.2)	97 (53.9)	<.001

NOTE. Data are no. (%) of patients, unless indicated otherwise. CDI, Clostridium difficile infection; CRE, carbapenem-resistant Enterobacteriaceae; LTCF, long-term care facility; MDRO, multidrug-resistant organism other than CRE.

^aMatching criterion.

^bBraden scores are missing for 4 community patients.

^cReceipt of antibiotics ≤24 hours prior to surveillance.

The prevalence of KPC carriage among LTCF patients was 8.3% (15/180), compared to a prevalence of 0% (0/180) among community patients (risk difference, 0.083 [95% confidence interval (CI), 0.043–0.12]; P< .001). The prevalence of KPC carriage among LTCF patients admitted to each of the 4 hospitals was similar: for hospital A, 4.3% (3/69); for hospital B, 10.4% (5/48); for hospital C, 11.9% (5/42); and for hospital D, 9.5% (2/21; P = .41). In contrast, the prevalence of KPC carriage among patients differed by LTCF subtype: for SNFs, 1.5% (2/135); for VSNFs, 27.3% (9/33); and for LTACHs, 33.3% (4/12; P< .001; Figure 1). The difference in KPC-carriage prevalence between community and SNF patients was not statistically significant (0% [0/180] vs 1.5% [2/135]; P = .18). Patients were transferred from 82 distinct LTCFs: 72 SNFs, 4 VSNFs, and 6 LTACHs. The 15 patients who carried KPC-producing Enterobacteriaceae were admitted from 7 different LTCFs: 1 SNF, 3 VSNFs, and 3 LTACHs.

FIGURE 1.

Average prevalence and 95% confidence limits of carriage of Klebsiella pneumoniae carbapenemase–producing Enterobacteriaceae among patients from specific long-term care facility (LTCF) subtypes, at the time of acute care hospital admission. SNF, skilled nursing facility without a ventilator unit; VSNF, skilled nursing facility with a ventilator unit; LTACH, long-term acute care hospital.

In the LTCF cohort, admission from a VSNF (odds ratio [OR], 24.9 [95% CI, 5.1–123]; P< .001), an LTACH (OR, 33.3 [95% CI, 5.3–210]; P< .001), or either a VSNF or an LTACH (OR, 27.0 [95% CI, 5.8–126]; P< .001) was associated with increased risk of KPC-producing Enterobacteriaceae carriage, compared to admission from an SNF (Table 2). Other clinical factors, including having a history of MDRO infection or carriage or infection with C. difficile, being in contact isolation, and having chronic indwelling devices, were also significantly associated with KPC-producing Enterobacteriaceae carriage in unadjusted analyses (Table 2). In multivariable analysis, after adjustment for propensity score that predicted LTCF subtype, patients admitted from a VSNF or an LTACH had a 7-fold higher OR (95% CI, 1.3–42, P = .022) of colonization with KPC-producing Enterobacteriaceae at the time of hospital admission, compared with patients from an SNF.

TABLE 2.

Bivariate Analysis of Risk Factors for Carriage of KPC-Producing Enterobacteriaceae at the Time of Acute Care Hospital Admission among Long-Term Care Facility Patients

Characteristic	KPC-positive (n = 15)	KPC-negative (n = 165)	P value	OR (95% CI)
SNF	2 (13.3)	133 (80.6)	Reference	Reference
VSNF	9 (60.0)	24 (14.6)	<.001	24.9 (5.1–123)
LTACH	4 (26.7)	8 (4.9)	<.001	33.5 (5.3–210)
VSNF or LTACH	13 (86.7)	32 (19.4)	<.001	27.0 (5.8–126)
Median length of stay in facility prior to acute care hospital transfer ± IQR, daysa	55 ± 121	210 ± 692	.11	Not applicable
Contact isolation	14 (93.3)	69 (41.8)	.001	19.5 (2.5–151.6)
History of MDRO or CDI	14 (93.3)	65 (40.0)	<.001	21.6 (2.8–167.8)
History of CRE	6 (40.0)	0 (0)	<.001	Not applicable
Male	9 (60.0)	73 (40.0)	.24	1.9 (.64–5.6)
Dementia	3 (20)	60 (36.4)	.26	0.44 (.12–1.6)
Incontinence	12 (80)	103 (63)	.18	2.4 (.64–8.7)
Braden score	10.8 ± 1.7	14.3 ± 3.4	<.001	Not applicable
Decubitus ulcers	11 (73.3)	75 (45.5)	.04	3.3 (1.0–10.8)
Gastrostomy tube	14 (93.3)	44 (26.7)	<.001	38.5 (4.9–301.4)
Tracheostomy	9 (60)	21 (12.7)	<.001	10.3 (3.3–31.8)
Chronic indwelling urinary catheter	9 (60)	23 (13.9)	.001	9.3 (3.0–28.5)
Current antibioticsb	15 (100)	124 (75.2)	.02	Not applicable
Prior antibioticsc	12 (80)	86 (52.1)	.04	3.7 (1.0–13.5)

NOTE. Data are no. (%) of patients, unless indicated otherwise. CDI, Clostridium difficile infection; CI, confidence interval; CRE, carbapenem-resistant Enterobacteriaceae; IQR, interquartile range; KPC, Klebsiella pneumoniae carbapenemase; LTACH, long-term acute care hospital; MDRO, multidrug-resistant organism other than carbapenem-resistant Enterobacteriaceae; OR, odds ratio; SNF, skilled nursing facility without a ventilator unit; VSNF, skilled nursing facility with ventilator unit.

^aValues are missing for 3 (20%) KPC-positive patients and 14 (8.5%) KPC-negative patients.

^bReceipt of antibiotics ≤24 hours prior to surveillance.

^cReceipt of antibiotics >24 hours prior to surveillance.

Only 6 (40%) of the 15 patients who carried KPC-producing Enterobacteriaceae were previously known to be infected or colonized with CRE. Fourteen (93%) of these 15 patients were already in contact isolation because of a history of MDRO infection or colonization at the time of KPC surveillance. Carriage of KPC-producing Enterobacteriaceae was not associated with increased in-hospital mortality (0 of the 15 KPC-positive patients died, vs 7 of the 158 KPC-negative patients; P = .99).

Seventeen KPC-producing bacterial isolates were recovered from 15 patients. Of these, 16 isolates (91.1%) were identified as K. pneumoniae and 1 was identified as Escherichia coli. All isolates were resistant to ertapenem, imipenem, third-generation cephalosporins, piperacillin-tazobactam, levofloxacin, trimethoprim-sulfamethoxazole, tobramycin, and amikacin. Two (12%) isolates were resistant to gentamicin; none was resistant to colistin or tigecycline. Fourteen K. pneumoniae isolates were available for PFGE analysis; pulsotyping of 13 isolates was successful. Pulsotypes of 11 of the 13 (85%) K. pneumoniae isolates were related to one another and to an isolate of multilocus sequence type 258 (ST 258). Both of the 2 remaining isolates were unique.

DISCUSSION

We performed a matched, microbiologic survey of patients who were admitted from an LTCF or from the community to 4 hospitals in metropolitan Chicago and found that immediate prior residence in an LTCF was associated with a significantly higher prevalence of KPC-producing Enterobacteriaceae carriage at the time of admission; 8.3% of LTCF patients were colonized with KPC-producing Enterobacteriaceae, compared to 0% of community patients. A nested case-control study that compared patients within the LTCF cohort who were colonized with KPC-producing Enterobacteriaceae to those who were not colonized demonstrated that the risk of carriage differed according to LTCF subtype, with admission from an LTACH or a VSNF (high-acuity LTCFs) being associated with 7-fold greater odds of KPC-producing Enterobacteriaceae colonization, compared to admission from an SNF, after propensity score adjustment. Although we identified a strong association between immediate prior residence in a high-acuity LTCF and KPC carriage, our study design did not allow us to directly attribute acquisition of KPC-producing Enterobacteriaceae colonization to any specific healthcare facility exposure.

Our results highlight the integral role of LTCFs in disseminating KPC-producing Enterobacteriaceae. Although LTCFs have been considered a reservoir for other MDROs, including methicillin-resistant Staphylococcus aureus,^20,21 vancomycin-resistant enterococcus,^21,22 and extended-spectrum β-lactamase-producing gram-negative bacteria,²³ LTCFs have only recently been identified as a significant source of CRE.^7-10 In the United States, prior information about the epidemiology of CRE in different patient populations has come from passive surveillance^7,8 or has been limited to single institutions.^24,25 Our study is unique in its description of KPC-producing Enterobacteriaceae colonization prevalence among patients in multiple hospitals within a large metropolitan area and its prospective assessment of the association of KPC-producing Enterobacteriaceae carriage with residence prior to admission.

Recently, Ben-David et al²⁶ used active surveillance to detect a 16.3% prevalence of carbapenem-resistant K. pneumoniae among LTCF residents in Israel. Our LTCF patient population differed from that of Ben-David et al in that ours comprised LTCF residents who were admitted to hospitals and therefore may represent a more ill group, compared to LTCF residents in general. Although direct comparisons of CRE prevalence cannot be made between the two studies, it is notable that similar to our results, the prevalence of CRE in Ben-David et al’s study varied widely, depending on the specific institution and type of ward, with rehabilitation wards having the lowest (3%) and skilled nursing wards the highest (25%) prevalence. Our study showed a disparity in KPC-producing Enterobacteriaceae carriage in patients from different types of LTCFs, with a prevalence of 1.5% among patients admitted from SNFs, compared to a prevalence of 33.3% among patients admitted from LTACHs.

Outbreaks of KPC-producing organisms within LTACHs have been described in Florida and Indiana,^27,28 but VSNFs have not previously been reported, to our knowledge, to be associated with CRE. This association is concerning because in many states, VSNFs are not subject to the same infection control accreditation requirements and may not have the same resources for infection prevention as LTACHs and hospitals. Although the Centers for Medicare and Medicaid Services require all LTCFs to establish and maintain an infection control program,²⁹ there is significant variability in practice. Surveys in Maryland and Michigan revealed that 50% or fewer SNFs employed full-time infection preventionists (IPs) and that only 44% of IPs at SNFs had basic infection control training, as compared to 95% of IPs at hospitals.^30-32 Although patient care needs vary widely among SNFs, VSNFs may care for a patient population that is similar to that of LTACHs and may represent an underrecognized reservoir of MDROs that may benefit from more rigorous infection control oversight.

Our study has limitations. First, the 4 hospitals included in the investigation were selected on the basis of professional relations among the investigators and may not have been representative of all hospitals in the region. Second, nearly 20% of potential study subjects declined participation, which may have introduced bias. Our findings are consistent, however, with recent reports that carriage of KPC-producing Enterobacteriaceae is currently sporadic among patients in hospitals in Chicago and more common in LTCFs, with epidemics in LTACHs.^10,12,28,33 Admission prevalence of KPC-producing Enterobacteriaceae carriage likely differs according to local epidemiology. Third, some community patients may have had distant prior contact with an LTCF. Our results suggest that these patients are uncommon or have a very low prevalence of KPC-producing Enterobacteriaceae carriage. Fourth, the reliability of retrospective data for antimicrobial and other exposures may have varied for patients from the community versus those from LTCFs. Finally, our statistical analysis of individual risk factors for colonization was limited because of the small number of KPC-positive patients. We therefore chose to focus on the main exposure of interest (LTCF subtype) and to create a propensity score to account for confounding due to other covariates in the multivariable logistic regression model.

More than 90% of patients were already in contact isolation for another reason when colonization with KPC-producing Enterobacteriaceae was identified (Table 2); active surveillance for KPC-producing Enterobacteriaceae changed the isolation status of only a few patients. This raises the question of whether active surveillance and isolation alone will be effective in controlling KPC-producing Enterobacteriaceae in this population. As cross-transmission appears to be a major driver of increasing KPC-producing Enterobacteriaceae prevalence in the Chicago region³³—11 of 13 (84.6%) K. pneumoniae isolates tested were genetically related—prevention of patient-to-patient spread should remain a priority. Knowledge of KPC colonization may be used to establish cohorts of both patients and staff who care for them at healthcare facilities.

In metropolitan Chicago, KPC-producing Enterobacteriaceae are concentrated in high-acuity LTCFs (LTACHs and VSNFs), with few cases in hospitals.¹² Coordinated, multi-center infection prevention efforts that focus on high-prevalence LTCFs and are coupled with improved communication of MDRO colonization status between facilities at the time of patient transfer have been shown to reduce the prevalence of MDROs—including CRE—in other regions.^34,35 These interventions and other control measures, such as reducing the bioburden of CRE on patients’ skin^36,37 or in their gastrointestinal tracts,^38,39 have the potential to halt wider regional dissemination of CRE. Future studies should evaluate the effectiveness of these control strategies and also investigate the role VSNFs play in the spread of CRE.